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.
Come on Keith, I never said nor intended to say that this was one of the most relevant drivers for the stock price over the coming two years. That's so stupid I can't even come up with a reason why you implied that I did. Instead I implied it might well make some, not insignificant, difference. Ofcourse server/mobile/desktop market share growth and preceeding design wins are far more important to name something. But that's all trivial, while what I tried to put into some context though less relevant (not irrelevant) for the stock the coming two years is way less trivial. It might not interest you but that hardly automatically means that it is irrelevant.
Regards,
Rink
Keith, re: and then what?
As you know: Difficult to tell. Currently I just have a feeling that AMD might be ahead in the process of setting up a cell-structure architecture when compared to Intel (AMD talked up the idea in the press, and Intel is currently busy working quite hard to close the architecture gap with AMD). This lead in cell-structure architecture might turn out to be true, and might turn out to be possitive (like being first to hypertransport, 64b, NX, etc... was and is possitive too). You're right that most of the potential benefits for the stock are further on the horizon than the announcement which way to go for cell. Announcing it before Intel would further enhance their image as leader. Continuation and expanding this image is important for the stock appreciation mid term. My wag is the announcement will happen end 2006.
Regards,
Rink
Though passion for technology is certainly a reason for me for investing in tech stocks I just happen to think that long term strategy bears on medium term stock price. To give you an idea, I've got Jan 2007 LEAPS as most relevant investment in AMD. I think that the currently potential AMD cell-structure has a high chance of being realized, and a reasonable chance of being announced before 2007.
Regards,
Rink
Keith, first I'll try to put some words to the four possibilities I mentioned for AMD to introduce Cell-alike structures into their cpu's, and at the same time mention some relationship to strategic advantages for the companies involved which add shareholder value.
1. Regarding Cell from IBM/Sony. This is a weak link. AMD has a process development relation with IBM. They might extent the relationship to licensing some cell IP. This might eliminate a lot of development from AMD's point of view (similar manufacturing process) and expand IBM's ecosystem. I don't think this option is the most likely as IBM might think they jeopardize their Power infrastructure.
2. Regarding cell from SUN (Niagara/The Rock). This is a stronger link. SUN is leaning quite heavy on Opteron because SPARC is lagging for a while already. If Niagara/The Rock fail they might have to go Opteron all the way. AMD/SUN might end up joining the best pieces of each others cpu architectures. SUN and AMD already have a partnership (together with Broadcom/Serverworks). SUN has shown it's willing to deal to expand their Solaris footprint. AMD has shown it's willing to work with SUN to gain access to their entire server portfolio. (AMD has also taken over some SUN engineers that previously came from API / Compaq / DEC. My speculation on that is that they might expand hypertransport to include proximity connectivity). All in all it's currently a strong and expanding partnership.
3. Regarding Patriot Scientific. AMD's interest in IGNITE is also interesting. IGNITE is a Risc-like 32-bit processor core with a strong SIMD component. According to Patriot, multiple IGNITEs can work together within the company's InFlame architecture to create a multi-SIMD processor not so dissimilar to the architecture Sony and IBM have been discussing of late as 'Cell' - the CPU technology that will power the PlayStation 3. http://www.channelregister.co.uk/2005/02/24/amd_patriot_investment/ It's not inconceivable that AMD will buy Patriot Scientific if this is what they'll use if they choose to go down the cell-alike route. Please see this in relation to my interpretation of Ruiz comments around complementing the x86 architecture via acquiring. Patriot IP and design can be used by AMD for a IGNITE-derived cell structure, or alternatively it's IP can be used in conjunction with another cell-architecture.
4. Lastly I think AMD is more than capable of designing their own cell, re-using and building upon some of their own IP and designs.
I think it's obvious that AMD has sufficient routes to make an interesting cell-alike processor happen. This is important as IBM/Sony/MS, and SUN have already done that, and Intel might do it too in the future although they have not mentioned it ever that I know of. Intel's choice however is inevitably one of the most important parameters for AMD to decide which route to pursue (because of compatibility benefits).
As for your comment that I included about every cpu developer in the world I can only say that long term projects that AMD might not have even decided upon itself include more than one option, and they'd be foolish not to explore all available to them. Obviously I'm intrigued to try to determine what the most likely resulting puzzle looks like and what pieces are going to be used to build and finish the puzzle in order to see what pro's and con's there are for my investment in AMD. Essentially I'd like to have a more or less idea the option that will be chosen before it is officially announced, which may take quite a bit of time still. For now I think AMD will not go IBM's cell route (1), and not use Patriot's design as is (possibly least strategic benefit - option 3). I think biggest chances are extent partnership with SUN (2) or go it alone (4).
Hope this is at least to some extent a relevant answer to your question.
Regards,
Rink
Keith, ok, you read the statement 'complementing our x86 architecture' as meaning: 'complementing our x86 platform architecture', while I'm reading it more like: 'complementing our x86 ISA'. Hard to tell.
Regards,
Rink
CNET: Are acquisitions something you're looking at?
RUIZ: I do think there's an opportunity for us to complement our x86 architecture. The possibility of either acquiring or doing things like that with some companies is something we'll consider.
Keith, yes: http://news.com.com/Why+100+computers+are+on+the+way/2008-1036_3-5684006.html?tag=st.num
See page two, last paragraph.
Regards,
Rink
Keith, regarding that statement, I'm trying to put this piece of the puzzle in the right place, so I'd be interested to know how you interpret that statement.
Regards,
Rink
Keith, I've previously mentioned four possibilities for cell structures in AMD cpu's:
Cell-like features might include either licensing some of Niagara/The Rock, or IBM/Sony cell, or using some of the IGNITE IP licensed from Patriot Scientific already, or develop a x86-specific cell structure from ground up. (IGNITE: http://www.siliconinvestor.com/readmsg.aspx?msgid=21198861 ). From here: http://www.siliconinvestor.com/readmsg.aspx?msgid=21316562&srchtxt=ignite
Please also see this in relation to this piece of my previous post on ihub:
Ruiz answering a question about possibility of acquisitions:
I do think there's an opportunity for us to complement our x86 architecture. The possibility of either acquiring or doing things like that with some companies is something we'll consider. http://www.siliconinvestor.com/readmsg.aspx?msgid=21269140&srchtxt=cell%20ruiz%20niagara
Regards,
Rink
Re: Do you have a link for [quad core vs. multi core]
Fred Weber: 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. ..., and this quote: 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
IDGNS: What lies beyond dual-core for AMD? Is there any limitation to the current architecture that would prevent you from putting more than four cores on a chip die?
Ruiz: It's hard to tell right now beyond four cores. The probability of having a four-core product is very high. There's a lot of work going on with our engineering teams and with our customers to try to figure out where do we go beyond that. There are two or three options that look pretty attractive. We'll be narrowing down our choices, working with customers to [see] what's beyond the four-core piece.
IDGNS: It's interesting that you didn't say that four-core is a certainty. Are you looking at different ways of improving performance, other than simply doubling the number of cores on a chip? What would prevent you from going to four cores?
Ruiz: At the end of the day, for us, it's going to be what our customers want from us. Making transistors is pretty trivial. We can make hundreds of millions of transistors. Figuring out what the hell to do with those transistors is the challenge. One could chose, for example, to have heterogeneous cores. You could have two cores that are different instead of the same. That opens up a completely different array of possibilities. http://www.infoworld.com/article/05/04/25/HNspansion_1.html?source=rss&url=http://www.infoworld....
Ruiz answering a question about possibility of acquisitions: I do think there's an opportunity for us to complement our x86 architecture. The possibility of either acquiring or doing things like that with some companies is something we'll consider. http://www.siliconinvestor.com/readmsg.aspx?msgid=21269140&srchtxt=cell%20ruiz%20niagara
Regards,
Rink
PS, not directly related but still interesting:
SUN's Proximity communication is likely to be used outside of Sun as well. "It's not a panacea, but it certainly helps. I think you'll hear more about it real soon," Krewell said.
http://www.siliconinvestor.com/readmsg.aspx?msgid=21288107&srchtxt=niagara
Re: what does this move say about Cell wrt. PCs?
Pravin suggested Cell with Opteron core (on SI thread)...
I did that on SI too 10 times or so during the last 18 months, although I called it Opteron with Cell elements, either derived from Niagara/The Rock or Sony's Cell. It looks though AMD will go quad core before implementing a cell-alike processor if the Inq is to be believed today.
EDIT: Chipguy, Keith, before saying this is nonsense, Ruiz mentioned a couple of months back that they were not sure at that point whether to go quad core first, or implement different cores on one die first (cell-alike). It's not exactly out of the question that AMD will still go that route. When is even a bigger question.
Re: And, when will IBM learn who their partners, and who their competitors are?
Good questions both of them. Apparently Cell won't be performing well enough for Apple (no low power version?, abysmal single thread performance) which might well be the main reason behind Apple's move towards Intel. IBM has pride, a weakness a lot of large companies are prone too. Long term strategy though requires some pride, although it also blinds sometimes where it counts. I don't think this is a bright moment for IBM, although with Sony / MS / Nintendo gaming + Cell for IBM workstations and servers must amount to rather good business for IBM. It's their choice and they might currently hate Job's guts, as it sure is a dent out of their recently expanding ecosystem.
Also this looks to be another indicator that Merom/Conroe/Woodcrest (in descending order of importance for Apple) apparently are quite good.
Regards,
Rink
(deleted; duplicate)
Chipguy, still think 2.6GHz at 90nm will be rather difficult.
Re: speed path tuneup
A guess: + 10% max?, Montecito has been tuned for quite a while already.
Re: and perhaps improvements in Foxton
Foxton does not improve base frequency, it improves max frequency, but we were not talking about that.
Re: the historic 130W family power limit
+15% max if the full 30W 'headroom' would be used? I'm not in the know whether Intel will use the 30W 'headroom', or if it'll even use more than that. I just think that's increasingly unlikely. I think they'd at least have felt quite a bit inclined to have used any power 'headroom' for Montecito already. Still 30W additionally should imho be considered. About Power: Power MCM's are a whole different animal to cool. I'd be interested to know what a Power 5 single chip, and a 4 processors MCM consumes, if you know those figures off hand that is.
Together, +25% max?, that would mean up to 2.5GHz which comes close to the 2.6GHz you mentioned, but with some reservations that it might end significantly below it too.
Regards,
Rink
Chipguy, I don't believe Montvale will get to the 2.6 GHz base frequency you mentioned, not on 90nm that is. They need some way to lower voltage, else the chip will fast try to get hotter than boiling water. I don't think Intel will have that technology in '06. New improved 65nm strained Si process would have helped reasonably significantly, and perhaps they'll use it at 90nm although I currently find that unlikely, but even that won't be nearly enough to overcome the increased heat caused by the increased frequency.
Just my 2c...
Regards,
Rink
Chipguy, so you're not disappointed that Montvale was dressed down and now will be made available at 90nm (instead of at 65nm as was the plan)? That's a rather weird point of view if true.
Anyways so far all your performance estimates or frequency estimates for different versions of the Itanium chips though interesting were too positive, at least all those I read. By contrast your K8 estimates for single chip performance were more accurate iirc. I can't believe though you're not disappointed about Montvale not making it now to 65nm.
Regards,
Rink
Buggi, Intel's Yonah may be delayed until 2006 http://www.theinquirer.net/?article=23677
So it starts to look like you were right.
Regards,
Rink
(deleted; wrong board; duh!)
Buggi, tx for the production comment. We'll see.
Re: 16Way - do this mean 8 sockets 16 way or 16 sockets 32 way?
Or 4 sockets 16 cores (8 sockets 32 cores) if Scott manages to delay Galaxy till 2007: http://www.siliconinvestor.com/readmsg.aspx?msgid=21316377
BTW, besides quad-core 'multi-core' can also mean two different cores (as in Cell and The Rock). The latter is in fact my (year old) guess. A dual conventional core + one or more cell-like cores per conventional core. Cell-like might include either licensing some of Niagara/The Rock, or IBM/Sony cell, or using some of the IGNITE IP licensed from Patriot Scientific, or develop a x86-specific cell structure from ground up. (IGNITE: http://www.siliconinvestor.com/readmsg.aspx?msgid=21198861 ).
FWIW, I think there's no chance that Sun goes beyond 8 sockets for first gen. Galaxy systems.
Regards,
Rink
Buggi, re: And Madisons can't be upgraded with Montecito processors.
Actually checking up on it again a bit because I wasn't entirely sure about that statement there seem to be two versions of Montecito (when I made that statement I thought there was only the one with the higher bandwidth that's mentioned quite frequently all over the web, and that I read was incompatible...). Apparently there might also be one version is socket compatible with current I2 and 8870 chipset, but with the slower 400MHz bus. That is in addition to the version I read more about and that will run with FSB @ 667MHz, and might provide a lot more bandwidth too.
From Paul DeMone: http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=3063&Thread=16&roomI....
From Intel: http://www.ewh.ieee.org/r5/denver/sscs/Presentations/2005.03.Montecito1.pdf
Maybe Chipguy is willing to fill in some more details...
Regards,
Rink
Buggi, IIRC Montecito is planned to be out towards the end of 05, and requires a new chipset (faster fsb with a lot more bandwidth). Also Intel stated that it'll perform between 1.5-2x better than 9MB Madison. And Madisons can't be upgraded with Montecito processors.
Regards,
Rink
Dacaw, tx for the info on SUN.
If Newton is heading NSG, what is Bechtolsheim's position? I thought he remained responsible for the Galaxy group and when he returned to SUN was reporting to McNealy?
Regards,
Rink
Kate, You are clearly out of your league, both for as far as knowledge and discussion skills are concerned. It doesn't help to deny that.
Just take a walk in the park before you try to post; it might take the edge of your posts and make them more balanced and factual at the same time.
Regards,
Rink
Kate, Intel typically but not always gets similar preferential treatment for putting a new fab in a certain state or country. That's normal; AMD is not unique.
Your remark that because of the German funding that AMD is "basically a German company" is wildly overstated and besides the truth, but maybe Kate the GREAT just likes to exaggerate.
Just curious, what exactly is it about you that you think is so GREAT?
Regards,
Rink
PS, you do realize that I'm teasing you a bit, right?!, but then I think you had it coming too.
Buggi, re: Flash
MB was only at 230nm until towards the end last year when the 110nm MB devices were introduced. There were major scaling problems, but they got resolved. I expect scaling problems to continue (which is why I said it would continue to have a scaling disadvantage) but not nearly in the same order as was the case until towards the end of last year. I hence expect that Ruiz is at least partially right in expecting much bigger MB growth this year.
I'd be surprised if MB isn't at 25% in Q4 rev. wise. In 2006 I expect it to grow further, and H2 06 I expect to see some significant contribution from ORNAND as well so that the combined rev might be closer to 40% end 06.
What are your estimates if I may ask?
Regards,
Rink
Buggi, you're right that the die sizes currently are only similar in size when produced at the same process node, and AMD is only at 110nm currently, while Samsung and Toshiba are at towards 70nm already, but then I read that AMD will move to 90nm in flash by years end. AMD isn't THAT far away from moving to 90nm to make a comparison of die sizes between 110nm and 70nm that relevant. I think comparing 90nm Mirrorbit with 70nm OneNAND is more relevant because that difference will remain available for a longer period. My idea also isn't that price proximity is reached within a year from now, but that it has to be met longer term. Over time quad bits per cell mirrorbit successor will overcome the process node (scaling) disadvantage. Well, maybe I was a bit too enthusiastic in how I formulated it... Thanks for the additions.
Regards,
Rink
Bobs, Re: Dragonfly.
FYI: http://www.theinquirer.net/?article=20871
Regards,
Rink
Buggi, Tx. That brightens up my picture. We'll see. (eom)
Buggi, re: 65nm
Yonah was said to become available end Q3, hence my comments. Was what you said related to Xeon's, or did it include Yonah too?
Regards,
Rink
Tx, DDB, that makes sense (that's in fact in line with what I thought but could not quantify).
Regards,
Rink
Doug (+ CJ), you might be wrong as the die photo's seem to show that the MC die space of the second core has been left more or less completely empty (empty spot occupying die space between the two cores). There might be die savings due to less space needed to attach copper wires on the die (the wires that connect the package with the die). Just guessing; I'm no expert here.
CJ, what's your take?
http://www.siliconinvestor.com/readmsg.aspx?msgid=21243369
Regards,
Rink
Maybe they interpreted "around same size as 130nm Opteron" as "around 190mm^2"?
Problems with this interpretation of their interpretation is that 130nm Opteron was just a tad bigger than 190mm^2, and that 90nm DC Opteron was projected previously to be a bit bigger than the 130nm Opteron. 190mm^2 would be nice though.
Regards,
Rink
Paul, you're reading it correctly, but from what I recall I think Stiller might be wrong in that current DC Opteron might internally have been refered to as K9, while the new cpu design scheduled for 2007 should be K10. By the way I was being a bit cynical in that Stiller can't know K10 (quad core) performance if he is not likely to know much at all about the design.
Regards,
Rink
Keith, IIRC: DC K8 = K9. 2007 is for K10. I agree with anyone that thinks it's a bit of a stretch to call a DC K8 a K9, but I think that's what AMD did.
Mighty impressive that Stiller knows K10 performance...
Regards,
Rink
Re: Fujitsu 32 way I2 + XT3 system sales announcements.
re: Fujitsu system being less advanced than expected:
I meant it's 32 way instead of 64 way that was mentioned in some PR's a couple of days ago (like the one I posted on this site). In addition to that I expected higher FSB as well, and reading RWT some other people were expected at least support for higher than 400MHz: http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=3304&Thread=3&entryI... Apparently that support is currently not publically known to be present.
Re: the two XT3 systems announced within two days:
I think that besides earnings related it's also DC Opteron related.
Regards,
Rink
Fujitsu launches 32-way Itanium system with 400MHz FSB: http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=3304&Thread=1&entryI...
Apparently slightly less advanced than we expected. They are however set to improve on this initial release some time in the future.
Regards,
Rink
Fujitsu to release 64-way Itanium server tomorrow: http://www.theregister.co.uk/2005/04/04/fujitsu_big_itanium/
Regards,
Rink
Chipguy, one step at the time. You clearly missed my comment about evolutionary pace.
Regards,
Rink
rlweitz, chipguy: Continental Airlines pushed the envelope when it moved its automated ticket-reissue application to an open-source software stack that included a 64-bit MySQL database server.
http://www.computerworld.com/softwaretopics/software/story/0,10801,100797,00.html
They moved to 64b Red Hat on HP Opteron systems. High availability through a 64b version of HP's ServiceGuard. Pretty good info on which hurdles Linux and Opteron have to take before it gets economically feasible for a datacenter.
Just an example. I know. Still I think it's not nearly as certain as you make it sound that the vast majority of the CIO's 'NEVER' will use Opteron during I presume you mean the coming year. I think it's more like that it's clear from the examples that Opteron continues to break enterprise ground but doing it at evolutionary pace. Your statement is also at odds with AMDs statement that they build an initial server presence with over 40% of the top 100 companies now. Continental is a good example of how an initial investment in development HP Opteron systems led to a full fledged mission critical system over time.
Regards,
Rink
CJ, anything particular you think customers will miss in zv6000 options that were previously available?
Regards,
Rink
Why are you assuming Galaxy is not based on Horus? The only thing we know is that Serverworks will be providing chipset, and there have already been some rumors / demonstrations of Serverworks chipset, which kind of parallels AMD;s 8131 / 8111. There has been no info about anything from Serverworks that would halp connecting more than 4 processors, so it is possible that Horus will be used.
Possible but not that likely because SUN hasn't mentioned Horus/Newisys once while they share in some way in the development costs of Serverworks chipset. So why are you presuming we have seen all chips from the Serverworks chipset already? Maybe there are more alternative solutions, but I'd find it weird if SUN would really use Horus (even though I agree with you that would help them scale much higher). Not counting out your suggestion, just currently find it not the most likely one.
Regards,
Rink
Keith, appreciate this fresh creative looking at all angles. (eom)