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.
Keith, yeah but my point was that the new Intel line replaces previously Intel based systems only. Or in other words even though the article doesn't mention it with one word there might still be media center AMD based HP systems in the future. I'd rather would have liked AMD being mentioned in this article too, it slightly bothers me, but I don't think yet it means no media center AMD HP systems in Q2.
Regards,
Rink
As far as I can tell neither of the old systems use AMD cpu's either, which is a shame too I agree.
Regards,
Rink
Aleph, re: I think I heard something about a Fab in Ireland being refurbished to the cost of 2 Bio$. Maybe that's been canned now that Ireland pulled the sponsorship worth about 350 mio.
No the 'refurbishment' of fab 24.2 to the tune of 1.6B euro's is not cancelled but is going ahead. The amounts that are mentioned in the media for the 'sponsorship' that got canned varies between 130 and 180M euro (whatever the amount it is not that significant really for Intel; but it is significant that AMD did get subsidized heavily both by Europe and German's State of Saxony). The new fab capacity Intel is seeking will cost $2.1-3B.
http://www.timesonline.co.uk/article/0,,2095-1512454,00.html
http://www.finfacts.com/irelandbusinessnews/publish/article_1000717.shtml
http://www.theinquirer.net/?article=21634
Regards,
Rink
wbmw, You conveniently forget all additional Intel fixed costs associated with unused fab capacity (hint: fab depreciation is by far the highest part of fixed costs for any chip). I'm not saying I know exact amounts here, I'm just saying I think this is the case, and am especially saying you don't know it and forgot to consider it in your post.
You also forget you don't know the details about Intel and AMD yields and bin splits.
You also forget Intel's mobile, desktop, and high end server cpu designs have almost exactly nothing in common when compared to AMD's mobile, desktop and mid range server cpu designs that have almost everything in common.
You also forget that using a new (slightly modified in that it uses different transistors) design for Turion not only increases costs but allows AMD to tap into a previously untapped market.
I can go on and on, but basically it would not be a bad idea to acknowledge you - like virtually all people here, incl. moi même - are in no position to conclude much about variable and fixed costs relative to AMD and Intel; we can have opinions, but lacking a complete overview of relevant facts it's rather hard make the kind of definitive conclusions you made, especially since you forgot to consider the above points in your post as well.
Regards,
Rink
wbmw, re: Turion == Athlon 64. No magic, no brand new design project or exclusive technology, just shared items with the exception of things such as deeper sleep states, which are fused into mobile parts only.
Close. Just to add to the new deeper sleep states: Turion 64 does use a different type of transistors too (i.e. previously unavailable lower power transistors that will not be used to the same extent in high performance K8 cpu's) as must be reasonably evident from those slides. It's hence a different cpu (die is different; functions are the same).
Regards,
Rink
wbmw, good point! Depending on the time you're either right or wrong about that.
Regards,
Rink
LOL! (eom)
Must be good vapor. Still Turion will rather much sooner (read now) become real than Montecito, and what's left from Turion vapors smells like Double Delight (if you know what that is)...
Regards,
Rink
DRBES/Keith, the link is up again (or rather a new article is with the same title). Nothing really new. The only way Intel can gain revenue from desktop strategy is from bundling their own chipsets with the cpu, no?:
Intel but not as chipzilla
Channel Brand new brands coming
By INQUIRER staff: Friday 11 March 2005, 15:54
INTEL’S CHIEF MARKETEER Eric Kim told Reuters that he plans to reshape the Intel brand name and create sub-brands for Intel chips for home, business and other markets.
It is working on a digital home chip brand for a Q3 launch this year.
Citing digital convergence as his motivator, Kim, who joined last year from Samsung, yesterday picked Interpublic Group of Cos. Inc to lead big changes in its attention grabbing. But don’t fear because Intel will continue to spend many hundreds of million of dollars on marketing.
Kim is quoted as saying that the slow moving PC market is not sufficient to drive chipzilla anymore so it will market itself around platforms such as Centrino.
(edited) Keith, tx for the info.
I remain mid term rather positive because of dual core Opteron, Turion, 90nm E-stepping, and nVidia SLI. All of these have enabled more design wins, and though 90nm volume is late, both dual core and Turion are on time.
Flash might not be good but the upside potential of cpu's does make up for it quite nicely imo. We're lacking solid information on how AMD can overcome mirrorbit scaling trouble, and higher speed NOR issues. Though I don't like that I still don't think it's going to put AMD in the red this quarter. Next quarter there's much more upside in cpu's, the quarter after that even more. So overall I think the story is quite positive.
I'm cautiously long term positive too, but for other reasons.
Regards,
Rink
Sanford C. Bernstein analyst Toni Sacconaghi apparently needs negative attention. It's completely inconceivable that Dell takes on Sparc/Niagara/TheRock development, Opteron systems, Solaris development, going cold turkey because of lack of Intel dope, etc... Extremely bad fit for their business model. SUN OTOH is rather unlikely to throw away Sparc/Niagara/TheRock + Solaris. Either this guy is utterly nuts or he just wanted a rather efficient exit strategy.
Regards,
Rink
Joe, you're free to think what you want. It's a bit weird though to hear you say you think that I think Bush is to blame for everything from the birth of civilization till present. That's not even a tiny bit true as I'm not pro nor really against Bush (not even for Iraq although his justification for it is completely flawed). He is just an example to make my point. According to the list I recall seeing in one article MS executives paid MORE (not exclusively but just significantly more) for Bush' campaign than Kerry's campaign (never bet on one candidate but do help the one with the most favorable view for your business to win; simple, no?) because they were sure Bush would help them lessen the fine (it was rather widely publisized that Bush's election would be helpful for MS). Also the result, after Bush's election, was that MS subsequently wasn't fined in any meaningful way, (except ofcourse by the EU but that's another matter).
Anyways, The principle, i.e. (more) campaign money for a candidate that can get you favorable policy, is simple, and therefore it also works. There are many many examples of US corporations trying to influence politics via campaign funding. You are free to see it otherways; I don't mind; this is just my personal observation. MS not having being fined in the US in a meaningful way despite being found guilty should, I thought, have rung some bells for you.
Against this background I've always fully understood why AMD tried to get EU and Japan to rule on Intel's monopoly abuse before they would try the same in the US. There are more reasons why I think that is logical but this is one of them (they would not stand a chance in the US).
Regards,
Rink
Yep, demand limited ... by Intel. I think it's hugely shameful that countries around the world consent to this unlawful competition killing and undemocratic practices. The US first as businesses are allowed to financially support the campaigns of those who can bail them out (e.g. Bush and MS) which makes the US system effectively corrupt.
Kind regards,
Rink
Pacifica (virtualization technique) is absolutely not needed for dual core operation. So that digit-life quote is awkward at best:
"The new dual-core Athlon 64 solutions are to be shipping in H2 2005, however there are still some issues regarding the software. In particular, Pacifica technology called to provide compatibility with single-core CPU software is still in development and the deadline is not known yet."
This quote suggests Pacifica is essential for dual core Athlon 64 functionality. That's incorrect. Also Pacifica is not needed to provide compatibility with single-core CPU software. As long as all software running on the cpu is running on the same OS no virtualization technique is required. So this comment is awkward and misleading.
Pacifica is needed for most specifically for servers in case they need to run more than one OS on that server. As comparison: Intel's virtualization technology for x86 servers (Silvervale) isn't ready until next year. Intel does have client virtualization tech that is ready (Vanderpool).
Virtualization is seen as a key technology to lower TCO of servers. Intel does not have it working for x86 servers either. It's by far not as essential for clients.
Regards,
Rink
wbmw, it's not that difficult to understand...
Smallpops wrote: I don't think Intel had problems with 90nm. Their problems were with the P4.
I wrote: Dothan was delayed +- as much as Prescott.
wbmw wrote: What does this have to do with suggesting that Intel has 90nm power problems?
Intel obviously had trouble with 90nm besides P4 design because Dothan was delayed about as much as Prescott, because leakage was higher even in a low voltage design then they had anticipated. Intel would not have delayed if they hadn't faced an issue with Dothan. From what I heard it was exactly the same issue (leakage) that delayed both. From another angle: Intel would never have created Prescott if they'd known about higher than expected leakage beforehand. Íntel's 90nm problems were magnified for all to see with Prescott, but they applied to their other products too (same process with higher than expected leakage) but in a lesser manner because voltage and frequencies (and hence leakage) are less than with Prescott. Same problems, but in a less excessive manner.
FWIW, AMD had the same problems too but again in a lesser manner because of lower frequency, and SOI (that doesn't solve the main leakage problem but does give some more breathing room by reducing total leakage).
Hope this explains to you what I meant.
Regards,
Rink
Dothan was delayed +- as much as Prescott (eom)
Same power when running unplugged is achieved by stopping one core almost completely and adjusting the frequency of the other according to cpu demand. TDP for as far as I heard IS higher for Yonah than for Dothan.
Regards,
Rink
Alan, see here: http://www.investorshub.com/boards/read_msg.asp?message_id=5597900
Anandtech has more details.
Regards,
Rink
8K from Yahoo: Guidance still "flat to slightly down"
http://biz.yahoo.com/e/050302/amd8-k.html (link via ajb/SI).
Regards,
Rink
That's not exactly a telling comparison. 800MHz was for a complete new cpu design, on SOI, but not on a new node. Pressler is not a complete new cpu design, is implemented on a new node (65nm) with slightly changed process technology, it isn't the first cpu to make it to 65nm, and it's predecessor already ramped to 3.8GHz.
So what exactly makes you think the ramp of K8 is in any way indicative for the ramp of Pressler?
Regards,
Rink
wbmw, per Anandtech: Pressler was demoed at 2GHz. http://images.anandtech.com/reviews/tradeshows/IDF/2005/Spring/Day1/Multicore/pressler1.jpg
So Doug's right. Intel does have a long way to go. But it still has a good year to do just that too. 2GHz at this moment isn't bad.
Regards,
Rink
You're right. That's why I deleted my post before you posted yours in the hope noone had seen it. Anyways...
R
(deleted)
Interesting. Would be nice to see an update to this article though - I'm guessing it's from spring 2004 (eom)
AAMMDD, that was > 1y ago. Better luck next time I guess. (eom)
AAMMDD, is it normal to get the rebates? I bought my wireless router + pcmcia card (both DLINK) in the US and let a US friend of mine sort out the rebates, but never got any.
Regards,
Rink
fwiw, in The Netherlands it is pronounced 'neesh' too (eom)
wbmw, Paul, biggest damage is losing face with consumers.
Regards,
Rink
wbmw, yesterday Rivet mentioned ones Fab36 is loaded there would be a lot of interest in Fab30 and that they were looking at all possibilities to keep Fab30 utilized. From how I interpreted it he also downplayed Chartered a bit which I find not surprising now Dell said no. All in all I think AMD does acknowledge it is likely that at least not all of Fab30 will be used for AMD cpu's once Fab36 reaches full capacity.
Regards,
Rink
One difference in 65nm is Intel's gate length of 35nm and AMD's gate length of 39nm. Shorter gate length iirc provides more speed and more leakage. Lot of slides on keeping leakage in check with various evolutionary techniques, among which improved strain as expected. Impressive presentation for someone that can't place the details (me). Tx.
Regards,
Rink
wbmw, re: I attended their Microprocessor Forum presentation last Fall, and they said 2-3 frequency steps.
OK then they must have said both, probably at different times. I'm really pretty sure about this too you see. Maybe it changed from 3-5 to 2-3 because of the renewed strained silicon. Anyways, if we can trust that single mention of 2.4GHz, 2-3 frequency steps down from max looks closer to the reality than 3-5 steps down.
I wrote: For as far as I know Intel hasn't solved increasing leakage at 65nm not with any new process technique that is which means there are continued limits to scaling (both the speed of scaling and the amount of scaling).
You replied: You can't "solve" leakage. Leakage goes up exponentially as transistor features shrink. The best you can do is improve it to go up linearly, rather than exponentially, as far as I understand it, and that's hard enough as it is.
OK, I'm Dutch and I used the wrong word (well duh!). My basic point, namely that Intel is not likely to use any new major technique to reduce leakage when comparing new 65nm with current 90nm process (a technique like SOI) remains.
Regards,
Rink
(deleted, redundant)
wbmw, I agree with most you say except:
- AMD did say 3-5 frequency steps below top binning K8. That was before they had improved strained silicon. Plus 3 steps down was for same power, whereas that Toledo uses more power.
- I'm not 100% sure about that 2.4GHz will be there immediately (even though that's what the article clearly implied). If it's true however we're rather likely to see a 2.8GHz at the same timeframe though. 2.4GHz might be more likely if Toledo is introduced end Q3 / start Q4, and that might actually be the case because Toledo was scheduled for H2 05 whereas the dual core Opterons were scheduled for summer 05.
Lastly, if Yonah does not have 64b it is not exactly an all encompasssing ultra-competitive product except in it's own market (multi-treaded notebooks, and quite possibly highly dense blades), and even though it's a powerful product you understand I think that is hyped too. Netburst continues at 65nm and at least until end 2006 for a good reason, i.e. high end performance (if Yonah would deliver that there would be a heck of a lot less emphasis on continuing a multitude of netburst cores on 65nm). For as far as I know Intel hasn't solved increasing leakage at 65nm not with any new process technique that is which means there are continued limits to scaling (both the speed of scaling and the amount of scaling).
Anyways, have you seen blades adds on HPs main page (one of three featured items so refresh until you see it). HP is consistently not underemphasing Opteron systems compared to Xeon systems.
Regards,
Rink
Tecate, re: Perhaps Intel has a lot of of things not announced - Dell stayed with Intel because they found Intel's roadmap compelling enough to not want to carry two lines.
First answer is Dell stayed with Intel because they always do.
Another answer could be that Dell is privy to 65nm performance at the very start of 2006 and decided to weather the time in between. There are a lot of possible answers but imo it's extremely unlikely that any of them includes Intel releasing unannounced cpu's.
There's always the not that unlikely marketing $$ story. Whatever is the real answer Dell going with AMD was not a wide spread expectation anyway.
re: Well you just as easily you won't see AMD blow away Intel
AMD dual core blows away anything Intel can offer in the x86 space for at least until start 06 but rather likely well into 06. Also AMD continues scaling 90nm to higher frequencies whereas Intel is stuck at 3.8GHz at least until end 06. Ofcourse AMD won't blow away Intel financially if that's what you mean, nor increase market share in desktops dramatically (meaningful market share increases in notebooks, workstations and servers are rather likely though). You're right that no Dell has a very significant influence in corporate acceptation, and that's a shame.
Regards,
Rink
Tecate, AMD still has a certain hyper transport advantage. Dual core advantage especially in server isn't that easy for Intel to tackle in x86 space as you think. Also dual core only really matters in workstations and servers, not notebooks and desktops. That might change yes, but only over the long term. 2.4GHz for dual core A64 is higher than almost any sane person expected after AMD's remarks half a year ago that dual core would be 3-5 frequency steps below top end (so I hope this number is correct, it might be if AMD compared at same power, if it is correct it is really sweet for AMD investors). Also it currently looks like AMD will 'only' be half a year late with 65nm which is not bad at all. Besides rather good looking dual core AMD also has Turion, Horus, Galaxy, coming. This isn't momentum that Intel just can blow away with 65nm provided AMD/IBM will remain on schedule.
Regards,
Rink
Duke, OneNAND is NAND with a NOR interface, while ORNAND is Mirrorbit with a NAND interface. NAND is easier to scale than NOR, and hence OneNAND is easier to scale than ORNAND. Mirrorbit does have a die size advantage but that is mostly countered by the fact that it's more difficult to scale. It's not that likely that ORNAND replaces NAND for high end. Nor is it sure that OneNAND will take a big bite out of the NOR market. Mirrorbit does imo have the potential to change the equation in two ways: It might give AMD a higher share of the NOR market eventually, and the lower Mirrorbit costs might be a factor that causes the decline of NOR market to go slower than CSFB thinks. But most people don't think this will happen because AMD didn't scale Mirrorbit that well. Lastly NOR and Mirrorbit retain their data better than NAND and are faster when accessed ad hoc (executing a program is better when from NOR than from NAND memory), while data is cheaper kept in NAND. CSFB bases it's claim of sharp NOR descend on new mobile phone design trends and extrapolates that to the future. This might well be reasonable, but not exactly fool prove. K, this should do for now.
Regards,
Rink
DRBES, this goes too far: You don't need to resort to writing articles to make your point!
Regards,
Rink
PS, well, actually the tone of how HP apparently did the announcement was not as clear to me before, so this article does make some difference (obviously I'm not yet entirely convinced yet HP is more pro AMD than pro Intel but they seem to be en route to get there which is good).
Tx, also for the virtualization training on SI. H1 is obviously better for BL45p (my Q3 comment was only a guess from memory). For as far as HP Proliant Server Blades is concerned Opteron will be just as well present as Xeon, which I think is pretty cool.
DL585 upgrade sounds pragmatic: 1GHz HT will really help in 4 socket systems, especially when using dual core. As will DDR400.
Regards,
Rink
Keith, "HP also looks set to update its popular four-way DL585 with new processors and will release a higher-end Opteron-based blade called the BL45p later this year."
http://www.theregister.co.uk/2005/01/24/hp_opteron_france/
Rereading the above comment I think it's really one new product: BL45p. The DL585 might well be updated with newer dual core Opterons, but that update might equally well also include a new mobo with newer chipset. We'll see.
Regards,
Rink
DRBES, everything is relative to ones point of view. Their feature story from yesterday was quite positive. Plus we're going to see at least two more Opteron products from HP (Q3?). The increasing percentage of Opteron lines with HP is the only argument supporting your notion that HP is pro AMD. A counter argument is that HP continues to move their high end customer base over to Intel's Itanium, and that it doesn't cancel any Xeon line (unlike SUN). Both these trends (increase both Itanium and Opteron systems) might actually provide for the optimum profitable situation for HP. Because I think that's the case I wasn't surprised at all that they're not porting their old flavors of UNIX over to x86. Anyway, all in all I can only agree with your statement that chipnostic means pro AMD if it also means pro Intel.
Regards,
Rink