AMD Quietly Announces K10 Launch Time Frame
AMD officially announced when K10 would be released and gave a small interview to IDG. I would have to say that the news is mixed. Let's look at what this announcement means.
The details directly from the AMD Corporate Virtual Pressroom are pretty straightforward:
AMD expects that the processors will begin shipping for revenue in August 2007, with systems from AMD platform partners beginning to ship in September 2007.
With planned availability at launch in a range of frequencies up to 2.0 Ghz, AMD expects its native quad-core processors to scale to higher frequencies in Q407 in both standard and SE (Special Edition) versions.
This does contradict the never ending stream of rumors that K10 would be delayed until September, October, November, December 2007, or even January 2008. However, it also contradicts the unofficial roadmap that showed K10 at 2.3Ghz at launch. There is little doubt that AMD is not happy with the 2.0Ghz launch speed. It seems common sense that AMD could have made this same announcement at Computex earlier in June. This is further supported by Anandtech at Computex who said:
We understand from the motherboard partners that AMD should hit 2.0GHz by September for the Barcelona launch
So, most likely AMD's plan was to give this news quietly after Computex to allow a few weeks for it to settle and expectations to tumble accordingly. Then they could start the July 2007 Analyst Day on a more upbeat note. Presumably AMD waited until they knew that 2.0Ghz was the best they could hope for. But, I'm sure AMD is aware that even a slow 2.0Ghz in August is still better than nothing until November.
Now we are left with the burden of rating this announcement. It is obvious that AMD's effaced statement was as much a mea culpa as if they had taken out an ad during the SuperBowl. So, AMD's rating is that they are not at the top of their game. Maybe they're trying but they aren't there yet. Still, there are other ratings besides AMD's. When Operon was first released back in 2003 it only hit 1.8Ghz compared to Barton's 2.2Ghz. If we apply this same ratio to K8's current 3.0Ghz then we come up with 2.45Ghz. Of course, Opteron and Barton were both single core. Curiously, bumping the clock speed down a grade to allow for the shift from dual to quad core we get 2.25Ghz which is about what the unofficial roadmaps showed at 2.3Ghz. So, it looks like AMD is down about 1.5 speed grades.
Certainly we have to consider if AMD has done a Prescott with K10. Although Prescott was theoretically launched in 2003, it took Intel until months into 2004 to get Prescott above Celeron speed. However, there is no indication that K10 is running hot. There is one clue from AMD's announcement:
This would be the first time AMD has made both standard and low power parts immediately available as part of a new processor launch.
Low power and for the first time? It sounds like AMD in its attempt to fit quad core K10 into the same power envelope as dual core K8 may have been a bit too aggressive. It sounds like the process has been tweaked a bit too much in favor of low power and that is why there aren't enough higher clocking parts. Of course, it could also be that AMD's process is just not quite there yet. We saw with Intel's C2D that it came up one speed grade with each revision. I'm sure AMD will fix this but the big question is when. They did say higher frequencies in Q4 but this could mean just a single bump from 2.0Ghz to 2.2Ghz. It could also be 2.3Ghz or 2.4Ghz.
Anandtech said: but according to AMD the partners will see at least 2.3GHz by the end of summer. Here, end of summer could mean the beginning of October which would still be Q4.
However, this wouldn't quite be enough. AMD would need 2.5Ghz for Phenom dual core in Q3 to match the already available K8 3.0Ghz speeds. This would also require a bump to 2.7Ghz in Q4 for dual core K10 to pull ahead of K8. The unofficial roadmaps showed 2.9Ghz for dual core in Q4 but 2.7Ghz would at least be enough to surpass the 3.0Ghz K8 speeds available today. If we just consider QFX, for example, AMD needs at least 2.5Ghz to match the performance of the current older 90nm 3.0Ghz chips. I suppose 2.5Ghz would still be worthwhile since you would get twice as many cores for the same power envelope. This would mean 8 cores total but I wonder if there are any games at this point that can use that many cores.
It is difficult to imagine just how frantic things are at AMD right now. AMD has DTX and mini-DTX motherboards coming out in Q3. The 65nm version of R600 (R650) is also due in Q3 along with the finished drivers. We know that Cray is waiting for Budapest (Barcelona with HT 3.0) in Q4. AMD has to also be planning for the R700 chipset for 2008, plus the new mobile chipset, plus the new mobile processor. And, all of this would be while working on the 45nm process with the new immersion scanners and getting DC 2.0 ready, continuing to ramp FAB 36, getting the bump and test facility up and running, and beginning the process of taking down FAB 30. We know that K10 already contains four HT controllers so it isn't clear just how much work is necessary to do DC 2.0. Presumably there must be something more to do or else AMD would release DC 2.0 with Barcelona instead of a year later. I would also imagine that AMD must be hard at work trying to duplicate the new Intel SSE4 instructions. Since AMD has not yet announced tapeout of Shanghai I would imagine they can fit in at least some of the new SSE4 instructions.
It is likely that AMD does have at least some 2.2Ghz chips. We know that AMD is not saving these for Cray since Cray is waiting on Budapest however Sun might be waiting on faster Barcelona chips for Ranger which requires about 16,000. Or AMD may simply be stockpiling faster chips for Q4 release. It is my guess that AMD will give the expected Q4 and Q1 K10 speeds when it gives the Q2 Earnings report later this month. There may be some information about future plans but most of this will probably be given at the Analyst Day at the end of the month. I don't know how likely it is that AMD can get the speed up 2 ½ grades from 2.0Ghz to 2.5Ghz. That is what would be required to equal the current top K8 speeds. This speed could also roughly match Clovertown. However, 2.5Ghz for dual core K10 would lag behind 2.93Ghz Conroe. Also, 2.5Ghz is only likely to match a 2.66Ghz Penryn quad core. A full set of benchmarks should show this more accurately when K10 is released in August but right now it does not appear that AMD will catch Intel's top speed anytime soon.
64 comments:
It is likely that AMD does have at least some 2.2Ghz chips.
Why is it "likely" that AMD has 2.2GHz Barcelonas right now? I see zero indication of that. If they had a stepping that produced 2.2GHz DVTs in any quantaties in hand, that stepping could be put into volume production and expect to produce 2.2GHz parts in time for September systems.
I think it is far more likely that they are not able to produce 2.2 in any volume with their current stepping. Otherwise, not launching 2.2GHz parts makes no sense whatsoever.
I've tried to figure out AMD's Announcement numbers:
AMD estimates that the new processors can provide a performance increase up to 70 percent on certain database applications and up to 40 percent on certain floating point applications
However, I have to admit that I don't really know what they are referring to.
If we just take the base 2.0Ghz clock and double the cores then divide this by the current 3.0Ghz clock then we get 133%. In other words a 2.0Ghz K8 with four cores and perfect scaling should be about 33% faster than a dual core K8 at 3.0Ghz assuming we have perfect scaling with four threads.
Unfortunately, scaling is a big factor and hard to estimate. For example, if we assume that K10 is 20% faster but we only have 90% scaling with four cores then we only get 44% faster. So, 30% - 50% faster would be about what we would expect. 70% greater database performance would require roughly 30% faster Integer IPC and this is more than I would expect.
I have no idea about the 40% on FP though. We know that according to the K10 Optimization Guide that almost all SSE instructions now decode in one clock as Fastpath singles instead of two clocks as they did in K8. The buses have also been doubled in size and the SSE execution units are twice as wide as well. So, SSE with perfect scaling would be twice as fast. AMD has confirmed this number on several occasions. For example January Geek:
that provides a 180% improvement clock per clock over today’s Opterons. Thus, moving to quad-core will see a potential 360% performance increase
360% is a per core improvement of 90% which is about the best scaling we could hope for.
So, I really don't have a clue what these 70% and 40% numbers refer to. There should be no way for Integer to increase to 70% or FP to only increase to 40%. 40% would be about right for total increase in Integer but 70% would be far too small for SSE unless it referred to the increase per core.
periander
"I think it is far more likely that they are not able to produce 2.2 in any volume with their current stepping."
I see you are getting confused about what I meant. 2.2Ghz now would be samples. 2.2Ghz in August would be more but a volume too small for general release. As I said, any small volume they might have in August they will either have to save until they do have enough or they might have some earmarked for special customers like Sun. 16,000 for Ranger would just a fraction of what they would actually need.
Good analysis as usual. I would say that the problem is the lower cash flow prior to the convertible bonds, and partly even after.
There is just so much specualtion and for some reason AMD is now the biggest liars ever.
They said June sampling, happened, July for some shipments, Aug for shipments and volume.
I don't know if they have pushed anything back but I am confident that Barceona will open a very large gap in TPC-H for AMD at 2.0GHz.
I mean four cores will LOVE a database even without the HT3 chipset (I thought Broadcom would have one by now or perhaps nVidia) which AMD should have done themselves if server parts were first.
I can imagine that they will get a limited number of 2.1GHz chips in September. AMD has always been good with their CTI method.
As far as the maturity of the process, thee are two ways to look at it,
1. the shrink will not be as efficient
2. Barcelona will take advantage of 65nm process improvements.
Anyway, even if they do not match clockspeed, the efficiency or HT and the IMC will make up somewhat for it.
Cloverton 2P sucks without the Dual FSB.
I will be interested to see how hot Caneland et al get (4 1600MHz FSBs). That will mean more in a datacenter than the clockspeed or perf advantages.
I figured that AMD would make some trade offs to keep in the Opteron envelope so it's not surprising.
I too was wondering what is taking Shanghai so long to tape out. It's just a shrink with more cache at last word but they could perhaps be updating SSE.
Again though this is all reminiscent of the Opteron launch. Doubts, doubts, rumors, rumors.
I see you are getting confused about what I meant. 2.2Ghz now would be samples.
That's what I said. DVTs=developmental samples.
The launch announcement was a clear indication that the current sample stepping has been placed into volume production for release, and are now undergoing the 10+ weeks or so it takes for the wafers to go from start to packaged chips. Thus, the announced timing; (end of) August shipping and September systems.
If the current sample stepping produced 2.2GHz chips in any quantity, then volume production of said stepping should produce said speeds in significant volumes. Since no such speed was announced, I see no indication that the current sample stepping is producing anything over 2.0 in any significant numbers.
What evidence shows they have 2.2GHz DVTs now? I see none.
christian
"There is just so much specualtion and for some reason AMD is now the biggest liars ever."
Well, they aren't technically lying. AMD said "mid-year" and later said "end of summer". If we divide the year up into 3 parts then the middle part is May, June, July, and August so I guess that would technically be mid year. And, it would fit the calendar date for Summer with begins at the end of June and runs through September.
Periander
It depends on whether AMD is actually launching in August or if they mean launching in September. If the launch is actually August (which is the way I read it) they would have to be shipping processors at least 3 weeks prior because AMD never does a launch without chips available. At least they've never done this with cpu's. R600 is a bit hazy since they technically launched the entire line with 2900XT but didn't provide the lower models for another month.
I'm assuming 2.0Ghz was the best they could do with maximum yield at this point. I think their biggest priority at the moment is volume for K10 at whatever clock speed they can manage.
Scientia:
If the launch is actually August (which is the way I read it) they would have to be shipping processors at least 3 weeks prior because AMD never does a launch without chips available.
You are kidding me here?
Brisbanes were launched on December 5 and none were anywhere to be seen for at least a month. The Brisbane 5000s which were supposedly launched with the rest of them didn't show up anywhere until May.
The BE parts launched almost a month ago. Where are they?
http://techreport.com/onearticle.x/12728
Then there were the 4X4s which took a while to show up.
http://techreport.com/onearticle.x/11378
We won't even mention the 2600/2400.
Has AMD not paper launched anything except the 2900 in the last year? What a joke.
So what white rabit can AMD pull out of their hat before the show is over ? I'm certain there are not few smart people inside AMD but unfortunately time and money are finite resources.
Can AMD license/purchase IBM's manufacturing technologies that supposedly allow 4 GHz Power 6, to be able to counter Penryn in Q1-07 ? Will they wait until .45 nm to match or challenge Intel from the manufacturing and performance points o view ?
The light asset strategy AMD is planning to employ suggests to me that they plan to survive and reduce casualties until .45 nm manufacturing arives.
So far the only thing that I've could dig on the net that could theoretically match Intel's high K process for .45 nm seems to be IBM's snow flake chips. (the snow flake chit name was given to it by Inq. not the linked site)
pop catalin
"So what white rabit can AMD pull out of their hat before the show is over ?"
They just need to keep increasing K10 speed. If they can manage a bump of 300Mhz each quarter that might be enough. They'll probably need 3.1Ghz to match Penryn at 3.33Ghz.
"time and money are finite resources."
As I've said, AMD can lose another $1.5 Billion. If they can reduce the loss by just $100 Million each quarter that would be enough. Remember, this would mean not breaking even until Q2 08 and AMD would still be fine financially. AMD would pull out of its decline with $500 Million to spare.
"Can AMD license/purchase IBM's manufacturing technologies that supposedly allow 4 GHz Power 6"
IBM doesn't have such a manufacturing technology. Power 6 clocks high because they stripped the Power 5 architecture.
"Will they wait until .45 nm to match or challenge Intel from the manufacturing and performance points o view ?
If AMD can only bump the speed 200Mhz each quarter then they'll be on 45nm before they are competitive. I would think that AMD could match K8's 90nm 3.0Ghz speed on K10 dual core by at least Q1 08 but I guess we'll see.
"The light asset strategy AMD is planning to employ suggests to me that they plan to survive and reduce casualties until .45 nm manufacturing arives."
No. This only means that AMD will keep making ATI parts at foundries; there is no other change beyond a slower ramp for FAB 38.
"So far the only thing that I've could dig on the net that could theoretically match Intel's high K process for .45 nm"
IBM has high K 45nm also but AMD won't use it until 2009 if it does. Again, we'll have to see.
"As I've said, AMD can lose another $1.5 Billion."
This statement is just flat out wrong. You ignore many practical realities:
1.) AMD's current cash balance (short term + cash) is only $1.2B
2.) Much of AMD's remaining cash balance will be restricted cash - meaning it is held overseas in China and India and repatriating it to pay expenses would cause AMD to pay much of the cash as taxes
3.) Financial distress and bankruptcy will not occur at a $0 or negative cash balance - it will occur at a much higher total based on suppliers no longer extending credit (reduction in payables = decrease in cash) for both expenses and capital expenditures.
4.) For AMD to stay competitive in terms of manufacturing, they need to maintain their $600M run rate in CAPEX which is greater than their $500M depreciation offset. In essense, an income statement loss of $500M = a cash burn rate of $600M.
The only reason that AMD is showing any signs of life currently is the cash they generated from working capital - notably, the decrease in accounts receivable which added $400M in cash last quarter. The main issue is that this is not repeatable and will have to reverse if AMD plans to increase sales.
All in all, it is pretty obvious that AMD will cease to be a going concern and the only question is if/who will buy them out. The K10 - even if it launched today, would not change that.
As far as losses go I estimated AMD could lose another $1.5 Billion because this would match the low point they were at in 2003. But consider this, what if AMD reduces the loss by just $100 Million each quarter? That doesn't sound like such an impossible goal. But look:
Q2: $400 Million loss, total $400 Million
Q3: $300 Million loss, total $700 Million
Q4: $200 Million loss, total $900 Million
Q1: $100 Million loss, total $1 Billion
Q2: $0 loss, total $1 Billion
So, AMD would pull out of its nosedive with $500 Million to spare. Then in Q3 08 I would expect FAB 36 to be running full capacity, 45nm production to start, the release of the new mobile chipset + new mobile processor + R700 series graphics processor. FAB 30 should also be converted to FAB 38 and ramping 300mm wafers by then.
Also, it is AMD's goal to have 50% of all Opteron sales as K10 quad core six months after launch. This would mean end of Q1 08. If 50% will be quad core then I assume the other half would be K10 dual core and K8 Opteron production phased out.
That is taken from AMDZone, I can't comment there so I'll do it here.
If you expect AMD to loose money until Q3'08, who do you expect them to convert all the fabs, Start pumping out 45nm, release chipsets, Release R700?
I said last year at the launch of Core 2 Duo that Barcelona for Servers will be towards the End of '07 and desktop versions may slide to beginning of 2008 and MY GOD was I attacked and ripped a new one. Now it's time for me to sit back and watch it all.
periander
Yes, I agree the tech reports article do suggest that AMD has been having supply problems since last December. However, I would expect this to improve since FAB 36 is now fully 65nm and Chartered too should be making 65nm chips.
AMD executed brilliantly, mobile, ATI purchase, DTX, expansions, only one problem they tried to sell yesterday's pie to customers while Intel was pulling a fresh hot pie from the oven.
I fail to understand why AMD has such low understanding of the markets psychology. K10 should have made an appearance 2 years ago. I was reading a couple of years ago around 2003 the Software Optimization
Guide for AMD Athlon™ 64
and
AMD Opteron™ Processors and even then when I've read that 128 bit SSE instructions are broken down into 2 64Bit instructions the thing seemd very odd, it was so darn obvious that this was a sure way to improve CPU performance that I new they were going to do it sometime in the future, but was expecting 2005 for a CPU arch refresh not 2007.
AMD is already late, 2 years late.
P.S. Oh yeah, X2 is not an arch refresh K8 already had an crossbar switch so no arch changes 2 years after K8...
AMD never does a launch without chips available.
Sci, can you please go into detail about that comment of yours keeping in mind what Periander said about Brisbanes and BE's and the 2X00 series?
You seem to be avoiding questions you do not have answers to, which in this case the best thing to do is never make claims you can't back up.
mo
"If you expect AMD to loose money until Q3'08"
No, I don't. I expect AMD to break even by Q4 07.
Since you (and some others) seem to be so pessimistic maybe you can explain why today would be worse than 2003.
By early 2003, K7 has slid so much that AMD dropped the Duron name and put K7 directly up against Celeron. In 2003 AMD began making K8 with the lowest yield they ever had with K8 dies twice the size of K7 dies. At 1.8Ghz K8 was not only substantially slower than P4 which shortly after released at 3.2Ghz but was also slower than 2.2Ghz Barton. AMD was still behind at 2.0Ghz although this speed did finally overtake Barton. It took AMD two speed bumps to 2.2Ghz to finally get a little ahead P4.
Also, AMD was behind with socket 754 because Intel had pushed the FSB from 400Mhz (which 754 was designed to match) all the way up to 800Mhz. This wasn't matched until socket 939 later in the year.
Do, you also recall the complete absence of supporting motherboards? For months the only motherboards available all used AMD's chipset. These were suitable for workstations but not for desktop. As I recall the first non-AMD chipset motherboard was released by VIA. It took quite awhile before nVidia and ATI started supporting K8. By Q1 04, K7 was still the dominant chip in production fully nine months after K8's launch. AMD was also a single 200mm FAB against Intel's multiple 300mm FABs.
So, why is the situation today worse than it was in 2003?
Today, AMD is still making Sempron. There are HPC systems waiting for K10 while there were no HPC K8 systems in 2003. AMD is now matching 300mm to 300mm. The K10 die is big but the yields are much better. The clock speed is down, however, unlike 2003, there is already support for K10. It took two quarters later until Q4 03 for AMD to get into a competitive position with K8. I would say that AMD should be much more competitive with K10 two quarters later in Q1 08.
Pop, I agree about the architecture. There was no real change to the architecture in 2005 with dual core. There were lots of changes in 2006 but none were performance related. They added RAS and virtualization and changed the IMC. The architectural changes in K10 were very much overdue. It seems to me as well that there could have been some kind of interim improvement in 2005. Maybe with the modular core they can avoid that in the future.
Why is AMD even in this situation? If it has happened to them in the past, don't you think they would have worked on their execution strategy to avoid such mistakes?
You think it's "ok" to be in this situation? Investors lost a LOT of money, but according to you, it's OK. Only because AMD has been there before. It is NOT ok in my opinion, It's never ok to repeat your mistakes.
Will AMD be in this spot everytime it releases a new architecture?
I have read your views and posts, You sci are a sugar coating machine. Everything that comes out of your mouth about AMD is thoughly sugarcoated and you have been proven wrong numerous times. Yet you keep making such false claims and AMD does not do launch without products. Yet AMD has been doing that since end last year.
2.0Ghz max for k10 is disappointing indeed no matter how much you sugar coat it.
How can you put Late August as Mid Year?
Scientia, you're a knowledgeable being but sometimes you just put a blindfold on knowingly or unknowingly and your credibility falls daily.
Break even in Q4? yet you're saying that at a rate of -100M, it would still post a loss of 100M in Q1'08.
So you're assuming that it will be posting lower than 100M losses over the next quater?
Your numbers don't add up as usual.
mo
"Why is AMD even in this situation?"
I don't know. It could either be because of a lack of time and resources or it could be due to a management decision. Remember that cpu design has a long lead time.
" If it has happened to them in the past, don't you think they would have worked on their execution strategy to avoid such mistakes?"
K10 will be supported at launch unlike K8. The yields should also be better. There are at least some improvements.
"but according to you, it's OK."
Do you actually read anything I write? I see these strawman arguments all the time on roborat's and 180's blogs. I guess it's easier to argue against me if you invent my side as well.
Why would you be so foolish that you could think that I've ever said that a $1.1 Billion loss in just two quarters was "ok"? I can think of several things I would have liked to have seen done differently. I can think of things for Intel as well.
"Will AMD be in this spot everytime it releases a new architecture? "
No. Modular core.
"I have read your views and posts"
Then why are you unable to accurately repeat them?
"you have been proven wrong numerous times."
Perhaps you could list a couple dozen from this apparently long list.
"Yet you keep making such false claims"
Maybe you could list these too.
" and AMD does not do launch without products. Yet AMD has been doing that since end last year."
I guess you missed my post where I agreed with this. Are you sure you read what I write?
"2.0Ghz max for k10 is disappointing indeed"
Yes which is what I already said. Did I lose you when I used the phrase "mea culpa"?
" no matter how much you sugar coat it."
You'll have to show me where I sugar coated it.
"How can you put Late August as Mid Year?"
Well, I would personally put mid year at late June or early July. Which is why I said the middle third of the year would "technically" be mid year. Do you really have trouble understanding this?
"Scientia, you're a knowledgeable being but sometimes you just put a blindfold on knowingly or unknowingly"
Well, then quote something from my article and show me where I had a blindfold on.
" and your credibility falls daily."
Would my credibility be higher if I cherry picked the news and allowed all the gutter rats to do troll and flame posts like 180 and roborat do?
"Break even in Q4? yet you're saying that at a rate of -100M, it would still post a loss of 100M in Q1'08."
Why are you trying to mix together math from two completely different statements? Why would it even make sense to you to do this? Try looking at each statement on its own instead.
"So you're assuming that it will be posting lower than 100M losses over the next quater?"
No. I didn't say that.
"Your numbers don't add up as usual."
No, not when you mix them up and rearranged them. Your mind seems to be a bit like a meat grinder which mangles and distorts what I actually say. It is interesting though that you attribute this to me instead of your understanding.
"Either prove to us that you work for AMD, or just accept the possibility that the yields suck."
Scientia works from AMD wheter payed or voluntary. Having and AMD centric (not biased) blog kind of makes you work for them. The bias problem is questionable even if you work or are just a fan. So no point in aguing about this. Actually Scientia's otpinions whould hold much more value if he was indeed AMD insider.
Anyway the yield statement is very odd, no one can possibly gues yields :) (especially whout any meaninfull data) there are the best keept secret, so either Scientia knows them or this is simply a trace of bias.
Anyway the yield statement is very odd, no one can possibly gues yields :) (especially whout any meaninfull data) there are the best keept secret, so either Scientia knows them or this is simply a trace of bias.
Very true! And for a good reason. If your yields suck, you do not want your investors to know about it. If they are great, you do not want your haggling customers to know about it :).
Sci, can you please go into detail about that comment of yours keeping in mind what Periander said about Brisbanes and BE's and the 2X00 series?
You seem to be avoiding questions you do not have answers to, which in this case the best thing to do is never make claims you can't back up.
It's very simple. Twice the amount of chips that come to the US go to the rest of the world.
There were people on forums from other countries who had Brisbane in Dec. At least one person from Australia verified it.
Also, AMD has more than just the channel, they have Lenovo, Dell, HP, etc and they sell PCs in the rest of the world too.
R600 turned about to be a 65nm bug in the low-end chips. I can assume they were counting on the problem being fixed in time.
mo
"But I'll do a quick search because the Great Sci believes he's never wrong."
I have no idea where you got this idea. I'll see if I can add some at the end that you missed.
I'm going to have to summarize your list because you repeat some of the same items two or three times.
1.) Availability: I'll take Tech Report's word for it that some AMD chips have not been available. It sounds like this started with 65nm. So, availability worse than I thought.
2.) 2.2Ghz: I was definitely wrong here. Apparently, the good yield is lower at 2.0Ghz.
3.) Q2 07 launch for K10: Since there won't be any chips in late Q2/early Q3 we can say that AMD has had some delay. I would guess that AMD required another revision so that would be about 10 weeks delay.
4.) 3.0 & 3.4Ghz speeds for K8L: That was back in 2006 when I had no expectation that the IPC would be increased. 3.0Ghz with no change in IPC is about the same as 2.5Ghz for the expected performance of K10. 3.4Ghz is about equal to 2.9Ghz.
5.) 7 week revision cycle: Earlier information I had suggested a 7 week cycle time. Apparently this was way off. I had also seen a statement on the INQ that the cycle time was 3 months. It now appears the minimum cycle time is about 10 weeks.
6.) Supercomputers expect K10: We now know that AMD is late with these deliveries.
gdp77
1.) AMD's X2 are still not priced below the lowest C2D; this hasn't changed. But, you are right that there has been some shift in prices.
Specifically the 4200+ is no longer equal to E4300. According to NewEgg the 4800+ is now in this spot. 6000+ which is AMD's fastest chip at 3.0Ghz is selling equal to E6600 which is only 2.4Ghz.
Obviously this gives Intel several price points above AMD. Since we're on the subject though I would price K10 dual core at 2.5Ghz higher than 6000+.
2.) Marketshare: Well, it looks like I was way too optimistic about this. Given the delays and slower speeds I'm wondering if AMD will be able to even recover its losses by end of year. And, similarly, it doesn't look like K10 will have the same effect for AMD in 2007 as C2D had for Intel in 2006.
3.) Clock speeds: Obviously, AMD is not hitting the unofficial roadmap speeds in Q3 so who knows what they can do in Q4. 2.5 and 2.9Ghz may have been too optimistic.
ho ho
1.) Speed bump: Yes, I generally refer to a speed bump as 200Mhz. You are of course correct that Intel uses both 200 and 266Mhz while K10 is capable of 100Mhz.
2.) TDP: No official information yet.
3.) No games using 8 cores, maybe 4 cores by Q1 08: Yes, that's what I was thinking too. That does cast doubt on the value of a QFX system for gaming.
4.) R650 bogus: You could be right. I've been trying to confirm this. I've only found one copy of the conference where R650 was supposed to have been mentioned officially by AMD but unfortunately the sound doesn't work. So, right now I can't tell if it was ever officially mentioned or not. I know they are suggesting that AMD will skip directly to 55nm in early 2008. If this is true then it wouldn't leave time for R650.
5.) R600 drivers: I know there have been some improvements but I don't know how much there will be by end of Q3. Without reduced power draw it looks like R600 won't really be competitive with nVidia.
6.) K10 not available until September: That does sound reasonable although wouldn't it take time for systems to be built versus just showing up at places like NewEgg? No, probably not since integrators would get their processors in sheets while NewEgg would need boxed sets. This probably is about the same amount of lead time. So, September.
7.) 2003 comparison: I believe AMD's equity ratio was about the same in 2003 as it would be now if AMD lost another $1.5 Billion.
8.) $600 Million loss in Q1: The operating loss was $504 Million. Why would I think that AMD could reduce the loss by $100 Million? I guess that depends on their expenses and what they sell. They could just as easily lose $700 Million in Q2.
9.) Ringinheat: Don't quote people who are being insulting.
Mo
Okay, ones you missed:
1.) I didn't expect C2D to increase its IPC as much as it did.
2.) I had suggested that Intel might remove micro-ops fusion from C2D because it seemed to mostly overcome the limited port problem in the Pentium M pipeline. This port problem has been fixed in C2D but micro-ops is still there.
3.) I never expected Intel to bring back HyperThreading with Nehalem.
4.) Q1 was much, much worse than I was expecting. Q1 was as bad as a one quarter drop in 2002.
edited from core2dude's earlier post
This speed could also roughly match Clovertown. However, 2.5Ghz for dual core K10 would lag behind 2.93Ghz Conroe.
On what benchmark? specfp_rate? I will believe it when I see it.
The bottom line is, the only benchmarks on which Barcelona has publically been evaluated are pov_ray and cinebench.
In abscence of any other demonstrations, all we can say is, Clovertown leads Barcelona 2-0.
Would my credibility be higher if I cherry picked the news and allowed all the gutter rats to do troll and flame posts like 180 and roborat do?
Case in point, you keep on claiming that Barcelona yields are great. While AMD's partners say that the yields suck.
Benchmarks
If someone chooses to believe the accuracy of some scores taken on the sly from a demo system I suppose that is their choice.
However, pouncing on a couple of benchmarks that were not taken on systems designed to show actual performance seems a bit desperate to me. We'll get some better benchmarks in, I suppose, August or September.
So, core2dude, we have your official prediction that K10 has no performance gain over K8. Are you going to stand by this in a couple of months when real benchmarks come out?
Yields
I knew what AMD's product demand was up until end of 2006. I haven't been able to figure it very accurately since then because of the split between 90nm and 65nm. For example, if 65nm is in a lot higher demand than 90nm then the problem could simply be that FAB 36 couldn't supply enough chips. However, if the demand for 65nm is not that high then it would be a yield problem.
As far as I can tell, these are the only two possibilities. Considering that AMD has suggested that yield is fine and also suggested that the 65nm transition was slower than expected and that their product mix didn't match what was in demand why exactly would you jump to the conclusion that the yields are bad?
It is looking to me like AMD may been caught with 65nm demand higher than what they could produce. Remember that in early 2007 FAB 36 was only half of production and not yet fully 65nm. Also, Chartered was not making 65nm chips yet.
So, what exactly are your arguments to show why you believe that the 65nm yield is bad?
ringinheat
"scientia doesnt like view points contrary to his own in this blog.....ask roborat..
It isn't about a different point of view; it's a matter of respect. I agree with very little that Sharikou says yet I have never attacked him personally the way that I have been. I agree with more that 180 and roborat say so I have even less reason to attack them personally. There has never, ever been a roborat or 180 bashing session on my blog.
However, I can't say the same for either 180 or roborat's blogs. I get routinely bashed in both places, often by anonymous posters, and at least half the time the opinons or statements attributed to me are simply invented. But, the quality of the comments are so low that it isn't worth my time to make corrections.
I suppose I could have been nicer to roborat but it is difficult to show unlimited respect to someone who shows me almost none. I would prefer that he would take more responsibility for his blog and stop instigating the personal attacks. A typical scenario seems to be that I write an article, he then writes a counter article linked to mine. His counter articles usually include personal attacks. These personal attacks are then continued in the comments by people who want to rant but don't seem to have enough confidence in their opinion to post here. Why would I have a lot of respect for a blog like this?
So, core2dude, we have your official prediction that K10 has no performance gain over K8. Are you going to stand by this in a couple of months when real benchmarks come out?
And where did I take that position? I have always said that I will believe it when I see it. In other words, if it indeed shows the performance improvements, I won't deny those.
But at this point, the only two benchmarks are pov_ray and cinebench.
And I don't care whether guys was sly in obtaining the benchmark, as long as he/she is an independent reviewer. Also, there is nothing sly about pov_ray. AMD showed it on their own, and then was sick with foot-in-mouth for a couple of months.
To sum it up, no, I won't deny AMD's performance advantages if they indeed show those advantages. However, at this point, I am extremely skeptical. It is simply not possible to have massive IPC advantage across the board and still lose in pov_ray, unless of course, they were running the demo at 1GHz...
AMD Barcelona performance promises revealed
quote: WHILE ENGINEERS ARE working full throttle at making the K10 (10h family) work, performance figures are becoming an integral part of marketing presentations and so on. Recently, we managed to see the sort of performance figures that AMD is promising for Barcelona.
These performance numbers were based on systems using tge Opteron 2356, or known to the world of us regular folk as AMD QuadCore at 2.3GHz. The orice of this processor is currently set at $796 USD in quantities of 1000, so we'll see what can Clovertown at 2.66 GHz (Xeon 5355) do about it…
So, what will Barcelona deliver?
For starters, SPECint_rate2006 and SPECfp_rate2006 speak the voice of doom for the 366 MHz faster clocked Xeon - and these are the reasons why Intel was downplaying performance of FP and praising the INT.
These figures were from a few months ago, so things may well have changed since then. But in the Integer test, a Barcelona 2.3GHz yields 21% higher score than Clovertown 2.66 GHz, but Floating Point test leaves a staggering 50% performance deficit for Clovertown, and this is not something 45 nanometre Penryn can solve overnight. Unless, of course the clock deficit for AMD is such that Intel speeds past.
When it comes to comparing Barcelona to Santa Rosa, a 3.0GHz clocked Opteron 2222, Barcelona is in general, around 65-70% faster than the highest-clocked dual-core Opteron. SPECweb99 yields a hefty 67% performance increase, SPECweb99_SSL offers 66%, while SAP-SD offers 70% performance increase.
The highest gain, of over 100% can be seen in SPECweb2005/Ecom benchmark, while smallest gain was in TPC-C SQL2000, where only 42% gain was marked. This was all based on 2P (dual-socket) systems, of course.
In the four-socket arena, Opteron 8222 versus 8356 offers an average of over 60% performance boost, which is not a small thing - given that clock difference is 700MHz. SPECweb99 and SPECweb99_SSL give 68% performance increase, SPECweb2005/Ecom yields another scaling dream (91%), while most modest increase is again, TPC-C SQL2000. Nevertheless, TPC-C Oracle test will yield 75% performance increase while Terminal Services will offer 71%.
Overall, these are very impressive score increase promises. All of the performance increases happened with a processor that has 30% smaller TDP. Opteron x222 is a 120W Max.TDP, while this Barcelona 2356 stays in the 95W range.
The only real question that now remains is whther AMD can execute. Sadly, the company's recent track record does not bode well, with constant delays of products and events. And we'll also wait to see real tests of the Barcelonas rather than paper promises. µ
Really interesting and it's good news indeed. =)
Scientia,
"If someone chooses to believe the accuracy of some scores taken on the sly from a demo system I suppose that is their choice."
I don't understand how it's "sly" when it was a demo setup by AMD. Are you implying AMD wanted to show their half baked CPU?
Like Core2Dude mentioned, all we have are the pov-ray and cinebench results. It's surprising that we are so close to launch and yet only thing AMD can provide are "estimated" projections of higher clocked K10s that won't even be available this year. I have already shown to you in your previous thread that per core, K10 will only have about ~12% IPC improvement. And this is according to Randy Allen's estimates that K10 will be 40-50% faster than existing dual core Opterons at 3GHz.
Scientia:
What's the main difference between AMD's SSE4a instructions and intel's SSE4?
Are these proprietary AMD extensions to SSE4 or are these some type of extensions to SSE2-3?
Also, is there any possibility for AMD to pursue their own multimedia extension (remember 3D-NOW) in a near future?
erlindo:
First of all, those are projections from chips that do not exist. HUGE difference between projections and Demonstrations.
second, they are from Feb, 07. about 4 months old. Lot has changed since then.
Those projections have absolutely ZERO merit and are of no valuable use.
Good try from the INQ though. They thought no one would notice their "source". I honestly believe New form INQ and FUDZilla should ba banned. I could make better guesses without even being in the industry.
Oups, this is the beginning of the devastating press coverage AMD will see soon ...
"I’ve seen benchmarks get cherry picked and twisted before but this is just outrageous. AMD is deliberately leaving out Intel’s best scores, leaving out Intel’s best products that shipped months ago, and putting in theoretical Barcelona scores for products that don’t even have a ship date. After Henri Richard (AMD executive) came in front of our ZDNet cameras to slam Intel for “un-ethical behavior” and promising not to do the same, we have caught them on four separate occasions behaving unethically. After this latest incident, it’s clear that AMD has no intention of behaving honestly or ethically."
Quote from previously mentioned article:
"Based on the latest certified SPEC.org results, AMD has a little more than a 1% clock-for-clock performance advantage in a dual-socket 8-core Server configuration but they have 50% clock speed deficit when the Barcelona finally launches in September"
I hate it when people show such lack of basic math skills. It's outrageous.
Correction: Barcelona at 2.0 GHz has a 33% clock speed deficit compared to Intel 3.0 GHz not 50%. On the other hand Intel has an 50% clock advantage over Barcelona at 2.0 GHz and not the way that the article implies. So means Intel 3.0 GHz will be ~50% faster than Barcelona at 2.0 GHh at launch if the article is to be believed that Barcelona only has an 1% IPC advantage.
George Ou is not the brightest crayon in the box but he does seem to to take great joy in bashing AMD. If K10 turned out to be twice as fast as Clovertown Ou's next article would talk about how Penryn would be better and that Intel had more volume. Sometimes I wonder if he has a lot of Intel stock.
I've been saying for quite some time now that those scores have nothing whatsoever to do with K10. They are simple projections of a theoretical quad core K8 (like it says in the fine print).
I have no idea why either Ou or Valich are talking about these socres now since they've been out for awhile.
Now, unless AMD has completely botched the architecture, K10's SSE score (SpecFP_rate) should be considerably higher. A good rough estimate would be about 70% faster per core and twice as many cores. This would give SpecFP_rate about 3.4X higher than K8 dual core.
Poke
What does the 12% refer to? What is your prediction for Integer and SSE?
Core2Dude
What is your prediction for Integer and SSE for K10? And, if your prediction doesn't match the Pov-Ray score (which shows no incease for SSE) then explain why.
erlindo
AMD only included a handful of new instructions with SSE4a while Intel added a few dozen with SSE4. AMD's SSE4a instructions were to compete with Itanium which already included these instructions. In some benchmarks Itanium had a 2:1 advantage. I'm sure AMD will copy the full SSE4 set at some point.
George Ou
"I’ve seen benchmarks get cherry picked and twisted before but this is just outrageous."
Then, George, why did you take AMD's example (which was specifically multi-socket) and rather than simply updating it you switched to single socket scores instead? Bait and switch.
"AMD is deliberately leaving out Intel’s best scores"
No, they didn't. Ed Stroligo at OC did his own comparison with current spec scores and they have about the same ratios.
Now, George, please explain why in your comparison you deliberately left out the SpecFP_rate scores which were AMD's best.
The bottom line is that George Ou is always against AMD and he doesn't mind cheating, bending, distorting, and twisting the numbers in Intel's favor even while he is screaming that AMD is not being fair.
Phenom X2GE
Unfortunately these aren't due until Q1 08. However, there are two things that seem interesting to me here.
The first thing is that these numbers seem remarkably close to Intel's. It could simply be that AMD is continuing the numbering system above the current 3.0Ghz 6000+. If these are intended to match up with Intel then we would have:
AMD 6400 (1.9Ghz) - Intel 6400 (2.13Ghz)
AMD 6500 (2.1Ghz) -
AMD 6600 (2.3Ghz) - Intel 6600 (2.4Ghz)
These ratios wouldn't make much sense because Intel 6400 would be 12% higher clock while Intel 6600 would only be 4% higher clock. So, I would assume that the naming similarity is only a coincidence.
The second thing that occured to me is that AMD has never pursued 45 watt chips like this before. So, it is looking like the BE series is intended for near-term DTX and that these GE Phenoms would also appear in DTX two quarters later. It makes me wonder if whatever BE's AMD is producing are being taken by manufacturers who are planning DTX boards later this quarter.
Scientia
Neither George's numbers or Ed's numbers are entirely up to date...
AMD - Intel 2PMB(2DC)
Intel 5160 SPECint_rate2006 - 65.0
AMD 2222SE SPECint_rate2006 - 60.4
Intel Advantage = 8%
Intel 5160 SPECfp_rate2006 - 45.5
AMD 2222SE SPECfp_rate2006 - 52.1
AMD Advantage = 15%
AMD 4PMB(2DC) vs Intel 2PMB(1QC)
Intel 5355 SPECint_rate2006 - 52.8
AMD 8222SE SPECint_rate2006 - 56.1
AMD Advantage = 6%
Intel 5355 SPECfp_rate2006 - 37.5
AMD 8222SE SPECfp_rate2006 - 51.3
AMD Advantage = 36%
AMD 4PMB(4DC) vs Intel 2PMB(2QC)
Intel 5355 SPECint_rate2006 - 101
AMD 8222SE SPECint_rate2006 - 108
AMD Advantage = 7%
Intel 5355 SPECfp_rate2006 - 60.9
AMD 8222SE SPECfp_rate2006 - 98.7
AMD Advantage = 62%
-Legend-
P=Socket
MB= Mother Board
DC=Dual-core
QC=Quad-core
-------------------------
I know we will have to wait and see, but how will DTX or MicroDTX stack up against Intel's MoDT using MicroATX?
enumae
Thanks for the update on the scores. Obviously, Ou was intentionally leaving out the SpecFP_rate scores because they would show AMD in the lead even on single socket.
To be fair, Penryn should help at least a bit because of the larger cache.
There are only two form factors that allow six boards from one blank: mini-DTX and mini-ITX.
If Intel hustles they may be able to match AMD with mini-ITX -- accept for price. As gdp77 pointed out earlier, Conroe is still priced much higher than X2. I suppose Intel could match if they do some further heavy price cuts to at least the 4XXX models. However, at this point in time there are no comparable Intel based mini-ITX motherboards.
Basically, mini-DTX is low price, lightweight, small case, quiet fan.
DTX is similar but a bit larger with more features. Theoretically Intel could match DTX but only with the smallest of BTX form factors, pico-BTX. People tend to forget that BTX is actually in the same class as Extended ATX, not ATX.
Micro-ATX is longer than DTX. Because of its length you can't get four from the same blank and therefore it's too big to compete with DTX. However, you can put a DTX motherboard in a micro-ATX case.
Scientia
Thanks for the update on the scores.
Not a problem.
---------------------------------
Just to be clear, the reason I pointed out MoDT is because of power consumption. MoDT is Intel's mobile on Desktop.
---------------------------------
I suppose Intel could match if they do some further heavy price cuts to at least the 4XXX models.
Well what about the Celeron 400, or the Pentium 2000's?
However, at this point in time there are no comparable Intel based mini-ITX motherboards.
A quick google search and I found a few.
---------------------------------
Like I said, well just have to wait and see, but I think Intel's MoDT will do well.
Here is a link to Intel's MoDT site, though alot of the systems shown are based on Core Solo and Duo or Pentium M, the market they will be placed in will most likely be satisfied with performance and size.
scientia
"I suppose Intel could match if they do some further heavy price cuts to at least the 4XXX models."
Haven't you heard of Conroe L and Pentium E series? First are single cores with prices starting from $50, second are dualcore with prices starting from $87. As Intel will have big movements in prices in a couple of weeks (E6700 for ~$180) I wouldn't be surprised to see those CPUs to get even cheaper.
Of course if everything else fails there are still the Netburst line of CPUs with even cheaper prices
If you are a gamer then those E2160/E2140 is not an option. As shown HERE those CPUs don't make competition to X2 3800+, hardly match X2 3600+ and E2160 fall short of Core2Duo E4300 by a good margin while having the same clock speed.
enumae
"MoDT is Intel's mobile on Desktop."
Most of Intel's MoDT motherboards are microATX rather than mini-ITX. Also, none seem to support C2D. You need at least Core Duo but I'm not showing any Yonahs on NewEgg. Has Intel stopped making these?
"Well what about the Celeron 400, or the Pentium 2000's?"
Celeron is single core. Mini-DTX supports X2. X2 3600+ runs only $63. The only dual core on NewEgg that is comparable is the 2.66Ghz Smithfield. However this processor would blow the power requirents. I think either C2D or Core Duo would work if Intel lowers the price.
"A quick google search and I found a few."
In this second set there are some mobile and desktop mini-ITX boards that support C2D. I don't think these would quite stack up against the AMD mini-DTX boards. However, if Intel wants to compete in this area they could catch up.
AMD has started to fight back flawed benchmarking ...
here
I really liked the presentation, this may come as a shock but I actually didn't knew AMD has specint_rate leadership in any benchmark over Kentsfield...and all that because of those darn Intel banners stating two digit performance leadership over Opteron. The reality seems to be a little different.
The reality is that for the time being, there is NOT ONE real-world benchmark that any AMD cpu can perform better than C2D or C2Q cpus. Whether u look for absolute performance or performance / watt Intel is the indisputable leader and will remain unchallenged for the next 1-1.5 years. (I believe that AMD will go BK in 9-12 months, so Intel will remain unchallenged for ever, but this is another story I guess).
Whoever believes that AMD can perform better than an Intel cpu is simply living in a dream world, or he is in the payroll.
I couldn't agree more with that:
"The firm has spectacularly failed to execute on every major product launch across the board, bringing slow products to the table late and earning the ire of journalists and punters alike. "
If I was the leader of AMD I would tell everyone to stfu and to bust their asses to make a better product. That is what AMD should do now and stop whining about benchmarks. This is another proof that they will go BK. They simply lack the understanding of what they should be doing atm. The target is there: It is called "Core2" architecture. When they stop the "witch hunting" they maybe (just maybe) escape BK and have the opportunity to make better products.
Scientia, could you move comment above this one to the Sharikou's blog please.
Scientia-
I'm very diasspointed that you deleted my comment as well. I offered constructive criticism, expressed some disappointment in the direction the blog is headed, and you felt that unworthy of being on the board?
I have always maintained a respectful tone here, have avoided falling into fanboy rhetoric on either side of the fence, yet you delete away? What is up with that?!
I would have sent this as a private e-mail, but there was no contact info on your main page- so feel free to delete this after you read it if it so humors you. Enjoy your one-sided world sir- not sure I want to participate anymore.
gdp77
"The reality is that for the time being, there is NOT ONE real-world benchmark that any AMD cpu can perform better than C2D or C2Q cpus."
This is not true. It is true that C2D is well ahead on single socket performance, however, AMD does win some benchmarks in dual socket and up.
giant
Why are you posting a secondhand reference to Ou's article? The original is here. Feel free to offer any opinion about it.
I am also baffled how anyone could defend someone so obviously biased in Intel's favor. Let's contrast what Ou himself says:
Ou: I’ve seen benchmarks get cherry picked and twisted before but this is just outrageous.
Ou: The fact that Intel is comparing twice the number of cores to AMD is ABSOLUTELY fair
Ou: It doesn’t matter that there is a compiler advantage for Intel
That sounds insane to me; does it make sense to you? Also Ou says:
Ou: AMD is deliberately leaving out Intel’s best scores
Ou then leaves out the SpecFP_rate scores which are AMD's best.
I have lost all respect for Ou.
Gdp77, Opteron does have the highest SpecFP_rate score if you compare with an equal number of cores for 4 cores or higher. AMD also wins some of the SpecInt_rate. Enumae already listed these scores.
Giant and Gdp77, AMD won't go bankrupt in Q2 08 much less any sooner.
AMD does seem to have a solid strategy with DTX. There is no indication yet that Intel is going to try to match. This should get AMD processors back into mid to low desktop systems late this quarter.
AMD's mobile is doing ok but it gets much more competitive in 08. There is no indication yet that Intel is responding to this.
The R700 series should be much more competitive on 55nm beginning Q1 08.
AMD's 690G chipset is ahead of nVidia's. There is also no indication that nVidia will create its own mobile chipset to match what AMD will have in 08.
AMD will have its own 45nm in 08. Also, this will be the first process where AMD has 100% of IBM's process technology. FAB 36 will be at maximum capacity by mid 2008 and FAB 38 should be ramping 300mm as well. There is a big jump in terms of effiency when AMD has two 300mm FABs. This means that AMD should have its production under much better control by late 2008.
What about these things suggest that AMD will go bankrupt in 2008?
Scientia after all these delays after delays in AMD's execution, what makes you so sure that R700 or 45nm will be presented on time?
Considering the BK arguments, let's wait and see Q207 losses and we'll talk again.
dr yield
" I think deleting Core2Dude's post on your past predictions that have turned out wrong lowers the quality of your blog substantially. You challenged him to produce evidence of past mistaken predictions- and he did."
I have a copy of every post that anyone makes including any that are deleted. I have checked over Core2Dude's posts but have not found any that referred to my predictions. As I recall, the exchange was with Mo.
However, I posted a compilation of Mo's, Gdp77's, and Ho Ho's comments about my predictions. These are the only one's I know of.
Core2Dude's (and other's) posts that typically get deleted include insults. Sometimes I edit these and repost; sometimes I don't. For example, in one post Core2Dude said that at the INQ Charlie was accurate but Theo was not but he did so in a much less appropriate fashion. More of his posts would stay up if he would leave out the words "fanboi" and "@asshole" and avoid personal attacks.
However, I deleted CK's post for the same reason which started with, "May I laugh at you Intel fanboys?". That was an obvious flame and was uncalled for. However, then I had to delete posts from Poke and Ringinheat that quoted this nonsense even though they didn't start it.
General rule: When someone is being insulting, do not quote them. Also, don't return insults. This only increases the number of posts that may have to be deleted.
gdp77
"Scientia after all these delays after delays in AMD's execution, what makes you so sure that R700 or 45nm will be presented on time?"
You've made a very good point; there is no way to know. R600 has problems because of high power draw (due to 80hs) and poor drivers. Barcelona is not only late but slower than I expected. AMD could always stumble and not deliver a 55nm R670 in Q1 08 and they could have problems with 45nm. It's always possible that they could fumble with their mobile processor or new mobile chipset. And, if Intel truly does have an advantage with high K then AMD will lag behind until at least later in 2009.
However, I'm not convinced that Intel's seeming indifference to Immersion is really about the quality of the process. I think it would be considerably harder for Intel to get up to speed with Immersion because of its greater number and geographically separated FABs. I think this is a headache that Intel is rightly avoiding. But I don't think we can say that it won't benefit AMD.
IBM also seems to be becoming more fully joined with AMD. IBM is going to support socket F and they will be complete partners at 45nm. Remember that back in 2002 there was no partnership at all and AMD paid IBM something like $10 Million for SOI technology. However, today IBM has three other process partners including Sony and Toshiba. It has been suggested that "asset light" also refers to the use of IBM's research faciities rather than having its own research FAB.
I really can't say that AMD's future offerings will be on time or top quality but it does look like their DTX strategy will help at least some.
"Considering the BK arguments, let's wait and see Q207 losses and we'll talk again."
I doubt Q2 will be much different. You'll need Q3 numbers at least.
However, I just don't know if bankruptcy is actually possible. I've been hearing some rumors about Microsoft. These include rumors that MS executives realized that if x86 were gone that MS's clout would about 10% of what it is now. And, a few other things like that MS realized that Intel was trying to phase out x86 when it introduced Itanium and that the x86 emulation was not as good as Intel had originally suggested.
Also, that MS realized that linux is getting better all the time. This is true both in terms of consumer level interfaces and installation and true in terms of server and enterprise level OS's and programming. Also, that MS's main advantage comes from the fast pace of hardware advance. In other words, fast turnover of hardware means more copies of Windows and Office sold.
So, in looking back there is the suggestion that if AMD hadn't created K7 then not only would Itanium have split the market and greatly reduced MS's influence but the slower pace of hardware would also have meant many Billions less in revenue to MS while having no effect on Intel's revenues. There have also been some rumors that Intel would have begun gearing its development in favor of Unix to further weaken MS's server OS influence.
So, it has been suggested that MS very badly wants and needs AMD to stay in the game, not only to counter Intel's efforts to increase its clout at MS's expense but also to maintain a fast pace of hardware advance so MS can continue to sell lots of software.
MS has tons of cash (a lot more than Intel) but the question would be whether they could actually use any of it to help AMD. I'm not an expert on finance but it seems like it would be possible for one of the wealthy MS executives to set aside some money which could be reimbursed by future stock options from MS. This money could go through some investment vehicles to ultimately provide security for further loans to AMD. This would have the effect of quietly maintaining AMD's solvency without obviously violating any laws.
Or maybe these are just baseless rumors and MS doesn't really care if AMD goes bankrupt.
core2dude
"You know Scientia, what I don't understand is, why people keep debating with you even after your most ridiculous and false claims."
I don't know. Comic relief perhaps. Of course, I do try to limit my claims to either false or ridiculous rather than both at the same time though.
"Why post compilation? Why not original comments? "
To avoid having long posts grow even longer. I would have to quote the originals and add my comments to them making them about 3X bigger.
I would also have to explain errors in context as well as other other errors and my replies would get even bigger. This then threatens to derail the comment thread. I have had occasions when I've had to delete half a dozen comments after I posted a new article because the comments got derailed right away.
"The way any reasonable person would interprete this is that, you deleted this assuming no one had read these, but then chickened out after a number of people posted that they had already read it."
Did you know that when I first started my blog I had active control turned on so that I had to approve the posts? And, did you know that during this time I did not approve posts that were simply complimentary? Someone would say, "Hey, great post scientia" but nothing more so although I read them I did not post them.
It takes quite awhile to do a compilation of three long posts plus responses so I was still in the process as the complaining posts about deleting were being made. I suppose I could have left the posts up until I responded but then I would run the risk of other replies to the originals.
However, I think you can use a bit more common sense than you have up to now. If I were only interested in deleting posts that insulted me why did I delete CK's post that only insulted Intel enthusiasts? Also, if I really had a problem admitting being wrong why would I add to the list? And, I have corrected my own posts many times at AMDZone. I have also edited my original articles here when people pointed out errors.
"As usual, you have zero credibility, and you are a lier. No other excuses!"
I'm not going to be called a liar on my own blog. This is your one warning. Don't do it again.
"You know Scientia, I never throw insults at people who engage in respectful conversations."
That may be true but the problem seems to be that you consider anything that disagrees with you as disrespectful.
"The problem with you is that you insult others at your will, and then expect respect in return.
When did I insult you? Or when have I made a practice of insulting roborat? And, when have I condoned this kind of conduct on my blog?
Yes, I have been critical at times of George Ou, Ed Stroligo, Anand Lal Shimpi, and Tom Pabst. I assume you are not one of these.
Further, I hardly think it is out of line to criticize someone who rants about deception and then argues that it is okay for Intel to compare its quad cores with AMD's dual cores. A more honorable (or perhaps intelligent) author would understand that that argument would be cost (because AMD would need two sockets) rather than performance.
By Ou's ridiculous logic AMD could post benchmarks of AMD 4 socket systems compared to Intel dual socket systems because according to Ou it is Intel's own fault that it doesn't have C2D on 4-way. However, if AMD actually did this you can bet Ou's next article would be screaming about AMD deception again.
Scientia
Further, I hardly think it is out of line to criticize someone who rants about deception and then argues that it is okay for Intel to compare its quad cores with AMD's dual cores.
The point is simple, AMD does not have Quad-Cores while Intel does, and Intel is going to market this heavily until AMD has a competing product.
With that said, business is not always about apples to apples or even being fair, it is usually about the best offerings from the competing companies, and in this case (currently) it is the 2P segment in which Intel has a distinct advantage.
In all honesty, if AMD had Quad-Core and Intel did not, do you believe that AMD would not capitalize on the oportunity to tell the world how well it performs against it's competitor, regardless if it's going up against Dual-Cores?
I can not believe that AMD is above making money, or capitalizing on a competitors lack of products.
-----------------------------
By Ou's ridiculous logic AMD could post benchmarks of AMD 4 socket systems compared to Intel dual socket systems because according to Ou it is Intel's own fault that it doesn't have C2D on 4-way.
That is not the same thing, Intel has a Dual-Core for 4P and up, it's not as if they don't have a product offering in which to compete, they just don't have a product that can compete (SPECint or SPECfp).
-----------------------------
However, if AMD actually did this you can bet Ou's next article would be screaming about AMD deception again.
I will not disagree with this statement at all, but it also doesn't change what I said above, I can not believe that AMD is above making money, or capitalizing on a competitors lack of products.
Both AMD and Intel are here to make money.
-----------------------------
If you needed computational power today in a 2P workstation wouldn't you compare a 2P AMD system to an Intel 2P system and while comparing the two, wouldn't you also factor in available Quad-Core processors?
It's not unfair, just available product offerings from competing companies.
Scientia
I have no intent to be rude, but do you have any thoughts to my post?
enumae
No, I would say that Ou's article is still nonsense. As I said before it isn't a performance issue but more of a cost issue.
It is entirely possible that you can get a Clovertown server for not much more than an X2 Opteron server with the same number of sockets but I'm sure it would be more. So, you can't separate out cost.
To simply claim that quad core versus dual core is fair because AMD doesn't have quad cores is absurd. However, if you are comparing power at the same price then that would be valid.
Logic and debate is a very old subject that goes back to the Greek city-states like Athens. If something is logically sound then you should be able to make the same counter argument without running into problems.
For example, if it is logical to compare A to B then it should also be logical to compare B to A. If B cannot be compared to A then why would it make sense to compare A to B?
Before the Soviet Union built its AN 124 Condor and then the stretched six engine version there was a rather absurd debate about what was the world's largest aircraft. The Guiness Book of World Records listed 3:
Hercules flying boat - wingspan
Boeing 747-400 - weight
Nasa Guppy - volume
The problem was that all of these were one way arguments. Like this:
Hercules wingspan > C5
747 weight > C5
Guppy volume > C5
But look what happens when you reverse the argument as you should be able to logically do:
C5 wingspan < Hercules
C5 weight > Hercules
C5 volume > Hercules
C5 wingspan > 747
C5 weight < 747
C5 volume > 747
C5 wingspan > Guppy
C5 weight > Guppy
C5 volume < Guppy
So, you can see that C5 wins 2 out 3 for each comparison with the three contenders.
So, again we have a one way comparison:
Clovertown performance > X2 Opteron
However, logially then AMD would be able to reverse this comparison and print its own graphs showing:
X2 Opteron power draw < Clovertown
X2 Opteron price < Clovertown
The flaw in your argument (and apparently Ou's) is that you would claim that this isn't a valid comparison because Intel has dual cores. However, if you can't reverse the comparison it isn't valid.
If you still don't understand this then try an unrelated example.
Fred claims that the banana is the perfect food because it is easy to peel and eat.
Jim counters that although the sweet potato has to be cooked it is actually several times more nutritious.
Fred then claims that it doesn't matter if a sweet potato is more nutritious because beans are nutritious too.
It makes no sense for Fred to try to inject beans into the argument because the only issue is the banana versus the sweet potato. So, what Fred is trying to do is artificially limit his comparison to only those things that favor his preference. In reality, all cross comparisons are valid.
Scientia
Ou's article was a lash at AMD, no doubt about it, but I am not here to defend Ou.
I am here because the way you view his comments leads me to believe that you feel Intel has decieved the public by comparing it's best offering against AMD's best offering, when in fact it is just marketing and AMD is not above this.
------------------------------
Anyone who has the means to buy a 2P server is going to look at AMD vs Intel, Quad vs Dual until AMD has a competing product.
Do you disagree with that statement?
------------------------------
In regards to The Guiness Book of World Records listed 3:, I can not agree with this view. AMD has Dual-Core, Intel has Quad-Core, each of which are the companies best offerings, to say you can't compare them is wrong.
Thats how the world works.
The basis for the presentation from Intel was available performance from Intel and it's competitor and as such they compared Dual to Quad, again its about marketing.
Can you claim AMD wouldn't do this if they had Shanghi (8 cores) before Nehalem (8 cores)?
And if AMD did do this, wouldn't you be able to understand that its all about marketing?
I will repeat this again, both Intel and AMD are here to make money.
------------------------------
However, logially then AMD would be able to reverse this comparison and print its own graphs showing:
X2 Opteron power draw < Clovertown
X2 Opteron price < Clovertown
The flaw in your argument (and apparently Ou's) is that you would claim that this isn't a valid comparison because Intel has dual cores.
However, if you can't reverse the comparison it isn't valid.
Your initial reversal was flawed because they have dual cores for 4P and up, and now you are going to claim new reversals and conclude that I would claim it is not valid...
I am sorry, but that is a very large assumption on your part. Why would that bother me, or how do you conclude I would find it invalid? It is marketing!
You have seen this haven't you?
You don't here me complaining do you?
------------------------------
The only thing you are showing is that your perception of my post is wrong.
I am making a very simple argument that Intel has a better product than AMD and there marketing it, and will continue to do so until AMD has a competing product and that it is simply about marketing.
------------------------------
Sorry if I have a slight hostile tone, but you making assumptions about me or my perception of marketing realy bothers me.
Look at what I found :)
Browse through the results and see if you notice anything.
While there not all outdated, some of the benchmarks are... George Ou needs to do the same thing to Intel that he has done with AMD.
Think he will?
enumae
What I'm saying isn't that complicated. It is invalid to compare best to best unless you also compare price and power draw. Performance alone is meaningless.
Glad to see your back.
----------------------------
Scientia
It is invalid to compare best to best unless you also compare price and power draw.
Ok, I may have misunderstood you initial P.O.V.
Performance alone is meaningless.
Not always.
--------------------------
I do have a question though, why didn't you answer my questions in your response?
Should I just not ask any?
enumae
What was your question?
BTW, why would you complain about that link. It is only to a utility to calculate money saved over time by having lower power draw. It doesn't even refer to any specific processor.
I suppose the questions could be somewhat irrelevant now that I understand your POV.
In regards to your question, like I said it really bothered me that you had a preconceived notion on how I would claim something as invalid, so my post had some frustration in it and there fore may not have been on point.
The one that does apply though is if AMD had Shanghi (8 cores) before Intel had Nehalem (8 cores) do you think AMD would be above comparing the their 8 core processor against Intel's 4 core?
enumae
"if AMD had Shanghi (8 cores) before Intel had Nehalem (8 cores) do you think AMD would be above comparing the their 8 core processor against Intel's 4 core?"
I don't know. If processors are simlar in a certain area then you can ignore that. For example if one processor draws 89 watts and another 95 then that is close enough to ignore. You couldn't ignore it though if one were 95 and the other were 120 watts. Similarly, you could ignore small differences in price like $300 versus $330.
You have to have some basis for comparison otherwise you could compare a single core mobile processor to a quad core and graph only on the basis of power draw.
It only makes sense to compare quad to dual core if they have similar power draw and cost. The same would apply to 8 core Shanghai with Intel quad.
Scientia
It only makes sense to compare quad to dual core if they have similar power draw and cost.
Ok, then that means we can compare the Opteron 2222SE (3.0GHz - 120W - $873) to Intel's 5345 (2.33GHz - 80W TDP - $890).
Does that seem fair?
The initial Intel comparison should be valid (according to your POV) on or around July 22nd when the price is supposed to be about $744 for the X5355.
Would that be correct?
enumae
If you could get a more powerful processor with lower TDP for the same price, isn't that what you would buy? That is fair.
Scientia
If you could get a more powerful processor with lower TDP for the same price, isn't that what you would buy?
Yes.
Post a Comment