Sunday, January 20, 2008

AMD/Intel Q4 2007 Earnings – Where Are We?

AMD's recent performance has been a study in disappointment. Earlier, SSE 5 and microbuffered memory had looked so promising for 2009. Now, it appears that these may not arrive even in 2010. Consecutive quarterly losses, low clocks, and the TLB bug have added to AMD's pain. In contrast Intel has rebuilt its revenues, upgraded C2D to the G0 stepping and delivered the 4-way capable Caneland platform. Intel has had some difficulty with 45nm Penryn but this hasn't mattered much with the G0 65nm quad cores available.

Sometimes things are a bit counter-intuitive. For example, I've seen many assume that AMD's fortunes began to boom in 2003 with the introduction of K8. In reality, 2003 was worse for AMD than 2002. 2004 was better but AMD suffered from volume limitations due to the larger K8 die. AMD's volume share stayed at about 16% from 2002 all through 2004. It wasn't until 2005, fully two years after the introduction of K8, that AMD began making strides in the market. We see the same pattern repeated by Intel where 2006 was far worse than 2005 in spite of the introduction of C2D. To understand just how well Intel is doing we need to compare with 2005, not 2006. The numbers show that Intel is doing well but hasn't quite recovered to what it had in 2005:

2005:
Processor Earnings – $28.1 Billion
4th Quarter Cash and Short Term Investments - $12.8 Billion
Average Gross Margin – 59.3%

2007:
Processor Earnings – $25.9 Billion
4th Quarter Cash and Short Term Investments - $12.8 Billion
Average Gross Margin – 51.9%

2008 Outlook:
Predicted Average Gross Margin - 57%

Intel's Outlook makes it clear that they do not expect to return to the Gross Margins of 2005. In fact, 57% doesn't even match the 58% Average Gross Margin of 2004. That point is quite puzzling. The common reasoning has been that as Intel switches to 45nm their costs will drop dramatically. This should easily be enough to boost the Gross Margin up from the current 58% to above the 2005 levels. But, Intel says this isn't going to happen and Intel's prediction of an Average Gross Margin of 52% for 2007 was right on mark. Apparently, Intel either expects that ASP's will drop or costs will rise enough to counter the decrease in costs from 45nm. This again goes against common reasoning which assumes that Intel's higher clock speeds insulate them from pricing pressure from AMD. This further goes against Intel's own statements of reorganization which were supposed to result in another round of layoffs in 2008. I think we now have to assume that Intel's reorganization has halted halfway through and that nothing further will be done. This however disagrees sharply with the inclination by many to label Intel as “lean”. Intel cannot be lean with only half of the reorg in place; Intel still has some love handles.

If we look at the facts instead of simply talking from our gut reaction this does make sense. AMD had gained a lot of server share in 2006. However, Intel took most of this share back at the end of 2006 dropping AMD's share by 40%. AMD's server share average for 2007 has been only 14.2% compared to the 23.5% average for 2006. Intel has held onto this share due both to the value of C2D dual core and having a monopoly on quad core. Unfortunately, the problem with being on top is that you can only go down. And, Intel now faces substantial quad core server pressure from AMD. AMD moved some 130K Barcelona server chips in 2007 and this will only increase. Secondly, AMD has substantially undercut Intel in terms of lower clocked quad core offerings. These lower clocked quads are also much more immune to Intel's lower 45nm power draw which is more dramatic as we increase clock. According to AMD, the high volume range for servers is 2.3Ghz. If this is true then Intel is not at all immune to server pricing pressure and we know that servers are Intel's highest ASP. I would expect that AMD will move back up to its 2006 server share in 2008. We also know that AMD's Griffin/Puma mobile platform will be out soon and this should put pressure on Intel's mobile segment which is the second highest ASP. It then begins to make some sense when we realize that the segment where Intel will still dominate will be desktop which is the lowest ASP. Even so, AMD did report some increase in desktop ASP in Q4.

The talk about an AMD takeover or bankruptcy is never ending. A takeover is technically and financially possible except that I can't think of a single company that wants to get into the frontline processor business and slug it out with Intel. Motorola was the last challenger but they divested their processor division as Freescale in 2004. Some might point to VIA. Well, VIA's parent company could pull off the finances but they've never had a frontline processor. Although Cyrix was initially competitive in 1995, Cyrix's position had slipped a generation by the time they were acquired by VIA in 1999. Secondly, VIA never manufactured Cyrix processors; they instead chose the Centaur architecture from IDT which was never frontline. Presumably if VIA had wanted to be competitive they would have worked on a new generation of the Cyrix design which would have been released around 2002 or 2003. This never happened so VIA as buyer is very unlikely.

Some have suggested nVidia or Samsung as a buyer but neither of these companies has any experience with processors. Samsung does memory products which is a long way from the logic intensive design of cpu's. These don't overlap at all in terms of manufacturing or design methodology. NVidia does chipsets and graphics and nVidia would be an obvious monopoly conflict due to AMD's ATI holdings. No doubt someone will mention IBM. However, IBM could never take over AMD's markets since IBM's use of AMD server processors would make it a competitor with its own customers. This would mean giving up the most profitable market segment of AMD's line and would probably impair volume with customers like HP, Gateway, and Dell that also produce servers. In other words, IBM would lose its current R&D support from AMD (the largest of all its partners), lose the technical support that AMD provides to Chartered, and lose AMD support for APM. The loss of volume would also make AMD processors cost more. Thus, IBM would turn a supporting partner and Intel hedge into a money losing niche processor division. Economically, this makes no sense at all. So, an AMD purchase is probably somewhat less likely than, say, striking oil in your backyard.

AMD managed to increase its revenues every quarter in 2007 as well as its Gross Margins. AMD finally managed to trim the quarterly loss to $164 Million. Now, I have to be honest and mention that AMD's current value per share of stock is worse than it was at the end of 2003. Adjusting for the change in number of shares AMD was comfortably higher by about $1.2 Billion but the $1.6 Billion goodwill charge has now brought this about $400 Million under. This is cause for concern if AMD's loss increases again in Q1. Supposedly AMD will break even in Q2. AMD's outlook though is oddly different from Intel's. Whereas Intel predicts a stagnant Gross Margin at a lower value than even 2004, AMD predicts that the current level of 44% will hold during the first half of 2008 but increase to 50% in the second half of 2008. So, why would this be? First of all, by Q3 AMD should finally have at least 2.8Ghz chips. This would cover nearly the entire range since it looks like Intel won't go higher than 3.16Ghz. There is also the question of 45nm from AMD. Statements made by AMD during the Earnings Call include:

Ruiz
“look forward to being able to ramp 45 nanometer aggressively in the second half of this year.”

Meyer
“Followed on in the second half of the year by 45 nanometer.”

Meyer
“we’ve got internal samples of our 45 nanometer microprocessors, we’re putting them through their paces currently and we’re on track to, the plans we talked about in the past which is to start our ramp in the first half of this year and ship revenue product in the second half of this year.”

I have seen some attempt to spin these statements to mean that AMD will begin producing 45nm in the second half with actual delivery in 2009. This, however, matches neither what AMD said nor common sense. Someone might mistakenly take Ruiz's comment to mean initial production however this is completely countered by Meyer's statement. Secondly, AMD first produced Brisbane samples in Q2 2006 and delivered product in Q4. If AMD has samples in Q1 then product delivery should by Q3. So, the most reasonable assumption is that AMD will begin production in Q2 and deliver some small amount in Q3. This amount is unlikely to have any effect on revenues but the Q4 volume should be more significant. Likewise, I doubt Intel's volume of Nehalem in Q4 will have any effect on revenues. So, adding in AMD's mobile and higher clocked quad cores in Q3 along with some 45nm in Q4 I could see an increase in Gross Margin in the second half. Finally, this begins to make sense. Right now, AMD overlaps the highest volume range and should increase its overlap during 2008. This would put pricing pressure on Intel with most likely some loss of ASP and this would offset the 45nm savings. On other hand, AMD's ASP's are very low so it can only move up. It also makes sense because 50% Gross Margin would still be significantly below Intel's 57%. This would be consistent with lower ASP's for AMD and higher costs due to a lower ratio of 45nm production. Some might also have noticed that AMD has scaled back its expected 2008 volume from 100 million units to 80-90 million units. This is undoubtedly due to the scale back in the FAB 38 ramp. This in turn has undoubtedly reduced AMD's target of 30% share for 2008.

I guess the bottom line is that as long as AMD can reach break even in Q2 it should be able to start gaining asset value again. And, it looks like AMD will finally get back to the processor revenues that it had in 2006. Judging from the current unit volume though I would say that AMD intends to start gaining again, especially in servers and Intends to get back to what it had in 2006. The really surprising thing is a comparison with 2001. AMD's volume share had been about 16% but this increased to 20% in 2001. With the release of the Northwood P4 and AMD's delay in getting to 130nm, AMD's share tumbled in early 2002. AMD left 2002 with roughly same 16% volume share that it had in 2000. It stayed at this level all through 2004 then creeped up to 18% in 2005. AMD's volume gain in 2006 was much more dramatic rising to 23%. If the 2002 pattern had been followed (as many predicted) then AMD should have lost this share again in 2007. When AMD's volume share tumbled in Q1 2007 many of these armchair experts gave themselves big pats on the back. However, they couldn't have been more wrong. AMD's volume share promptly rebounded and AMD has held onto its 2006 average for the last three quarters. What these self styled experts failed to take into consideration was that when the crash occured in 2002 AMD had only been at its high point for two quarters whereas when the crash came in 2007 it followed five quarters of good volume. In other words, the two are opposites. The high in 2001 was temporary whereas the dip in early 2007 was temporary. According to IDC AMD lost 0.4% unit share and is now at 23.1%. This is about equal to AMD's 2006 average. Overall, I expect AMD to gain back its 2006 server share and go above its previous mobile share while holding onto and increasing its current desktop share. This would mean an increase in unit share from the current 23% to perhaps 26% in 2008. This seems doable but falls quite a bit short of AMD's previous target of 30%.

Intel in 2008 should finally pull ahead of its previous processor revenue high of 2005. It's Gross Margins may not reach what they were but they will still be quite good and I'm sure Intel will continue increasing its cash and doing stock buybacks. By any account this should be good performance and a reasonable rate of growth. Pricing pressure in the second half of 2008 as AMD moves up above 2.6Ghz should produce some good values for buyers. We can look forward to seeing how well K10 scales and whether or not 45nm Shanghai produces any change in speed or any reduction in power draw. We should also be getting reviews of Nehalem in Q4.

103 comments:

Scientia from AMDZone said...

One thing I could mention is the timeline for 45nm. AMD began 65nm testing in Q2 2006 and delivered 65nm in late Q4 2006. Their goal was to begin 45nm in 18 months or 6 quarters after 65nm. The point of this faster schedule was to reduce Intel's lead which was fully 1 year with 65nm. An 18 month schedule would reduce Intel's process lead to 6 months.

AMD began testing 45nm in Q4 2007 so this would match the 6 quarter goal. However, AMD is not expecting to begin actual production until Q2 2008 which is 7 quarters after 65nm production in Q3 2006.

So, you can look at this either positively or negatively. On the negative side AMD has slipped its 45nm schedule by one quarter while on the positive side AMD has gained one quarter on Intel reducing Intel's process lead to 3 quarters.

Scientia from AMDZone said...

chuckula

I'm trying very hard to take your comments seriously but you have so many mistakes in what think I said that I'm left wondering what blog you've been reading.

Intel 45nm is not yet generally available well after the supposed launch. A quick look at NewEgg shows only 3 items currently available: E5440, E5450, and QX9650. No mobile.

There is no doubt that Intel had trouble with 45nm. They had originally planned to have two FABs producing 45nm in Q4 but so far all the 45nm chips are from D1D. Secondly, I don't have any idea why you seem to think that 45nm is Intel's Holy Grail. The performance is only slightly better than the 65nm chips and the power draw is just a little better than the G0 stepping. The main advantage is supposed to be cost but with a Gross Margin of 58% in Q4 Intel is doing fine with 65nm. It can afford a little slack with 45nm.

I'm still scratching my head over your quad core comment, "Intel's quad-core shipments are so small that they don't even count!"

Intel has been shipping quad cores since 2006. I'm sure they've shipped a considerable number by now. AMD did not have quad cores at all until Barcelona so I can't imagine why anyone would bother comparing numbers.

You also make claims about how many units of 45nm Intel shipped. I haven't seen this given anywhere. If I were to guess I would estimate somewhere around 3X-4X what AMD shipped of K10.

howling2929 said...

Do not count Samsung out yet. Remember, they were the second supplier of alphas in the long gone DEC Times. Thety are manufacturing the ARM Guts of the iPhone, and they want a second avenue of revenue away from the (comodity) memory market.

Aguia said...

Scientia,

Again excellent analysis, your Samsung IBM points are perfect.
I was already putting Samsung out because they didn’t have x86 license.
IBM has but won’t use because of your major points. Besides IBM have gotten rid of the Mobile/Desktop line but I can’t see do the same with servers since it’s their profit market.

AMD won’t break even in Q1 or Q2 because the quarters are weak. PC components are so low I can’t see they rising the revenue share.
Unless the rumors that Ati is going to release the 3890, 3860, 3690, … parts where ATI could charge more.
PS: the 3890, … rumors are the basically the same parts with more clock speed.
3890 850Mhz…, 3860 720Mhz, …
If this “new” products allow margins going up, then maybe who knows.

enumae said...

Scientia
...and the power draw is just a little better than the G0 stepping...


XbitLabs

TechReport

A 25-30% decrease in power consumption under load is not "a little better".

Dick R. Pimmel said...

Don't count out Google so quickly. They can easily afford to swallow AMD and they need INSANE amounts of computing power. Bying AMD could come them cheaper in the long run than always buying CPUs. My prediction: 50 years from now, noone will know what computers are, but everyone will have a gomputer on his desk.

Real said...

2005:
4th Quarter Cash - $12.8 Billion

2007:
4th Quarter Cash - $7.3 Billion


First, it makes no sense to look at cash for a free cash flow positive business. Second, you need to look at cash + short term investments (including trading assets) and you will see that the real numbers are:
Q4'05 = $12,8B
Q4'07 = $15.4B

So, Intel is accumulating cash currently, but as I said, this is meaningless. Intel's cash position is mainly dictated by whether or not they do stock buy-back, which is a financial decision and is completely decoupled from profitability, which you seem to be implying. For AMD, cash position is very important as they are free cash flow negative so it is useful in determining when they will require additional funding.

Second, the gross margin guidance is pretty disturbing. You have to back up and realize that Intel has jettisoned low margin businesses with significant revenue - such as Flash. By ridding themselves of these boat anchors, margins should be much high - especially compared with 2005 when flash had significant revenue. I think what is being hidden is a drastic decline in the profitability of chipsets. I think the CPU business is healthy - margins there are much more dictated by ASP's, which are stable, than by cost. I think they real issue is on chipset ASP and unit cost, which is likely the true culprit for suppressing margins.

Scientia from AMDZone said...

chuckula

I can't publish your comments because you still can't seem to understand that you are not writing on a junior high school bathroom wall.

However, I checked and the E8400 has appeared at NewEgg since yesterday along with the E5410. This is well short of the items launched but it is incouraging. We are missing E8500, 8200, and 8190 as well as Q9550, 9450, and 9300. We are missing the Xeon X3360, 3350, 3320 and 3110. We'll see if the other items appear over the next couple of weeks.

Scientia from AMDZone said...

Dick

I hadn't considered a computing user like Google as a purchaser of AMD but I don't think they would be very likely. Google likes to keep its options open so that it can buy either Intel or AMD hardware, depending on what is the best deal.

Scientia from AMDZone said...

enumae

Yes, I'm going to have to agree with you. The 45nm E8200 at 2.66Ghz does draw 1/4 less power than the G0 stepping 65nm E6550 at 2.3Ghz. And, these seem to both be in the same 65Watt class. That is pretty impressive. So, yes, 45nm is going to be important for Intel for high density server applications. Intel could probably get in the lead if they were willing to scrap FB-DIMM. However, I think Nehalem needs FB-DIMM.

Erlindo said...

This is what "intel-fanboy" Ed Stroligo had to say about any AMD takeover:
quote:
This is a Patent Cross License Agreement between AMD and Intel. This allows AMD and Intel to use each other's patents and copyrights. It was signed in 2001, and will remain effective for at least another ten years.

Unless . . . .

If you like to read legal contracts, click on the link above, and look at clause 6.2 of the agreement.

If you don't, what clause 6.2 appears to say is that if AMD gets taken over or goes bankrupt, Intel has the right to end AMD's right to use Intel's patents and copyrights after sixty days notice.

This would seem to mean AMD couldn't make x86 processors anymore.


What an irony:
AMD was the one leading x86 design when intel wanted to kill it with their Itanium dinosaur.
We can all agree that x86 is intel's "unwanted" child after all. =)

Scientia from AMDZone said...

Real

Thanks for the correction. I had overlooked that because Intel used to report Cash and Short Term Investments as one item whereas today they are separate. Okay, I corrected the article so that both years include Short Term Investments.

Yes, unfortunately they don't give the profitability of chipsets separately. I wonder though because Intel chipsets and motherboards combined are only about 1/3 of the revenue of processors and we know they are less profitable.

I checked nVidia's earnings report and it looks like they make 30% profit on geForce GPU products while only making 9% profit on nForce MCP motherboard products. This still wouldn't seem to explain why Intel would see no growth in Gross Margin later in 2008 as more of its products become 45nm. I'm still thinking that reduced ASP is more likely.

Scientia from AMDZone said...

Erlindo

This has been talked about before. Supposedly AMD's license to make X86 derivative processors cannot be transferred to a new buyer. National Semiconductor had such a license but sold Cyrix to VIA. I don't think VIA has a license. VIA bought both Cyrix and Centaur but I don't think Centaur had a license either.

This is why some people keep mentioning IBM as a buyer since IBM still has a license. But, as I've already mentioned I think an IBM purchase would change AMD from a valuable partner to a money losing division.

I looked at the rumor about a failed nVidia takeover but I couldn't figure out how a company with $2.5 Billion in assets could raise the $4.1 Billion necessary to buy AMD.

howling

Samsung has the money to buy AMD with no trouble. However, the current ARM is nowhere near a frontline processor. The best ARMs today are 2 full generations behind. This would be in the K7 or P4 Northwood class. Frankly I can't imagine though why Samsung would want to buy a company that is losing money. Secondly there is the issue of Intel. A Samsung purchase of AMD would sour relations between Samsung and Intel and my guess is that Intel products are Samsung's largest revenue base and Samsung is only about half of Intel's size. I doubt they are looking for that fight.

Aguia said...

Scientia,

Good article:
Process Technology Advancements

That might show why Intel is ramping 45nm slower than a turtle. Intel might be having issues in manufacturing 45nm CPUs.
I wonder how much time will take the 45nm CPUs OC to 4.0Ghz to melt...


enumae,
It’s very interesting that Intel while having great power consuming decrease doesn’t bother to change the TDP of those parts and keeps all of them at 65W TDP. Some of them could have 45W TDP easily.
Also the temperature doesn’t improve much.

Pop Catalin Sever said...

IBM has all the reasons in the world to buy AMD, opterons don't compete with Power CPU's, they are not in the same market segment. Some of the reasons why IBM should buy AMD:

- a big slice of the x86-x64 server market share
- a whole new line of chips starting from embedded to desktop to server and graphics chips for mobile desktop and embedded
- IBM are system integrators, vendors and designers with strong presence in various markets ranging from consoles to PC's to servers and HPC.
- they currently produce CPU's for both XBox 360 and playstation 3, buying AMD will also give them the graphics segment for XBox
- the increased fab capacity will greatly help them justify their R&D costs.
- greatly expanded potential and business opportunities for the future.
- there won't be a better time to buy AMD than now. (and if AMD writes off more of it's debt next quarter they are making themselves a too good buy opportunity)
- they can afford it, without even denting their balance sheet too much.

If IBM buys AMD all they need to do is to invest in a superior .32 nm fabrication process, build the NY fab and they've opened themselves to a totally new range of opportunities for the following decade.

enumae said...

Scientia
I don't have volume share information yet so I can't completely nail down the current position.


IDC Unit Share

Scientia from AMDZone said...

enumae

Thanks.

On an overall unit basis, Intel earned 76.7% market share, a gain of 0.4%. AMD earned 23.1%, a loss of 0.4%. These shares are identical to the shares of 2Q07.

Scientia from AMDZone said...

Okay, with the volume share information I had to make a few tweaks to the article. For example, I think it is pretty clear that AMD will start gaining volume share again but I don't see any possible way that they can hit their previous target of 30% in 2008. I think they will fall well short of this with perhaps 26%.

Mo

I skipped your comment about NewEgg and a similar one from Enumae because you said the same thing that Chuckula said. My response to him should have covered yours as well (unless you just wanted credit for saying it too). Yes, E8400 is indeed readily available as opposed to say E8500 and E8200 which are also supposed to be available but aren't yet.

Scientia from AMDZone said...

pop

You are quite mistaken when you say that AMD's processors don't overlap with Power. IBM does use Power for its mainframes because of their higher reliability but the rest of the lineup does overlap. Again, IBM cannot buy AMD because they would lose almost all of AMD's current server market. There really is no way to get around this. Do you honestly think that Sun would want to buy AMD processors from IBM who is a direct competitor?

How about an IBM purchase of nNidia?

amw said...

I am trying to buy an Intel E8500 at the moment and they are in very short supply, in fact I cannot source one. E8200 and E8400 are ok to some extent, however I do take it as an example of issues with Intel 45nm production.

Are the E8XXX being still made at Oregon, or are these from the Arizona FAB?

Scientia from AMDZone said...

amw

You and Enumae have both asked about 45nm production at Intel. Here is the official Intel FAB 32 announcement from October 25, 2007.

I have had some people here argue with me and try to claim that D1D is a high volume FAB. It is not. D1D is capable of real production but only at a fraction of Intel's regular FABs. This is clearly stated by Intel in referring to FAB 32.

"Intel Opens First High-Volume 45nm Microprocessor Manufacturing Factory"

"Intel first produced 45nm processors in its Oregon development facility, called D1D, in January and is now moving into high-volume production with the opening of Fab 32."

The fastest that Intel could get Copy Exact running on FAB 32 would be 30 days later. This would be November 24. Then it takes 90 days for these chips to be manufactured and distributed. This would be February 22 at the earliest.

However, a more likely scenario is 60 days from opening for one test batch then an additional 90 days for actual production and distribuation. This would be March 24. This is why I've said that all production so far is from D1D.

Aguia said...

How about an IBM purchase of nNidia?

I don’t know who said that, pop or scientia?

That scenario would be great.
I hate monopolies.
With that would leave:
INTEL vs AMD vs IBM vs VIA.

All the four can build complete CPU + Chipset + GPU.
VIA is weak but hey there is market for everyone.

Who the hell wants a single TV brand manufacture? Or DVD player?
More brands, more choices, better prices.

enumae said...

Scientia
The fastest that Intel could get Copy Exact running on FAB 32 would be 30 days later.....


"The first 45-nm production wafer rolled off the line several weeks ago." John Pemberton, the Fab 32 plant manager during a presentation officially opening the new facility.

Extreme Tech

I tried to post this on October 28, 2007, but I got no response from you.

enumae said...

Aguia
Could you be more specific with your point, because I sure didn’t understand...


Scientia is saying "This would be February 22 at the earliest."

Meaning, product availability, (to OEM's, the channel etc...), from FAB32 would be no sooner than February 22.

But, if production wafers were already finished prior to the opening of FAB32 you could then subtract...

1. 30 days for copy exact. Why would they start running wafers the were not the same as FAB D1D?

2. The amount of time to complete the wafer. That would only leave having to cut/test-bin/package.

I believe my terminology is incorrect (describing what happens after wafer completion), but the point should be clear. FAB32 production should already be with OEM's and the channel etc...

Scientia from AMDZone said...

enumae

That statement at Extreme Tech was clearly wrong. If that had been accurate then FAB 32 chips would have been available in November and they were not nor were they available in December.

I think the mistake was in his use of the word "production". I believe what he was trying to emphasize was that FAB 32 had already qualified its 45nm process. You notice that he never said anything about FAB 32 producing Penryn dies (which is the way you are interpreting it).

In other words, FAB 32 was ready to run 45nm production wafers. You might recall that when FAB 36 came online in early 2006 AMD only ran 90nm tests for the first quarter.

enumae said...

Scientia

In looking a little deeper I have found an interesting piece of information that would reinforce your position about products released in 2007 being from D1D.

Intel will ship its first 45 nanometer microprocessors on Nov. 12 from a smaller demonstration Fab in Oregon. The first chips from Fab 32 will be delivered in the first quarter of next year.

AZStarnet

Aguia said...

What a coincidence me talking about VIA and how "weak" they are, they just got upset :)

VIA releases a brand new core

It seems pretty good on paper specs. Very Core 2 Duo like.
We really need more choices.
I hope it’s fast enough to give them more market share.

Pop Catalin Sever said...

Rumours link IBM and AMD

IBM may look to form a deeper partnership with AMD, according to industry speculation yesterday.

AMD’s share price leapt by nearly 11 per cent following a Financial Times report claiming Big Blue’s microelectronics arm may strengthen ties with the chip maker. This could even include a merger, the article said, citing industry sources.


IBM must have read my comments on your blog and realize that there's plenty to gain from an AMD buy/merger :) .

But still let's not get hasty, those are only rumors so far.

enumae said...

Scientia

I think that when we have been discussing FAB32, we have forgotten about building an inventory for launch.

I believe the statement about "production wafers" is correct, and it would take some time to build an inventory for the Jan 20th launch just like they built an inventory for the server parts (Q3 conference call "not qualified for sale").

So, I think you may be correct in the sense that most if not all of the 45nm parts in 2007 were from D1D, but I believe that you are wrong when it comes to your comments regarding product availability from FAB32.

Would you think it fair to assume that D1D = Server, and FAB32 = Desktop and Mobile?

Aguia said...

More info about the new VIA processor:

Via launches Isaiah: 64-bit low-power, high-performance processors

VIA Isaiah Processor Architecture Preview

Jeach! said...

Scientia, I don't necessarily agree with your point of view on why IBM shouldn't buy AMD.

IBM could do so and create it's own opportunities... greater than what they would possibly loose.

But anyway, I just wanted to add a few points in regards to the rumors that IBM 'may' merge with AMD.

First, I would like for AMD to be taken out by IBM, but unfortunately, this is totally the wrong time!!! In two years maybe!!

AMD is currently rectifying various problems it created for itself. Subsequently this has reduced it's overall value (shares dropped). If AMD waits a bit, they will be more competitive and it's value will be higher (thus better for its shareholders).

Second, AMD stood alone against Intel in regards to the Antitrust situation. Now that things look positive for AMD why should a company such as IBM come along and benefit from this. If AMD walked alone during war time, they should walk alone for a while longer.

Third, AMD is worth a hell of a lot more than its current market cap! Since IBM has been massively profitable during the years, they could apply AMD's loses against their future revenues. Since AMD has about $2.5 Billion in accumulated losses in the last 5 years and barely any profit all of these will benefit IBM. This should easily add another $1 Billion to the price tag!

In all, I wouldn't vote for a sale to IBM unless IBM paid $9 Billion to $10 Billion (minimum). That's what I believe AMD is worth NOW (includes 2 year potential)!

Scientia from AMDZone said...

jeach

What opportunities are you talking about? IBM would lose Sun, Dell, and HP as server customers. IBM has been getting rid of hardware. It only keeps its IC division because it has to make Power cpu's.

Scientia from AMDZone said...

enumae

It takes at least a month for chips to get into circulation once they roll off the assembly line. Chips appearing at the end of Q1 could already be fabbed.

Pop Catalin Sever said...

BM has been getting rid of hardware

That's a non argument, IBM did sold some of its hardware business, but on the other hand it keep the Power division and it still has an R&D department and fabs.

"IBM would lose Sun, Dell, and HP as server customers"

Why would it loose them? Opteron is still a competitive CPU with market share, if Sun, Dell and HP stop selling opteron servers they are the ones that loose the market share. Even if this will happen (IBM loosing Sun, HP, and Dell) IBM will be the only ones left capable of selling competing x86 servers which is not such a bad thing.

But neither Sun, HP or Dell will give the market they have with opterons away that easy.

enumae said...

Scientia

Would you think it fair to assume that D1D = Server, and FAB32 = Desktop and Mobile?

Jeach! said...

I do acknowledge that IBM reinvented itself as a services company and that they got rid of their PC business which had become a low-margin division.

But on the other hand 'Pop Catalin Sever' is right in saying that where will HP, Dell and Sun go? Only with Intel? Might as well buy Intel stock right now cause their margins would hit near 80% (if not more).

They would have no choice but to keep buying chips from IBM/AMD.

Besides, look at it this way... the same way that desktops became a low-margin, commoditized business, so will laptops, workstations, enthusiast, etc.

The future is SoC... selling a single chip which has everything built into it, producing phone-size computers.

The future is (low-powered) parallelized servers (a.k.a virtualization)... how much native virtualization has IBM integrated into their Power/Cell chips? I hope they did otherwise IBM will loose server market share (and so will any server chips without this technology).

If IBM keeps the Opteron line for servers and Fusion for such hand-held PC gadgets, they could still sell Athlons to Dell, Sun and HP for laptops, desktops and workstations.

SoC is the future for Intel and AMD... regardless of anyones past success.

Whomever can release a SoC that has several x86-64 cores along with graphics that can handle wireless and full multimedia will dominate for the next decade.

I believe that IBM is seeing this right now and they want in.

Scientia from AMDZone said...

jeach

SoC has been around for 30 years; why do you think it is suddenly the future? I have the book "From Chips to Systems" and I don't think much has changed since it was written. SoC can be the cheapest but they are also the least flexible.

You may really like the iPhone but these type of sub-notebook devices have serious problems with tiny screens and no keyboard. I don't know about you but I can't even begin to imagine surfing the web on a palm device. My screen already seems tight at 1280 x 800. I can see making phone calls, snapping low resolution pictures, and playing mp3's on an iPhone but nothing else. A cell phone cam is not going to compete much with my digital camera with 12X zoom and 5 Mpixel resolution. The last I heard the iPhone demand was essentially filled in the first two weeks after they came out.

pop

IBM got rid of printers, laptops, and desktop systems.

You can't compare Opteron's position today under AMD with a future position under IBM. There is no doubt that if IBM owned AMD that they would use it to their advantage. They would make sure that they had to best chips, they would take advantage of the lower cost, and they would almost certainly be less receptive to other users. An erosion of market share would be inevitable.

Scientia from AMDZone said...

Source: No Truth To IBM-AMD Acquisition Rumors

Widespread talk of an IBM-Advanced Micro Devices merger, credited for giving AMD (NYSE:AMD)'s share price an upward kick of nearly 11 percent for a few hours Wednesday, is nothing but speculation run amok, a source close to AMD said Thursday.
"A deal could see IBM (NYSE:IBM)'s microelectronics division merge with AMD at some point, possibly in the near term," reported Mergemarket's David Zielenziger in the online version of the Financial Times Wednesday.

But the M&A rumor mill often churns out scenarios involving Sunnyvale, Calif.-based AMD, many citing Armonk, N.Y.-based IBM as a possible buyer, the source said. In this case, the source wondered whether published speculation by a financial analyst might have been misinterpreted and then passed on as fact by the "industry sources" cited by Zielenziger.

****************

Fancy that.

Jeach! said...

Scientia

I understand that SoC may have been around for the last 30 years.

Allow me to explain my point of view:

When the laptop was invented people were saying "why would I want to carry my computer around". Today, most people are replacing their home PC's with a laptop.

Eventually the desktop will eventually die and only laptops will be sold.

Right now there are some 4 to 5 Billion cell phones being sold around the world. Their processors vary in many different ways from cell phone to cell phone. So does their OS and the language used to produce their software.

Then there are smart devices like the iPhone which I believe uses a risk based Samsung processor.

In a few years when SoC from Intel and AMD are readily available for cheap it won't make any sense to spend billions of dollars on alternate OS's and software which only work on such and such processors.

Unit manufacturers will see a huge benefit of having x86 at their disposal. Many people have told me in the past (long ago) that RISC servers were perfectly ok and in no danger of CISC servers (x86). Well look at where RISC-based servers are now... still loosing market share and will eventually die!

The same thing will happen with these devices processors... x86 will come and eventually eat their lunch.

It's just that Intel and AMD have never really focused on that segment before (ok Intel tried and failed).

But now that BOTH companies are putting all their energies on SoC, which happens to be targeted for that market... well watch out!

The main idea here isn't the fact that they are SoC! The idea that it has just about everything within the chip will just be a HUGE bonus.

Intel's and AMD's business model have always been to have high ASP's! I believe that will eventually change to 'low ASP and high volume' business model.

If you could sell 5 Billion units at an ASP of just $10, you get $50 Billion rather than a $120 ASP for low volume.

Why? Because even processors are becoming commoditized and always trying to capture that 'premium' is too hard to always achieve. The 'Premium' business model will die and the 'High-Volume' will thrive.

On x86 SoC, you'll be able to run Windows Vista on your cell pone and use Outlook, etc. And I'm not talking about cut-down, embedded-type versions of those software either.

You could even install the latest PC game and play it without having any speed or hardware constraints.

Now I do agree with you about the screen size... actually that is the reason that I personally hate all kinds of hand held gadgets.

But I'm also assuming that you've heard of the breakthroughs they've made in LCD's such as LG's flexible LCD (the size of a sheet of paper):

here and
here

Xerox is also working on something different, but could be applied similarly.

My boss told me that he saw a demo of a product where the guy unrolls his monitor/screen (like a rolled up dollar bill). I'm not too sure what that's about, I'd have to ask him.

History has shown us that all hardware that isn't open-architecture and x86 compatible die in time.

RISC (servers), Rambus (memory), Apple (closed architecture), Intel (Itanium), etc, etc.

Aguia said...

VIA's New Centaur Designed Isaiah CPU Architecture

Well the chip runs Crysis which is amazing.
There will be SSE4 and dual core versions.

I think Intel and AMD can but the Silverstone and Bobcat in the trash.

I will try to find out more info about the CPU and will post the links here.

Aguia said...

VIA VIA VIA VIA VIA VIA VIA VIA VIA VIA VIA

The VIA Isaiah Architecture at extremetech

Isaiah revealed: VIA's new low-power architecture at arstechnica

A look at VIA's next-gen Isaiah x86 CPU architecture at techreport

Henry says Centaur will go to multiple cores if it needs that level of performance or if Intel convinces people they have to have it.

The VIA Isaiah Architecture G. Glenn Henry

Glenn Henry on the Isaiah architecture

Interesting interview to Glenn Henry, he talks about Cyrix and Transmeta and why the failed.

He also resorts into names calling to Intel and AMD so you should all read.
^Just kidding ;)

VIA VIA VIA VIA VIA VIA VIA VIA VIA VIA VIA

Scientia from AMDZone said...

Note: I'm going to do the best with your post that I can. It would be much easier if you could simply do a post without trolling. The editing however shouldn't effect any of the factual content.

*Edited* lex said:

Let me help you *****.

Go look at the table that "Aguira" referenced in a recent summary of silicon technology from IEDM says it all *****.

I've always told you guys that INTEL at any given node is better then others. The data there is clear. THus if and when AMD gets to 45nm it won't be close to INTEL 45nm. Now whether it materializes in a real product performance is a matter of design.

Go look at page 6. Mr Wang did a nice job of summarizing for me the claims Scientia has always asked me to prove.

Look agross a given process node for IBM using their "claimed" superior SOI and compare to the same process generation data from INTEL or other. Go down to the row that reports the drive current ( strength/speed ) of the transistor. Then next see what leakage ( power ) its measured at. YOu can see that at any given generation say the 130nm node or the 90nm node INTEL transistors are far faster and less leaky then anything else out there. Thus at any given time in princple at the transistor level INTEL has had between a 10-30% transistor advantage, and a 2x density for about 1 out of every 3 years as they transistion to a new node faster. That is a huge advantage in leakage, power, cost and output.

What has always puzzled me is how with this huge advantage technology advantage in the end the INTEL circuit designers couldn't produce far better CPUs then they do. Imagine what AMD designers could have done with transistors that where 10-30% faster and getting onto a new generation node a year earlier? I have to agree with you all that AMD has done a great job and INTEL designers have done a sorry ass job given what this table shows. Its like being given a better more powerful engine with better gas mileage too, but you end up with a slower and worst performing car many times.

Regardless of how crappy INTEL design may be you can't escape the fact that currently INTEL can't be beat. INTEL with their superior process, mystery designers, huge R&D, huge bankacount, huge factories and now doing this tick tock. Unless they return to their Itanium and Prescott ways they will beat AMD to a pulp. AMd is competing with their clever designers, poorer process, behind a year schedule, billion dollar debt, minimum money to put into R&D to catch up.

This year AMD may fix their TLB issue, fix their speed issue, get some early volume on 45nm. But lets not kid ourselves AMD doesn't have the money to invest for the next 3 years starting last year at the billion dollar level in the silicon R&D to catch INTEL 32nm let along get their 45nm to match INTEL. They will alway be behind a bit. IBM and the consortium has enough money to keep the 2 year cadence but not the need nor motivation ( money and greed ) to catch INTEL. Thus AMD will forever be behind on technology. I am not a designer but I will wager that AMD design is clever and desperate, but NOT cleverenough nor that much better then INTEL design to make up for this huge technoogy gap.

Now the orginal article there was a discussion about would anyone buy AMD for their design expertise, CPU brand. I see this similar to asking if you saw Burger King faltering would anyone choose to buy it to compete with McDonalds. Whom ever that buys it would have to be willing to throw billions of dollar a year at the investment in addition to the outright front up money to buy it. THey would have to see it as long-term being a strategic advantage to them to buy, and also that the billions to buy and invest represents a better growth opportunity then investing those tens of billions in some other market.

Don't know about you but Samsung is so vertically integrated that I don't see how throwing tens of billions at a slow/mature niche market makes sense. Today the hot things for samsun are their CE products, CE silicon ( they are in iphone and ipod ), why throw billions into a new market you know nothing competing against a huge and well positioned competitor. Forget the little detail about the x86 license too.

Now IBM, here is a once proud technology company; invetory of the DRAM, Harddrive and some many other things, but manufacture of little anymore. You really think that the top brass belives buying AMD and then having to invest billions every year in silicon and manufacturing is worht it? They have a verynice powerpc server line running, they get their silicon R&D subsidzied by AMD, Toshiba, Samsung and others. It is competitive for the past 10 years on a generation behind technology, they make a ton of money on software and services. Why would they change this model? Just for the ego to compete against INTEL?

Lastly, some say IBM microelectronics merge with AMD. NOPE won't happen. IBM sees it as a self paying bringing in money from partnerships, still gives the the chance to try and claim they are leaders ( they are NOT). The can't give up critical manufacturing of PowerPC core and mainframe CPUs mainframes. Those are huge cash cows thru software and service that IBM can't risk outsourcing to TSMC and others. Beside AMD doesn't have money to buy, nor does IBM have motivation to finance a good deal for AMD for all the above reasons.

How does Google make any sense, how does a company that is focused on information management leadership to sell you ads interested in hardware. All of googles competitive advantage lies in the software and infrastructure. Anyone who thinks because they have billions is a *****. Put yourself in the GOOGLE CEO's seat, what does he get when he buys AMD that helps grow his core business or his infrasructure, NOTHING!

People who speculate about AMd purchase do it from a fanciful dream that because they don't want to see AMD die as a competitor and innovator against INTEL that it makes business sense for someone else with money to buy. I'm sure if it was your money you wouldn't waste, but you are all ***** and won't every have that business decision in front of you. The biggest decision you guys make is whether you'll by that 500 dollar AMD or INTEL cpu.

Now will AMD survive? NO not if but only when. Even huge companies like TI have bowed out of investing in silicon and getting out. THe irony of why AMD can't get out is they compete with INTEL and silicon and manufacturing is the competitive advantage in this high volume business. THe only hope is that the CPU of the computer is no different the the CPU in your phone or iPOD or smartphone then the leading edge CPU doesn't need leading edge silicon to be competitive. But these days it takes money AMD doesn't have. THeir current market share simply won't give them the cash flow to invest back. They will go out of business. If INTEL hadn't had their Itanium, Prescott, Tejhas disaster AMD would already be gone. In some perverted way actually if INTEL is successfull in its latest strategy which is low cost IA everywhere then AMD may have a small chance to survive as performance matters less and less as you drive to CPUs with prices in the teens and volumes in the billions. Thus the more INTEL moves to grow the CPU business by moving to higher and higher volumes at cheaper prices their is a glimmer of hope that AMD may last a bit longer. But in the end AMD will die and their will be one, or we'll see AMD, Via and then the big one INTEL.

Oh as to margins drop, I think you forgot a few things, pricinge pressure with AMD rebound, general 45nm ramp ( I suspect are higher then INTEL will admit ), expecting lower margins from Silverthorne. What I find most interesting is how everyone buys AMD's story hook line and sinker about recovery and strong roadmap to continue to compete effectively. But again INTEL could always pull another Prescott.

SOC, the value of this is form factor and cost. For certian applications like iPHONE it is far simpler to put all this on a single chip, but you gotta sell it to apple for cheap. I got news for you, but analog, and RF doesn't scale well. Margins are low here and that will be the domain of the Samsungs and others. INTEL is too fat and spoiled by 50% margins that they won't ever survive here. SOC is here but isn't what will make ultrmobile devices work. I don't know about you people a phone is a phone, if I want to really watch a movie I want it on a 15" screen or something. Sure watching some stupid ass TV show or news on a 1" screen is useful but once we have real broadband we'll all want that big thing.

*****

Scientia from AMDZone said...

lex

What surprises me the most is how much you agree with me. You've stated almost word for word what I've been saying. I think about the only thing we disagree on is that AMD will go out of business. Okay, let me go over each point.

Let's start with the IEDM chart.

If I am reading the chart properly then we want a high ldsat and a low loff. If memory serves, ldsat is the drive current and the higher this is the faster the gate. And, loff would be the leakage when the gate is off. Now, if this is the case then Intel is not dominant for the entire chart but has been moving ahead at 65nm and 45nm.

130nm
NMOS
IBM ldsat 1100, loff 100
Intel ldsat 1300 (+18%), loff 100

PMOS
IBM ldsat 625, loff 100
Intel ldsat 660 (+6%), loff 100
Slightly more speed for Intel.

90nm
NMOS
IBM ldsat 940, loff 100
Intel ldsat 1450 (+54%), loff 400

PMOS
IBM ldsat 535, loff 100
Intel ldsat 800 (+50%), loff 400

This is quite a speed advantage, 50%, but these circuits would run hot due to the very high loff leakage. This also seems to be backed up by the fact that vdd is only 1.0 for IBM but 1.2 for Intel. This would seem to match what we saw with Prescott in 2004 and 2005.

65nm
NMOS
IBM ldsat 1259, loff 200
Intel ldsat 1210 (-4%), loff 100

PMOS
IBM ldsat 735, loff 200
Intel ldsat 710 (-3%), loff 100

Okay, no difference in speed but Intel has gotten loff down while it increased with IBM. Speed should be similar but Intel would draw less power.

45nm
NMOS
IBM ldsat 1364, loff 200
Intel ldsat 1360, loff 100

PMOS
IBM ldsat , loff
Intel ldsat 1070, loff 100

Unfortunately, the IBM data is absent but if I were to estimate I would say that Intel has about a 16% advantage plus lower power draw. It is notable that IBM's low power bulk 65nm process is identical to Intel's but comes 2 years later.

I should also mention that when power draw is at a limit then lower power draw can allow a faster speed.

Looking at the SRAM size we see that Intel was larger at 130nm, the same at 90nm, but smaller at both 65nm and 45nm.

Feel free to make corrections if I am not reading this correctly.

Okay, continuing with the points:

10%-30% faster transistors. I'm not seeing this. I wouldn't think a 16% gain in drive current would give 30% more speed. I would say you might get 20% at the top end due to power limits. Secondly, it wouldn't be a year earlier if Intel delivers a few chips in Q4 while AMD delivers a few in Q3 the following year; that would be 9 months.

Intel does seem to have an advantage at 45nm. And, AMD did take Bulldozer off the roadmap. So, this too would seem to give Intel an advantage if Nehalem is a good design. If Nehalem is a good design then AMD may have a hard time holding onto share in 2009.

R&D cost should not a problem. The cost is shared among IBM, AMD, Toshiba, and Sony.

But, we seem to agree that there is no obvious buyer for AMD. Neither Samsung, nor IBM, nor Google appear to be good candidates.

I do not agree that AMD will inevitably decline to VIA's level. VIA showed no interest in having a frontline processor when they purchased Cyrix. AMD has been fully pursuing this goal and has succeeded. You seem to be forgetting that in 2002 AMD barely appeared at all in server processors with Athlon MP. This was in sharp contrast to Intel which had been in servers since Pentium Pro. Athlon MP was greatly surpassed by Opteron. We seem to be seeing a similar situation with Turion and now Puma. I think AMD will continue to gain share during 2008. The big question is whether AMD can hold onto it during 2009.

Finally, the volume issue is a red herring. AMD is currently below the volume that it needs however it has never been closer. Nothing that Intel did in 2007 has kept AMD from holding onto a 23% volume share and AMD only needs about 26% to get the volume that it needs. If we compare this 23% volume with AMD's almost fixed 16% from 2000 to 2005 we can see that the last two years have been quite different. AMD has a shot at getting 26% in 2008.

Scientia from AMDZone said...

Oh, a small correction:

"It is notable that IBM's low power bulk 65nm process is identical to Intel's but comes 2 years later."

IBM's ldsat is the same as Intel's however Intel's loff power draw is only 0.1 compared to IBM's 1. Obviously Intel would have substantially less power draw at idle.

lex said...

Scientia, you like most are blinded by your desire to see why AMD can survive vs. taking the hard look at thd details to see how uncompetitive it is and realize the only reason AMD survives is because

1) INTEL choose to allow AMD. INTEL at anytime right now can choose to let margins drop to low 40's still make a couple billion instead of 8 billion and over the spand of 3 years gurantee AMD goes BK. NOTHING amd does can prevent that. Will INTEL do that? Of course not, it would have the US govt. all over them, but no law would be broken because INTEL would still be making money.. just chooosing to do so a little less.

2)Or INTEl doesn't execute Tick/Tock and pulls another Prescott. As you already admit IBM process is worst and late. Its impossible to field a competitive offering and make the money AMD needs to compete across the board and protect their 23% MS. AMD45nm will be inferior to INTEL's thats a given. Barcelona is a dud, AMD can only hope that follow-ons are fixed, but again they are hampered by 10%+ performance and 20-30% power handicapp. To AMD's advantage the INTEL designers seem to be clearly less efficient as Penrym isn't a large an improvement as I'd expect given the huge technology advantage. We'll see if Nehalem Tock is as good as the Core2 tock, but COre2 was easy, INTEL was copying AMD, lets see if INTEL designers got the balls to really innovate or note.

3) CPUs get commodized and anyone can build a low margin business using generic TSMC or trailing edge silicon.


In the end bottom line is AMD will wilt away unless INTEL screws up again. As I've said before that isn't a very solid business plan nor a team I'd be rooting for. Its like rooting for the Giants this week. YOu count on Brady, Moss, and the rest to be off to win. Sure there is always that small possibility and why you play the game as Eli can't like Dirk can't just get up and leave their team, but in the end over the long haul they WILL lose and wither away.

Giant said...

AMD 3870 X2 fragged by 15 month old 8800 GTX:

http://enthusiast.hardocp.com/article.html?art=MTQ1NCwxLCxoZW50aHVzaWFzdA==

We need to look at the performance in comparison to the closest competition for this video card, a single GeForce 8800 GTX. In every game except Half Life 2: Episode 2, the GeForce 8800 GTX delivered higher framerates. Keep in mind we are simply using a stock clocked GeForce 8800 GTX, so this doesn’t take into consideration a partner overclocked video card or the 8800 Ultra.

Scientia from AMDZone said...

lex

Thank you for doing a post that I didn't have to edit.

Actually, the difficulty is not caused by a blindness on my part but by the fact that you don't read what I actually write. You read about half and simply make up the other half. I'm not surprised that you would consider this half real and half imaginary scientia to be blind.

Let's look at what I actually say for a change. I've already stated that AMD's asset to stock ratio after taking the goodwill charge for ATI is lower than its low point in 2003. However, there are very few players left who understand leading edge IC manufacturing and even fewer who have Billions in extra cash to do the buyout plus have enough for operations and upgrades for the newly absorbed AMD. There is IBM and ... nobody. IBM is the only company large enough and technically saavy enough to absorb AMD. But, there is no indication that IBM wants to move in this direction. The fact is that AMD would be ripe for a buyout if there was a buyer.

The second issue is bankruptcy. AMD's outlook has improved slightly by reducing the losses in Q4. However, AMD needs to continue doing this. AMD can't take another year of $2 Billion in losses. I'm not sure they could stand another $1 Billion in losses. Again, this goes beyond anything in AMD's history. Just how thin can the ice get before it starts cracking under foot? I'm sure AMD's accounting department has sent out queries about additional financing and credit. They and other top AMD executives are probably the only people who actually know how bad it is and I doubt they will be sharing this information.

The third issue is whether or not Intel could take substantial share from AMD and push them back down to, say, 16% again. To do this Intel would have to be able to supply AMD's share and given Intel's current inventory status this can't happen soon. I would guess that all of Intel's older FABs (D1C, 11X, 12, and 24) are running at capacity and therefore can neither be expanded nor taken offline for upgrades. If this is the case then Intel would have to wait until the two new FABs, 32 and 28, ramp enough to give Intel some slack. I would assume this would mean that toward the end of 2008 Intel could begin a more aggressive capacity upgrade if that is what it wanted.

I realize that it might seem strange (or even AMD biased) to think that Intel wouldn't have enough spare capacity to take AMD's share right away. But, I seem to recall that Intel canceled some of its FAB projects in 2006 to reduce capital expenditures. And, it also moved chipset production to newer 300mm FABs to avoid the shortages that plagued it in 2005. Thus, becoming leaner while also ensuring chipset availability has meant that Intel has no spare capacity.

So, Lex, perhaps you can explain where the blindnesses are in these statements?

******

Now, let me point out some your blindspots. For example, K10 is a dud? K10 Phenom chips have a higher ASP than K8 chips. K10 server chips have also been sold but have been limited by the requirement to patch the server OS. So, perhaps this isn't as positive as it might be but is still a positive for AMD.

Secondly, have you looked at Intel's ASP's versus its Gross Margins? I understand that Intel is making Billions but you are going to have to toss out the myth that Intel has lower costs than AMD. A lot of Intel fans are claiming that Intel is now leaner than AMD. This is false. AMD's costs are still lower than Intel's. Intel is very much dependent on its higher ASP's to absorb its higher costs. Imagine that that ASP's for both AMD and Intel moved to an average of where they are now (midway between AMD and Intel). Which company do you think would fair better? Hint: it wouldn't be Intel.

Aguia said...

I don’t know how lex can say K10 is a dud.
K10 is:
-Native quad core.
-Improved server performance and $$$ margins.
-Top line performing server chip design.

Too bad toms didn’t compare Phenom with two cores enabled VS Athlon with two cores enabled. They did a comparison of one (K10) core VS one (K8) core:
AMD Phenom vs. Athlon Core

A dual core K10 would improve performance (would also cost more to manufacture) but would put the AMD in a different league against Intel at server, and at desktop would lower the difference between the two.

Scientia from AMDZone said...

Giant

As it says on the first page of my blog I will never knowingly say anything false about either AMD or Intel. I'm not sure you can make the same claim.

"AMD 3870 X2 fragged by 15 month old 8800 GTX"

Your use of the word "fragged" is quite dishonest. What HardOCP actually said was, "a video card that almost matches a GeForce 8800 GTX in performance". The phrase used to describe the 3870 X2 was "very high level of image quality in games".

And, I'm not sure why age would matter unless nVidia has something newer available at the same price.

"We need to look at the performance in comparison to the closest competition for this video card, a single GeForce 8800 GTX ... so this doesn’t take into consideration a partner overclocked video card or the 8800 Ultra."

The GTX and 3870 X2 both run about $450. The ultra is not comparable because it costs $150 - $200 more.

Now let's also look at what you left out:

During our evaluation processes AMD made some breakthroughs in Crysis with the 3870 X2 driver. We are using one of those driver builds they provided for us in our testing, which smoothed out performance compared to the driver build we had started with.

Now that patch version 1.1 is out, both AMD and NVIDIA can work on their drivers to improve multi-GPU performance, and we do expect performance increases to come.

As driver updates are released it is a safe bet performance will also be increasing, especially in Crysis.

******

The 2900 was not comparable to GTX and we all know that the power draw of dual SLI 2900's was approaching scary.

The 3870 dual card is a great improvement. The power draw is a bit higher than GTX but nothing like it was for 2900. And, the price of 3870 X2 is typically lower than 8800 GTX.

Azmount Aryl said...

Giant said...
AMD 3870 X2 fragged by 15 month old 8800 GTX


Your spreading FUD and lies. GeForce 8800 GTX and Ultra lose to Radeon 3870 X2 almost in every test. In many cases 8800 Ultra is as much as 25% slower than 3870 X2:

http://techreport.com/articles.x/13967/7
Obviously, AMD has captured the title of "fastest single graphics card" with the Radeon HD 3870 X2. There is much to like here. The X2 is generally faster than the GeForce 8800 Ultra, and it has HD video decode acceleration that Nvidia's older G80 GPU lacks. In all, the X2 looks to be a pretty good value in a single-card, high-end solution at $449.
-Scott Wasson

Scientia from AMDZone said...

azmount

Read the Editor's note on the last page of the HardOCP 3870 X2 review.

Editor's Note: You will see here today that our evaluation of the gaming performance produced by this video card does not track with some other big sites on the Web, and the simple fact is that those sites did not measure "gaming performance." Those sites measured frames per second performance in canned benchmarks and even some of them went as far as to use cut scenes from games to pull their data from. I have been part of this industry for years now and we are seeing now more than ever where real world gaming performance and gaming "benchmarks" are not coming to the same conclusions. Remember that when we evaluate video cards, we use them exactly the way you would use them. We play games on them and collect our data.

Another thing to think about is this. Do game developers want to provide built in benchmarks that show their games running slow? Or would the game developers rather put a game "benchmark" in that shows their game hauling ass? Do you think that slow benchmarks equal more sales?

The "3dfx way" of evaluating video cards is DEAD. It did have its time and place, but we are beyond that now. Any person using those methods to influence your video card purchase is likely irresponsible in doing so. You might even consider them liable. And I think that is going to come bubbling to the surface more and more as the industry matures.


I agree with the editor; pure synthetic testing is not the best way to judge the quality of game play.

Scientia from AMDZone said...

Another minor correction:

"the power draw of dual SLI 2900's was approaching scary"

Obviously, this would be CrossFire on ATI's 2900 rather than SLI as on nVidia's products.

Aguia said...

Remember that when we evaluate video cards, we use them exactly the way you would use them. We play games on them and collect our data.


That HardOPC guy is funny.
He does some of the weirdest reviews in the web and says he does the right way, the other do not.
He say we use them exactly the way you would use them which is bald lies.

Since when in some games I use the 1920x1200 resolution and with others the 2560x1600?!?!

Also most gamers have 17'/19' display screens which does 1024x768/1280x960 (4:3), the wide screen guys 1440x900 much lower than his real way to use them.

Even the 42" plasma/LCD guys no all have those huge resolutions.


Most web review tests are flop, they say they want to test GPUs at high resolutions because they don’t want (or want) the CPU or GPU to be bottle neck, however in the end the bottle neck is their unrealistic testing methods and choices.

In the end what they want to show is that:
- you really need one QX6700
- you really need a Geforce 8800 Ultra

Where:
- One low end dual core (X2 4000)/Pentium E2160 is more than enough.
- One 8600GT/2600XT is more than enough.

Randy Allen said...

In the end what they want to show is that:
- you really need one QX6700
- you really need a Geforce 8800 Ultra

Where:
- One low end dual core (X2 4000)/Pentium E2160 is more than enough.
- One 8600GT/2600XT is more than enough.


When you test high end hardware, you use high resolutions (1920x1200, 2560x1600). When testing mid-range cards like an 8600 GT or 2600 XT, resolutions like 1024x768 and 1280x1024 do make sense.

Aguia said...

When you test high end hardware, you use high resolutions (1920x1200, 2560x1600). When testing mid-range cards like an 8600 GT or 2600 XT, resolutions like 1024x768 and 1280x1024 do make sense.

No it does not make any sense to think you or him knows exactually what display you have and say you are going to play with those settings, that doesnt really make any sence.

And why high resolutions and not high AA settings like 8X AA or higher?
To think the buyers of 8800GT and 3870 have high resolution displays like 2560x1600 is plain nonsense.
Besides games have various detail settings that have nothing to do with display resolution.

Also take a look at this list at new egg:
Newegg 2560x1600 displays

How many do you see there?
Yes I didn’t believe either.

Mo said...

Penryn annihilates in the Dual Core field.

The E8200 (priced LOWER than X2 6400+) walks all over the 6400+.

There is not a single benchmark in exhibits test where penryn fell from top 3 spots.

and to top it all off, the penryn consumes more than 100W LESS than the AMD solution.

The power of Penryn is beyond what was expected.

Are you surprised SCI?

http://www.xbitlabs.com/articles/cpu/display/dualcore-shootout.html

Scientia from AMDZone said...

NewEgg

Q6600 2.4Ghz Kentsfield - $275

9600 2.3Ghz Phenom - $225

9500 2.2Ghz Phenom - $190

It looks like AMD has successfully undercut Intel's lowest quad price. I seem to recall that some here claimed that this wouldn't happen, that Intel would just lower its prices to match.

Scientia from AMDZone said...

With such a big gap:

Q6600 2.4Ghz $275
Q6700 2.66Ghz $540

and 2.4 and 2.6Ghz Phenoms due late Q1 my guess is that Q6700 Kentsfield prices will tumble. Late Q1/early Q2 should be a good time for anyone waiting to purchase a quad.

Scientia from AMDZone said...

Digitimes:

According to the article, E8000 series processors are in short supply. Apparently Intel is doing this on purpose because there are still a lot of the older E6000 series processors in inventory. But, apparently this isn't working:

Despite the CPUs being short of demand, sales of E6000 series models have not shown any improvement, the sources said.

Not an ideal situation for Intel.

Scientia from AMDZone said...

giant

You quoted a ton from the same article that mo referenced. The relevant section is this one:

So, the results of our today’s dual-core processor shoot-out indicate clearly that Intel processors win the “Best Buy” title in every single price segment. And it will remain this way until AMD reduces the prices on its Athlon 64 X2, which keep rapidly losing their appeal.

There is no doubt that the old K8 based dual cores are not much of a match for Intel's C2D cores. This is why AMD has not been using the FX designation on its top end chips. If AMD really thought they were competitive the 6400 would have an FX designation. I doubt this review was a surprise to anyone who reads this blog. Was this really such a surprise to you, Giant and Mo?

This is a lot like early 2003 when AMD's K7's were trailing Intel's P4's but K8 was only out at low clocks.

The situation may also change if they launch revised triple-core and dual-core processors on Phenom-like architecture, which have a chance of become more competitive against Core 2 Duo.

Again, I agree. When faster K10 based chips appear the prices should be more even and the triple cores may also make things more of a match.

Information Week:

Sources close to AMD tell me that the new B3 stepping, which corrects the problems via a silicon fix, is ahead of schedule and things are looking good.

It's about time AMD did something right.

Engineering samples of the new B3 silicon, which fixes the TLB bug in silicon so that there's no performance hit, already are in the hands of some key AMD OEMs.

Production B3 Barcelona parts will start rolling out toward the end of the first quarter (by the end of March), with volume shipments ramping up as planned in Q2.


This is a little sooner than was expected.

In addition, two new, faster Phenoms are coming in Q2. These are the 2.4-GHz Phenom X4 9700 and the 2.6-GHz Phenom X4 9900.

amazon has 9700 listed as usually ships in 3-6 weeks but does not have 9900 listed. Perhaps the 2.4Ghz version at least may be available in Q1.

Perhaps AMD could have 2.8Ghz parts by mid year.

enumae said...

Scientia
I seem to recall that some here claimed that this wouldn't happen, that Intel would just lower its prices to match.


I believe the comments were referring to Intel lowering the price of a quad-core to match that of the top Tri-core.

------------------------------

If the 2.4GHz and 2.6GHz models come out in April, they look to be facing an Intel price cut that is rumored to be on April 20th. Lowering the price of the Q6600 to $225 and the Q6700 at $266.

As it has been shown in reviews, the Q6600 fairs well against the 2.6GHz Phenom (slight advantage) and the Q6700 (clear advantage) at $266 will only hurt AMD. This is all without Intel's 45nm parts which only worsen the situation for AMD.

I think we may see a price cut from AMD in April or when the new Phenom's launch.

----------------------------------

Scientia
This is a little sooner than was expected.


Actually it's not. Here is what Derrick Meyer said in the Q4 conference call...

"So as Hector said we’ve got B3 material out of the Fab and we’re putting it through its paces internally and in a couple of weeks we’ll be providing engineering samples to our customers and follow that quickly with production samples later in the quarter. We’ll be working with our server customers very diligently to accelerate as best possible the availability of their systems and the marketplace, you’ll start to see that very late in Q1 and most of the systems early in Q2 and later into Q2."

It's not ahead of schedule, it's on schedule.

Scientia from AMDZone said...

enumae

"Lowering the price of the Q6600 to $225 and the Q6700 at $266."

These are being replace with the Q9300 and Q9450. Dropping prices of EOL items is no big deal. Supposedly the Q9300 that replaces the Q6600 will start at $270. This would not appear to match tri-core.

lex said...

Superior process with competent design is a much better situation then great design on a inferior and lagging technology node. Penrym will always have an inherent advantage of Barcelona and Phenom. Even in 2009 when AMD is on 45nm their process will still be inferior to INTEL's 45nm.

THe lagging process node is simply something to great to make up, and AMD is facing this situation for a good year out of every 4 years. Think about it between 65nm and 45nm you get esentially double the transistor density to work with for the same die size or you can shrink the die those are huge advantages that allow INTEL to throw more cache and be inefficient and still win. Is it no wonder AMD losing billions and will never catch up?

I still always wonder what Opteron design team could have done if they had INTEL's 65nm or 45nm instead of the dud they have.

AMD can't win nor can it keep up. All you dreamers can only hope Nehalem and follow-ons are more Prescott like then Core2 like.

Barcelona with TLB fix on B3 is LATE, LATE, LATE, damm its a B3 think of it A stepping plus dashes, B stepping plus dashes and finally getting it right on B3. Its still a huge die that they need to get all four cores at the same frequency for topbin parts.

Tick Tock Tick Tock by the time they got that squared away on 65nm and start focusing on 45nm INTEL will be there with Nehalem at 45nm and still have Penrym pulling from the bottom. AMD will not see much profit this year and no cash to fund the big spending required for 45nm HighK/Metal Gate. Also they've picked the wrong approach too!

enumae said...

Scientia
Dropping prices of EOL items is no big deal.


So when is the EOL of the Q6600 and the Q6700? All products have an EOL, but Intel, as far as I have read, has not announced an EOL for those products, so your comments is null.

This would not appear to match tri-core.

The prices for the Q6600 and the Q6700, after the projected price cut, were shown for a comparison against the Phenom prices, not as the actual product that would be competing against the Tri-Cores.

In regards to the Tri-core competition, I feel we could possibly see a new Quad-Core with less cache.

--------------------------------

No comment on what Derrick Meyer said in the Q4 conference call?

Scientia from AMDZone said...

lex

Let me see if I understand your reasoning. Intel was unable to take back share from AMD (except in servers) after 18 months of being ahead with C2D and having a monopoly on quad core. Yet, through some type of magic, a simple core refresh to Penryn is going to do more? This seems to be nothing more than wishful thinking on your part.

Now, if Nehalem is another big leap ahead (unlikely but possible) then AMD could be facing a difficult road in 2009. BTW, in spite of your glowing description of Intel's 45nm process the indications are that Intel's laptops will pull more power than AMD's Puma.

enumae

"So when is the EOL of the Q6600 and the Q6700? All products have an EOL, but Intel, as far as I have read, has not announced an EOL for those products"

Q9300 was supposed to arrive in this quarter. So, are you saying that Q9300 is late, the yields on 45nm are still poor, or that Intel has ramped up enough volume on 45nm yet to produce a mainstream product? It is common sense that 45nm Q9300 would be cheaper for Intel unless the yields are still bad. But, as I recall Intel has already stated that they have satisfactory yields at FAB 32.

Aguia said...

Enumae,

So when is the EOL of the Q6600 and the Q6700? All products have an EOL, but Intel, as far as I have read, has not announced an EOL for those products, so your comments is null.

When you have new products (Q9xxx) that are:
- faster
- use less power
- and cost about the same
Who the hell is going to buy the Q6xxx?

Tell me if by releasing a new product that’s better than the older one isn’t an EOL what the hell is?
New market? New segment?


I believe the comments were referring to Intel lowering the price of a quad-core to match that of the top Tri-core.

Intel didn’t even do it when AMD released the quad core parts how the hell is Intel going to lower the price of their quad to match the AMD tri core parts?

enumae said...

Looks like Q4 for AMD 45nm parts.

-HKEPC Link-

Erlindo said...

AMD to lower power usage of 45nm processors

Quote:
AMD recently notified partners that its 45nm-based processors will be launched on time by the second half of 2008. The company highlighted that the new processors will consume less power and have better performance, according to sources at motherboard makers.

AMD's 45nm processors are expected to see core frequencies greatly increased, while L3 cache sizes will grow from 2MB to 6MB. Power consumption of the cache will also see a significant drop, noted the sources.

The initial launch of AMD's 45nm processors will be AM2+ socket-based with memory support up to DDR2 1066. The company will launch AM3 socket-based processors with a built-in DDR2 and DDR3 memory controller in the first quarter of 2009, which will support memory up to DDR3 1333.


At last some good news from the AMD camp. =)

Jeach! said...

lex said...

Superior process with competent design is a much better situation then great design on a inferior and lagging technology node.


That is total BS! Put it this way; I was recently given an SMP Opteron server. I thought it would make a perfect little Linux server I needed to experiment on. But every time I go near it I can't get out of my mind that its 130nm processors in that machine... which bugs me :(

And that is my point! Process is only for geeks like you and me. At the end of the day, Ruiz is right... they sell packages of technology regardless the processes it was manufactured on.

Think about it between 65nm and 45nm you get esentially double the transistor density to work with for the same die size

Now is it really double? I thought it would have been 31% (45/65-1)?

or you can shrink the die those are huge advantages that allow INTEL to throw more cache and be inefficient and still win.

But if they would go by that logic, then they would keep eating their economies of scale! Moore's law would slow to a crawl (or stop)!

That isn't innovation and luckily Intel doesn't think that way!

Barcelona with TLB fix on B3 is LATE, LATE, LATE, damm its a B3 think of it A stepping plus dashes, B stepping plus dashes and finally getting it right on B3.

We'll all remember what you just said when Intel 'attempts' at releasing (on time) their new K10 copy-cat architecture.

AMD will not see much profit this year and no cash to fund the big spending required for 45nm HighK/Metal Gate.

Assuming AMD looses 'real money' in 2009 and runs out of cash. There will ALWAYS be someone there to back them up! Companies like that just don't turn over and die. What would that do for shareholders? My guess is NOTHING because AMD can't declare bankruptcy (I'm assuming this includes chapter 11 also) without loosing their x86 license. So they would get cash infusion before filing... so share holders are relatively safe! But only a guess!

Scientia from AMDZone said...

enumae

"Looks like Q4 for AMD 45nm parts."

HKEPC only talks about desktop units. The 45nm server chips should be Q3.

Scientia from AMDZone said...

enumae

If you want to keep talking about Q6600 then explain if you think Q9300 is delayed or if you think Intel will keep making Q6600 along with Q9300.

enumae said...

The Q9*** series of processors could be delayed, I don't work for Intel, but, their does seem to be something that is keeping Intel from launching the parts, be it a 4 layer motherboard issue, etc... I don't know.

-----------------------------------

In regards to the Q6600, I do believe that they will continue to make them just as they continued to make the E6600 for about an additional 7 months since the launch of the E6550. The EOL for the E6600 just happened in January 08.

People still have motherboards for these parts, so why stop offering a product to, while few, people who may like to upgrade to Quad-core while on an older chipset.

----------------------------------

Bottom line, it is all speculation unless you work for either AMD or Intel, but the problem I have with some of your post is that you make your statements as though they are fact and when asked for a source you have a tendency to ignore it.

An example of this is...

"However, from what I've seen it looks like Puma may actually have better battery life than Centrino."

This is from AMDZone, and I am still waiting to know what you have seen?

Scientia from AMDZone said...

Lex, I don't understand why you can't post without all the insults and trolling.

Edited: Lex said:

What logic or what BS from *****. You can censor this if you like *****

For those that don't know the move from 130nm to 90nm to 65nm to 45nm is a numerical description of the technology node. If you look more closely the scaling is ~ 0.70 per generation which if you do math 0.70x0.70 is 0.49 going from 65nm to 45nm. If you look at what IBM/AMD, or INTEL publish for the size of their SRAM it also is about 1/2 the size. *****

Note: what Lex is fumbling with but not explaining very well is that 65nm is a linear measurement. However, dies are two dimensional; they get smaller in both length and width. Therefore to get the proper relationship you compare (45 X 45) / (65 X 65) = 0.48. So the basic process shrink is half.

As for Moore's law if it wasn't for them geeks you'd still be using a 25 Meghz 80386. No industry has every driven cost and functionality like those silicon geeks. That has enabled the low cost electronics everywhere. *****.

I recalled in the Prescott days there were those that claimed absolute AMD superiority by most benchmarks. AMD as a result kept prices high, gouged the customers and made some money and grew a few points in MS.

Note: this is of course completely false. From 2003 when K8 was introduced until late 2005 AMD's volume share never rose above 16%. However, AMD's volume share increased by 50% in late 2005 and early 2006. AMD has held onto this volume share, about 23% for two years now.

Funny, that INTEL still retained dominate market share and made billions every quarter. Why didn't AMD capture more MS?

Note: again this is obviously wrong. Intel's processor revenues shrunk dramatically in 2006. Intel's revenue's in 2007 have almost caught up with the 2005 numbers.

Finally, the comment about "dominant market share" borders on psychotic. AMD has never had two fully ramped and operational FABs and this is what would have been required to reach 50% share.


Now lets spin forward 2 years. INTEL is ahead by a bit on performance, AMD has signficant installed based and track record. Their CPUs are marginally worst, a bit overpriced in may segments and of questionable value versus INTEL. Yet I get asked why INTEL didn't get all its old MS back and some.

Today INTEL is a generation ahead between Penrym and Nehalem. INTEL will have better cost structure with more transistor density and more performance for a good year. AMD simply can't compete. Hector makes lots of noise about Marketshare. Sure MS is important but if you get it by pumping twice as much silicon area and selling it at lower ASP and lose billions is that a viable business, yeah right!

Note: Lex is confused again. AMD cut its loss in half in Q4 apparently by pumping out the same outdated silicon and selling it at lower ASPs. It is also clear that AMD would have broken even if they had had a volume of $2 Billion. In other words, AMD's profits are related in part to volume.

I think a little about Chrysler with their clearly inferior cars across the board; Trucks, SUVs, sedans everything. Yet people buy them by the millions. Their marketshare didn't evaoporate overnight, nor did Honda, Toyota and others go up like crazy? Why is this a mystery to anyone.

Note: I'm confused about this comparison. My brother is actually a plant engineer at the Chrysler factory at Kokomo, IN. Chrysler's biggest problem was being bought by Daimler Benz who interferred with advertizing and product developement.

THe big difference? INTEL has made money and continues to invest positioning itself to lead by more thru superior technology, new designs and huge capacity.

Note: Intel cut back on some FAB expansion back in 2006 which is why today Intel does not have the capacity to dramatically take back share anytime soon. This is evident from Intel's inventory levels.

AMD doesn't have a strategy at all. AMD is more like Chrysler in that regards simply not big enough to compete across the board against the big company and puts out inferior products. Do you wonder why Toyota doesn't have 100% of the minivan market though the Sienna is far superior to the Chyrsler offering? There are people some driven by ***** they still buy the inferior product. It makes perfect sense why Sun, HP, IBM, Lenova and others buy a little AMD. It is pricing control in negotiation with INTEL. It isn't because AMD has some superior offering or value. I am sure if any company went exclusive INTEL they would not LOSe one % of MS.

Note: This is just a silly rant. Neither I nor anyone with any common sense would suggest that Intel's products are worthless. They are in fact quite good. But there is a big difference between being able to understand that Intel has some advantages and thinking that Intel has every advantage.

Look only at Apple to see that, the danger is if you are big and only have INTEL there is some supply risk that you negotiate from a pretty weak point. Apple is unique and leverages its relationship very well with INTEL.

Note: Apple isn't much of an example since it didn't get processors from AMD before it went with Intel. However, it is a certainty that if not for AMD, Apple's position with Intel would be far worse.

Don't put up this stupid ? about why doesn't AMD marketshare just dissapear, its obvious. The thing you can't understand is that yes B3 may fix AMD's short-term design issue, but AMD simply has a broken business model. It can't compete across the board or even at any segment of the CPU business.

Note: There is that rant again. Lex cannot seem to understand that AMD is in fact the second largest cpu manufacturer in the world. The highest that Chrysler ever got to was third behind both General Motors and Ford and was at one time fourth also behind Honda. Apparently Lex feels that AMD got to number two purely by accident.

AMD isn't like IBM where they can fund the CPU business and or hide a generation behind by getting huge revenue and profits from the software services side. AMD competes on chips along and against INTEL and needs to make that business stand on its own, and they can't

***** You are clever in how you choose to spin things and avoid others. Bottom line AMD business like I've said over and over, not sustainable, it will fail and they will fold as an indepdending company. Not if, but only when, and long before INTEL and before Moore's law hits the wall.

Note: Yes, Lex, you've repeated it many times and if repeition is a factor in how AMD performs then you may be right.

AndyW35 said...

As a slight aside, and I hope the change of topic to some extent is Ok, I bought an Intel E4600 a month or two ago and as it was retail came with the Intel HSF. Now I have finally got an Intel E8500 and that also comes with a heatsink fan. Here is a picture of the 2

http://www.zen141854.zen.co.uk/hsf.jpg

The E8500 is the smaller of the two. I think this shows how good the Intel 45nm process is, a 3.16GHz cpu using something with half the metal of the 65nm 2.4GHz.

Whilst testing the E8500 a lot of peoples are hitting walls around 4.6GHz and the way they overclock is interestingly different from the 65nm. You can bench 3D a lot closer to the 2D max value for instance.

bk said...

enumae
Bottom line, it is all speculation unless you work for either AMD or Intel, but the problem I have with some of your post is that you make your statements as though they are fact and when asked for a source you have a tendency to ignore it.

An example of this is...

"However, from what I've seen it looks like Puma may actually have better battery life than Centrino."


In Scientia's defense, not that he needs anyones help, what you have just used as an example is absurd. This is clearly an opinion and not a statement of fact. If Scientia only commented on facts, this blog would be very boring. So I want to hear Scientia's opinion even if there are not facts to back everything up and it is simply speculation. Which by the way I think it is fairly easy to decipher one from the other.

Scientia from AMDZone said...

Edited: lex said:

I will say again, to compete in leading edge high end CPUs without the subsidies of software like HP, IBM, SUN and others requires you compete on the performance of the chip and platform. To do this you need leading edge silicon, top end design and billions to throw at it year over year.

Why I say billions is that the leading edge CPU technology takes signficant design resoursce, validation, test, and then basic silicon. To miss on any and you have Barcelona like execution in schedule and Barcelona like performance and die size.

AMD doesn't have enough money to drive the silicon, do the design and all that stuff. Its silly purchase of ATI distracted the management even more and whats worst caused slips and distraction at ATI too. Now both parts aren't excuted, are distracted with integration. Here 1+1 is < 2 for sure. Plus AMD is loaded to the gills with debt.

Scientia from AMDZone said...

andy

Thank you for the picture. Your conclusion however seems to be incorrect. The two cpu's have the same power rating, 65 watts. The larger HSF in the picture looks to me like the stock HSF that Intel introduced back with Prescott. That HSF should be capable of handling a lot more than 65 watts (more like 130 watts). The second HSF appears to be matched.

Scientia from AMDZone said...

We had an odd situation with Turion versus Pentium M. Although Turion drew less power than Pentium M, the actual Centrino drew less than AMD mobile systems because the 3rd party mobile chipsets were not as good as Intel's.

Today, AMD finally has a proper mobile chipset so that is no longer an advantage for Intel. However, there are indications that Intel's Penryn will not be able to conserve power as well in spite of lower power draw from 45nm. In other words, it looks like Penryn will draw less at full power but more at more typical loads.

Scientia from AMDZone said...

lex

Intel's R&D Budget is about the same as it has been:

2006: $5.9 Billion
2007: $5.8 Billion
2008: $5.9 Billion (Intel's Outlook)

AMD's R&D Budget is probably going to be similar from 2007 to 2008:

2006: $1.2 Billion
2007: $1.8 Billion
2008: $1.9 Billion (projected from Q4 2007)

Obviously, AMD's budget increased by 50% in 2007 due to adding ATI. Okay, so Intel has about 3X AMD's R&D budget. The big question is just how bad this ratio is. Does Intel really spend 200% more on processor design than AMD?

These are estimates I came up with for the R&D portion of areas other than cpu and chipsets based on costs at places like Spansion for Flash, etc.

$1 Billion - Itanium R&D
$ 0.9 Billion - Flash and other R&D
$ 0.5 Billion - Motherboard R&D

First Estimated Total - $2.4 Billion

5.9 - 2.4 = $3.5 Billion, 84% advantage

The problem is that this would put Intel's chipsets and processor R&D at 84% more than AMD's. Since the other estimates of R&D are based on 1:1 ratios, a ratio of nearly 2:1 for X86 cpu's and chipsets would be unusual. For example, this would mean that Intel spends less on R&D for Itanium than AMD spends on processor R&D. This is probably not the case.

So, if I bump the initial estimate by 25% from $2.4 to $3 Billion we would probably be closer. This would leave $2.9 Billion for X86 processor and chipsets. Assuming a 1:3 ratio with chipsts with Intel as we saw with AMD we would have:

Intel chipsets $967 Million versus
AMD $633 Million, 53% advantage

Intel processors $1.93 Billion versus
AMD $1.27 Billion, 53% advantage

It wouldn't surprise me if Intel spends 50% more for X86 R&D than AMD. However, we can see that this is by no means "billions more" as lex suggested. It's about $660 Million more per year.

Scientia from AMDZone said...

Edited: lex said:

Lets get right down to the point.

Note: when people say that they are getting right to the point they don't usually ramble on for five paragraphs. I will try however to sort some real points out of the verbage.

Tell me flat out, do you think AMD has a sound and viable business plan that enables them to be competitive with INTEL.

It is clear that you have no real intention of seeking an answer. But, you are going to get one anyway. I don't know of any specific reason why AMD could not be competitive. I don't know of any specific reason why AMD could not be viable.

That means competitive both in terms of their stated goal of holding on to some given Marketshare percentage, competing across the board in Quadcore+, dual core servers, quad-core - dual core desktop/mobile, and chipsets, etc.

Exactly. I wouldn't consider AMD to be a success if they ended up being marginalized as happened to Motorola, ARM, MIPS, and VIA's processors. AMD needs to have a frontline processor and cover server, desktop, and mobile.

By viable I don't mean whether you see Barcelona at 2.8 GHZ, and AMD breakeven by end of the year. THose are really meaningless short term goals. I'm talking about an AMD that has put in place all the basic capability to meet the cadence that makes them competitive thru 32nm and INTEL's next Tick-Tock cycle.

This is an odd question because AMD looks better at 32nm with Bulldozer in 2010 than it does in 2009.

Do they or Don't they have this all aligned. DOn't tell me about their roadmap. I want you to tell us readers why is it in plain english that makes you think they can execute to the roadmap, that the IBM consoritum will deliver faster cadence and better silicon to compete against INTEL, how they will fund the billions in capital to put the manufacturing in place for 45 and 32nm and also the HighK/Metal gate transistion at 45nm they keep hinting about.

First of all, AMD should be fine with their current FAB 36 ramp 45nm. This however does mean that they've scaled back their capacity ramp to a volume share of 28% max by end of 2008; they had originally wanted 30%. We still don't know if AMD intends to do this on 45nm or wait until 32nm. However, you seemed confused because there is no additional cap ex cost for moving to high-K.

Your estimate of "billions in capital to put the manufacturing in place for 45 and 32nm" is way off. That is a fair statement for Intel. Intel's cap ex outlook is $5.2 Billion for 2008. AMD spends about $1.6 - $1.9 Billion each year for cap ex. However, it looks like that with FAB 36 built, the bump and test facility up and running, and FAB 30 only needing modifications that AMD may not spend as much in 2008 and 2009. This may be down to around $1 Billion. However, I haven't yet seen an official AMD outlook to confirm this.

The process technology should be fine since it is a group effort. If AMD had to depend on its resources alone then I would not bet on their keeping up with Intel. However, IBM is the only game in town; so it means either making the IBM partnership work or not keeping up. IBM has a big incentive to do this as well since Power needs to stay competitive with Itanium. Again, I can't think of any specific reason why IBM/AMD couldn't keep up with Intel's process technology.

I assume that what you are doing is projecting the present into the future. Unfortunately, this is just a guess and nothing more. There have been times in the past when Intel's process was behind. Specifically, Intel lagged by more than a year in moving to copper interconnects and Intel had problems with 90nm. That Intel's process is better at 45nm is no indication that it will be at 32nm.

Finally, there is the factor of the cost of upgrading. One of the things you've overlooked is that the same immersion scanners will be used at 32nm. This means that the money that AMD spends now on 45nm will save them money later at 32nm. This also means that the money that Intel saves now at 45nm will cost them more money later at 32nm. This makes AMD's cap ex outlook much better at 32nm.


I'm all eye's as to how you will tell me they can overcome one year of being behind on 45nm, and once they get on 45nm being another year behind on getting the leakage/power and performance benifits of HighK/Metal Gate.

Well, you are leaving out an awful lot with this simplistic statement. First of all, Intel carries almost nothing from 45nm to 32nm. This is because a different chemistry is required to prevent the deposition surface from reacting negatively with the immersion fluid. This is no minor hurdle.

Consequently, you have this backwards. AMD's experience with immersion at 45nm is directly applicable to 32nm whereas Intel's dry experience at 45nm is no benefit with immersion at 32nm.

Secondly, AMD has an obvious fall back position if 32nm is late since they can shift to 45nm with high-K and still see improvements. No, this does not guarantee that AMD can catch up in 2010 but it does show that AMD's position is not nearly as bad as you suggest.


You are in the end saying AMD can compete with inferior process and larger dies size and 1/4 the market share and make enough money that they can re-invest forever.

No, this is where I get annoyed when people simply invent my position and ignore what I've actually said in favor of weaker arguments that they can more easily refute.

An inferior process is not a benefit but how much of a negative is it? The degree of advantage is not measured in watts or heat or speed but in how this translates into money. For example, if a 5% advantage in process translated into a 30% advantage in money then that would be very bad but if a 20% advantage in process translated into a 5% advantage in money then that wouldn't so bad at all. I understand that you can recite endlessly Intel's process advantage in terms of speed, power draw, and heat but I don't think anyone can yet say how this effects the market.

Die is similar. Smaller die size is good assuming of course that yield is also good so that this translates into lower cost. I've noticed that you have avoided mentioning that Nehalem has a larger die size. You've also avoided mentioning that Intel's necessity of double patterning makes their process slower. Faster is also lower cost.

We can say however that AMD does not currently have enough market share at 23% with the current ASPs (and I doubt ASPs will change much over the next year). I think AMD needs at least 26% to start making money again and probably 28-30% to be in good shape. It is not currently possible for AMD to gain this much share because FAB 36 is running at capacity. It could be possible at the end of the year as FAB 36 continues to ramp and converts to 45nm. It becomes more likely in 2009 as FAB 38 ramps but this could be attenuated by Intel advantages over Shanghai.

Scientia from AMDZone said...

enumae

You can keep asking me over and over about my statements about AMD's Puma platform. However, I'm not sure what the point is. The early indications are that Puma may have longer battery life than Intel's mobile platform. However, like other early indications, suggestions, lab tests, and rumors, this could be wrong. Puma should be out soon and we can see how things actually compare.

Scientia from AMDZone said...

An interesting rumor I've seen lately is that AMD is rushing 45nm. I would not have expected this actually. I would have expected AMD to continue to 2.8 and probably 3.0Ghz on 65nm before Shanghai would catch up at 45nm. Recent rumors suggest that AMD may stop the 65nm parts at 2.6Ghz and jump directly to 45nm. I don't know how likely this would be.

Randy Allen said...

An interesting rumor I've seen lately is that AMD is rushing 45nm. I would not have expected this actually. I would have expected AMD to continue to 2.8 and probably 3.0Ghz on 65nm before Shanghai would catch up at 45nm. Recent rumors suggest that AMD may stop the 65nm parts at 2.6Ghz and jump directly to 45nm. I don't know how likely this would be.

According to AMD's own website, they plan on releasing up-to 3GHz Phenom:

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_15331_15332%5E15344,00.html

Q: Will AMD be introducing additional frequencies on the processor side in the near future?
A: In addition to the 2.2GHz and 2.3GHz AMD Phenom processors introduced on November 19, we plan to offer a 2.4GHz and 2.6GHz processor in Q1 2008 and AMD Phenom quad-core processors with expected frequencies up to 3GHz in the first half of 2008.

Scientia from AMDZone said...

randy

"According to AMD's own website, they plan on releasing up-to 3GHz Phenom"

I think you need to read AMD's statement (which predates Nov 2007) more carefully:

we plan to offer a 2.4GHz and 2.6GHz processor in Q1 2008

There won't be any 2.6Ghz processors in this quarter. So, there is no reason to think that AMD will have 3.0Ghz available in the first half of this year. Secondly, AMD makes no mention of whether these would be 65nm or 45nm.

Aguia said...

I don’t understand how you guys (Scientia included) expect AMD to release anything faster then 2.6GHz in the next months (if not year).

It’s pretty obvious that the engineer team has warned the marketing team that they couldn’t do anything faster than 2.6 due to TDP or clock problems.
So the marketing team has placed the model 9900 has the top at 2.6Ghz otherwise the 9400 (2.2Ghz) would be 9100 or 9200 leaving the 3.0Ghz for the 9900 model.

Unless you can explain me what’s higher than 9900 with 4 numbers?

Aguia said...

UPDATE:
At amdcompare.com with the TDP of 45W the fastest is clocked at 2.3Ghz so this is very inline with I said before:

2.3Ghz dual core with 45W TDP = 90W TDP quad (K8)
Add the “extra” stuff to K10 and there you have Phenom 9600 at 2.3Ghz with 95W TDP.

Aguia said...

According to amdcompare.com the fastest 65W CPU is clocked at 2.7 GHz. I don’t know what was in the AMD expectations when they saw those higher power consuming numbers in their first 65nm K8 processors.

Since Barcelona has 4 cores I assume 2X power consuming.
So the best case scenario for an quad core CPU (K8) at 2.7Ghz is with 130Watts TDP.
But since Barcelona besides the stuff inside K8 has more like L3 cache, more complex northbridge, higher clocked HT bus…
It’s easy to assume that adding all that this would lower the “available” GHz, so realistically 2.4 GHz is not that bad. Having quad core at 2.6 GHz with 130W at 65nm is a huge feat for AMD.

Scientia from AMDZone said...

Aguia

"According to amdcompare.com the fastest 65W CPU is clocked at 2.7 GHz."

Close. The actual fastest 65 watt Brisbane is the 2.8Ghz 5400+.

So, I don't think a 2.8Ghz quad Phenom would be so unlikely. I tend to agree that a 3.0Ghz quad would require a revision. The big question I have is whether AMD is really going to abandon 65nm at 2.6ghz and then how long it would take them to exceed that with 45nm.

I'm still thinking that AMD needs a 2.8Ghz 65nm Phenom in Q3. Historically, it has taken AMD a couple of steps to match previous speeds so I don't expect a 3.0Ghz 45nm quad in Q4.

Pop Catalin Sever said...

AMD can't fix the .65 process now, that's why it needs to drop it asap in favor of a better .45 nm process.

The thing is the .45 nm process still won't be better than the one Intel has, but it will close the gap nonetheless.

Considering the fact that Intell will release Nehalem I really don't see any way AMD can retake leadership till 2010 when they should roll out their .32 High-K based process along with possibly a new architecture.

But until then I'm really hopping that I'll be able to drop a Phenom in my AM2 motherboard, meaning AMD will solve it's compatibility issues, if not well... Intel's quads are better anyway...

Scientia from AMDZone said...

pop

I can't really say if it's AMD's process that is the problem. Indications I've seen lately suggest that a lot of AMD's quads have one core that is slower. This wouldn't have been an issue for Intel with its MCM packaging since they would have simply paired the faster dies with other faster dies.

Secondly, I've seen some indication that AMD may be running the voltage higher than it expected and this is why it is hitting the thermal limits early. I have no doubt that AMD will get these problems fixed but it is looking more and more like this can't happen any faster than 45nm would be available. So, this again brings up the question of whether AMD is going to fix the 65nm chips or just try to move quickly to 45nm. I guess we'll see.

Scientia from AMDZone said...

Fudzilla:

The sources very close to Intel have confirmed that Intel managed to ship over a million 45nm chips.

This number includes all quad cores, dual cores, and probably the bigger portion of the grand total were sold as server parts.


I had to laugh when I read this. The posters over at roborat's blog insist almost everyday that I have no idea what I'm talking about and that I just make up things to try to make AMD sound better. Hmmm.

When it was reported that AMD sold 400K K10's my only comment was that even if these sold for $200 apiece this would have only been a tiny fraction of AMD's Q4 revenue and therefore not really a signficant money maker yet.

On the other hand there were Intel fans that jumped on the 400K number as "proof" that AMD was doing badly. So, I stated that Intel probably moved a similar ratio of 45nm chips which would be about 1.2 million. From what Fuad is saying, it sounds like I got that right. I also said that only D1D was producing 45nm chips in 2007 and this too was attacked as incorrect. But again common sense would suggest that if Intel needed two FABs to produce just over 1 million chips then they must have as serious problem with yield. Not likely. So, again both my volume estimate for Intel and my statement that the 45nm chips only came from D1D was correct. I guess I can't be wrong all the time.

Aguia said...

Yes scientia I completely agree with your previous post.

Here are some of my thoughts that I posted at sharikou:

AMD says they shipped 400,000 quad cores this quarter, Intel sold 1.2 million LAST quarter and did 57% more business in the quad core market this quarter.

Since AMD has less 5X market share, doing basic math:

AMD
5 x 400.000 = 2.000.000

Intel
1.200.000 / 5 = 240.000

So by using an X size ratio Intel shipped less quad cores than AMD. AMD has a much bigger chip size, bad yields (rumor) and bugs in its design.
I wonder what would happen with a design without bugs and with good clock speeds (2.6Ghz)?

Scientia from AMDZone said...

aguia

"Since AMD has less 5X market share, doing basic math"

I'm sorry but I don't understand your math. AMD's average volume share for the last 2 years and its current volume share is 23%. This would give Intel 77%.

400,000 / 23 * 77 = 1,339

I was using 25% for AMD because this puts Intel at 3X AMD's volume and makes for some quick calculations. There is little point trying to get down to the last decimal point when we have a number like 400,000 which is almost certainly rounded up or down. Anything from 1.1 - 1.4 million chips for Intel would be reasonable and still at about the expected ratio.

Aguia said...

Right.
I forgot AMD recovered and retained its market share.

xbitlabs

Pop Catalin Sever said...

Hey Scientia you might wanna read
http://www.tomshardware.com/2008/02/14/amd_merge_or_not/page6.html
because it's one of the best articles about AMD's situation (from my point of view) in a long time.

Scientia from AMDZone said...

pop

Well, nVidia may have a market cap of $14 Billion but it only has about half of AMD's value in actual assets. nVidia would have to borrow a couple of Billion to buy AMD and apparently that is not going to happen.

Sure IBM could buy AMD but there would be a number of problems. For example, Theo mentions that, "all three consoles feature IBM processors, and the two leading consoles have ATI GPUs inside."

But apparently he doesn't see this as a monopoly issue which it is. He also doesn't seem to understand that IBM's purchase would severely hurt AMD's server market share. On top of this it would lose AMD as a contributing partner in process development. I don't see any way that an AMD purchase could prove profitable for IBM.

Aguia said...

Scientia could you calculate the "value" of AMD?

If they paid for Ati $5.4B not too much long ago.
With each factory valuing around $2B to $3B each.
More the AMD own value.
That “real” AMD value would be around what, $15B?

Scientia from AMDZone said...

You can check the facts yourself if you like. Just look at AMD's Q4 07 Earnings:

On page 2 they list the assets and liabilities.

Current inventory, cash, and accounts receivable is: $3.8 Billion

Current accounts payable, owed compensation and debt interest: $2.6 Billion

This puts AMD ahead by $1.2 Billion.

FABs and equipment are worth $4.7 Billion

However, long term debt is $5 Billion

There are lots of other items but after you add everything up AMD is $3 Billion ahead.

Interestingly, nVidia is only $2.5 Billion ahead. Yes, the company that people keep claiming can buy AMD actually has fewer assets than AMD.

Jeach! said...

There are lots of other items but after you add everything up AMD is $3 Billion ahead.

Interestingly, nVidia is only $2.5 Billion ahead. Yes, the company that people keep claiming can buy AMD actually has fewer assets than AMD.


Now if looking on "Page 2" was that easy and exact, Warren Buffet wouldn't have made half his fortune in buying undervalued assets?

Scientia from AMDZone said...

jeach

True enough. We are only talking about current assets, not future assests or stock value. If AMD continues to lose money then their assets will shrink but if they start turning a profit that $5 Billion in debt shrinks. If AMD begins turning a profit then the stock could easily double in value and be a good investment. However, if AMD continues to have losses then maybe not such a good investment.

Ho Ho said...

scientia
"So, there is no reason to think that AMD will have 3.0Ghz available in the first half of this year."

I wonder where I've heard that story before. Oh right, I said it months ago when everyone were saying how K10 will quickly overpower Core2 by fixing the bug and crancking up its clockspeed :)


"The big question I have is whether AMD is really going to abandon 65nm at 2.6ghz and then how long it would take them to exceed that with 45nm."

Well, if AMD is supposed to start selling 45nm CPUs "soon" (this year) I see little reason to spend development money. Of course if 45nm isn't going that well it might make some sense.


"Historically, it has taken AMD a couple of steps to match previous speeds so I don't expect a 3.0Ghz 45nm quad in Q4."

My expectation would be that 45nm will be pretty much the same as 65nm: almost nothing availiable before newyear and clock speeds aren't exactly great.


pop catalin sever
"The thing is the .45 nm process still won't be better than the one Intel has, but it will close the gap nonetheless."

I wouldn't say it closes the gap, just gets a bit closer. Remember that Intel isn't sitting still either and it wouldn't be that surprised if we have another G0 on 45nm soon.

Erlindo said...

Hey Scientia:

What are your thoughts about this:
AmTech Outlines Possibilities of NVIDIA/AMD Merger (NVDA, AMD)

Scientia from AMDZone said...

erlindo

The nVidia merger story is a joke based on some very fast and loose reasoning by a single analyst, Doug Freedman at American Technology Research.

An actual merger is 100% impossible. This is true because neither top AMD nor nVidia execs would be willing to give up control. ATI was not a merger; it was a purchase. ATI has now essentially expanded and replaced AMD's existing chipset design center. ATI also brings new consumer technologies and discrete graphics which AMD was not involved in (except for the tiny presence of Alchemy).

An nVidia merger is impossible because nVidia would still want to operate as an independent company. The original plan had nothing to do with a merger; it had to do with the head of nVidia trying to gain enough additional financial support that he could purchase AMD outright.

The market is simply reading too much into a vague and poorly considered "what if" scenario.