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.
If you can put any stock in the Xbit article, it looks like Intel's 90nm is ok, and that the problem is specific to Precott--either SS3, cache, and/or clock speed. This goes nicely with what was said in Intel's CC about it.
P4EE easily beats AFX is many benchmarks
AFX beats P4EE in many more, and a higher speed grade may be available by the time Intel actually releases it.
It's supposed to be faster than Prescott? Says who?
Let's just handle this with logic. Intel announces a 1000$ P4EE chip available--maybe--in November. Ergo, Prescott is going to be really really late, really expensive, and/or significantly slower than P4EE. With extra pipeline stages and concomittant hit to IPC, along with less cache, there's good reason to think that P4EE will be faster clock for clock. Last I heard Prescott was to be introduced at 3.4, about a 6% improvement over P4EE, so who knows, maybe 3.4 will be about even with 3.2 P4EE, but I doubt it. And then all the down-throttling it's going to have to do...
See, I didn't even mention the Inquirer.
AMD's 90nm process on schedule? Which schedule?
Mid 2004 for volume production. This timeline has been very consistent for quite a while now.
Based on past performance, the chances of Intel delivering are far greater than AMD.
Based on Prescott delays and wattage rumors, it does not seem that Intel's 90nm is going smoothly. Unless, as Otellini suggested, Intel intended to make 100+ Watt desktop chips... That's not to say they won't fix it, but Prescott and Tejas delays buy AMD time--no way around that.
As far as AMDs 90nm switch, I am fully aware of the risks there, which is why I put that condition into my comments.
Seems she's a little light on her computer history. Microsoft has had OSs for Alpha, MIPS and ARM.
She probably meant non-Intel x86.
Everytime Intel bumps their roadmap up in frequency, AMD needs to update two pinouts to keep pace.
Actually, in recent days it's been Intel who has had to keep pace. Given the superiority of AFX over P4EE (which is supposed to be better than Prescott), there is little reason to think this is going to change soon. Maybe with higher-clocked Precotts, but then only if AMD's 90nm process is not on schedule.
Do they share a common die? If not this is going to be hell trying to forecast and build to.
From what I understand, they do have a common die. 940 and 754 do, for sure. 939 might require an update to the memory controller--anybody know?--to make dual nonregistered DDR work, but there's no reason to think that they all couldn't use the same die.
Imagine having to actually forecast and build "value" processors rather than just downbinning them from the "performance" line in response to market demand.
The value line for near future is Socket A. The XP will become a value line (maybe with a Hammer core, maybe not) on socket 754 in 2004. However, that does not mean that socket 754 is an all-value platform, anymore that socket 478 is an all-value platform because it supports Celeron. Socket 754 will support high-end and low-end desktops, sockets 939 is for the highest-end, workstation-class processors.
What impact does this have on their motherboard partners who are trying to figure out the same thing with a volume to split amongst them of <1/6th the Intel compatible volume? Now in addition they need to figure out the split between performance and value footprints? How can they possibly make any money?
Intel dominates the Pentium chipset market, wheras AMD stays out of its own chipset market except for servers. That means that Via, SiS, nVidia, ALI, ATI all get a chance at complete participation and greater return. Not to mention that AMD64 boards are going to be cheaper to design and manufacture given the integrated memory controller.
As far as socket differentiation, AMD will be able to tell them well enough in advance, because, as I've already explained, 754 is not just a "value" socket--there is flexibility in this platform, almost as much as in socket A or 478. Since most (all?) mobo companies that produce for 939 will also have a 754 product, it will only require increasing production on one line, decreasing on the other--not a big deal. Design requirements for the platforms are identical except for the socket itself and the memory traces--trivial stuff.
Writing all this, it's occured to me how well AMD has improved on its market segmentation with the new generation of chips. Ruiz really is much more a businessman than Jerry.
EDIT: Although, Jerry was a better trader ;)
Nice wish projection there buddy.
You have it all backwards. As an AMD long, my greatest hope is that Intel continues to dump its money, time, and reputation into Itanium. Then AMD would have no competition.
Intel's silence on Itanium sales (opposed to their disclosure of Xeon % increase) pretty much says it all.
Herb Greenberg is a fool, his opinion on AMD aside (even a stopped clock tells the right time twice a day).
Remains to be seen if AMD will finally have the last laugh, but that's Hickey's bet. And before you go saying that Hickey is a washed-up perma-bear, just remember this: That's what I used to hear when I would quote him in the mid-to-late '90s.
Now, there's a *big* difference between being bearish in the mid-90s and being bearish in the (very) late 90s, equivalent to the difference between a 72 Pinto and a late-model Ferrari.
Anyways, Greenberg just gets fed recommendations from hedge-fund managers often with vague, wrong, or non-existent rationale. IMO, he's a shill for these guys to attract sellers so they can get out of their big short positions without causing a squeeze.
Hector: "We don't need 300mm in 04. We need it for 65nm node in 2005 for ramp and 2006 production. We will be public to partners on a fab by years end."
As has long been speculated, Ruiz all but confirms AMD will be building a .65 300mm fab with IBM, TBA in the next couple months.
Congrats to everybody who's held on. I expect alot of revisions and upgrades, institutional buying should run AMD up nicely from here. Moody should de-junk the bond rating soon.
Meanwhile, Sun's keynote had a similar theme, but their solution was to solve the problem using software (i.e. Java). That seems to me a better solution than using x86 (or in AMD's case, their self-serving AMD64 advocacy), when it clearly doesn't come close to meeting the low power and high embedded performance of ARM/MIPS
To be fair, Java is a pretty inefficient solution, particularly for resource-starved embedded processors, and we have not yet seen an AMD64 embedded processor. Certainly the Hammers aren't going to work in a PDA, but that does not mean the instruction set cannot be deployed in a stripped-down, low power design.
Both ARM and MIPS cracked a joke a Webber's expense when they took the podium for the high performance embedded portion of the conference
Not surprising, since Weber basically was proposing to make their companies, expertise, and designs obsolete. Whatever you think of likelihood of Weber's proposal, you gotta admit he's got some balls to announce it at a forum attended mostly by non-x86 people.
From the K9 article someone already posted (thanks):
Separately, in a keynote speech at the conference, Weber sketched out a future of computing in which most computers, handhelds, cell phones and other devices will rely on processors based on the so-called x86 architecture. The overall x86 architecture is at the heart of chips made by Intel, AMD and a few other small companies, with K8, K9 and Intel's NetBurst architectures being x86 derivatives.
The reason for x86's rise, he said, comes down to history. Most software is written for chips based on the x86 architecture, which is also the oldest microprocessor architecture. There is also a great diversity of x86 chips.
"The time is right for x86 processors in all form factors," Weber asserted. "Because of its dominance on the desktop, it has grown to have the vast majority of software."
http://zdnet.com.com/2100-1103_2-5091845.html
Somebody here mentioned that they thought Alchemy might be eliminated. It sounded strange given that they've just announced a reference platform, but it seems like there might be something to it. Though I think x86 moving up the chain into high-end systems is the way of the future, I am skeptical about moving down.
I know Combjelly and maybe others here work on embedded stuff, any thoughts?
Elmer, They'll promise profits are right around the corner and you'll believe them again. It's the battered wife syndrome.
I can quit anytime I want.
AMD is going down this morning. I wouldn't be surprised if a disappointment is in store, at least compared to what has already been factored into the price.
Now this would not surprise me. I've taken a little off the table today--shares I've held more than a year--in expectation of a selloff and to free up some cash. Regardless of what happens in the coming months, we've have a nice runup.
I'm holding the larger part of my position at least till January for cap gain reasons, and because things still look very good for AMD. I do expect to see gains following any post-CC selloff. I do not expect a stellar CC that will send the stock through the roof, though I'll sure benefit if it does.
I'm trying to make money and it looks like I'm winning.
Exactly how do you know this? Are you stealing my mail?
You guys don't need good reasons. It said so in the inquirer!
Weak, Elmer. I accept this as your concession of defeat.
Elmer,
Then Intel would have done a 90nm shrink of NW and enjoyed a nice ~10% performance boost.
Another good reason to suspect that 90nm is in trouble.
Your scenario doesn't make Intel look any better. You're basically saying that Intel is screwing over its third party chipset and mobo people along with its OEMs *again* after already revising the socket 478 power specs once before. It's bad business to hold off on the launch of a major product until after Christmas just to make things smooth for a chip that might come out 6-12 months later. Can't buy this--they could easily revise the spec then when it's needed and not sacrifice now.
That is, unless Prescott is embarassingly slow now and they can't make it sellable without exceeding the already heightened power spec. Then your scenario would work :)
Bad analogy, logicguy. Ruiz has never claimed that losing money was a financial goal. The point is, Prescott TPD is an obvious screw-up, but Otellini just won't admit it. It looks worse when you make feeble excuses. Change of "thermal target," come on. If he'd just admit, I would believe that Intel's 90nm is on track now, but with that answer, I don't know.
"The (90 nanometer) process is very healthy. We ended up changing the thermal target for systems for Prescott as we got silicon out and ran it against our models and what we think the likely speeds are going to be on that product," Otellini said.
http://news.com.com/2100-1003-5091113.html?tag=cd_top
I'm reminded of Peewee Herman falling off his bike and declaring, "I meant to do that."
HP getting pretty desperate?
In a white paper that HP began circulating last month, engineers at the Palo Alto, Calif.-based company said AMD's Opteron misses the mark in key processing requirements and that HP continues to favor the Itanium processing platform it co-developed with Intel.
"A microprocessor's virtual address space defines the maximum amount of information the processor can reference," HP engineers wrote. "So a much more useful definition of 64-bitness requires a 64-bit data path and 64-bit virtual address space." The paper said that "the virtual address space of Opteron is limited to 48 bits."
http://www.channelweb.com/sections/AllNews/article.asp?ArticleId=45004
48bits=256TB of RAM. If that's the worst criticism they can come up with, then there must be a big fight going on inside HP over Opteron.
No, he is definitely NOT referring to Itanium.
Care to explain exactly what you mean?
There's no reason to totally degrade the 64-bit capabilities of Athlon 64. A chip can still be AMD64-compatible even if it can't be connected to more than 4 GB of memory. The logical address space is still 48 bits (or whatver AMD64 says), while physically there's only enough address pins for 4G of memory.
Yeah, but so few desktops will use >4gb of RAM in the near future that this is almost useless as a market segmentation gimmick, because there wouldn't be any real performance hit. I suppose it would work if they only "degraded" the slowest-binning hammers, that way you would have a speed and memory limitation. Of course, they should still market them like a Duron64, not Athlon XP, in order to encourage consumption of ported software.
sgolds, Given your three options, you're probably right, but..
If yourbankruptcy's comment has something to it, and AMD does plan to contract out k7 productions, then the scenario makes more sense to me. They can't contract out A64 to anyone but IBM, but some version of K7 should be produceable by one of the Tawainese foundries by now. Maybe even UMC--there was never an official break up. That would free up Dresden for 100% AMD64 production. If demand cranks up, they go to IBM. With all the money they make, they build a new fab or two and everyone is happy but Intel.
My major worry about AMD going forward is capacity, this seems to be a way they could manage it. Demand and supply are not independent. If the OEMs don't think AMD can supply, they will not demand...
evidence is that they will produce 32-bit processors as degraded socket 755 A64s.
I realize this is for market segmentation, but how smart is it when it does not seem there will be much saved on die size? Shouldn't AMD be trying to seed the ground with as many 64bit chips as it can? Especially since these will be socket 755, it seems a wasted oppurtunity to me.
I think the Prescott New Instructions will be marketed as a rival to AMD64, even though it will not allow true 64bit addressing (that would kill Itanium). Until they switch to socket 775 at 3.8ghz, they don't even have enough pins for 64bit. But whatever it ends up being, I would bet almost anything that Prescott is *not* AMD64 compatible, and therefore would not be able to run Windows64 for AMD64. Details of the instruction set have only been available for a year, and Intel would have had to design a completely new implementation in order not to violate AMD patents. Not realistic, I think.
By Dothan, I think you mean Tejas.
If true, it would explain the delay of prescott into q1 04 for it to coincide with the release of windows64
Apart from not making any business sense, this is a little too generous to Intel. The delay has more to do with the 100+ watts (starting to look like a quantispeed rating) and a new 90nm process.
On another note, have the inner workings of AMD64 been public long enough for Intel to have already produced (in Prescott) a clone?
Funny little banter between EP and Windsock regarding Opteron
Reminds me of an old joke. Two old ladies are sitting in a restaurant. One says, "The food here is terrible!" The other replies, "Yes, and such small portions!"
Some excerpts from BW/AOL chat with S&P tech analyst (AOL keyword BWtalk).
***
BusinessWeek Online
Transcript of Oct. 7, 2003, investing chat
TECH STOCKS TO TRY
Thomas W. Smith, group head of the Standard & Poor's analysts covering technology stocks, addressed questions on how S&P sees that investment area now. The chat was co-hosted by BusinessWeek Online and S&P.
***
Tom Smith: Our view on the technology sector is positive, and we are overweighting the sector, as we believe the American economy is in the midst of a multiyear turnaround. We think information technology equipment -- both hardware and software -- will grow faster than the overall American economy through 2005. And therefore, we generally see good things ahead. Valuation for tech shares in particular, and the market in general, is always open to a lively debate. However, we feel that generally low interest rates and the prospects of an economic recovery help to justify a generally high valuation level for the market.
Overall, we are bullish on the market. I might note that S&P's investment policy committee, which has a recommended allocation by asset class, is recommending 60% equities, 25% cash, and 15% bonds. As of Oct. 1, the investment policy committee's yearend target for the S&P 500 was 1085.
***
Rontan 226: Any pickup in IT spending?
JackBW: I gather you expect it to come -- has it begun?
Tom Smith: My sense is that it's beginning, but has not accelerated in a strong manner so far. As I listen to the semiconductor companies on my beat, one senses that their end markets have been stabilizing for several quarters now and a better-than-seasonal pattern for PC sales appears to be at hand, according to reports from Intel (INTC). However, we look for a more decisive improvement in information technology equipment sales in 2004. Quickly grabbing the S&P Economic Outlook for August -- so I'm maybe a month behind the latest estimate -- our economist, David Wyss, shows real GDP moving from about 2.4% estimated for 2003 to 4.1% estimated for 2004 and 3.9% estimated for 2005.
Now, within that, equipment investment, which is something of a proxy for technology equipment sales, will turn positive in 2003 and grow at 4%, clearly higher than the 2.4% estimated for the economy by Standard & Poor's. And equipment investment is projected to improve to 9.5% in 2004 and grow at 8.7% in 2005. Both of those figures are more than twice as fast as our projection for the overall economic growth for the U.S. economy. I feel good about the prospect for tech spending generally. That said, the pace of tech sales in this upturn is not apt to be as strong as it was in the great heyday of the late 1990s.
***
We also have a 5-STAR (buy) recommendation on Intel (INTC), which has delivered consistently strong results for this time in the semiconductor cycle. It seems to be leading the way up for the industry. We think that they have partly made their own luck by focusing on the Centrino technology for wireless notebook computers. The gross margin story has been attractive for Intel, in our view. And we think they should do pretty well as American corporations start to hire again in 2004 and 2005 and have a need to replace PCs that may have been bought back in 1999 for Y2K compliance reasons. The typically useful life for a PC is about three years. So as we begin to move four and five years away from Y2K, we think there will be an interest in upgrading PCs generally, and moving on to wireless notebooks, in particular. There are many other semiconductor companies recommended as 4-STAR accumulates.
http://businessweek.com/technology/content/oct2003/tc2003107_8905.htm
Q: What do you think about Advanced Micro Devices?
A: Well, we expect AMD to do better sequentially in the September quarter, and better again in the December quarter. Clearly, the PC sector is recovering. Particularly for AMD, we expect ASP [average selling price] to rise noticeably. We think its new Athlon 64 and Opteron chips will be big successes. The company is well ahead of Intel (INTC ) in launching 64-bit desktop and mobile chips. Microsoft (MSFT ) has strongly supported AMD's 64-bit efforts, and someday that could translate to marke-share gains for AMD.
There are a few negatives: AMD lost some share previously in the Asian and European markets that it needs to regain. Also, the company's balance sheet isn't as strong, and the breakeven point is high.
I thought A64FX was single-processor only. Did I miss something?
"Also on display was the “Black Widow” system featuring a dual-Athlon 64 FX system on the MSI motherboard seen earlier."
http://www.amdmb.com/article-display.php?ArticleID=261&PageID=10
http://www.amdmb.com/images/computex2003/day1/via_widow_big.jpg
Since Hammer integrates a northbridge, why aren't all Hammer chipsets "single chip solutions"?
Bribery: "anything given or promised to induce a person to do something against his wishes"
This is a strange definition. My Scribner-Bantam dictionary does not contain it.
The way I see it, anyone taking money or other compensation in a quid pro quo arrangement is, by definition, doing exactly what he wishes--he wishes to take the money, and is willing to trade duty, integrity, etc. to get it. It's perfectly consensual, both ways.
Short of coercion (which is not bribery) how can anyone be induced to do something against his wishes? Particularly an economically-motivated decision in which personal emotions are not a significant factor--I doubt Sun will be able to keep many customers by applying the old guilt-trip.
Anyway, if you buy the premise of this definition, then HP *would* be guilty of "bribery" as would every advertiser who ever printed a coupon or gave a rebate or a free sample. . . I just don't accept that people or corporations should be viewed as so weak and morally helpless.
Designing and validating a dual core Opteron is the easy part, considering that Opteron was developed with this in mind. The problem is manufacturing a die that big. They have to wait until the process is mature enough so that producing them is economically feasible.
This way they can keep Opteron as just a processing unit inside the server. The way Intel bus works, technically and legally, is that cpu becomes the core the whole computer is build around. There is nothing that can prevent Sun from having anyone to clone their computers, as they never will have more control on the cpu bus than anyone who has a license.
I've never heard of this, could you clarify?
The way I see it, SUN can make Solaris run only on proprietary Sun harware (or those machines from other vendors which are specifically qualified by Sun) if they like. The DMCA would protect them, and no legitimate business would circumvent. Are you saying that Intel forces its custumors into some EULA that would restrict this?
Right. AMD goes out of their way to encourage the third-party support it depends on. Buying SUN would alienate their best and most high-profile partners and customers--MSFT, IBM, HPQ, etc. This is completely counter to Ruiz' strategy.
No, better to win customers than buy them.
Yes, they could do exactly what you suggest, they could kill AMD in short order. That would be the actions of an unethical company on a vendetta instead of an ethical company with the shareholders interests in mind.
That is basically what Walmart does. Sometimes they even price their goods below cost to eliminate competitors. The vast majority of this is legal, and when it's not, someone screwed up (rather than intentionally broke the law) or the law was fuzzy to begin with. It's been a very successful strategy over the years, to say the least, particularly for the shareholders.
Following your logic, you must believe that either Walmart is an "unethical company on a vendetta" or Intel's management is much less capable than Walmart's.
Or they're both wrong and there are other reasons besides incompetence and the boy scout pledge for why Intel does not compete more aggressively against AMD.
Intel has had countless opportunities over the
last two decades to twist a knife in AMD's heart if it was so
inclined. But AMD is Intel's never fail get-of-out-monopoly-
charges-free card.
I'd be interested to hear a few these opportunities you see Intel as having passed up. It's not that I disagree; I suppose Intel has the money to hand out processors for free for a couple of quarters if they wanted too. I don't see, though, that they could do it without damaging their interests in the long term as such tactics would increase pricing pressure even with AMD out of the way.
Can someone give me a link for AMD short interest? Open options, too if you have it. Thanks
Well, the rumor says this is Tejas/PV which would have a different socket, pinless I think. But, yes, it does sound odd.
I think Stroligo's got the idea. I doubt Prescott will be 64bit (or at least that this feature would be activated), but Intel's and MS's comments certainly suggest that Tejas will be. If Tejas does sample in Q1 (and they're having enough problems with Prescott!) and Intel manages to start selling 6 months later (optimistic, probably, given the novelty of what they're trying to do), then that puts initial release right at the point around Q3-4, or right when AMD64 has really ramped up and conversion to 90nm is happening.
I agree, too, that Intel will couple Tejas with an IPF co-processor rather than AMD64 or Yamhill--because neither of those would require enough extra silicon to make splitting up the part sensible. If Intel is serious about pushing 64bit to the desktop, this is going to be the only way to ensure backwards compatibility without a performance hit.
Of course, this is going to be a ridiculously expensive way of going about it, either for consumers or for Intel (if they subsidize it, which seems to be the IPF business model so far). Assuming that AMD64 will be priced closely with Intel's 32bit offerings, it will be a hard sell to get customers to buy two processors to get 64bit. Then think about the power profile of the Prescwatt successor plus an Itanium in the same box. No notebooks off this combo...
From what I'm reading, this is going to be a hard sell, unless Intel leapfrogs AMD in performance or unless AMD really botches 90nm.
From Intel's perspective, it would be a way--maybe the only way--of delivering a 64bit desktop that wouldn't kill Itanium, nor Itanium's "profit" margins.
We can almost declare Yamhill dead. It would already have developers working on it and rumors all over the place if it were really on the map for 2004--NDA or not. And AMD64 compatibility is just not Intel's style.
And this is all assuming that the rumor is true :) But we should expect to hear alot more from Intel re:64bit desktops soon in any case.
Funny, they answered a question you didn't ask, and didn't answer the question you did ask.
Now we have AMD's dogged development of x86, both features and process, and the legacy of Intel's market presence behind AMD64. Will AMD do to Intel what Intel did to IBM?
If they don't trip up too badly, they have a good shot. And I think MS is working overtime on the WinXP port as much in order to help them succeed as to stave off competition.