Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
When they originally expanded the cleanroom, they put it on top of the original building while it was in full operation...
Could this be a factor in AMDs alleged yield problems? If so, it would not be one anymore.
The thing is, I don't think many people were that enthusiastic about AMD roughly 13 months ago - that's why the stock was at $3.
Only the really fanatical, thoughtless AMDroids were blindly loyal enough to buy in at $3. We are lucky that so many here--you among them--want to save us from similar mistakes by offering objectivity and balance.
Had those fools who bought in at 3 only listened, they would have understood that SOI was unworkable, that Hammer was unproduceable, that Quantispeed was part of an elaborate Satanic conspiracy, that Jerry Sanders lights his cigars with AMD stock certificates, that Athlons randomly burst into flames, that AMD chips were inherently unstable and would never be considered by 1st tier server vendors...
Too late now for them.
Btw, I think today we really have a reason for more than one high-fiver.
Yes, but a butt-slap would be excessive.
I know everyone is enamored with taking over the high end and giving up the low end, but remember microsoft won the war from the bottom up, not the top down. So did intel vs. traditional risc.
These are not great analogies. Opteron is the bottom of the server world and has the price/performance that made Intel win out over risc. As far as desktops go, that whole market is low end, and I don't see what would be gained by giving away chips, particularly when AMD has a hard-won performance advantage. I can't see how AMD could do better with the hand it has, do you?
Dell is squeezing them with Xeon on the low end and HP etc are squeezing them with IPF on the high end. Embracing Opteron will help them with the first problem but not the second.
My own opinion is that while Opteron aimed at the Xeon market, it will hit alot of the big iron market at the same time. There's never been a 64bit x86, so it's pointless to cite existing market segmentation. It's also pointless to argue that Opteron has only been successful in HPC, as other high-end markets take much longer to get into--as they have for Itanium.
But wouldn't Yamhill, because of its x86 compatibility be worse competition for Sun than Itanium?
Yeah, it is a strange quote. The only way I can make sense of is that Sun is more worried about ISAs than particular chips. From that point of view, Yamhill and Opteron would not "compete" (if it is true, as Sun implied, that they are software compatible chips). The other thing is that AMD64 is supposed to be for the Xeon segment, which they already serve anyway, whereas Itanium aims right at the Sparc breadbasket. In any case, it's not hard to see why Sun, forced to choose the lesser of two evils, would back the chip with the more open standards and at least potential competition (from Intel) to keep prices low and designs responsive to their needs.
I'm not sure if Sun's recent deal with Fujitsu shows them to be serious about the future of Sparc or not.
As it stands now, Intel customers who need 64bit must go to Itanium, regardless of whether they are interested in all the wonderful features you enumerate here. Intel would love to keep the whole market segmented this way, because Itanium is much more expensive than Xeon. Opteron messes that up a little, Xeon64 would destroy it. Intel would effectively be competing with itself and endager the still fragile IPF ecosystem. How significant this effect would be is debatable, but your assertion that there is *no* overlap in these markets is wrong.
Personally, I think Intel is waiting for a certain level of Itanium acceptance before unleashing Xeon64, probably AMD64 compatible and latent in upcoming cores. They are quiet about it because they do not want to FUD their own product, IMO.
EDIT: The advantage of being AMD64-compatible is that they take advantage of the OS/Apps that are already ported or being ported. If Yamhill were being released in 2004 and not AMD64 compatible, then we would already know about it since they would have to give info and development rigs to MS and app makers.
This is exactly the shortage of demand that Intel had anticipated back in Q1, except it's three quarters late.
Now, I do believe the Flash shortage is good news for Intel too, but this is an extradordinary bit of spin. Are you really blaming the shortage for not coinciding with Intel's prediction of when it would occur? Intel lost a noticeable amount of money and market share because of this bad call. In the grand scheme of things it is not a big deal, but come on...
I am perfect market timer, it's just that the market is often several quarters late.
I think it has more to do with pride
I agree; I just thought I should point out what the article seemed to say.
I think it's very significant that it's going to be called "Windows for 64 bit extended systems" and not "Windows for AMD64." No conspiracy theory needed, MSFT has a reason to encourage INTC to develop on this architecture.
If Intel goes with AMD64, expect it to be called by a more generic name except maybe in technical documents. If socket 775 is really for AMD64, it might well be that most consumers will be unaware that this is AMD technology. It would be a loss of face, but if Intel executes correctly, only the geeks would remember or care anymore than people cared that the Athlon used the EV7 bus. That said, every month that goes by without a Precott64 makes it less and less easy for Intel to pull it off.
I really hope AMD will ramp their marketing in conjunction with their production volume. I saw the little AMD64 display at Fry's last week, it looked very lonely, just empty processor boxes and some info posters. They need demo systems, 3d games set up, all that... something to generate some excitement.
On another topic, way to go Ruiz on calling the Flash shortage! I think the Street will begin to get the idea that he's not an idle boaster like Jerry was. Credibility on the Street alone will raise AMDs valuation, not to mention the seller's market in Flash. All the ingredients that propelled us to 47 are in place and then some. But no predictions here, other than up.
Glaskowsky seems to think Intel has no AMD64 license:
As he points out, Intel will have a lot of trouble licensing AMD64 because the pill is too bitter for Chipzilla to swallow.
Not sure if this is just an assumption on his part. Sometimes these guys do know stuff...
What Sun article are you guys talking about?
I feel strange siding with Elmer, but as I understand it, defect density shouldn't be different between SOI and bulk. It has to do with foreign matter (dust) not the kind of wafer--same fab, same air, same defect density. If there are problems peculiar to SOI, they would not be defect density.
Unless there are using bad wafers...
EDIT: Well, process size should matter, but they are both at 13nm here too.
No doubt that the Atlon PR has them in a hole wrt Celeron. It would look tacky to change PR schemes late in product life, especially to explicitly tag it to an Intel product (PR is rated against Mustang, not p4, remember? ;) . However, no point worrying too much about XP while it is slowly being phased out. A64 already is PRed to Pentium4, probably good against Prescott, too, but we'll see. I would rather they get out of the PR racket altogether... and they are doing this with their premium product. I guess the thinking is that sophisticated buyers don't need it and the general public still does. Probably right, but Centrino will change that gradually.
Why bother making something to compete with Transmeta when it would be simpler, maybe even cheaper, to just buy the company?
On second thought, I think that between Geode and low-power k8, they may be able to spackle over the Centrino-sized hole in their product line. I'll be interested to hear what next-gen Geode plans will be.
As far as TMTA, I think it might be better if they use that money to develop their own product, assuming they had the money/credit to buy them out.
why wouldn't AMD just continue making Athlon XPs?
Even if 64bit is removed instead of just disabled, Paris would still have mem controller (for socket 754 board compatibility) and SSE2 support--both would boost performance and garner higher PRs at equal clock.
I'm assuming that it will be socket 754, otherwise it would ditch the mem controller.
13nm bulk has hit its speed limit. The good news is that AMD has shrunk the k7 to 90nm SOI (and added a bunch of other stuff). It's called k8. Why would they do it again just to make a value chip that will just serve as an ASP placeholder? Better they should work on k9, sse3, low-power... The "New" AMD is going after the high-end, high asp markets. To protect those markets and appease mobo makers, they'll castrate some low cache chips, no big deal, this is not going to be the flagship big volume product.
As far as selling points go, I agree that 64bit will be a big one. It would be dumb not to make people pay for it by giving it for free in a Duron-equivalent.
Slides stated that AMD's mobile focus is on Desktop Replacement segment through '08. I'm taking this to mean that there is nothing on the roadmap that will compete power/performance with P-M or TMTA, too bad. I don't like this, but maybe it's the best use of resources for the time being.
Elmer and other people who know process...
What do you make of this?
http://images.visualwebcaster.com/17861/19392/slide10.jpg
EDIT:
Disturbing, it looks like there really are AMDroids:
http://images.visualwebcaster.com/17861/19392/slide31.jpg
KeithDust, it sounds like rationalization. If you can add 64 bitness for next to nothing, why not. Applications not apparent today will develop for the chip. If nothing else, you can say why buy a Celeron/P4 when you can get an Athlon and get 64 bits for free.
I used to be on your side on this, but I've been convinced that it is a bad idea. AMD cannot meet demand for all segments. They don't want to steal $100 Celeron sales, they want to steal 1000$ Pentium 4EE sales. A64 is a compelling product that people are willing to pay more for. AMD should let them. Well, I guess make them is more to the point.
You can see what's happening now with the Athlon XP competing with Celeron instead of P4. The clock to PR pricing is identical, even though the XP is clearly superior. Now, if there were a value line like Duron, Athlon XP prices might have held up much better against P4.
The other reason I've changed my outlook on this is because I am more and more convinced that Yamhill will be AMD64 compatible if/when it comes. Saturating the market with AMD64 only makes sense if AMD is trying to dig in against an upcoming Intel-x86-64 chip.
"He said the company has a site selected for a 300 mm factory"
I did not hear that.
I did
It was also in the q3 CC. Ruiz said they would announce plans to build a new 300mm 65nm fab before the end of the year with an unnamed partner.
Tim Fowler,
On this version of the new road map its in Orange like the other 64 bit chips. The 32bit chips are in purple.
I guess they use orange for SOI process.
I thought the orange was for K8.
But that also gives Intel more time before they have to knife Itanium, and reduces the pressure, if it exists, to adopt x86-64.
It could also be that AMDs (alleged) release of a 32bit K8 is a sign that they don't need to push the porting issue, i.e. that Intel will be supporting AMD64 in the near future. I don't know if I believe that, but it's a possibility.
Exactly. You said he didn't say it and I said he did.
Not me, I think that was Borusa. I just didn't understand the argument.
I guess your point is that we don't know for sure (i.e. from an Intel press release) whether IPF was to become a desktop solution. My response is, so what?
Borusa--I don't think he meant that x86 was going to die anytime soon, just that Intel's plan for its (64) future was/is epic
I'm having trouble seeing the difference between the two claims, except that one uses a harsh-sounding word.
Umm... what's an i860?
I don't beleive in mainstream IPF. No more than to marcian invasion.
I don't either. But you don't think that Intel would love to do it if they thought it would succeed? It would kill AMD, giving them complete control over pricing and architecture. Now there are alot of reasons why they wouldn't go with this strategy now, especially with Hammer out, but it would be naive to think it wasn't/isn't being considered very seriously.
Do you honestly believe that MS is generating an AMD64 OS that has nothing to do with Intel?
You sure seem to think so:
Intel? Compatible? Are you joking?
A superset? At least.
Totally different? Why not?
In case you're not getting this, Intel can only use Windows-AMD64 if their chip is AMD64 compatible.
And as for this exhange:
<Do you think it is IA-64 compatible?>
I don't even understand this question. Why would it be? That wouldn't compete with AMD at all in this market. IA-64 is a totally different market.
I don't think it will be either, but if you can't see why Intel might want to mainstream IPF, you shouldn't be managing your own money.
blauboad, chipset + cpu revenue was $503 mil. Say, $480 mil were processors. If you think volume was significantly below 8 mil, do the math and see how crazy ASP will go.
I don't have a dog here, I just want to know where people are getting these numbers, since I'm not aware that they are reported by anyone. Ok, so you estimate ASPs and product mix based on Pricewatch or other and then divide reported revenue. I assume this is what Elmer is doing too.
This is an interesting way of speculating, but it is inherently imprecise. We don't know, for instance how many and at what price AMD is selling Durons. That could throw off alot of numbers...
Elmer, we know AMD made over 8 million chips in Fab30 in Q3
Elmer won't tell me how he gets his numbers. How do you arrive at yours?
sgolds, great analysis.
It looks like Intel's response will come in the form of Tejas in early 2005. Certainly the initial versions of Prescott are not likely to have active 64bit capabilities (though they may activate them in conjunction with socket 775 mid-late 2004). That said, I would not be surprised to see some attempt at a FUD64 strategy, i.e. an entirely marketing-based "64bit" that will preserve premium in the consumer market while keeping Itanium safe. SSE3-"64" or something similar. Bottom line: AMD will most likely have time to ramp 90nm before Intel enters the ring.
If intel goes x86-64, it will be AMD64. They already have a proprietary "x86 compatible" instruction set called IPF that they would love to get on desktops. There is no point or advantage in developing their own x86-64 other than killing off AMD. According to the Intel boosters here, Intel won't do that for anti-trust reasons. Not to mention that with a proprietary Yamhill they'd have to wait for software support which would take another year and kill the whole point... 3D-Now had nowhere near the committments that AMD64 has, and for other reasons the comparison is not valid.
Under this scenerio, Intel's opportunity comes before AMD ramps up 65nm/300mm. They have a one year window - 2005, to be exact - to turn on the volume and reduce prices, squeezing AMD in that year.
So, short term, an Intel entry would cause demand for AMD to skyrocket and AMD would be the toast of the town. Medium term, 2005 could be a problem as Intel manufacturing kicks in. AMD would have to tread carefully, once again pulling a rabbit out of the hat or becoming toast (I do hope they get royalties from Intel). Long term, if AMD opens a new 300mm/65nm factory in 2005 for 2006 product, simultaneously with maintaining Dresden for mature process, then it could be a whole new world.
Remember that AMD and IBM will share the same process. IBM would be thrilled to make k8s for AMD until the new fab comes on-line. It will come at a cost, no doubt, but the ASPs will be higher than what AMD has been used to, especially if AMD steals the prestige and innovation premium that Intel has enjoyed. Even at no net profit (NBL-not bloody likely) it would be worth it to keep share and market presence up.
I would expect that the switch to 90nm, possible contracting to IBM, and then new fab at 65nm 300mm will be accompanied by increased committments to the big OEMs. Assured supply in sufficient quanitity is what these guys really need. This is why the HP support seems weak now while AMD is producing in the 1000s. That will change as we produce in the millions...
On the earnings side, the fiscal displine that AMD has learned through the lean years will make it shine in coming better times.
And today's gossip implied Yamhill is AMD64-compatible. That would be an endorsement, not a bombing.
It's better for AMD if it is AMD64 compatible, but it is still bad news if Intel gets into the x86-64 game any way you look at it. At best, we go back to the Athlon days of selling more or less comparable chips, probably having to undercut on price, etc. At worst, Intel muscles the industry toward its own x86-64 and squeezes AMD out of the mainstream market.
The bomb that everyone's waiting for from MS or INTC is definite word of Intel's dektop 64bit strategy.
I think MSFT naming it "Windows for 64bit Extended Systems" is a hint of what Intel is going to do. P4 Extended Edition anyone?
7-8 million processors
Yes, this is the number I'm looking for. How did you get it?
What number are you using for Fab30 sellable output?
Ok, but I meant where are you getting these numbers to plug in. Die size is about the only thing we know for sure, and even there we don't know the product mix of different die with much certainty.
How do you estimate the wafer capacity of Dresden?
How do you know how many saleable die are produced? A few ways pop to mind, but they all seem very imprecise.
I have a feeling I'm going to regret asking this, but from what data do you figure low output from FAB30? How much do you think should fab of Dresden's size and process be producing? And, is this the only reason you have for believing in low yields?
I'm asked "can I get you to verify your address"
It's even worse when it's to obtain a shipping address for an order (vs. for security/identification purposes). Then the phrasing suggests that the schmuck on the phone *really* knows where you live, and you're just taking a wild guess.
Thanks, Elmer. Another thing that's going to irritate me from now on... ;)
You are splitting hairs over the use of the subjunctive. You stated that my usage was incorrect. I have shown it was not.
No, you're quite wrong. The link Doug provided would show that, were you to look at it.
I'll venture that there is no conceivable situation in which the phrase "If I was" would be grammatically correct. It is certainly not "present subjunctive" as you claimed.
Sorry for the OT posts; this has hit a pet peeve of mine.
Doug re: subjunctive...
As someone who cares about language, thanks for bringing up the topic of the subjunctive case. If I were to hear or read the phrase "If I was" one more time, I would be just... well, I would... be really mad one more time. You even see it in newspapers.... ach!... People whose job it is to write do not even have a basic grasp of the language...
Now, on TV, I've come to accept that kind of thing, but I still can get shocked at how inarticulate those guys are. I just saw a guy on Fox News report that President Bush "was all about Iraq" in recent statements to the press. Between that, and half of my state on fire, it makes a person look at the big picture.
If the frequency ramp is going along as well as it should be, 256k cache models could replace the 3000+ and 3200+ models, freeing up the 1meg models for Opteron and top-model Athlon. Not sure what they would do with the inevitable sub-3000+ 256k cache chips, though... except maybe sandbag for the upcoming socket 754 Duron/XP line. They would have to be confident in their process. The die size savings might make it viable to dump the old XP cores altogether in that case, which I think is the plan.
Are you saying that lower cache is less of a performance hit on FX than single channel?
And I'm not sure whether the 3.2Ghz P4 is absolute standard that AMD has to measure themselves against. I see no reason for them not to release chips that are slower--they will have a market. Getting AMD64 in as many boxes as possible while raising ASPs ought to be AMD's goal for 2004, not biting their nails about P4 3.2g.
As far as manufacturing numbers, didn't some additional capacity come online, or is scheduled to come on-line at Dresden? I recall reading about an add-on the fab there. Might help to understand the numbers, or not.
That big volume day after the CC was interesting. There's alot of new blood in @14. Should be pretty strong resistance. With luck, it will hold until the institutions start buying...
chipguy, on possible Prescott delay reason
>>My own opinion is that the purported delay in Prescott
is mainly due to a desire not to undermine Xmas sales of
Northwood P4s by announcing a superior next generation
device that couldn't ship in any quantity for the Xmas
buying season.<<
Doubly, if not more, true if you consider that Prescott may require a slightly different motherboard. Intel's hands in the chipset/motherboard market would be severely burned.
Isn't this just a way of putting a nice face on it? I mean, why can't they ship in any "quantity" before Xmas? Problems in process and/or in design. And that takes us back to the original question...
My comment about it being design problems was predicated on the Xbit labs piece that there are a bunch of 3ghz 90nm Precott-Celerons being stockpiled waiting for adequate supply of Pentium-Prescott. If that really were true, then the 90nm process would be more or less ok and the problems mostly in Pentium-Prescott design. Intel CC mentioned having to redesign parts of the chip.