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.
The X2's are just the hottest high end mainstream cpu's. They bring together the best of two worlds: Best single thread performance of dual core (close to that of single core) and best multi thread performance (generally speaking). They are the future. Everyone wants a piece of that. Just not everyone can afford it. They're probably in short supply (because the prices are stable; EDIT: just saw BUGGI confirming very low supply into the channel).
Regards,
Rink
SGI tries to gain a foothold in the 1-2 processor market with their new entry level Itanium server: http://www.itjungle.com/breaking/bn071105-story01.html
Regards,
Rink
Chipguy, my view of Cell design influencing long term chip design has nothing to do with being neutral enough in discussing Itanium's future. What you're doing now amounts to imaginary mud slinging. I never said I was sure about my own view concerning the cell design, but instead always used words like 'might', 'could', 'if', 'I think', etc... (I can't be sure because I'm not an expert and about everyone knows that including you, and even if I was I could be wrong too). Though I'm hardly sure I still continue to think the cell design will influence long term x86 processor design (I won't mention the arguments because it's another discussion). Also I never hesitate a second to admit that I'm wrong if it seems even only a bit more likely to me that I'm wrong than not; I do this because I'm interested in getting my views balanced as much as possible and because I know I'm not a chip expert. When you would know me you might just find an integer person.
Your accusations about Ashley's integrity are imo out of proportion because you're rather unlikely to know him as a person and partly because of this it's unlikely to know the reasons behind his personal views. Because of this you might want to be a bit more reserved in accusing him. Everyone deserves some respect until one knows him enough to know otherwise. The fact that Ashley is at the opposite end (pro-Itanium vs. contra-Itanium) of the opinion scale compared to you manouvers you in a less neutral position as to judging whether he is 'bashing' or not.
My question hence was if you consider yourself neutral enough to judge him, and you did not answer it. Nor did you answer the questions I asked about things that if I recall correctly you have been wrong about with regards to Itanium. Instead you start throwing imaginary mud on another subject, using emotional language once again.
I have shown you respect, and will simply continue doing that for now, even though I disagree with you accusing people (including myself as of your last post) because I find that cheap and respectless.
Regards,
Rink
Chipguy, I'll let your lack of answers to my questions speak for itself.
Regards,
Rink
Chipguy, re: I predicted that IPF would continue to grow sales and enjoy OEM and ISV support.
Yes you did and you were right, but you were wrong a lot of times as well. So what? I asked you if you are neutral enough to judge Ashlee. You chose not to answer that question directly. Your answer instead included heart-felt emotion. Why?
re: I know the size of the server market. And I also know that ...
Yeah you know now but I asked the question in past tense. You answered at least partly in present tense. Now, you didn't know at the time that Itanium wouldn't take a single digit cpu sales away from mid to high end Xeon market, did you? In fact you thought the opposite, namely that it would take part of that market away from x86 Xeons, for the reason that x86 wouldn't gain 64b capabilities any time this soon, isn't that right? Let's not get into the exact IDC estimate but it was based on the exact same reasoning, right?
Just to be complete I think you know that I find some of your posts highly informative. I have for a couple of years now hoped for even more balance though.
Regards,
Rink
Are you sure you're neutral enough to judge him? Bashing Ashlee is all too easy from your point of view. So easy in fact that I wonder if you had a good look in the mirror last couple of years. Didn't both Itanium 1 and 2 miss your both your performance and release estimates? And wasn't it you that expected Itanium to pick up some mid to high end Xeon business as well, i.e. beyond the HP RISC replacement niche that's exploited by SGI as well, e.g. because there would be no viable 64b x86 alternative? At least Ashlee is stating a fact. You appear to be bashing him just because you don't like hearing that fact again once each quarter. What's more from your point of view it can hardly be but conceivable that at the time you believed that very same IDC estimate.
Regards,
Rink
Dacaw, if you want to have some fun you might consider joining this class action against?: Intel: http://www.theinquirer.net/?article=24482
Regards,
Rink
Dan, UpAndDown, Spansion does not produce RAM chips, and hence the fab capacity they ordered from TSMC won't be RAM either. It won't be chipset chips either because Spansion does not produce any chipset chips. TSMC will produce flash for Spansion. The only question is which kind. Here's my reasoning as to why I think TSMC will produce older floating gate flash: http://www.siliconinvestor.com/readmsg.aspx?msgid=21477390
Regards,
Rink
Maybe you forgot to post a link that Tad recommended to buy AMD at $24.
Regards,
Rnk
Klaus, re: In performance terms, probably yes.
Why?
Regards,
Rink
Buggi, re: I'm really not complaining ...
LOL! Me neither.
Regards,
Rink
Buggi, tx, Intel though is adding a second core with 65nm for mobile to the Pentium M line. AMD will not be adding much at all to their DC's but just shrink it pretty straight forward. K8 power consumption went down a lot with the move from 130nm to 90nm (and later again with rev E mainly because of SGOI and I think also 11 metal layers). I am pretty much convinced that power consumption will be reduced with any staight shrink when going from 90nm to 65nm (both for AMD and Intel).
As for Intel: Because of below two reasons they are able to have Yonah (at 31W iirc) at all:
1. power savings going from 90nm to 65nm (for straight shrinks), and
2. the fact that they'll intro significantly more potent strain (compared to their current version) with 65nm. One graphic I saw showed double benefit compared to the benefit provided by their current version of strain. Strain can be used both for better performing, or lower power transistors.
However once Intel has used this advantage they'll have rather limited further headroom on 65nm, meaning Merom has to use more power than Yonah. AMD has a 90nm DC now that consumes at desktop levels, and might use the shrink + lower power version to get to the same point. It all adds up: I really think it might prove to be that simple, especially after reading your post.
Regards,
Rink
Keith/Buggi/CJ, do you think I'm far off thinking a shrink of X2 is all it might take to get AMD a worthy competitor for Merom? Like between a quarter and half a year after 65nm introduction?
Reasoning being more or less that AMD will intro lower power dual core Opterons beginning of next year (iirc). And the shrink to 65nm (presuming no new stuff will be added) will result in further power reduction. This compares to Merom that might well use a not insignificant amount of more power than Yonah as it adds 64b integer units, more and enlarged registers, changed scheduler, etc...
For now I'm guessing AMD will start selling 65nm products in Q3 06 and could possibly start selling a Merom competitor around the end of 06. So AMD might not be too far behind at all with a Merom competitor provided the above reasoning isn't flawed. For comparison: Merom is thought to appear around Q3 06.
Regards,
Rink
Cell phone market moving to 2b/cell flash:
http://www.siliconinvestor.com/readmsg.aspx?msgid=21459368
Regards,
Rink
Not worth commenting further (eom)
re: With the new FX 57 at $1037 and the old FX 55 still stuck at $827 where is has been for over 6 months wouldn’t you think that AMD’s manufacturing cost for these products is simply to high? It appears that AMD is announcing chips it can not make in numbers just to create an illusion of having superior technology.
OK, let me be a bit more precise:
1. Manufacturing costs are not high for FX (especially considering the extreme margin) as it uses the same San Diego core as the mainstream high end A64's. See AMD's roadmap if you don't want to believe me. San Diego's die size is comparable with that of a normal higher end desktop cpu, and it is manufactured by AMD in high volume. Consequently costs are not outrageous by any measure.
2. My previous post pretty proved AMD has in fact superior technology in this niche market (high end pc gaming). I'd say it's even beyond possible to prove the contrary. Hence it's not an illusion. It's a fact. Ofcourse the superiority of the A64FX series enhances consumer perception of the whole of the A64 line, especially for those consumers that like gaming. Perception in this case is based on facts, not illusion.
3. It should be clear that because of complete lack of competition in this niche market from Intel they can price the new FX upwards.
4. AMD not making it's numbers is rather clearly due to flash, not cpu's - i.e. completely contradictory to your comments above.
Based on the above what exactly is not at least naive in your comments?
re: A naive comment? Just curious what do you think AMD’s asp was last quarter?
AMD's ASP for all it's x86 cpu's has rather little to do with your one but last post on this subject (see first quote above) as those comments related specifically to the A64FX series, and not to A64, and definitely not to all AMD cpu's in general (which include Sempron, Opteron, and even Geode too).
Regards,
Rink
Herb, that's one naive comment. FYI here's a qoute from the conclusion of the HardOCPs FX57 review that just might change your stance:
Why did we not take the time to compare an Intel CPU in our comparison? The simple fact of the matter is that while Intel would have you believe that they do make "gaming" processors, the AMD Athlon FX-57 simply kicks the crap out of anything Intel makes in the realm of gaming. We also figured that anyone that considers themselves a hardcore gamer in this market is very likely already using an Athlon 64 CPU and that is the reason we supplied you with the 3200+ scores. We were hoping to give you a good basis of what sort of power an upgrade from that class of CPU might give you. All of this is not to say that Intel does not make Pentium 4 and Pentium Extreme Edition processors that are capable of gaming. Most new processors today have more than enough power for the casual gamer, but in the super high end gaming realm, Intel does not currently compete.
http://www.hardocp.com/article.html?art=Nzg3LDU=
Regards,
Rink
Haven't seen them yet at dixons in NL yet. Dixons does carry Gateway from time to time. Previous time that eMachines AMD laptops were in fashion they were only sold in the US but at that time eMachines had not been acquired yet by Gateway (over a year ago). What I tried to say is that eMachines laptops just might become available in those outlets that sell other Gateway systems.
FWIW, some of the eMachine systems are available at Dixons and PC World in the UK.
Regards,
Rink
I2 volume is low. Maybe they disable L3 and reuse the same die. That would make the transistor count the same. Especially initial products are more likely to come from one die. Just my 2c.
Regards,
Rink
Paul, All of them? (eom)
Chipguy, I ofcourse meant 1.8GHz max base frequency. Sorry for apparently not being specific enough for you.
Anyways, I would still highly appreciate it if you could answer my polite question in my previous post (maybe you don't know?, you at least sounded like you would know...)
Regards,
Rink
Dixons and PC World carry some of Gateway's lines in some parts of Europe (I'm dutch).
Regards,
Rink
Chipguy, do you know if the 1.6GHz Montecito that was demoed by HP/Intel was using 12/18/24MB L3?
BTW, at the ISC at Heidelberg in Germany they are not expecting any Montecito higher than 1.8GHz at the beginning, and Intel does not commit to anything regarding clockspeeds apparently (see Heise.de article below).
BTW2, not that important but I was ofcourse expecting a range of frequencies which is why I asked if you had expected "1.8GHz maximum". I thought you capable of concluding I might have meant that there would be chips clocked with lower frequency than maximum as well.
Anyways...
Regards,
Rink
Ref: http://www.heise.de/english/newsticker/news/61008
Chipguy, didn't you once say you expected Montecito at higher than 1.8GHz maximum?, or is this what you expected?
Regards,
Rink
Quad Montecito HP system demoed running linpack at 46GFLOPS at 1.6GHz: http://www.heise.de/english/newsticker/news/61008
For comparison current quad dual core Opteron systems run about 30GFLOPS according to the same article.
I thought Montecito could be demoed by now at higher clock speeds. The article says 1.8GHz is expected at the introduction date but Intel doesn't commit to anything. I think even 1.8GHz is lower than some people were expecting if memory serves me well.
Regards,
Rink
Dan, keep yourself in check pls (eom)
Joe, Doug, I'm somewhere in between you and Petz I guess: While I agree with your interpretation I also keep thinking that while it might not mean much Charlie did not retract his INQ story about a possible quad core in Q1 06, nor did he say anywhere on the internet that I know off that that story was bogus. At the least it means he has since not received any good info to the contrary. I don't know, Charlie might easily be wrong as the official AMD reading was more like that quad core is more of a 2007 thing.
Maybe a couple of points from that analyst presentation:
#68: Coming soon: Multi-core + scalable SMP architecture, HT v3.0, DDR3 + FBDIMM, etc...
This means e.g. that possibly all of the HT 2.0 speeds ranging from 1.0GHz to 1.4GHz are already available now but are not turned on e.g. because of power / performance equation currently is the reason that 1.0GHz apparently is OK for now. It also means that Q1 06 parts might possibly include HT3.0 (ofcourse the time frame here is not exactly a given). Multi-core refers to quad core imo (this will however not necessarily be used at the start as dual core was included in the first K8's but only got implemented with second generation 90nm). This multi-core mention supports Groo's stance to some extent though. Scalable SMP architecture might include improvements to how the cores work together (actually I don't know enough about SMP to really make even this comment; sigh...).
That same slide #68 contains information about longer term future: FPU extentions to AMD64, Throughput architecture (related to cell architecture stuff imo), On-chip co-processors (e.g. for FPU), HT4.0, DDR4 + FBD2, etc...
Regards,
Rink
Joe, I agree that two mem channels is probably going to be the maximum for a while, and while I also agree that a 4th HT channel might help in a 4 socket system I don't think that will happen in Q1 06 - I think AMD will relie on gradually improving HT speeds and bandwidths instead (1.2GHz by Q1 06 is e.g. not out of the question for e.g. socket F parts).
Regards,
Rink
Re: do you happen to have sales numbers of DC Opteron and X2 at this point?
They are according to sourcing talking with The Register (yeah I know the WSJ is a better source but at least this is a small indication) are "exceeding expectations by leaps and bounds": http://www.siliconinvestor.com/readmsg.aspx?msgid=21445278
Regards,
Rink
Re: Rink has been talking about a fake quad [two low power dual core Opteron dies in one package].
I'm not anymore. As I'm inclined to believe Groo when he mentioned this that Opteron xbar has already been updated to handle quad core: http://www.siliconinvestor.com/readmsg.aspx?msgid=21445168
Regards,
Rink
Kpf, though you may be right, Doug is right too in saying that the official public 65nm schedule is H1 06, and 90nm from fab 36 early 06. So you know you might be equally well wrong here.
Regards,
Rink
re: What I wonder is how quickly the remaining 50% will fall. I think rather quickly.
I think the demise of big tin systems will be more slowly than you think unless:
1. HP drastically changes their habit of supporting Itanium, something I don't expect soon.
2. Intel somehow joins EPIC and x86. X86 will need to get a heck of a lot additional registers though to be able to emulate I2 reasonably efficient. This is not something I expect soon either.
Both above options are not impossible however.
Some reasons for slower pace:
- Cray is having less of a field day than SGI.
- Power is still evolving at good pace (and Cell contains interesting concepts some of which IBM will leverage for Power too - my prediction only ofcourse)
- Montecito while not exactly exciting looks rather OK. Even though I don't think it'll ever take over not even the x86 high end Xeon area like Chipguy was thinking it's good enough to continue forward.
- HP has given Itanium a lot of rather good credit (they are making this work like they did previously with their own chips, and until they can't do that anymore they will continue leveraging it as the only significant alternative to Power).
- Momentum of the Itanic and equally gigantic Powertanker can't be changed but gradually, and even if currently at snailspace will take years to stop. It's not even that impossible that both continue to exceed for decades in some form or another.
Regards,
Rink
IIRC Cray also had problems a while back getting the communication chips that are manufactured by IBM. The comm chips are what allows Cray to put the Opteron nodes in a 3D torso.
Regards,
Rink
fpg, maybe you're interested in this thread as well: http://www.siliconinvestor.com/readmsg.aspx?msgid=21435860
Regards,
Rink
You're right, only a small decline for IPF.
It seems though that lots of Xeon and Itanium systems are marked to be erased from the next list (numbers 400-500 contain only one Opteron system). A good percentage of Opteron systems on the list will get a boost from dual core upgrade, and more are bound to get on the list. I think dual core Opteron will get us a couple more systems on that next list too. Still quite amazing how much more Xeon systems are currently on that list.
Regards,
Rink
Re: Typical Dan, sees things 180 degrees from reality.
Yep!, and the growth of Itanium in the latest top 500 supercomputer list from 84 in the last list to 49 now is 180 degrees from reality too.
Regards,
Rink
(Numbers are graciously gleaned from Smallpops earlier post on this subject)
Mmoy, retorical question? In case it's not here's quoting from Mike's post (look at the last column of his table). I've deleted a couple of columns to reduce the information available.
P4 820 (2.8G DC 2x1MB) - 268.30
P4 820 (2.8G DC 2x1MB) Tray 244.26
P4 830 (3.0G DC 2x1MB) - 347.13
P4 830 (3.0G DC 2x1MB) Tray 320.46
P4 840 (3.2G DC 2x1MB) - 571.77
P4 840 (3.2G DC 2x1MB) Tray 537.58
Turion notebook with HP Biometric Fingerprint Sensor. Cute!
Re to your other post: Nice notebook comparison as well. Higher MHz or lower power Turion's should provide for better comparisons still.
Regards,
Rink
Joe, I think SIMD core shared by several K8 cores is what Fred Weber had in mind with this quote from him:
Fred went on to say that for future microprocessors he’s not sure if the K8 core necessarily disappears and that in the long run it could be that future microprocessors feature one or more K8 cores complemented by other cores. ...
However, Weber did mention that there's interest in sharing parts of multiple cores, such as two cores sharing a FPU to improve efficiency and reduce design complexity.
http://www.siliconinvestor.com/readmsg.aspx?msgid=21184935
They might e.g. start with one shared SIMD core, and later expand that to more. AMD's crossbar architecture should lend itself with slight adjustment to this. The design of the currently hypothetical shared SIMD core is a much bigger question. It's conceivable that they expand their relationship with Cray for this purpose.
Regards,
Rink
OK. That clarifies your conclusions. I can explain why end 2006 might be a good potential announcement date but think we've talked more than enough about this.
Though you might indeed lack vision you make well up for it by being soundly realistic for which I obviously will continue to appreciate you .
Regards,
Rink