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.
wbmw
I think AMD would have to ramp faster in order for Barcelona and derivatives to give them a competitive advantage, at least as far as the customers buying these servers.
I think they would if they could. But working overhours during pregnancy does not deliver the baby sooner. :)
In the bigger frame, after listening to AMDs song for the last quarter i believe the question i was unsure about who is acting and who is reacting, in what segments and how can be answered now: AMD apparently goes for unitshares, sacrificing gm, asp and maybe even earnings. Folks clearly took initiative in this respect, in particular in mobile and desktop low end. It makes sense to me - because it is what they can do and need to reach critical mass to stay in the game without being dependant on subsidies and helping hands, eventually. They will concede servershares (units and rev) to Intel big time this year. But frankly i see no means to avoid this anyway, so they gotto live with it and let it happen. I see them passing a quarter of unitshare this quarter, which will make a lot of headlines as soon the numbers are out. The market forgives red ink as long as companies grow.
Bottom line, i like the way folks are playing the game. But I think Mike Masdea has it right: Better to watch the race from the cheap seats for now. (btw i appreciate mike has been trained to lower his pace of talking, it apparently is now balanced with his pace of thinking and produces nice aphorisms like the above :) )
K.
wbmw
Server offerings, limited etail and retail availability, and pilots of OEM-design this year. OEM-Volume in 08.
Iaw, your 7,5% for 07 is on the optimistic side in my book. However, single digits might be the ballpark.
K.
wbmw
taped out K8L some time between June and September,
Second half of August. And i am in the same camp for timeline. It's a complex design. Rather 18 month than 12 from t/o to volume. Samples by mid-year ain't a problem. Neither is a formal introduction - with immediate availability [for reviews]. Shipping for revenue translates to shipping for benchmarks. What's new? Having said this, I might add I don't mind. We probably should just get used to it. Development and production blend into each others more and more these days, apparently, so there is nothing wrong with it.
K.
wbmw
Intel claimed that they are already shipping >50% dual core finishing Q4
Naaah. Here is what Intel said verbatim:
"The company shipped more than 70 million 65nm microprocessors during 2006 and ramped dual-core technology to greater than 50 percent of fourth-quarter shipments."
Very easy to misunderstand when not listening carefully, admittedly. But then, what's new?
K.
Joe
I see what you mean, and basically agree. We both have seen the Timna-approach a decade ago, and folks at AMD have seen it as well. Hope to have an opportunity to talk to folks before the decision is due which way to go. Which I believe would be around mid-year.
K.
Joe
I would still assume that the approach will be modular.
Sounds right, although i don't know what you mean. :)
Could you elaborate on what you mean by "modular" pls?
K.
Joe
Major one is upcoming after that.
yes. But probably not in core-µarch. Rather next integration-step. Makes more sense.
K.
wbmw
Just a thought. Good thought. :) K.
metropick
Tx for posting. Is Don Clark still with WSJ?
K.
p.s: Nice picture.
Comb
Just for a suggestion: How about opening your own shop with a "Booting on Alpha-Silicon" board? Make sure to hyphenate the use of endless loops in the original posting. :>)
On a more serious suggestion: I don't mind folks looping this endlessly, but could you guys pls tag further contributions to this thread with "BOOT"? It's not OT, but i began to get bored of it a while ago already... Many tia.
K.
elmer
The interview was 05 iirc.
Combjelly mentioned 250µm ball pitch. Sounds to be in the right ballpark (sic!). This would allow for 1600 pads less edges per sqcm, if my math ist correct. Intels NB has 1439 balls iirc. Do you have its current diesize handy?
K.
elmer
That would be Paul Otellini's dreams. He explicitely said so, in an interview (with Hans Mosesman iirc), quite a while ago.
K.
Chipguy
Many thanks. Interesting. Do i interprete your remark right NB and SB for one CS generation are on different processes? (I never looked closer at Southbridges).
K.
On second thoughts about it, a single-die NB/SB chipset solution should ease pad-limitations: No i/o pads required to connect the two chips.
However, AMD still plans for a two-chip-solution for its 7xx generation. Probably for diecost-considerations implying this from a dfm-stance, and flexibilty as well. Comments appreciated if i miss something else here.
K.
Comb
That would be either a very small chip or a lot of I/O on the package.
The latter. Ballcount fourteenhundredsomething. Guess the limitations are not for i/o but rather for supplying as many power-planes you want for lower power-consumption.
K.
Comb
Hard to be pad limited for flip chip.
Pads can easily become a bottleneck when you integrate and shrink. Pads do not scale very well. Intel is already pad-limited at its chipsets.
K.
alan
Fun aside, i don't expect much change in shelf-space wrt Intel/AMD in the US and Central Europe for a while. Battlefields seem to be asia/pacific, Eastern Europe, South Americas etc., currently. China in particular.
K.
alan
I suspect much fewer Pentium D systems for one trend
Probable fewer. However i am not sure you keep your smile becoming aware what will occupy the space for another trend.
K.
wbmw
Naah. Just clearing out stock to make room for Vista-books on the shelves.
K.
paul
An Intel-Server offering would certainly not mean to drop AMD. Just adding. I am not as certain as Keith and others about it. Plausibility does not necessarily make a rumour to materialize. No immediate denial does not mean anything. At times SUN stock doing well Scooter usually is out for a round of golf or two at the weekends. :)
K.
paul
Essentially this is nothing else than Phil said. "Production" in AMD-terms is when the design is in the fab. Which is tapeout plus the time it needs to have the first mask. Was ever so.
K.
joe
rather than just integrating existing design on the same die.
You theoretically could do this on-die, but it would not make much sense imo. While it could possibly make sense as an MCM, with some glue (piece of silicon). (That's why i mentioned MCP as a possible intermediate step of integration). Not sure it offers a value proposition worth doing it, though. I'd expect to see MCP-approaches rather from Intel than from AMD.
K.
joe
Time to market absolutely sucks if it indeed is end of 2009.
Consider it makes no sense doing it before 45 nm, and it also makes no sense to release it as first 45nm silicon from a manufacturing point of view, you end up at H2/09 (still optimistic in my book). Besides, it is not only the manufacturing side preventing Fusion to arrive sooner: Two years for design-development of such complex arch and another year for getting silicon ready for prouduction is an agenda as tight as possible at all, imo.
Wrt to the AtI-merger, whether Fusion arrives by end of this decade or beginning of the next is beside the point imo. The point is whether they can stem this at all within this timeframe. Without AtI it would be close to hopeless to get there before mid of next decade.
In the meantime, there is options for intermediate steps: Taking grafics onto the chip by means of MCP to allow for integrating Chipset/SB to a single die, e.g., if pad-limitations allow for the latter to take ten dollars cost out of the platform. Maybe this die could be taken onto the MCM as well, in the next step.
Bottom line, this industry has ever been about integration, which will eventually end in a SoC. Not the question if, but when. From this stance, the question is not whether AMD could afford taking Ati, but rather if they could have afforded not to take it.
K.
joe
The same as what AMD does - keep upgrading to the latest process technology, and if some of the existing equipment can't keep up, sell it off.
This works for AMD. It is eventually practicable for Intel as well as soon as Moore's law has a different shape in the next decade and the nodes get longer with it. In the meantime, i am not sure it would work for Intel.
but that's proably way lower on the list of concerns of the investors.
I don't think so. People with long worry lists do not invest in technology. I think it is far easier. Silicon investors do not care all too much about numbers for the past. But they want companies they are invested in to paint a sunny day picture for the future. Which Intel obviously did not deliver.
K.
joe
I think Intel should dump NAND as well
Any idea what Intel should do with the fabs after CPU, then?
Chipsets are not enough to utilize these, and the competition is trending to leading-edge nodes with chipsets already.
Foundry-ops? Hard to compete against asian fabs.
They try to sell one currently. Lessee if and how that works out.
I am a little suprised by the negative reaction on WS.
Intel does not talk about mss anymore. And it plans to ramp one fab less in the next node than in the current node. I don't intend to ignite a discussion on diesize-economies, i just mention it because i believe the street connects these nodes in way to explain the market reaction.
K.
chipguy
proverbial rooster who egotistically thinks his crowing makes the sun come up.
Didn't know this proverb. I like it.
I disagree on the rest. Intel has demonstrated it always prefers PR-inventions over computer-arch-inventions for keeping its business alive when it can in the lack of competition. Quite understandably, words are pretty cheap to manufacture compared to silicon, so this practice is in its own and its shareholders best interest. I'm not saying it did not invent arch as well. But this can fail. Think Itanium. Marketing inventions are not only cheaper but also less risky.
On a sidenote, i read you since couple years and am aware you know a thing or two in some fields. However, as long you keep your tone along the line of "complety and utter nonsense, wtf", i will continue to do my very best to respond in a suitable way.
K.
Chipguy
I see subtle hints don't work for you. Here:
http://www.investorshub.com/boards/read_msg.asp?message_id=16290480
K.
chipguy
The perpetual need to keep users interested in upgrading their systems on a regular basis is far more important to Intel than losing/gaining a few points of market to/from AMD.
No dispute. But it does not take a new arch to achieve this if there is nothing better in the market. Intels marketing people are at least as inventive as Intels arch-designers.
More generally, as wbmw and tecate got my point already there is hope you arrive there within days.:)
K.
Katie
What i was trying to get as is Netburst at 3GHz would be competitive today without K8-arch in the market, and Intel had no need to replace it anytime soon. :)
K.
wmbw
Then it's also fair to say that AMD wouldn't be where it is today without a company like Intel to aim at.
Certainly.
The rest is beyond the frame of a posting to dig into it. The underlying question whether demand is created by supply or vice versa is very very complex.
K.
mas
Intel's best new architecture introduction and ramp in decades
I think it's fair to say it is more due to AMDs management than Intels this arch materialized.
K.
eracer
U r right. Could be Fudo misinterpreted a snippet picked up from CES about bundles. In fact, AMD is offering bundles to OEMs.
K.
eracer
I understand Fudo talking about OEM-business, not channel.
K.
Chipguy
I didn't lose you with this intricate analysis of the
situation hmmm?
Don't worry, i can easily deal with the degree of complexity of your analysis.
As long as no numerals are involved, that is. :>)
K.
ced
Whenever a sugar-cube appears short before earnings, you better prepare your organism for the bitter medicine to be dropped onto... ;>) K.
ced70
If the earnings are bad I am not for sure what effect if any it will have on AMD.
The street has a habit to crucify AMD for Intels misses. And the street is not known to brake easily with its traditions.
K.
mike
As far as dinners go, we can do that at any time.
Sure. That is not the point. The point is the fun of the imagination paying the bill in the restaurant with money you avoided to pay for an unnecessary brand premium.* Try it. It will make you smile, you bet. Your spouse will notice - and like - it. (Beware of talking what you are smiling about. It would badly spoil the party) :)
K.
*Which you only are aware of because you are one of very few with personal experiences with both systems. All others depend on brand-awareness from benchmarks and media.
chipguy
Idealogically driven preferences aside, why would anyone buy a K8 system over a C2D system?<i
Because systems purchased are not used for benchmarks, and to use the 100$ note saved for a romantic dinner, e.g.
K.
drjohn
It's not even clear DX 10 will launch with vista.
Interesting. DX-10 for Server 03 is RTM since a while already. However, i could imagine DRM-issues preventing a release with Vista.
K.