Will Hammer be delayed until 2H 2003?

[mrman3k]

Okay, I am a little confused, the other day I read that Hammer would be delayed to 2H 2003. Then today I take a look at the Inquirer only to find them saying that it will be released Q1 2003. Can someone please explain to me what is going on, who is right, and who is wrong.

Thanks!

 

[Pink0]

They're both right and they're both wrong. Here's how it went down:

In the conference call the Vice President and Cheif Financial Officer of AMD said that AMD will focus its hammer efforts on the server space which means sledgehammer. Why? They state that the hammer "will be the best server offering." Basically it comes down to this: They are having severe yield problems and can only make so many hammers. Heck, they're having severe yield problems with T-bred 2 let alone SOI hammer. Anyway, say you can only make a small ammount of hammers. Now, are you going to sell them as clawhammers and make very little money on them or are you going to sell them as sledgehammers in the server space where they stand to win the most mindshare because that's where their best performance can be realized and where they will have the higest profit margins. The most profit is to be made in server space so this is where they will sell their limited quantities. Consumer confidence is down during the recession and it is best to wait until it's over before releasing clawhammer so they can demand a higher markup as they did with the original K7. They retailed for $700Cdn.

AMD's PR specialists have been quoted as saying that desktop hammer will be available in 1H 2003. So many people think this means that they are talking about the Athlon DT (clawhammer) They didnt' specify which hammer though. While the PR hacks aren't lying they aren't telling the whole truth either.

Now, I'm going to trust the Vice president and financial officer of AMD more than some PR hack. AMD will make and sell hammers in 1h 2003 but they will be sledge hammers. To joe blow, hammer will be delayed until 2h 2003 since he's only interested in clawhammer.

Sledgehammer 1h02, clawhammer 2h03.

 

[mrman3k]

That sucks. That was what I was afraid of. I don't really care about Opteron, I mainly care about Athlon DT and I just hope that in the long run this move will be the best. I just would hate to see the end of AMD when they were about to come out with an excellent product.

In this case, I think it would be best if AMD were to totally skip 130nm Athlon DT and go straight for 90nm DT to compete with the by then aged Prescott.

 

[Pink0]

Yes, ED from overclockers seems to believe that it's entirely possible that AMD might delay the AThlon DT until .09um because it simply woulnd't be effective otherwise. The 1.6 Athlon DT is alleged to only have a PR rating of 3000 which, considering that it would launch in 2h03 wouldn't be competitive with Intel's offerings. Other rumors peg it at 2.0ghz and 3400+ which by 2h 03 will still not be competitive therefore the procssor would not make enough of a markup to be profitable.
The important thing is for AMD to start making money on their processors again. AMD has been intentionally devaluing their stock and liquidating palaminos so they can do what they just did. In the conference call, AMD's Vice president was able to drive the message home that AMD is going to focus Barton on the consumer market for 2003 and hammer on the server market. AMD tried to spend their way out of tough times and how that the R&D is done, it's all a matter of timing. Sink or swim time. End of AMD or surge forward similar to that which happened with the introduction of the K7. Promises of old stock being liquidated, huge R&D cuts and layoffs caused AMD's artificially depressed stock price to rise by 25%.
The principle here is that AMD is weathering tough times through spending a fantastic ammount of R&D on Barton and Hammer and by liquidating old stock. Now that they've done this they will sell the fruits of this expenditure to the server market where the markup is large enough to put AMD back into black ink again. Barton will hold AMD through 2003 in the consumer market.
Once this has happened, you might see the Athlon DT at the end of 2003. It simply won't be competitive enough if they release it now to charge any kind of profitable margin on so it will go to the server space. Don't feel bad about missing hammer for 2003 though. Barton is very capable and it is definitely in the company's best interest.

 

[mrman3k]

All I hope for is for AMD to survive and pull through the tough times. It would be horrible if they went under. Could you imagine paying $800 for a craptastic CPU?

 

[CrazySaint]

Originally posted by: mrman3k
All I hope for is for AMD to survive and pull through the tough times. It would be horrible if they went under. Could you imagine paying $800 for a craptastic CPU?

AMD isn't go anywhere. They're finally getting smart and focusing on the extremely lucrative server chip market instead of the extremely non-lucrative low-end value consumer market. Companies like Dell are already interested in making Opteron-based server lines and according to AT's "Inside ATI and NVidia" article, NVidia is very interested in using the Opteron in their server farms (which, if they do decide to use the Opteron, that would be literally many thousands of chips right there).

 

[First]

Originally posted by: CrazySaint

Originally posted by: mrman3k
All I hope for is for AMD to survive and pull through the tough times. It would be horrible if they went under. Could you imagine paying $800 for a craptastic CPU?

AMD isn't go anywhere. They're finally getting smart and focusing on the extremely lucrative server chip market instead of the extremely non-lucrative low-end value consumer market. Companies like Dell are already interested in making Opteron-based server lines and according to AT's "Inside ATI and NVidia" article, NVidia is very interested in using the Opteron in their server farms (which, if they do decide to use the Opteron, that would be literally many thousands of chips right there).

Pink0 is only half right. While AMD's server Hammer CPU (Opteron) is scheduled to arrive in 1H 2003 like Pink0 mentioned, the desktop Hammer CPU (ClawHammer) is not scheduled for 2H 2003. Here's AMD's official line on Hammer:

AMD's guidance regarding the roadmap introduction schedule for the hammer processors has not changed. We expect system availability for the desktop hammer to be in late Q103 or early Q203. Likewise, the upcoming AMD Opteron processor is on track to be available in the 1H03 for the server and workstation markets. Strategy-wise, AMD is focused right now on the server space as Opteron will give us a clear performance advantage vs. the competition in the enterprise server market.

As far as Opteron interest is concerned, NVIDIA isn't the only company that has shown a heap of interest in Opteron. There are many others. It's understandable that AMD is focusing on the Opteron K8 platform for these reasons, and less so on the ClawHammer platform when you factor in the good possibility that Barton will be competitive enough with Intel latest and greatest HT P4's in the near future, leaving little need for ClawHammer until later on in 2003.

 

[Duvie]

I agree with crazysaint...

I also think amd still has a product in the tbred rev b's and future barton that should keep it competitive and can definitely still be the best bargain around....The hammer and its 64bit marketing scam for the joe blow consumers doesn't make sense. In the server environment it translate into real world performance but this item is just too far ahead of the software and sorry to say it is going to really take intel to implement first to get it to really take of with software writers...I mean how can you argue with 80% of the PC market???

AMD needs to concentrate on actually delivery products when they are launched...Heck if the 2800+ was actually available I would say it has the crown slightly but by a mile when one figures cost....The barton with the 333mhz fsb and the 512kb cache can extend its platform. Let us also not forget some of the chips have opc'd well so it seems with correction of yield problems they should be able to push barton line up to 3200+ at least....

 

[Pink0]

AMD isn't go anywhere.

That may or may not be true. AMD is going through some rough times right now. AMD is making the largest gamble that they have every made this quarter and if it fails they will go under. reliable sources have it (from AMD employees) that the Tbred-2s are only getting yields of around 50% as of a few weeks ago. This would seem to be accurate. This time last year, AMD was producing exactly twice as many chips (same source) as they are this time. However, they have exactly the same ammount of fabs and production going into it. So if you have exactly the same ammount of chips attempting to be made but you're only making half that....hmmm....well, you be the judge. Now, the XP 2800+ is a "limited production" processor meaning that when barton comes it will be a 2800+ but will run at a lower clock frequency. If they are having this much trouble with .13u how much more trouble might they have with Silicon on Insulator on Hammer? Who knows. It's all speculation. Maybe T-bred 2 and barton are the only things suffering. I doubt it. AMD is flopping around like a chicken with its head cut off, "it will be SOI...no it's wont'...yes it will...now it's cancelled! now it's going to carry the entirety of 2003 for their consumer market." you don't just switch design and production overnight These are acts of desperation.
That's me being devil's advocate. I'm saying that if AMD has as much trouble with Opteron as they are having right now they will not be able to break into the server market and they will surely disappear without saviour funding. For all we know, Opteron is their one great project and they will make a killing with it.
Only time will tell.
Good Luck AMD.

 

[Pink0]

Evan Lieb, that is AMD's PR which is confusing everyone which I referred to. It does not specifically say Athlon Hammer. What it says is "desktop hammer" which could also mean a 2 way workstation opteron. This is what I said. AMD's PR has said "desktop hammer" but that could (and probably) does still mean opteron based on what AMD's VP and CFO stated. I think his statement that "The Athlon with the name Barton will carry the consumer sector through 2003" is very self explanitory. I could be wrong on this. I think it's important (always) to look past AMD's (or any company's) PR to get the real story. Looking at trends in AMD and their production difficulties as well as the company's VP's damning statements for Athlon DT in 2003 I would tend to discount AMD PR as just that: PR. You might have better sources. I don't know. These are just my speculations

 

[Pink0]

Transcript of conference call where AMD VP and CFO clearly states desktop hammers not available until second half:

"Soundview Technologies Analyst: Can you talk more about the Hammer schedule milestones and sort of where what the confidence level is right now on keeping that schedule?

AMD: Relative to the schedule, we're currently looking at mid-first half '03 to be our intial shipments of product. We are focused right now on the service space, we think the service space gives us the best opportunity in particular to break into the enterprise class customers which we want to break into and also gives us the area that's absolutely the clean kill . It by far will be the best server product available on the market. . . . "

...

"Prudential Securities Analyst: On Hammmer, it seems that your strategy has changed a little bit in that you're going to address the server market first and foremost. Why is that the case? Is there an issue with the performance of the chip, or is that a more attractive market Are you going to focus first on the Opteron family rather than on the Athlon Hammer core?

AMD: The first product we're bringing to market is actually based on what we recently called the Sledgehammer die. We think the biggest opportunity for this product is in the server space, and with the Sledgehammer die, that's really what the market is going to serve first. Also with the . . . putting Barton into our roadmap, which starts to ship in the first quarter, we believe that with the Athlon XP lineup based on that K7 core called Barton that we're going to be able to have a competitive performance product in the desktop space and the mobile space through most of '03 with that product. So the real focus on getting the benefits of Hammer is going to be driven largely by the server space and the workstation space to some extent and as we go through the year we'll get into the high-performance desktop space and it doesn't really become a player in the volume desktop space until sometime in the second half of next year. "

 

[First]

Key word there is "volume" Pink0. ClawHammer won't be in full production (in the millions) until later in the 2H of 2003. That's certainly believable, but who says that there won't be limited availability of ClawHammer CPU's in Q1/Q2 2003 as AMD has said? Correct?

 

[Pink0]

Key word there is "volume" Pink0. ClawHammer won't be in full production (in the millions) until later in the 2H of 2003. That's certainly believable, but who says that there won't be limited availability of ClawHammer CPU's in Q1/Q2 2003 as AMD has said? Correct?

Correct. Point well taken. I fear, though, that it might be the same way that the athlon xp 2800+ is launched right now. I certainly hope not. We'll certainly see. Good Luck AMD.

 

[CrazySaint]

Sure, AMD is having a tough time financially right now, and with product yields, and yes, they're pretty much betting everything on the Hammer chips. However, I believe that once they get the bugs worked out in their 0.13um manufacturing process, the future looks bright for them. And really, I think that's what's really key for them. They've obviously shown that they're having no difficulties designin very competetive chips, the problem so far has been in producing them in sufficient quantities. I can only assume that they'll have most of their manufacturing bugs worked out by early Q1 '03 and should be in good shape with the Barton in the desktop performance market, and with the Opteron in the low and mid-end server market. Even if the Opteron is only a moderate success in the server market, it should still significantly improve their bottom line.

 

[Pink0]

I think the largest problem that AMD is having is with their outsourced fabs. They seem to be having trouble making up their minds about which designs to run and this may be part of the delay in converting fabs. I'm sure that the ability (or lack thereof) of fabs has also affected roadmaps directly. Don't forget though, once they have gotten production of .13u up to par, they then have to retool the fabs for a much larger (and therefore more likely to have bad yields just because it has more transistors) dies in the barton core. Then AMD has to master production of SOI for the hammer series which may be more problematic for them than .13u. They may have mastered it already. Only time will tell.

 

[Wingznut]

Pink0, to run multiple product lines, it is not necessary to "retool" a fab.

For instance... On the exact same tools we've run Northwood, Tualatin, Gallatin... All very different sized dies, all ran back to back. Now when you change processes (say, from .13µ to .09µ), that's a different story. Most tools can still be utilized, but some cannot.


But to answer the original question... Imho, I think Hammer will be "released" by late H1, but won't be available in any decent quantities for a few months later.

 

[Duvie]

Oh I see...the amd 2800+ xp is a limited launch??? If that is like the hammer s going to mean that means none for us here...maybe the a few for the review sites and that is it!!!

I remember when we were bitchin about the 2400-2600+ paper launches at the end of august and ppl were saying the oems were getting them first...hmmm...that whole time and I never saw one pre-made system by anyone out there...Just this last week the 2400+ really started coming out yet I still haven't seen anyone post results here....

I can see the paper launch and "limited launch" of the hammer right now...sounds like a sad replay of the 2400 to 2800+ xp chips....

 

[Pink0]

Pink0, to run multiple product lines, it is not necessary to "retool" a fab.

For instance... On the exact same tools we've run Northwood, Tualatin, Gallatin... All very different sized dies, all ran back to back. Now when you change processes (say, from .13µ to .09µ), that's a different story. Most tools can still be utilized, but some cannot.

I did not know that. Thanks for the correction. There is still a larger likelyhood of reduced yields due to increased transistor count with barton though, is this correct?

 

[Pink0]

Oh I see...the amd 2800+ xp is a limited launch??? If that is like the hammer s going to mean that means none for us here...

Yes, I'm not sure if that's public knowledge of not Oh well, anyway, the 2800+ was, in no uncertain terms, PR. It was meant to show that AMD can remain competitive with Intel but not meant as a permanent part of the Athlon line. It will be a limited run and will be quickly replaced by the Barton 2800+.

 

[Duvie]

Hence why it is rumored not to hit shelves until q1 03??? I mean that is the time frame we seem to be hearing for the barton...

Will the 2800+ barton be at the same speed as the 2800+ xp or will it be slower with the fact the 512kb of l2 cahe should give it a possible 5-10percent performance gain agianst same speed chip???

 

[Pink0]

As stated in my earlier post the Barton will have the same rating of 2800+ but will run at a lower clock requency.

 

[CrazySaint]

Originally posted by: Duvie
Hence why it is rumored not to hit shelves until q1 03??? I mean that is the time frame we seem to be hearing for the barton...

Will the 2800+ barton be at the same speed as the 2800+ xp or will it be slower with the fact the 512kb of l2 cahe should give it a possible 5-10percent performance gain agianst same speed chip???

I would really hope that the PR ratings on the Bartons would maintain the same PR per clock rating as the XPs in order help counteract the P4s forthcoming HT advantage, otherwise Bartons won't stand a chance against similarly "rated" HT-enabled P4s if HT turns out as well as its hyped.

 

[Wingznut]

Originally posted by: Pink0
There is still a larger likelyhood of reduced yields due to increased transistor count with barton though, is this correct?

Yeah, but not so much with the increased transistor count. Yes, more cache adds transistors... But the cache transistors aren't as vulnerable as the ones in the more critical logic sections.

Now, how it could hurt yields is in defects with the larger die.

(Using very hypothetical numbers...) Say, with T-bred they get 200 die per wafer, and with Barton they get 170. If a defect affects 5 die, that's obviously a larger percentage on the Barton wafer than the T-bred wafer.