Friday, May 11, 2007

How We Got Here -- The X86 Industry In Perspective

In my last two articles I argued both sides of the Intel/AMD aisle to get a measure of the people who comment. With that perspective in hand it's time to get back to real posting. I'm going to look in detail at X86 development over the past twenty years to see how we got here and where we can go.

AMD has a very noticeable four year develop cycle. AMD released the RISC 29050 processor in 1990. AMD then reorganized the RISC team and started working on a new X86 processor. It appears that mostly what AMD did was take the 29050, strip down the register set, and put a micro-code X86 instruction translator on the front end. This effort took five years and the K5 was released in 1995. The performance was not spectacular but for AMD's first completely new X86 design it was a reasonable result. In fact, the K5's biggest problem was not its performance but its inability to achieve higher clock rates. This appears to be more related to process technology than chip architecture. Four years later in 1999 the K7 was released and four years after that in 2003, the K8 was launched. We are now looking at the K10 four years after K8. Clearly, this is a four year development cycle. K6 is an exception in 1997 but that was because it was created by the separate Nexgen design team. The pattern since K5 is a four year major architecture cycle with a smaller team doing updates in between.

It only took Intel three years to get from 80386 to 80486 but the bar since then has also been four years. It was four years from 80486 in 1989 to Pentium in 1993. However, the extraordinary thing about Intel is that it was running not one but three separate microprocessor design teams. The second team was working on the iAXP432/i960 line and the third team was working on the VLIW i860. There is no doubt that the i860 team was put on Itanium development. Presumably, the i960 RISC team was reassembled into the Pentium Pro team. We know that Pentium Pro had a RISC core and the development would have had to start in 1991 which was during the time that Pentium was still under development. The four year pattern at Intel picks up again with Williamette. It is four years from Williamette in late 1999 to Prescott in late 2003 but then five years to Nehalem in late 2008.

Williamette was an exception in that it appeared five years after Pentium Pro but, curiously, it wasn't really complete. Williamette's performance suggests that it was rushed, yet there should have been more than ample time in the five years to get the design right. Since Northwood was a good design we can probably work backwards four years to late 1997 which is fully two years after Pentium Pro was finished. The best guess is that Intel expected Itanium to take over the top end on the desktop. When they realized in 1997 that Itanium was going to be late they hastily assembled the Williamette team and rushed it out. Since Northwood's launch probably coincided with the 130nm process it is probable that the Williamette team was assembled in early 1997 and that Intel decided to skip 180nm for Northwood. Northwood just two years after Williamette

The Williamette team went on to create the Northwood and Prescott designs but then presumably abandoned two years worth of work on Tejas in 2005. This cancellation would explain the extra year to Nehalem which presumably would have been 2007 if Tejas had not been canceled. I have a very strong suspicion though that any of the Tejas work that was useful was applied to Nehalem II. Obviously this wouldn't be the original Nehalem but this could indeed mean that C2D version of Nehalem could be more than just a two year refresh of the Conroe Design. Apparently, a third (or fourth) design team was assembled in Israel in 1999 to work on Timna which was a cost reduced version of PIII. This was a good design but was sunk by the onboard RDRAM controller in 2001. This team apparently then began working on Banias which was released in 2003 as Pentium M. Banias too has a look of being rushed. The most logical conclusion is that the Haifa team was able to use a lot of the Timna information to put together a reasonably good design. There is no doubt that Banias is an upgrade of PIII but it reached a better state of polish with Dothan.

However, Conroe is a mystery. The quality of the design suggests the same four year cycle but this would imply that work began on C2D in 2001 when Timna was canceled. I've seen some suggest that Conroe was an upgrade of Yonah but this is clearly not possible; the gap between Yonah and Conroe is far too large and the time between releases is far too short. I've seen others credit the Conroe design as a derivative of Banias but this too is suspect because both Williamette and Banias had deficiencies when rushed and nothing about the Conroe design stands out as a deficiency. There is however one clue in Intel's history that is peculiar. We know that Intel specifically avoided patenting all of the tricks that went into Banias. Looking back at this fact I'm skeptical that the successor to the mobile versions of PIII would cause this much concern for Intel. It is a certainty that Intel was aware of Prescott's problems by 2002. However, it seems more likely that after design work on Prescott had been going on for more than two years that Intel was having doubts in 2001 and that the work on Banias was folded into a larger design project including Conroe. We could assume that Intel realized the problem with Prescott in 2002 and was impressed with the completed Banias design and then began working on Conroe however this wouldn't seem to explain Intel's unusual move in hiding the patentable characteristics and, again, the Conroe design seems far too polished.

Since the separate Penium M line has now been abandoned in favor of the C2D Merom, we would seem to need one less design team. However, it is hard to believe that Intel which has always had three design teams is now down to just two: the Itanium and the C2D line. We either have to assume that the added items like CSI are taking an entire extra team and two teams are working on C2D or perhaps that the third team was put on special projects like the TeraFlop chip. I find it unlikely that Intel cut out the third team as a cost saving measure since dropping this many engineers would surely have been noticed. If the third team is indeed working on special projects this could mean that Intel will have new surprises waiting in the wings to be added to X86 designs in the next few years. Also, as I've already mentioned, Nehalem may indeed have extra tricks in its design that have been inherited from the aborted Tejas work. An onboard memory controller shouldn't be a problem since Intel has already worked on this for Timna and obviously has been making full blown memory controllers for years in its chipsets.

In addition to the single main design team that AMD has maintained they have had another design team doing chipsets. It was four years from the 750 chipset to the 8000 chipset with the upgraded 760 and 760MP chipsets in between. The importance of chipsets for AMD is difficult to overstate. AMD's K6 processor was fully capable of dual socket operation. However, no chipset existed that would allow K6 to do this. It is also clear that without AMD's 750 chipset that it would have been nearly impossible for AMD to get support for K7's Slot A. It is also clear that without the 760MP chipset that Athlon MP's dual socket capability would have gone for nothing as it did on K6. And, there is no doubt that support for Opteron was nonexistent without the 8000 chipset. I think it is clear though that the effort of designing additional chipsets exceeded AMD's abilities and this ultimately led to the decision to purchase ATI. We know that a lack of a proper chipset prevented dual socket motherboards for K6 and that it was only AMD's in-house 750, 760MP, and 8000 chipsets that allowed proper support for K7, AthlonMP, and Opteron. AMD could have skipped the ATI purchase but we would have to wonder if AMD would be able to keep up with new chipsets to keep future chip capabilities from falling by the wayside. I think a similar argument can be made that Intel's primary motive in switching to an onboard memory controller is the geometrically increasing difficulty of designing the Caneland quad FSB chipset.

Finally, we have to wonder just how well it can work for both AMD and Intel to switch to two year design schedules instead of four. For every design that was newly created in the past there was a period of upgrade. Obviously one team had to be working on the new "clean" design while another smaller team concentrated on the upgrade. There is no doubt that for engineers the relative clean slate of a new design was more exciting and offered fewer limitations. It is also the case that a new architecture will demonstrate the largest improvement. In this regard it can be seen that the upgrade team is the perpetual underdog with far greater restrictions and less improvement. The two year design cycle requires dropping two separate teams and blending the upgrade and new design teams together. This makes the upgrade cycle more rewarding because there will be larger improvements. This does create greater restrictions for the new design team because they both have to work within the existing architecture and because not all ideas can be incorporated in the shorter cycle time. However, no really good idea needs to die because each good idea can potentially be included in the next upgrade or the one after. Apparently Intel has already moved in this direction while AMD has responded by making the CPU core modular. AMD's modular die allows some "black box" modification. In other words, AMD can always assign engineers to trouble areas and these engineers can make the necessary changes to that module without worrying about the rest of the die. This ability to isolate problem areas saves a great deal of iterative design work. There is no indication yet of modular design from Intel but then Intel probably has the manpower to deal with this and may prefer the greater flexibility of a non-modular design.

How we got here is one of the important ways of seeing where we are going. I have confidence in the K10 design because it matches AMD's historic four year design pattern. Likewise Nehalem at five years beyond Prescott matches a similar pattern. Obviously though, until Nehalem appears we won't know if any of the Tejas design work made it into Nehalem, nor do we know much about it other than the Integrated Memory Controller and CSI. However, given Timna and Intel's great experience with chipset memory controllers I can't imagine that the IMC would be a problem. Likewise, Intel's PCI-e initiative in 2006 should generate the necessary hardware to make CSI a reality. I suppose we will have to wait until AMD's K11 to see just how well the new modular approach is working.

My final thought is that the doubling of SSE speed by both C2D and K10 can't be very good news for the Itanium design team. Clock speeds shot upward from K6's 166Mhz to K7's 1000 MHz in just two years while Intel matched pace with Pentium II and then Pentium III. I believe there is no doubt that this more than anything caused Intel to miss its initial performance objectives with Merced Itanium. One has to wonder if a Nehalem with SSE4, IMC, and CSI might be a serious threat to the next generation Itanium. Nor is Itanium alone. Eventually, even IBM's mighty Power will come under siege from powerful but cheaper X86's.

86 comments:

enumae said...

Scientia

I have a question, why is it that most people look at Itanium as a failure?

And how much more expensive is Itanium than a comparable X86 system?

Just a thought, but if Itanium was capable of reaching 2.0GHz or 2.2GHz...

Itanium at 2.0GHz would be about 7% faster at SPECfp, and SPECint.

Itanium at 2.2GHz would be about 18% faster at SPECfp, and SPECint.

The results are based on using 100% scaling from results found on SPEC.org for the 1.6GHz/24MB 9050, 16 Cores.

And comparing those results against projected numbers of K10, +10%, and at a clock speed of 2.9GHz.

Your thoughts would be appreciated.

Joerg said...

I do not really believe that cheap x86 will substitute Sparc or Power. Compared with the RAS features of high-end processors x86 looks like "best-effort" computing.

InTheKnow said...

I think you have missed something fundamental by assuming that Intel has eliminated a design team, or even has the option of eliminating a team. The whole "tick tock" model is based on having 2 design teams that are 2 years out of sequence. That gives you a full 4 year design cycle but puts out a new architecture every 2 years. Itanium remains an independant entity with it's own design team.

Since Nehalem is due out 2 years after Conroe, it would seem that the "tick tock" process actually started with Conroe despite being announced at about the same time. With the changes being incorporated in the Nehalem design it would seem reasonable to expect the design to have started in 2004. Which matches up well with the Conroe design starting in 2002. In light of the huge roadmap shifts and the numerous misses by Intel in 2005 it might be possible to argue that Conroe design began in 2003 and was responsible for those changes and misses as resources were reallocated to pull the introduction in by a year without sacrificing the quality of the product.

In defense of this speculation I offer comments form a report on Otellini's announcement of the "tick tock" model below.

Intel has been watching the growing popularity of AMD and knows that the latter means business now. Rather than embark on a head-to-head processor pricing war with its rival, Intel must return to the quick development habits it used when producing its Pentium family of chips, Otellini said. Intel backed off that pace after producing the Pentium 4, and soon began to lose marketshare when AMD launched the Opteron chip in 2003.

“We're doing product refreshes every two years, which is the model we invented, and then stopped doing after Pentium 4, shame on us,” Otellini said. “We fell off it-mea culpa, we screwed up-and now we're back on that pace.”


http://www.dqchannels.com/content
/global/107040604.asp

Scientia from AMDZone said...

enumae

"I have a question, why is it that most people look at Itanium as a failure?"

Probably for three reasons:
1. Intel hasn't paid for Itanium's development cost yet.

2. Itanium's sales have missed Intel's projections by a factor of 100.

3. Rather than gaining in the market, Itanium has been pushed out of the workstation market.

"And how much more expensive is Itanium than a comparable X86 system?"

That is why Intel is making Itanium more socket compatible with Nehalem. That is also why IBM is making Power socket compatible with Opteron.

Scientia from AMDZone said...

Joerg

"I do not really believe that cheap x86 will substitute Sparc or Power. Compared with the RAS features of high-end processors x86 looks like "best-effort" computing."

That is true today but not for the future. RAS features have been added to X86. As these increase we will eventually reach a point when X86 is as good as any heavy duty chip. And, it's much greater volume will make it cheaper.

Scientia from AMDZone said...

InTheKnow

"I think you have missed something fundamental by assuming that Intel has eliminated a design team

I think you misread my article.

The whole "tick tock" model is based on having 2 design teams that are 2 years out of sequence. That gives you a full 4 year design cycle but puts out a new architecture every 2 years.

Yes, I said in my article that it wouldn't make sense for Intel to eliminate a design team and that the second team could be working on C2D as well.

However, on the issue of tick tock your understanding of this is incorrect. Tick tock is an enhanced upgrade model rather than a design model. In other words, you boost the upgrade beyond what you would normally do rather than (as you claim) fitting in 2 development cycles in 4 years. This would not actually work since waterfall would destroy your design efforts in just two or three cycles.

"Itanium remains an independant entity with it's own design team. "

As I said.

"Since Nehalem is due out 2 years after Conroe, it would seem that the "tick tock" process actually started with Conroe despite being announced at about the same time."

Yes, I agree except Nehalem might benefit from some of the canceled Nehalem work.

"With the changes being incorporated in the Nehalem design it would seem reasonable to expect the design to have started in 2004."

Doesn't this contradict your earlier claim that tick tock was still on a 4 year design cycle?

" Which matches up well with the Conroe design starting in 2002."

No, my bet is that the Conroe design started in 2001. This would give Conroe 4 years of development.

"In light of the huge roadmap shifts and the numerous misses by Intel in 2005 it might be possible to argue that Conroe design began in 2003"

You could argue this if you know nothing about processor design. 2003 is absurd.

" and was responsible for those changes and misses as resources were reallocated"

No. In 2003, Intel was running 4 design teams:
1. Itanium
2. Nehalem
3. Conroe
4. Whitefield

plus 2 upgrade teams:
1. Dothan/Yonah
2. Prescott/Smithfield/Tejas

The Whitefield team was Indian so it didn't pull any resources. Apparently, the Tejas upgrade was delayed a bit and became Tulsa.

"and was responsible for those changes and misses as resources were reallocated to pull the introduction in by a year without sacrificing the quality of the product."

What do you think was pulled in by a year? Even if we assume that Core 2 Duo wasn't started until Banias was finished that would be 2002, not 2003. And, 2002 would only have given three years development time which doesn't seem like enough. As I said in my article I'm thinking that Intel avoiding patenting the Banias architecture because they had already started developing Conroe in 2001.

"Intel backed off that pace after producing the Pentium 4,"

Not true at all. Northwood was 2 years after Williamette and Prescott was two years after Northwood. C2D was just a little over 2 years after Prescott. So, Otellini's statement is bogus.

“We're doing product refreshes every two years,"

Yes, notice the word "refresh" rather than the phrase "entirely new architecture".

" which is the model we invented, and then stopped doing after Pentium 4,"

Actually, it was done with Northwood but also done with Dothan and Yonah.

abinstein said...

I don't think you get the right ideas on P-pro, P-4, and Conroe.

Conroe's microarchitecture is very similar to that of P-pro (p6), except added with a lot of small better stuff. There's no doubt that it is clearly an evolution from Yonah. I'd say Conroe is about as much different from Yonah as Prescott is from Williamette.

InTheKnow said...

However, on the issue of tick tock your understanding of this is incorrect. Tick tock is an enhanced upgrade model rather than a design model....

I refer you to intel's own documentation.

Cadence refers to Intel’s strategy of introducing a new microarchitecture linked with a new generation of silicon process technology approximately every two years.

download.intel.com/technology/eep/
cadence-paper.pdf

You decide whether or not they are telling the truth, but that is what they say.

"With the changes being incorporated in the Nehalem design it would seem reasonable to expect the design to have started in 2004."

Doesn't this contradict your earlier claim that tick tock was still on a 4 year design cycle?

No, Nehalem is due out in 2008. 2004 to 2008 is 4 years the way I learned to count them. :)

"In light of the huge roadmap shifts and the numerous misses by Intel in 2005 it might be possible to argue that Conroe design began in 2003"

You could argue this if you know nothing about processor design. 2003 is absurd.

The design came in in 2006. That is a 3 year design cycle. With enough resources I don't see why this is an absurd assumption. Though I will admit my knowlege of the design process is sketchy.

For the record, my background is process engineering across a number of different fields and more years than I want to admit to.

Would you object to stating your credentials in the design area? I just don't want to be in the position of the blind man being lead by the blind.

"Intel backed off that pace after producing the Pentium 4,"

Not true at all. Northwood was 2 years after Williamette and Prescott was two years after Northwood. C2D was just a little over 2 years after Prescott. So, Otellini's statement is bogus.


Unless he is considering the model to be a new architecture every 2 years as is stated in Intel's own document.

I am not going to uphold Otellini as a paragon of virtue that would never mislead the masses. However, you would have to at least provide a reasonable explaination of what Otellini gains by such a misleading statement if you want your dismissal to count as more than wishful thinking.

Fujiyama said...

I really don't understand this belief than Conroe was designed from 2003 or more than a year.
This is just a redesigned Yonah - nothing more or less. Look at Athlon64 vs. Yonah test here.
This chip with faster FSB, bigger cache, micro-ops, higher frequency matches 65nm Core2Duo design.
So it is difficult to believe that AMD was not prepared for such design or they don't pay enough attention (this is bad) to competitive designs.
Yonah was a great chip and 65nm conversion helped to achieve almost 3GHz to beat K8.

Regarding design approach, AMD should become a leader introducing DirectConnect 2.0 architecture, splitting design teams into small groups responsible for CPU "components".
It is already done IMHO.
Barcelona is complete.
The question is which chip: Shanghai or Fusion will be the first DC 2.0 chip with modular design.

Scientia from AMDZone said...

abinstein

I don't know who your remarks were for.

Pentium Pro was cost reduced for Pentium II, then PII was upgraded to PIII. Timna was a cost reduced version of PIII. The work from Timna was used to develop Banias which was then upgraded to first Dothan and then Yonah.

It is clear that C2D is in the PIII family (which decends from Petium Pro) however it is completely impossible that C2D is an upgrade of Yonah. C2D is derived from PIII, Timna, and the very early work on Banias. C2D's development paralleled Dothan and Yonah but it wasn't derived from them.

The gap between Yonah and C2D is far greater than the gap between Williamette and Northwood or the Gap between Pentium Pro and PIII. Again, C2D is not derived from Yonah.

Scientia from AMDZone said...

intheknow

"Cadence refers to Intel’s strategy of introducing a new microarchitecture linked with a new generation of silicon process technology approximately every two years."

This doesn't mention interleaved design teams or a four year development cycle. It simply mentions development in parallel

Yes, as I said in my article I believe Nehalem is the last of the four year architectures.

"The design came in in 2006. That is a 3 year design cycle."

Absolutely not. The design is set aproximately 12 months before release. If the design were started in 2003 then that would be only 2 years of design until the design was fixed in 2005.

" With enough resources I don't see why this is an absurd assumption. Though I will admit my knowlege of the design process is sketchy."

The problem is that changes in the first design ripple over into the next and these then disrupt the second design and cause delays. If you have to wait until the first design is nearly fixed then your overlap is very short. In practice it is impossible to do overlapping clean slate designs. Instead, each successive design is derivative. However, because each design is an upgrade if you boost each upgrade cycle then you end up with a similar increase after two upgrades as you get with a four year cycle. Think of each two years as an enhanced upgrade rather than as a clean slate design.

"Unless he is considering the model to be a new architecture every 2 years as is stated in Intel's own document."

No, it says "new generation", not substantially new architecture. Terms like upgrade, refresh, enhancement, architecture, design, generation, and revision are all very subjective. Was PII really a new generation? Was PIII? Northwood was indeed a signficant upgrade to Williamette and then Prescott was a new architecture. I think what Otellini is trying to say is that P4 ran out of steam before he had intended it to. But, this is a function of the underlying architecture and not the design process itself.

The only examples we have of real delay are Williamette and Tulsa.

Scientia from AMDZone said...

fujiyama

"I really don't understand this belief than Conroe was designed from 2003 or more than a year.
This is just a redesigned Yonah - nothing more or less."


Clearly you are not familiar with either cpu architecture or cpu design. The difference between C2D and Yonah is vast. There is as much difference between C2D and Yonah as between K7 and K8.

"Regarding design approach, AMD should become a leader introducing DirectConnect 2.0 architecture,"

Yes, I would say that AMD has a good chance of holding onto the top server spot in 4-way and up. Single and 2-way is more competitive once Nehalem is released.

" splitting design teams into small groups responsible for CPU "components"."

Yes, this is what a modular core is about.

"It is already done IMHO."

I'm not quite sure what you are saying is already done. If you mean the teams then yes. If you mean the design work then we are close; the tapeout on the first modular core should be within the next two months.

"Barcelona is complete."

Barcelona was nearly complete last year.

"The question is which chip: Shanghai or Fusion will be the first DC 2.0 chip with modular design."

I was thinking that the modular core would appear with DC 2.0. However, AMD changed the schedule a bit when they moved HT 3.0 up to 2007. At the moment I not certain whether Shanghai is the modular or just a shrink of K10. I'm certain that AMD will clarify this in June.

lex said...

A nice extraction of the history of x86.

The bottom line is INTEL with its far greater number of design teams and infinite well of money, tons of manufacturing and being ahead by a year and half can afford to throw lots of silicon at multiple things.

The result... AMD will be a little player even with Barcelona.

Too little to late.

All will be absorbed by x86 including the power business its only a matter of time and money of which INTEL has lots.

gdp77 said...

The result... AMD will be a little player even with Barcelona.

Postponing desktop barcelona for Christmas means that AMD is facing complete extinction imo. There will be a 1-2 month windows until Penryn arrives, which won't allow AMD to regain their lost market share.

So there is a very good chance that AMD won't even succeed in being a "little player". There is a good chance of being a "non-player". I hope I am wrong though...

Scientia from AMDZone said...

Penryn won't have nearly as much effect as you think it will.

Even with Penryn Intel will trail on quad core 2-way and up. It will trail on 4-way and up with either quad or dual core.

This leaves the areas where Intel might be competitive with Penryn as single socket dual or quad core and dual socket dual core.

In other words, mobile Penryn (45nm Merom) should be competitive.

Wolfdale (45nm Kentsfield) might be competitive.

Yorkfield (45nm Clovertown) will not be competitive.

However, Nehalem is another issue. We have no idea how big of a bounce Nehalem might get.

Real said...

"No. In 2003, Intel was running 4 design teams:
1. Itanium
2. Nehalem
3. Conroe
4. Whitefield

plus 2 upgrade teams:
1. Dothan/Yonah
2. Prescott/Smithfield/Tejas

The Whitefield team was Indian so it didn't pull any resources."

This is incorrect according to my understanding. Whitefield design was begun in Israel and transferred to India midstream. The processor that was transferred was an unworkable deisgn which is why the Israel team was happy to be rid of it. The Indian team of course stumbled with completing the project and took the blame for it. However, the Israel team was then reassigned to help with C2D.

Secondly, I believe your statement about C2D not being a Yonah derivative is mistaken. I believe even the Israel design engineers will tell you that. As for Timna, bringing up the name now is an attempt to deflect blame for a project that was an utter failure. I doubt the learnings extended much beyond "don't make a product the market does not demand" despite the current attempts to spin is as the forerunner of Yonah.

gdp77 said...

Even with Penryn Intel will trail on quad core 2-way and up. It will trail on 4-way and up with either quad or dual core.

This leaves the areas where Intel might be competitive with Penryn as single socket dual or quad core and dual socket dual core.


Apparently u haven't realized that it was Desktop superiority + low prices that allowed Intel to get in one quarter the market share that took AMD 2 full years to gain.

AMD needs desktop superiority, something that K10 could give if it would ship in 3Q07 like it was scheduled.

Also, in my opinion, failing to ship desktop K10 on time, means that AMD will also fail their 18 month die shrink cycle.

By the way, I would like some info regarding the failure of AMD to ship desktop K10 in Q307... Is server K10 on schedule or there will be delay there also?

gdp77 said...

Sorry about double posting, but I was just informed about the upcoming R600 failure... R600 will not be able to compete against G80, while drawing 100 more Watts of power that its rival chip...

so

1) AMD lost all their market share that they fought so hard to gain
2) Their depth is unbelievably high with no way in the near future to recover
3) AMD failed to deliver K10 on schedule
4) AMD failed to release a competitive gpu

Conclusion : Intel will be a monopoly sooner than some of you think...

This is my perspective on the x86 industry atm.

enumae said...

Scientia

I am not sure how you have drawn a conclusion that Penryn will not be competitive in the 2P space against K10.

I have run several scenarios (based of published results) in which Penryn quad-cores will still be faster in SPECint, but slower in SPECfp, as is today, and as you somewhat acknowledged in an earlier post.

4P and above, well we have not seen any Tigertown benchmarks with its quad FSB, theoretically it should be about double what we have now, and would then allow AMD to possibly take the performance lead in both SPECfp, and SPECint due to better scaling.

But all in all, it will come down to price/performance. Like I have said before, as have others, if K10 is not significantly faster they will continue to lose money, and no, DTX and the lackluster performance of ATI graphics will not be enough to keep them from losing money.

The latest news from AMD is Barcelona will launch late July or early August, while it has also been said that Penryn dual-cores will launch in the third quarter.

Volume will be tough for both Intel and AMD, but I would have to believe that Intel could ramp faster, but that is just my opinion.

My main question, and why I am posting is what is AMD going to do against the news that Intel plans more price cuts on July 22nd considering that they are already loosing money with current prices and they will have to respond?

Thanks.

enumae said...

Scientia

Thanks for answering my questions about Itanium.

Scientia from AMDZone said...

real

Whitefield is derivative of Banias so it was originally part of the Haifa work. Essentially, Whitefield would have been a true quad core Dothan with the full version of CSI. This failed primarily because they could not get CSI to work. Without CSI, the Whitefield design overlapped too much with Cloverdale so it was canceled. Since Whitefield the specs for CSI have been gutted and what will eventually be released will only be a fraction of what the Indian team had planned. This is why CSI will only be roughly as fast as HT 1.0 rather than competitive with HT 3.0.

Again, it is impossible for C2D to be derivative of Yonah. Less than six months separated Yonah's release from C2D. In fact, while Yonah is simply a dual core Dothan, C2D is not derived from Dothan either. C2D's design goes back at least to the original Banias design but probably was begun before Banias was actually finished.

I think you are misunderstanding about Timna. Banias was not Timna without the IMC. However, Banias did make use of the design reviews and research that had been done for Timna. Timna also makes it clear that C2D could have been released with an IMC if Intel had chose.

Scientia from AMDZone said...

Okay, how to answer this.

gdp77

I'm not aware that K10 is late. The release timeframe that I have always seen is late Q2 for launch on server with perhaps low volume. This seemed to be primarily to fill early HPC contracts.

Q3 was supposed to expand the volume and delivery of server chips in both quad core and dual core. I expected the FX chips to be released here too. Q4 was supposed to be desktop. I believe AMD is still on this schedule.

Also, as far as I know, AMD is still on schedule for 45nm at mid 2008. This still gives them until mid 2007 to tapeout Shanghai. I believe Shanghai should be DC 2.0 but perhaps not. I suppose AMD will say more about this in June.

In terms of market share it isn't clear that AMD has lost anything yet. Markets tend to be elastic for up to two quarters. In other words, whatever AMD has not gotten back by Q3 is a genuine loss. I'm sure you understand that customers don't really flip their preferences every 3 months.

BTW, in spite of the fact that I am routinely called an AMD fan I said the same thing back in the first part of 2006 when everyone else was saying that AMD took share from Intel. I said that you can't tell in just one quarter; you have to wait and see if the change remains over time. I was concerned that AMD would only have a temporary bump in share like they did back in 2001. Things like this that I've said are always quickly forgotten by those who have weak arguments and want to make ad hominem attacks.

As far as the R600 goes I'm not sure what you are talking about. The last that I heard, AMD's release will not match the top nVidia card but will match everything below it. It will take another release to match the top. This gives nVidia a little breathing room in the top spot.

I agree about low prices but this is why AMD launched the DTX standard. This should allow lower prices in desktop boxes without changing cpu or any other prices. It isn't clear that BTX can match this. The desktop is going to get squeezed very hard in late 2007 and into 2008.

I don't agree with your pessimism about AMD at all. AMD should have profit again and be competitive both in terms of graphics and servers in Q3. AMD should be very competitive on the desktop in Q4 and should make a pretty healthy profit. Why would any of this be reason for concern?

Real said...

AMD should be very competitive on the desktop in Q4 and should make a pretty healthy profit. Why would any of this be reason for concern?

Launching a low volume part with minimal design wins will not impact AMD's Q4 - Hector himself said there would be no impact on 2007 financials. You are being overly optimistic thinking that AMD will have profits in DT in Q4 when likely 90%+ of their sales will be the old cores which today hemorage money. AMD's only chance at profitability in 2007 would be if server shipments of Barcelona are high enough to offset the weakness of the rest of their line-up in server (very unlikely) AND can offset the continuing losses on the desktop front. The chances of this happening are slim to none and slim just left the building.

Look forward to more of the same for the rest of the year - I told you before AMD would have massive losses in Q1 and Q2 and that they would need to raise cash - I am 2 for 2 and Q2 being a massive loss is pretty much a slamdunk at this point. Barcelons is too little, too late. It should have launched by now if it was going to save the company but we still will have to wait 3 more months for server and desktop is unlikely before the last week of xmas sales. On the bright side, that new 125W AMD chip will make winter a little less chilly....

Scientia from AMDZone said...

enumae

"I am not sure how you have drawn a conclusion that Penryn will not be competitive in the 2P space against K10."

You haven't yet seen real benchmarks for K10. Look at Anand Lal Shimpi's last attempt to get more information out of AMD and see how unsuccessful he was. The numbers you've seen are conservative estimates only.

Now, how is this going to shape up?

For dual core I'm not seeing any real advantage for AMD until you get to 4-way. For quad core I'm not seeing any advantage for Intel at all. The simple fact is that native quad will beat MCM unless Intel can increase the speeds much faster than has been suggested.

So, what does this mean?

This means that Kentsfield and Clovertown prices are going to be slashed by Q4 because they will slip from being faster than K8 to slower than K10. Intel should release some 45nm server chips but there is no indication that these will clock any higher than 3.0Ghz and this will not be fast enough to keep up with Barcelona's if they can clock to 2.97Ghz.

Dual core isn't so clear but Intel will fall behind here too if they don't release 3.2/3.33Ghz chips.

The latest news on K10 is not pessimstic. K10 was only supposed to launch at 2.3Ghz and the engineering samples are already running 2.4Ghz. Still, it don't mean a thing until AMD actually ships it.

"The latest news from AMD is Barcelona will launch late July or early August"

If this is the volume launch of K10 then this would match previous timelines.

" while it has also been said that Penryn dual-cores will launch in the third quarter."

I've never seen this claim. Where did you see it?

"Volume will be tough for both Intel and AMD"

What do you mean?

"My main question, and why I am posting is what is AMD going to do against the news that Intel plans more price cuts on July 22nd considering that they are already loosing money with current prices and they will have to respond?"

I would have to see what prices. I'm actually planning to get into this more if I can ever get ahold of volume information.

Scientia from AMDZone said...

real

"Launching a low volume part with minimal design wins will not impact AMD's Q4 - Hector himself said there would be no impact on 2007 financials."

Now that was slick. However, you can't substitute the word Q4 in place of 2007 and then claim that it was said by Ruiz.

" You are being overly optimistic thinking that AMD will have profits in DT in Q4 when likely 90%+ of their sales will be the old cores which today hemorage money."

Perhaps, or perhaps not. Again, I'm not really getting into financials yet because I need to see volume information. However, there has been some suggestion that Q1 sales will show up on the books in Q2. Unfortunately I don't have any way to check the validity of this.

We'll know more about AMD's profitability in Q2 and Q3. For example, if AMD lost another $600 Million in Q2 then I would be pessimistic too. However, if Q2 breaks even or makes a small profit then I would say your outlook is wrong.

Real said...

However, there has been some suggestion that Q1 sales will show up on the books in Q2.

This is not happen unless someone plans to end up in jail....the revenue recognition rules are very, very clear. What Hector said is that design win traction was better which should result in higher unit shipments in future quarters but this does not have anything to do with 'Q1 sales' showing up in Q2 which would land someone (probably Hector/Rivet in jail given they sign off on the financials under SOX)....

We'll know more about AMD's profitability in Q2 and Q3. For example, if AMD lost another $600 Million in Q2 then I would be pessimistic too. However, if Q2 breaks even or makes a small profit then I would say your outlook is wrong.

If AMD breakseven in Q2, I will admit that I was wrong and never post here again. AMD will post another huge loss - there is absolutely nothing that changed in Q2 which would indicate increasing profitability. Q2 will likely be down in terms of revenue $'s from Q1 and margins will be just as bad with some slight upside possibility with 65nm reducing cost and valuation of Barcelona wafer starts (versus direct expense).

Roborat, Ph. D. said...

For example, if AMD lost another $600 Million in Q2 then I would be pessimistic too. However, if Q2 breaks even or makes a small profit then I would say your outlook is wrong.

AMD Q107 earnings report:
"In the seasonally down second quarter, AMD expects revenue to be flat to slightly up".

AMD reported first quarter 2007 revenue of $1.233 billion which gave an operating loss of ~$500M. Why do I get the feeling nobody listens to what AMD says?

gdp77 said...

How many more quarters can AMD keep losing money until they go BK ?

enumae said...

Scientia

The numbers you've seen are conservative estimates only.

I know we talked about this before, but I do not believe they will be more than 10-15% faster than the claimed 50% and 20%.

If they are that would be great for AMD.

...higher than 3.0Ghz and this will not be fast enough to keep up with Barcelona's if they can clock to 2.97Ghz.

If the demos are running at 3.3GHz and 3.2GHz, why do you believe they will be released slower than the demos?

Still, it don't mean a thing until AMD actually ships it.

Agreed, and the same goes for Intel.

I've never seen this claim. Where did you see it?

Take this with a grain of salt, but someone once explained to me that AMD doesn't lower prices unless it had a new top end processor to make up for the lost revenue, so the same should be applied to Intel.

The article I read that said it could launch was here.

I do not know the site or it's reputation, but looking at the price cuts and applying what you had explained, along with the two articles at TGDaily talking about the NDA and also the final testing, it would seem to like its possible.

What do you mean?

Both Intel and AMD are about to launch new products, they will both be ramping and volume may be difficult.

I would have to see what prices.

HKEPC

Scientia from AMDZone said...

gdp77 & real

Okay, see if you would agree with this:

Having another loss of $500 Million or higher in Q2 would bring up very real questions about AMD's future. You would have to wonder if AMD would still be losing in Q4. Very poor outlook.

I suppose a $400 Million loss would still be bad. It's improvement but too slow. Bankruptcy still possible. Poor outlook.

I think AMD needs to at least cut the loss in half in Q2. $300 Million isn't a great number but it would demonstrate progress. Bankruptcy not likely but still not a favorable outlook. Neutral outlook.

I would say that a $150 Million loss would show fairly good progress in reducing costs. Slightly favorable outlook.

To get a good outlook I'd say AMD needs to have a loss of no more than $75 Million in Q2.

Scientia from AMDZone said...

real

The suggestion was that AMD delivered processors in Q1 with payment due in Q2 and that is why it would show up in Q2. As I've said, I have no way to determine if this could be true or not.

enumae

"I know we talked about this before, but I do not believe they will be more than 10-15% faster than the claimed 50% and 20%."

No, you don't understand. These are scores for select benchmarks. The overall scores would be lower. So, even with higher performance the overall scoring might still only be this good. Okay, think of it this way. At 15% AMD only needs 2.6Ghz to match 3.0Ghz. If AMD can get 20% then it would need 2.8Ghz to match Intel at 3.33Ghz.

"If the demos are running at 3.3GHz and 3.2GHz, why do you believe they will be released slower than the demos?"

Okay, I have my doubts that AMD will clock higher than 2.6Ghz in 2007 and I have doubts that Intel will clock higher than 3.0Ghz in 2007. I suppose 3.33Ghz on 45nm for Intel in Q4 is possible. But, perhaps 2.97Ghz is possible for AMD. I'm very certain of 2.6 and 3.0Ghz; I'm less certain of 3.33 and 2.97Ghz in Q4.

"Take this with a grain of salt, but someone once explained to me that AMD doesn't lower prices unless it had a new top end processor to make up for the lost revenue"

Tsk, tsk. If you keep misquoting me I'll start thinking that you are getting desperate.

What I actually said was that if a price cut were accompanied by the introduction of a higher speed grade then there was no actual drop in prices because the new top chip simply replaced the old top chip at the same price.

"Both Intel and AMD are about to launch new products, they will both be ramping and volume may be difficult."

Yes, I would say K10 and 45nm will both be low volume.

Okay, I would say that if Intel can get real volume of 45nm out the door in Q3 then that will be a grade of A+. Everything that I have seen suggests low volume even in Q4 with 45nm accounting for a lot less than 1% of the desktop.

Yes, I would tend to wonder the same thing. Why cut prices in July if you aren't introducing a new chip? It doesn't really seem to be in response to AMD since the server chips and FX will be released first and shouldn't really effect Intel in Q3.

gdp77 said...

Why cut prices in July if you aren't introducing a new chip?

Because u want to completely destroy your rival and never let him stand up again and because u can afford it to. I agree with your "if / then" outlook scientia. I still see a big loss for AMD in Q2 and even bigger in Q3 and Q4 (since there will be no desktop K10 and R600 is a complete failure).

I am afraid that in 2008 AMD will seize to exist (possible buyout by another big company) and I am afraid that in the future we will see no real innovation in the cpu market (since Intel will be a monopoly).

This future becomes more real, each day that passes.

gdp77 said...

I'm not aware that K10 is late. The release timeframe that I have always seen is late Q2 for launch on server with perhaps low volume. This seemed to be primarily to fill early HPC contracts.

Q3 was supposed to expand the volume and delivery of server chips in both quad core and dual core. I expected the FX chips to be released here too. Q4 was supposed to be desktop. I believe AMD is still on this schedule.



Oh come on!!!! We were discussing for months about K10 server arrival in late Q2 and desktop parts in Q3... And now you say "I am not aware that K10 is late"??? Please don't insult us... Want me to post links about that?

enumae said...

Scientia
Tsk, tsk. If you keep misquoting me I'll start thinking that you are getting desperate.

I really hope you are kidding with that comment.

What I actually said was…

What I said amounts to the same thing, and I was not trying to quote you, just pointing out the basis of your explanation, release a new processor to account for the drop in revenue.

Dr. Yield, PhD, MBA said...

Scientia, you 'splained:
I suppose a $400 Million loss would still be bad. It's improvement but too slow. Bankruptcy still possible. Poor outlook.

I think AMD needs to at least cut the loss in half in Q2. $300 Million isn't a great number but it would demonstrate progress. Bankruptcy not likely but still not a favorable outlook. Neutral outlook.

I would say that a $150 Million loss would show fairly good progress in reducing costs. Slightly favorable outlook.


So the question is: How do they get there this quarter.

Guidance for revenue is flat to slightly up--> no help there

Layoffs? Expenses up for the quarter they occur in, down long haul--> no help there

Capex reduction of $500M over the year--> ~133M benefit this quarter?

Acquisition related charges? Last quarter was $113M, this quarter estimate is $80M+$25M additional tax bill--> again, no help here

Employee stock comp is estimated to grow from 28 to 30M for the quarter.

Cost savings from 65nm transition? Might help, might not- cost/die may be lower, but if they are piling into inventory (most of Q1 build was 65nm according it the conf.call), it doesn't help that much. 65nm equip depreciation is probably full bore now as well.

I just don't see where the Q2 recovery will come from. Topline prediction is flat (that covers the pricing environment), and expenses are only significantly impacted by capital expense push out. My 2 minute analysis (admittedly not my best work) would guess a best case scenario of $350M loss (small revenue growth+ 1/3 of capital pushout). I hope for AMD's sake they pull a rabbit out of the hat- they could always attempt to push out more of the caputal expense from Q2 vs Q3/4.

Also, they better figure this out quickly- according to a recent presentation they made, their debt payments grow from 125M this year, to 250 next, to 341 in 09... see http://www.amd.com/us-en/assets/content_type/DownloadableAssets/sderhsehsdt.pdf

Scientia from AMDZone said...

gdp77

"Because u want to completely destroy your rival and never let him stand up again and because u can afford it to."

I don't really want to get into this right now because I want to discuss it when I do an article on the Q1 revenue. However, I can tell you that for several reasons this is not possible.

Okay on the launch, let me say this as simply as possible. If AMD does not release any K10 sever chips before July or August then it sounds like their schedule has slipped. Except, schedules don't slip by 1 month at this stage. At this stage a schedule can normally only slip by one production cycle. In other words, you run off a revision hoping for a B0 and you end up with an A2 and then you have to run off another one to get a release chip. However, the indications so far have been that AMD did not need an extra revision.

That's my confusion. For example when Intel pulled its 1.13ghz PIII it took them six months to replace it. If AMD slipped one cycle this should be about 3 months. So, perhaps this could mean that AMD originally planned to release K10 at the beginning of May.

Dr. Yield, PhD, MBA said...

Our host opined:
Except, schedules don't slip by 1 month at this stage.

Actually, it is possible to slip only a month with a design rev, particularly if the silicon is close.

For example, assume B1 silicon is currently inline. Critical erratum is found (they happen to everyone) in test, and the design team determines they can patch it with a fix to M4-M6. All the silicon inline pre-M4 is still good-> it holds for the M4+ mask sets (assume 7-10 day turnaround for BEOL), and production is back on. At 1.5-2 days/ML, you are looking at a little over a month after the new masks are delivered.

Of course if the fix needs to happen in FEOL, that theory is shot- more impacted silicon, more validation, more masks, and more time to get revise die out the door.

Scientia from AMDZone said...

dr yield

I don't know. I simply don't have that kind of detailed information about AMD's finances. I'm just saying that more than a $300 Million loss is going to look bad for AMD.

Scientia from AMDZone said...

enumae

"What I said amounts to the same thing"

No, it doesn't. You were implying that I claimed that AMD only drops prices when they release a new chip and I never said that.

Now, your innaccuracy of what I said aside, it could be accompanied by a new release.

Dr. Yield, PhD, MBA said...

I don't know. I simply don't have that kind of detailed information about AMD's finances. I'm just saying that more than a $300 Million loss is going to look bad for AMD.

Scientia, that's the problem. Everyone is taking a recovery on blind faith. The simple fact is that AMD's public (and therefore must be legally truthful) statements have indicated nothing that would allow for a conclusion that shows recovery in Q2. They have given enough information that barring some miraculous reversal on the topline, Q2 is going to be another loser.

Beyond Q2 is still open for speculation- but I think "fans" should steel themselves for another quarter rivaling Q1 in terms of results. Probably a bit better- but not a lot.

enumae said...

Scientia

No, it doesn't. You were implying that I claimed that AMD only drops prices when they release a new chip and I never said that.

Thats not my intention.

My intention was to apply what you had said while you were explaining how the drop in prices is not a drop in revenue due to the introduction of a new processor...

If both AMD and Intel introduce new chips and leave the price per rank the same (even if an individual speed of chip changes price) then there is no competitive change in price structure.

Then apply that to Penryn as being the new chip.

I was not trying to apply this only to AMD, and I did not make that part clear.

My apologies.

PS: The grain of salt comment was meant to go in front of the source of the article talking about the Q3 introduction of Penryn.

Real said...

"Capex reduction of $500M over the year--> ~133M benefit this quarter?"

This is not correct. CAPEX shows up on the income statement as depreciation - depreciation in the semicon world is usually spread over 4 years. By pushing a the $125M in CAPEX ($500M for the year / 4 = $125M in Q2), the max savings on their income statement would be $125M/16 = +$8M (16 = 4years * 4 quarters/year). So, pushing CAPEX can help AMD, but $8M is a drop in the bucket on a $500M loss. BTW, I believe the analyst consensus estimate = -$0.85/share = -$470M. Maybe Scientias will be correct, but I am guess he is wrong just like he was about Q1, Q4, etc......

abinstein said...

"By pushing a the $125M in CAPEX ($500M for the year / 4 = $125M in Q2), the max savings on their income statement would be $125M/16 = +$8M (16 = 4years * 4 quarters/year)."

I don't understand your math here. Why do you divide the depreciation of a certain quarter into four years then (again) into four quarters? To me, $500M less capex for this year means they'll spend $500M less this year, and thus in average $133M per quarter. Didn't AMD say they'll have more capex reduction next year? So are you going to divide that reduction by 4, and then what, add back 1/4 of this year's reduction? Seems like messy arithmetics to me.

abinstein said...

"Actually, it is possible to slip only a month with a design rev, particularly if the silicon is close."

Of course it is possible. It is also possible that the chip never comes out if say the fab is blown up and AMD is bankrupt right away. Many things are possible but they are no likely, to the point that they're effectively impossible.

First you have to assume the B1 stepping has some critical errors.

Second you have to assume the error was not found in the months before but in the first 3 or 4 weeks after rev. B1 tapes in.

Third you must assume the fix for this critical error only involves the later stages of the fabrication - which would limit it to essentially signal routing.

Fourth you must assume the production fab actually accepts such on-the-fly mask change.

Fifth you must assume the design team are confident of the chip to be released after fixing a "critical" error on-the-fly.

Lets say you get 20% chance on each. Take 5th power and the possibility is less than 1/3000. Well, it is possible...

Real said...

don't understand your math here. Why do you divide the depreciation of a certain quarter into four years then (again) into four quarters?

This is pretty simple math and accounting given we do not know exact details from AMD. In the first division ($500M in CAPEX reduction for the year), I am dividing by 4 making the assumption that they will make an equal reduction in each quarter for the entire year given this info came out with the Q1 warnings which likely means the first part of this CAPEX cut was already seen in Q1. Once we have CAPEX per quarter, it is pretty easy to see the impact for pushing $125M in CAPEX out of Q2. The quarter depreciation generated by $125M in CAPEX = $8M (this is the $125M/16 quarter lifetime step). So, by pushing $125M in CAPEX, the impact in Q2 will by a savings of $8M as opposed to Q1. BTW, there is a multiplier effect such that Q3 savings should be an additional $8M over Q2 and Q4 will be an additional $8M over Q3. However, as the discussion focused on Q2 impacts, the relevant number is the $8M savings over Q1.

bk said...

Why is the 500 million reduction significant?

I don't think it is because the depreciation will be less. I do think it is because the cash flow/spending will be reduced by 500 million.

When investors are concerned about bankruptcy, they are more concerned with cash flow than depreciation.

Real said...

When investors are concerned about bankruptcy, they are more concerned with cash flow than depreciation.

That is absolutely correct - however, the $500M loss in Q1 and project $470M loss in Q2 are significant as it calls into question whether AMD's business model is viable long term. The analysts from Goldman Sachs has stated the business model is fundamentally flawed and the data seems to support him. $500M CAPEX elimination is not a repeatable event and not sustainable long term for AMD as equipment does wear out and does need to be upgraded. The business model question is the most important as once more people come to that conclusion, AMD's access to the capital markets is eliminated and bankruptcy is only a matter of time.

bk said...

Real
That is absolutely correct - however, the $500M loss in Q1 and project $470M loss in Q2 are significant as it calls into question whether AMD's business model is viable long term.

For your hypothesis to be valid you need to show that AMD's losses will continue. No one is arguing that AMD is in trouble if they lose 500 million every quarter. I don't agree that this will continue long term. There is no evidence to support such a claim. It is pure speculation.

Scientia from AMDZone said...

enumae

I've finally gotten a handle on those July price cuts for Intel. They are not real price cuts. The HKEPC article left out the new top speeds.

Apparently, Intel is rippling its prices down a notch on quad decktop with the intoduction of the 2.93Ghz QX6800. The new top chip, the QX6800 still has a price of $999.

Dual socket quad Xeon is similar. The top chip, the new 3.0Ghz X5365 still has a price of $1172.

abinstein said...

real,

I am not an accountant but I don't think it takes an accountant to see the logic error in your calculation.

"... given this info came out with the Q1 warnings which likely means the first part of this CAPEX cut was already seen in Q1."

Why? At the end of Q1 AMD gave warning of revenue shortfall and claimed it will cut cost by reducing capital spending. This seems a response to the bad quarter. Do you see $125M less capital spending in Q1's balance sheet?

"The quarter depreciation generated by $125M in CAPEX = $8M (this is the $125M/16 quarter lifetime step)."

This is wrong, because you have 4 quarters of depreciation reductions, each span 16 quarters. Thus, according to your theory (which I do not agree either, more on this below), the $500M should effectively reduce expense $125M/16*4 = $31.25M per quarter, and such reduction will last for 4 years (so total is still $31.25M*4*4 = $500M).

"BTW, there is a multiplier effect such that Q3 savings should be an additional $8M over Q2 and Q4 will be an additional $8M over Q3."

I have no idea why you account in such messy way. A reduction in a certain year is a reduction in that year. AMD's plan for 2007 capex is 2 billion; do you divide that into 4 years? AMD's 1st quarter capex were at $586M; do you divide that into 4 years, too?

"However, as the discussion focused on Q2 impacts, the relevant number is the $8M savings over Q1."

Actually, AMD said itself the cost reduction would be effective the second half of the year. The effective on Q2 will be small, but the point is it has nothing to do with you $8M, which is just wrong.

enumae said...

Scientia
They are not real price cuts. The HKEPC article left out the new top speeds.

Again, I should have been more specific, I was referring to Desktop.

Well, then I guess I will never understand this, look at this, and if you could, please explain how this is not a price reduction.

Thanks.

Real said...

I am not an accountant but I don't think it takes an accountant to see the logic error in your calculation.

Clearly, that is one of the causes why you can not understand the math or accounting involved.

At the end of Q1 AMD gave warning of revenue shortfall and claimed it will cut cost by reducing capital spending. This seems a response to the bad quarter. Do you see $125M less capital spending in Q1's balance sheet?

Let me let you in on a little secret. When you give god awful earnings, you look for any ray of sunshine you can possibly give to Wall Street so you stock is not butchered. One way is trying to show big numbers for cuts. AMD said that they would cut $500M in CAPEX - that means that they are reducing their former CAPEX plans by $500M, not reducing a 'balance sheet' line item by $500M, but we can look at if AMD spent as much in CAPEX in Q1'07 as they did in Q4'06...

Q1'07 = $586
Q4'06 = $666

So, we do see a reduction vs. Q4'06 but as I mentioned, we would need AMD's projected CAPEX before the reduction to be 100% sure. However, it is pretty obvious some of the $500M reduction in CAPEX occured in Q1'07.

I have no idea why you account in such messy way. A reduction in a certain year is a reduction in that year. AMD's plan for 2007 capex is 2 billion; do you divide that into 4 years?

We are not talking about the year impact. we are talking about the impact on Q2 earnings. As such, the number that matters on the income statement is the depreciation. As such, we can make the assumption that Q2 CAPEX will be $125M less than it was originally planned (as explains again above). The next impact ON Q2'07 EARNINGS is 1 quarter of depreciation on Q2 earnings - which is equal to $8M higher earnings. Read the logic again, if you still don't get it, I can't help you but the numbers are 100% correct given the available information on the CAPEX reduction - take it from an accountant.

Actually, AMD said itself the cost reduction would be effective the second half of the year.

Yes, as I posted, the next effect of the savings will be $16M in Q3 vs. Q1 and $24M in Q4 vs. Q1 and $24M/Q for the next 12 quarters after that. This is pretty basic, please don't ask me to walk you through it again.

gdp77 said...

"No one is arguing that AMD is in trouble if they lose 500 million every quarter. I don't agree that this will continue long term. There is no evidence to support such a claim. It is pure speculation."


How about "the situation hasn't changed since Q1"? K10 is not here (and it won't be until late Q4), C2D still beating everything AMD has to offer and Intel prices are even more aggressive. In addition to that R600 is a complete failure and it is going to stay at the shelves of stores . So, nobody is buying AMD... (same thing happened in Q1).

Isn't it logical to assume that losses will be the same also?

Axel said...

Scientia

If AMD does not release any K10 sever chips before July or August then it sounds like their schedule has slipped. Except, schedules don't slip by 1 month at this stage. At this stage a schedule can normally only slip by one production cycle.

Looks like the Barcelona launch schedule may have slipped by at a couple months, according to motherboard manufacturers:

http://www.digitimes.com/mobos/a20070516PD217.html

So much for any hope of AMD posting a profit this year.

Scientia from AMDZone said...

enumae

It looks like for Intel on the desktop you have an average price reduction of 7%. This is a very modest overall decrease in prices.

However, the bulk of this 7% comes at the very top. The top bin quad core takes a hit of 17%. The top speed dual core slips down to second bin and also takes a hit of 17%, however, there is no new dual core to replace the top bin.

This may seem like a lot but remember that the top bins are very low volume so this won't have much effect at all on Intel's sales. The overall effect should be no more than 2%.

Scientia from AMDZone said...

gdp77

"R600 is a complete failure "

Where did you get this silly idea from? There is nothing wrong with R600. The only thing missing is the top speed bin. Everything from 2nd bin down matches nVidia. How is that a failure? Or do you judge an entire graphic line by only the top bin?

Scientia from AMDZone said...

axel

I'm not quite sure how to respond to the digitimes rumor. If it is true and AMD begins ramping K10 in September then that is going to hurt. This does conflict with earlier rumors that AMD was at the B0 revision and with rumors that AMD's SE chips were running 2.4-2.6Ghz. Like I said though, it don't mean a thing until it ships.

gdp77 said...

"R600 is a complete failure "

Where did you get this silly idea from? There is nothing wrong with R600. The only thing missing is the top speed bin. Everything from 2nd bin down matches nVidia. How is that a failure? Or do you judge an entire graphic line by only the top bin?




U were saying about the same thing about the AMD X2 vs C2D performance some months ago. U were wrong then (Q1 results proved that) and u r wrong again.


AMD's graphic cards, are a)slower b)more expensive c)draw much more power cards d)they have awfull drivers e)the linux support sucks compared to nvidia's.

Therefore, nobody will buy them.

Wise lnvestor said...

gdp77 said...
R600 is a complete failure. a) b) c) d)... Therefore, nobody will buy them.

Young man, you shouldn't rush into conclusions.

Here a apple to apple review from Legion hardware .

Radeon HD2900XT does show promise. Hope you enjoy the review as much as I did.

Scientia from AMDZone said...

gdp77

"U were saying about the same thing about the AMD X2 vs C2D performance some months ago."

You seem to be confused. What I said about X2 was that it was as fast as a C2D that was two speed grades (400Mhz) slower in terms of Integer. In other words, it would take a 2.4Ghz X2 to equal 2.0Ghz Conroe.

In terms of R600, all of the released products are competitive, the only thing missing is the top of the line card.

" U were wrong then (Q1 results proved that)"

The Q1 results show that AMD is making less money and selling fewer chips; it does not show that what I said (and everyone else is saying) about performance is untrue.

" and u r wrong again."

Really? Does any review back you up on that? Let's see what Wise Investor's link says:

"AMD's graphic cards, are a)slower"

one thing that remained fairly constant and that was the Radeon HD 2900XT’s lead over the GeForce 8800 GTS graphics cards.

" b)more expensive "

for $400 US this product is quite good value.

"c)draw much more power cards "

True.
Power consumption is a major issue that cannot be overcome at this stage, the 65nm versions of the Radeon HD 2900XT may fix this problem, but we will have to wait and see.

"d)they have awfull drivers "

Well, they didn't say awful.
the drivers probably are not as polished as they should be

Here's what they actually said:

to be honest we were expecting a complete dud, but it just isn’t!

at $400 US you get a good performing DirectX 10 graphics card

Should the Radeon HD 2900XT mature quite well over the next few months, it could become a real threat to the GeForce 8800 GTX, particularly given it costs significantly less.

gdp77 said...

scientia I could link a lot more sites like hardocp, anandtech, toms that describe R600 as a failure. But this is irrelevant. The question here is : Will R600 be a nice selling product for AMD so it could help reduce the losses? My personal opinion is "No", because is reviewed all around the world with conclusions like "flop", "failure", "power hog" etc. I am sorry u don't see that.

If u believe otherwise, please make a statement that "this product is good and will be a good seller". I believe u won't do it.

The truth is (and believe me, I don't like it) that AMD has nothing of value that can sell or help reduce the mass losses that it is going to have in Q2 and Q3. Thus, I see AMD BK before the end of this year. I am sorry they delayed so much K10. Intel has the money to tolerate failures and weaknesses. AMD doesn't. A gpu (that won't sell) and a CPU (that will never ship on time) are signs of a company failing all the time. Bankruptcy is inevitable.

bk said...

gdp77

When an acorn falls on your head, do you think the sky falling?

Yes, AMD has had some negative things happen lately, but I do not believe they are nearly as bad as you proclaim.

I think there is a major flaw in much of the logic about AMD, not just gdp77's logic. Everyone looks at the sales from a standpoint of individual parts and not the system as a whole. When HP or Dell buy parts they are looking at the whole system and how well it will sell at the price point.

AMD is still around because they provided parts to build a well performing system at a good to excellent price point. I believe this will continue.

Erlindo said...

gdp77 is a well-known intel fanboy at THG's forumz. You guys are wasting precious time answering his worthless posts.

GDP: you can always go to your beloved intel forum (or should I say THG) and continue with your doom and gloom over there. You're hihgly appreciated over there. ;-)

gdp77 said...

actually I am not even registered in tom's forums. I owned AMD since my first pc. My first 8088 was AMD and I stayed with AMD ever since and my current cpu is a 4600 x2 system. I have only bought a Intel system for my brother's pc, since he needed to upgrade and any AMD offering atm is a waste of money for someone who wants to overclock.

AMD is leading the cpu technology for quite some years now. But this is irrelevant with the question of whether AMD will go BK or not.

Erlindo, my opinion is that AMD will not survive this battle this time. Does this make me an Intel fanboy? Not being able to tell your opinion and being characterized by that is called fascism.

If u have anything to say about my arguments, then please do so. Otherwise stop characterizing me please.

gdp77 said...

I was wondering if u can answer to me these questions :

a) for K10 to be ready on Christmas, when the production must start?
b) How much more losses can AMD sustain?
c) What will be the impact to AMD's transition to 45nm from the current economical situation?

Erlindo said...

gdp77 wrote: actually I am not even registered in tom's forums
You don't need to have the same user name to be registered at forumz. ;-)
Also, weren't you banned from AMDzone for the same annoying behaviour??

gdp77 wrote: Erlindo, my opinion is that AMD will not survive this battle this time. Does this make me an Intel fanboy? Not being able to tell your opinion and being characterized by that is called fascism.
Well, you can live in your world thinking so, but saying that AMD will go BK is just absurd and wishful thinking from intel fanboys.

gdp wrote: If u have anything to say about my arguments, then please do so. Otherwise stop characterizing me please.
Indeed, your typical arguments are: AMD won't survive, AMD's R600 is failure, AMD will go BK, A;MD lags behind intel, AMD..., AMD..., AMD... You already sound like a scratched disk. :-D

Sorry, but you don't have valuable arguments worthy of any serious response. ;-)

enumae said...

Scientia

I have revised the chart to show what you have described.

Looking at the new prices, this makes the top bin (when Core 2 Duo was introduced) a mainstream part.

$266 for an E6850, or $266 for an Q6600...

What is AMD going to offer if the recent news about desktop 10H (K10) processors not being released until the fourth quarter is true?

Also, if AMD has to lower prices again what kind of impact would you expect on there earnings?

Thanks

abinstein said...

"What is AMD going to offer if the recent news about desktop 10H (K10) processors not being released until the fourth quarter is true?"

I believe the price cut from Intel is well expected by AMD, and we will see further price cuts on K8.

But if the new Intel pricing is true, then Q6600 at $266 would be a real good purchase for single-socket quad-core systems. OTOH, Q6700 is plain overkill for an MCM and at 1066MHz FSB it will be only marginally faster than Q6600. E6800

gdp77 said...

You don't need to have the same user name to be registered at forumz. ;-)
Also, weren't you banned from AMDzone for the same annoying behaviour??


Mr big brother, I was banned from AMDzone because my posts weren't the "AMD owns everything" style. Many AMDzone forums users expressed the opinion back then that my ban was not justified. Using the same nickname here proves that I have nothing to hide or to be afraid of.


Well, you can live in your world thinking so, but saying that AMD will go BK is just absurd and wishful thinking from intel fanboys.

.............

Sorry, but you don't have valuable arguments worthy of any serious response. ;-)


My arguments have been posted. Apparently u missed them. I will repost them for you:

AMD in Q1 lost 600M $ . Nothing has changed in Q2, therefore it is logical to assume that the same reasons that caused Q1 loss, exist in Q2. Intel has even more aggressive pricing, C2D is still the superior mainstream and high end cpu (price wise and performance wise) and R600 will probably not be a good seller according to the review results. So I don't see how AMD will recover. In addition, K10 desktop will be delayed until Christmas (at best)so AMD atm is having nothing which could use to respond to Intel.

Apparently my arguments are not valuable and not worthy of any serious response, but regardless of your opinion about them, these are probably the reasons that will lead to AMD BK.

I repeat my question: How much loss can AMD sustain? 1B$, 2B$, 3B$?

enumae said...

Scientia

More pressure coming against AMD's Barcelona's launch.

Erlindo said...

enumae wrote: More pressure coming against AMD's Barcelona's launch.

I really don't see presure to AMD's Barcelona launch. I think this is a sign that intel is scared of AMD's future offerings. Now its AMD's turn to set prices just as intel did with C2D.

Seems like history will be repeating itself.

enumae said...

Erlindo
Now its AMD's turn to set prices just as intel did with C2D.

They can only dictate price if they have the better product.

Also that dictated price can only be what someone is willing to pay for the performance advantage it gives.

Question:

If AMD is even on Integer performance, but 40% ahead on FP can they dictate prices?

Axel said...

Scientia

In terms of R600, all of the released products are competitive, the only thing missing is the top of the line card.

No. You seem to have missed the fact that the only released product is the HD2900XT. The others were paper launched: not available until late June at the earliest, and pushed back farther if AMD has issues with fabbing them at 65nm.

Now, looking at the HD2900XT alone, I don't see how anyone can conclude that it's competitive at the current price. Most reviews out there show that the 8800GTS 640MB is slightly faster in most cases, while also being cheaper and consuming less power. The HD2900XT, released seven months after the 8800GTS, is a flop in its current form pure and simple.

Axel said...

Scientia, any idea why Puma (AMD's answer to Santa Rosa) will be based on K8 and not K10, despite being launched in mid-2008?

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2992

http://arstechnica.com/news.ars/post/20070520-amd-sics-puma-on-intels-santa-rosa.html

This doesn't make sense to me. It must mean that AMD intends to ramp K10 quite slowly, not leaving sufficient volume for mobile. It also means that Puma doesn't have a chance of competing with Santa Rosa in either raw performance, and probably not performance-per-watt either.

Aguia said...

alex,
the “new” K8 will have some K10 tech inside but not all that’s why you won’t see it sooner.
Besides most of the performance you need in a notebook is right there on the K8 core.

Most of the notebooks sold where Celeron a year ago. That must say something.

I would prefer a slower, quieter and less hot apple macbook than the current fast hot and noisier. A better IGP would help too.

abinstein said...

"Scientia, any idea why Puma (AMD's answer to Santa Rosa) will be based on K8 and not K10, despite being launched in mid-2008?"

Starting at K8 rev.G, AMD's mobile and server lines begin to diverge. Thus Kuma as well as Barcelona both are based on K8, but not on each other. There are some server/workstation additions in Barcelona that are not as important on mobile platforms.

gdp77 said...

Now, looking at the HD2900XT alone, I don't see how anyone can conclude that it's competitive at the current price. Most reviews out there show that the 8800GTS 640MB is slightly faster in most cases, while also being cheaper and consuming less power. The HD2900XT, released seven months after the 8800GTS, is a flop in its current form pure and simple.

Considering the 2900XT I am reconsidering my previous negative opinion that it is a flop. First of all, most reviews show that 2900XT is slightly faster than 8800 GTS 640MB and at least where I leave (Greece) it is cheaper than 8800 GTS.

Sapphire 2900XT costs 339€ while the cheapest 8800GTS costs 359€. Maybe after all (just maybe) AMD manage to make some money out of this product...

Azmount Aryl said...

HD2900XT at $419.99 has less memory and cost more than 8800GTS at $349.99. Those are the cheapest prices in U.S.

bk said...

Azmount Aryl
HD2900XT at $419.99 has less memory and cost more than 8800GTS at $349.99. Those are the cheapest prices in U.S.

Your note that the HD2900XT has less memory implies that it is inferior to the 8800GTS. The review from legion hardware that wise investor linked compares these exact two gpu's and the HD2900XT comes out on top the majority of the time and sometimes by quite a substantial margin.

Greg said...

while the hd2900xt does score well when you look at raw fps on average or low settings, the 8800gts does outperform it in terms of playable settings. HardOCP has a review on it where they show it's not as good as the 8800gts. However, most websites are not looking at playable settings, but raw fps. I'd also like to point out that the score trends show all the classic signs of poor driver optimization. However, it's actually a stable driver, so they've at least got a leg up in that respect.

Greg said...

Also, the low end versions are already doing very well (despite not being released). All the big pc manufacturers can't get enough of them apparently. I guess having a certified driver early enough is kinda a big deal.

Pop Catalin Sever said...

http://amd-04.developer.amd.com/pgi.jsp

Check out AMD's latest announcement, kinda funny!

abinstein said...

There's a good reason why AMD promotes PGI. Most compilers optimize for only one microarchitecture - and you know which uarch it would be. You want to distribute one executable to your users, and you want 75% of your users happier, don't you?

However, PGI adopts a special approach, where it can combine two separate codes for both uarch into one unitifed binary. This way the executable can choose the portion that's optimized for the underlying uarch.

abinstein said...

Forgot to link the source of information: Compiler as referee.

I'd like to note that AMD and Intel chips has not been 100% binary compatible since the day of Pentium (contrary to what some clueless THG users believe, when they "discussed" - more like babbled - on this topic).

abinstein said...

More information on the PGI compiler topic.

While the article seems to be a PGI PR, one thing is clear: the compiler is in no where favoring AMD (or Intel for the same matter) if it claims to compile for Core 2 Duo better than Intel's compiler.

It gives users freedom of choice, something Intel with its monopolistic mentality does not like, and many users/fanbois do not want.

Scientia from AMDZone said...

pop catalin

"Check out AMD's latest announcement, kinda funny!"

I've read the announcement a couple of times but I still haven't found anything unusual, particularly notebworthy, or funny. This is a routine announcement of a compiler upgrade. What about this is funny to you?