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.
Kate, re: May Hang.. get it. Bios workarounds are common and you know it. Yawn.
A bios workaround that shuts off ALL of Speedstep will not make it wanted, especially by corporations. Without Speedstep, and considering the already extreme max power it will now need to function at constantly, it's virtually sure that Intel will hardly sell any significant volume of this first iteration of Presler.
Yes bios workarounds are very common as you noted, but this 'workaround' is so crude that it rather significantly devalues Presler, especially for corporations; so much that it wil be shunned by most especially by corporations but also by the knowledgeable consumers. This thermal Presler fluke probably will drive continued pressure on their 90nm fabs because Prescott will need to replace Presler is most cases. Both in server and desktop lines.
I think no Intel person actually said 'yawn' because of the consequences for sales - I frankly think most used the F-word instead. Even more importantly Intel's image is not exactly enhanced by the continuing string of P4- and Itanium related flukes.
Regards,
Rink
Morrowinder, isn't the amount you know less than the owner and CEO of Voodoo PC indicative for the extent you are willing to discredit yourself, considering you don't take any of his comments even remotely serious?
Regards,
Rink
Re: statistical significance of $619m per quarter of [IPF server] sales
Isn't that only a bit over $100m for IPF processor sales?
Regards,
Rink
Chipguy, Maybe you think IDC and Gartner are lying about x86 and IPF sales or their respective growth rates too? Go stick your head back in the sand.
You know much better than most how exactly HOW wrong IDC and Gartner have been for as far as Itanium growth projections have gone. There hasn't been a single multi year projection that was even remotely low enough for IPF. This amounts to pretending your head is in the sand while it isn't.
Re: That is part of the story but is far from complete. HP's total HP-UX sales are steady while PA-RISC sales shrink but HP-UX represents less than half of IPF sales. Do the math genius.
I wasn't talking about HP but all of Itanium sales, again as you well know. As for your smart ass remarks, they aren't.
Re: x86 holds less than 10% of all servers costing more than $10k and less than 5% of all servers costing more than $25k. Try to rise above your PC myopia.
% of x86 servers costing more than $10K/$25K isn't declining which invalidates your point that Itanium is eating into high end x86 servers. Not in any pragmatic sense.
Lastly you apparently can't refute my main point which is that Itanium hasn't slowed the x86 server growth curve, something I pretty much am sure you said would have happened years ago.
Regards,
Rink
Chipguy, You know that I think that your argument that Itanium is growing by a factor amazing, and x86 only growing gradually is in fact BS. A more pragmatic way of looking at it is that Itanium is replacing some RISC/MIPS architectures. Sure Itanium sometimes is sold with Windows too, sure it wins over a previous x86 account. x86 however does the same. It's sold into very high end taking over previous UNIX accounts, and is sold with non Windows OSs too (although mainly LINUX, some Solaris, still non Windows support is only growing). And as you know the integration of virtualization into x86 hardware (on top of 64b) will help very nicely with selling more into high end. Main point is that the growth rate of x86 servers did not show a decline as the result of Itanium. Your Itanium projections have mostly been consistently wrong (either frequency, or due date, mostly both, from years before Merced until now). Because of this it is not more than logical that your remarks of x86 servers being eaten by Itanium from the top many years ago already have been off as well.
Regards,
Rink
NH. That's quite a bit to the north. Gets colder there generally than in Holland. You probably have the joy of a ski resort reasonably close by though.
Pension all over Europe is under pressure. General direction is that people need to work longer and pay more for mandatory pension. Instead of the additional pension schemes some people are getting into I thought it better to invest cash in stocks with growth potential. I don't trade much. I got into investing in my 20s but had to learn quite a lot. That's why I like your idea of supporting your kids with a bit of real money to develop that skill.
Regards,
RInk
Chipguy, re:They probably listened to droids who said Intel wanted to replace x86 with IPF.
As did you, right?! I think I remember your postings and even an article I think (on RWT) mentioning that it was likely that top x86 servers would gradually be eaten into by Itanium years ago. None of that happened or is likely to happen any time soon even now.
Regards,
Rink
Michael, re: ...and give him a bunch of money to play with on trades. ... to give him an important life skill.
Great idea. Sounds like something I might try myself when my kids are older.
Is your winter that bad? Whereabouts do you live if I may ask?
Regards,
Rink
Rupert, re: I am suspicious about the Piper Jaffrey opinion. By adding such precise estimates for AMD's percentage of DELL's sales for both 2006 and 2007 the analyst went to some lengths to achieve credibility and, therefore, an impact on the market. But its those very details which don't ring true. How could he know?
Same here. Taiwan subcontractors are known to leak details once in a while. Dell may have placed a cancelable order (they did so once before iirc) that they'll cancel the moment Intel comes through for them. The proceeds via Intel discounts may well be more than the fine of canceling the orders.
Most likely this Dell rumor will run out of steam within weeks e.g. when Dell itself squarly denies it; just to complete the cycle once again.
Regards,
Rink
Boy you sure know how to broker a deal with your kids :)
I don't know if I would do it better myself as though my kids are still quite a bit younger than I think yours are they already get me doing plenty for them.
Regards,
Rink
Avatar, a couple of the other times the rumors had good legs too, but the rumors never became real. Absolutely no reason why this time would be different. It doesn't make sense either: Just when Woodcrest is supposed to become available Dell would start shipping Opteron systems?
Yeah I know, maybe only 4+ sockets but that market isn't that big so it's not the reason. In Q3 there is simply no reason anymore anywhere nearly big enough for Dell (I mean current rather large discrepancy, and previously DC Opteron vs. SC Xeon have not been reasons either).
It's my observation that these rumors usually start surfacing when Intel is delaying something. Then they ramp up the back up plan. That has two advantages: If Intel really screws up a lot more than Dell initially thinks is possible they have the back up plan ready for production or even release. If Intel scrambles and more or less gets a remotely nice solution together they can and therefore do leverage their Opteron solution to pressure Intel and get better prices. Intel knows this but plays along not willing to risk even a low percentage chance of Dell running one single product line with AMD processors inside. Dell is triple satisfied with Intel because the partnership allows cost of processors, inventory, and production all to be as low as possible (economy of scale). It comes at a cost but Dell has shown it's willing to take even reasonably high costs of loosing a bit in high end server and gaming markets to not mess up their main current reasons for existance and growth. This only makes sense. There are plenty of precedents too. With what we know Dell releasing Opteron at the time of Woodcrest does not make sense. Yeah, that simple.
Regards,
Rink
(Don't get me wrong. I'd LOVE to be wrong. I don't see how I can be though.)
Voodoo Turion notebook with 2.5h battery life (ML34 @ 1.8GHz, 2kg, 12', small battery): http://www.theinquirer.net/?article=28795
It's a bit weird to see ML34, 2.5h, and small battery in one line.
Regards,
Rink
FPG/CJ, here you go: http://www.siliconinvestor.com/readmsg.aspx?msgid=22017108
Regards,
Rink
It sounds like Intel appologizes to Sony, Toshiba, and the like for not being able to prevent further distribution of their trade secrets.
Regards,
Rink
Wbmw, re:So you're saying that AMD is going to retrofit this technology at 90nm?
Exactly right. Here's one snippet (not the best one but it'll do). I did see several though. IBM is set to use it directly with 65nm produced soon, while AMD will use it first in 90nm:
Andy Wei, a member of the technical staff based at AMD Dresden, described a new process that AMD will first retrofit [SiGe] into its 90-nm microprocessors, and then use it in its 65-nm designs going into production in the second half of next year at AMD’s new 300-mm fab in Dresden. http://www.eetimes.com/news/semi/showArticle.jhtml?articleID=174901200
Regards,
Rink
Wbmw, re: I don't expect much more than 1.8GHz out of DC Turion (assuming 35W)
I presume that in the above statement you forgot to include the impact of much improved strain by added SiGe at 90nm SOI in fab 36??
For Conroe, 2.7GHz will be more like 2.66GHz to conform to the FSB multiples. If that is the case because of the 65W envelope, I can believe that. However, Woodcrest and the Extreme Edition will both have higher power envelopes, and I think this will translate to bin n+1 or n+2 in frequency.
Agree.
Regards,
Rink
Chipguy, tx for your anwswers. Much appreciated.
Regards,
Rink
Chipguy, thanks. That was exactly what I was looking for.
Do you believe Intel rumors that Conroe will intro at 3.0GHz? I just had a discussion with Keith where I thought that ~2.7GHz would be more likely and would appreciate your opinion as well.
Regards,
Rink
Ibanker, Conroe has twice the power envelope of Merom/Yonah. As Keith said earlier today this is similar to difference between high end SC Athlon 64 and Turion.
Regards,
Rink
Chipguy, what do you think Conroe IPC is vs K8? Highly appreciate your thinking.
Regards,
Rink
Keith, re: do you expect INTEL to continue to trail AMD on both of those vectors [performance/power] with the new offerings? Or what else?
Until Conroe/Woodcrest AMD will continue to lead in both. In T&L Yonah leads in both until Merom.
I'm not sure how much SiGe will do for AMD. It think it's significant and might lead to roughly 10-20% higher speed or lower power (a wag ofcourse), so ~3.1GHz for DT/Server and ~2.5GHz SC T&L / ~2.1GHz DC T&L by Q4 '06. I think this is quite reasonably cautious.
So provided IPC remains roughly the same (as currently with Yonah and X2) after Merom/Conroe/Woodcrest I think that AMD will continue to lead in performance in DT/Server (except for Virtualization where it will lead by a large margin), and only slightly underperform in T&L. It might only slightly underperform in power consumption.
Hence my wag in another post today that I think that though Intel might do less worse than this year with the arrival of Yonah and Dempsey/Bensley, after Merom/Conroe/Woodcrest I think that both AMD and Intel will do well.
I actually think that only continued level of strongarming by Intel will be able to succeed to put AMD back in a corner after Merom/Conroe/Woodcrest, but I think that chances that that will happen are decreasing with the way the lawsuit is going currently. AMD apparently thinks so too because it is said to announce another new fab summer '06 already (production by '08?).
Regards,
Rink
Buggi, we all know that for years now in all cpu lines Intel has not lived up to hype or even internal goals for as far as frequency is concerned. I'm not sure if past mispredictions are indicative for Merom/Conroe/Woodcrest though.
Yonah's intro frequency is in line what I expected although I too thought that one bin higher was not exactly out of the question at launch (while some others thought that would be even likely because they heard so directly or indirectly from Intel itself). I didn't/don't fear Yonah that much because of several reasons:
1- It's 32b meaning it won't run the highest volume mainstream version of Windows Vista. This probably means that Turion will keep a niche but probably not grow market share anymore during most of 2006, ASP will suffer though.
2- It's relatively soon to be outdated both by Merom and DC Turion. Osbourne effect.
3- Turion doesn't have a large share now.
What I do still fear is Merom/Conroe/Woodcrest. 'Still' because their performance is unknown (32b->64b performance delta similar to K8?, not precisely known frequencies), and because AMD's rev F on a SiGe process performance is not entirely specified either. Too much unknowns yet. As you know I'm cautiously optimistic though because of very competitive cpu designs (especially DT en Server) in combination with higher fab capacity and lower costs per die because of 300mm. Intel's highly vaunted trio is late (on schedule but way behind K8). AMD's 65nm is late too, but it'll see a second mid life kicker for 90nm in added SiGe and will see the 200mm disadvantage shrink. My wags are that AMD will do well next year Intel won't do as bad as this year. Both AMD and Intel might do well in 2007.
Question: I'm not sure yet if it's likely that added SiGe is implemented from the start at fab36. Does anyone here know?
Regards,
Rink
Keith, well, maybe you're right again (eom)
Keith, first you are wrong that Conroe will be restricted to a 65W
power envelope. That is for mainstream parts, but not high-end.
Thanks. That explains it partly.
As for Merom, that one is restricted to a much smaller power envelope than even mainstream Conroe, and is thus not relevant to
determine how high Conroe will clock at its debut (that is btw. supposed to be later than Merom). The fact that Turion is currently restricted to 2.2Ghz doesn´t exactly mean that DT A64 will not get to 3Ghz (or 2.8Ghz already today). Actually, if you add 600Mhz to Merom´s speed, you get to 2.93Ghz, and that is what I expect for Woodcrest. Conroe could be close or slightly higher clocked.
Conroe and Merom are based on the same design and hence Merom's debut frequency is relevant to get an idea for Conroe's frequency (I think you agree with this). I'm not sure if I can expect Merom mainstream to fit in the same power envelope as Yonah mainstream (31W). I am quite sure however that Conroe power headroom vs. Yonah mainstream can't be exclusively spend on frequency increases as Conroe includes the 64b functionality that Yonah is lacking. Ofcourse the designs are different but even more importantly for power consumption is that they share the same process (Merom/Conroe might have slightly more advanced power saving techniques but the logic part of the cores is likely to be somewhat significantly bigger than Yonah's). So in conclusion I still doubt if you can add 600MHz to Yonah or Merom clock speed to get Conroe mainstream version max clock frequency.
Also I heard that Woodcrest TDP would be between 70-85W. Do you know if that's correct?? If it is I expect Woodcrest to clock just as high as Conroe mainstream.
Regards,
Rink
Keith, interesting. Why?
Sure it's a next gen product compared to Yonah. But it is based on the same 65nm process and has 64b functionality that impacts power consumption of ALU's, data paths, scheduler, etc... and needs to fit in 65W power envelope (so not all of the power headroom for the desktop part Conroe can be spend on frequency increase). 3.0GHz is quite a lot above the rumored frequency of Merom, i.e. 2.33GHz. Based on rumors and the above I think it's unlikely that Conroe will debut at 3.0GHz. So I'm very interested to know why you disagree.
Regards,
Rink
Wbmw, HailMary, I've seen some unsubstantiated rumors claiming that Conroe will debut at 2.93GHz:
Yeah but as you know x86-secret's among the utmost unreliable sites out there. Here's a quote from another albeit only slightly pro Intel site THG that strongly implies differently:
Presler, however, is unlikely to live long enough to see any significant upgrades, since Intel's next-generation processor architecture "Conroe" will already debut in the second half of 2006. Consuming less power than the current Pentium 4 architecture and carrying up to 4 MByte L2 cache, the chip will enable Intel to lift processor performance to a level of a comparable 3.6 GHz dual-core Pentium D with Presler core at debut. Conroe will merge into the Pentium D 900-series as 940, 950 and 960 models, actual clock speeds have not surfaced yet.
http://www.tgdaily.com/2005/07/14/more_clock_speed_for_pentium_d_with_arrival_of_900/index.html
Another angle somewhere in between, possibly a bit more in the direction of x86-secret, is from yesterday's preview of Yonah (part 2): Honestly, as it stands today, if Intel can get clock speeds up, the only area that they will need to improve on is gaming performance to be competitive with AMD. We wouldn’t be too surprised if the comparisons we have shown today end up being very similar to what we encounter at Conroe’s launch: with AMD and Intel performing very similarly at the same clock speeds, but with AMD’s on-die memory controller giving it the advantage in gaming.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2648&p=14
Anyways for the moment I think it's unlikely that Conroe debutes at ~3GHz.
For comparison Merom was said to debut at 2.33GHz which I believe is reasonably realistic. My WAG for Conroe isn't more than 2.7GHz at intro.
Regards,
Rink
FPG, tx, forgot about that (eom)
Suddenly the deal to produce Geodes in China makes a lot more sense...
I think you may have this wrong. No Geodes are made in China, and none are due to be made there either fafaik.
Currently Geodes are fabbed by TSMC and UMC. AMD only has fabs in Dresden Germany (fab 30 & 36). It has a flex arrangement with Chartered (fab 7). Plus has a small stake in the new to be build SemIndia fab for which it also delivers process tech. None of this is in China fafaik.
Regards,
Rink
Drbes, Kate who calls herself 'The Great' (after Alexander) might have called herself 'The Blond' as well never mind that being true or not. Then again greatness is in the eye of the beholder as is truth for some.
She brings virtually no content or relevant knowledge, but rather a flurry of close to personal attacks, daft one liners, and an attitude that she would know anything relevant at all.
And now you invite her to SI. Have you gone completely balmy???!!!
Regards,
Rink
CJ, All I know is that Intel has been talking about shortages for more than a year. How long does it it to qualify a company for low end chip sets? Because low end is always the area they identify. Their recent move to drop royalties on certain chipsets rings hollow.
Exactly.
Regards,
Rink
Keith, still I'm not entirely confident that Intel's capacity problems are the entire issue. They used to be able to focus on high margin in these situations. Not the case this time. In an increasing manner Intel seems boxed in. With nowhere to escape this Q. Sure Q1 will bring relieve but isn't likely to reverse moves set in motion long ago. Yonah is good; it's not powerful enough. Merom/Conroe/Woodcrest are good, the mystical question though is how good vs. rev F and slightly later vs. AMD 65nm.
Regards,
Rink
LMAO. Tx! (eom)
Got a good laugh from that one, tx!
Regards,
Rink
Kate, do you call justifying buying an Intel VIIV system just for the well marketed logo cheering, or a sound investment driver?
Here's what I'm refering to: http://www.investorshub.com/boards/read_msg.asp?message_id=8790078
Regards,
Rink
LOL (eom)
Kate, just curious, have you been a cheerleader?
Regards,
Rink
Imho, fully agree (eom)
Smooth, you're right about marketing ofcourse but that's it. You still don't see the difference and that's fine too.
Regards,
Rink
imho, re:I think it was more the marketted notion that the Centrino "package" and not just the processor was the differentiator.
True to an extent ofcourse. Still low power was the only feature that couldn't be matched by AMD and therefore the most important one.
re: HDTV is an important driver, just like .11b/g was for Centrino. Unlike the Centrino, Intel has partnered (e.g. TiVo) for the ViiV campaign. This is one difference that is more important, I think, than the difference you mentioned(what CPU is inside). I suspect AMD is trying to find partners for its campaign as well.
HDTV is an important driver. I think it can be matched by AMD though, but admit I haven't given thought yet to the question with whom they'll partner. Any ideas?
Regards,
Rink
Chris, great chart. To think there were actually people here that believed those predictions EVERY TIME...
Regards,
Rink