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.
Sure, I was laughing at his suggestion that SiS could better design a memory controller, which also implies that an off-die SiS MC would provide better system performance. What a hoot!
Sounds like the Kealia Opteron boxes are going to be quite impressive.
"We believe that Sun may publish benchmarks that show its servers (both pizza boxes and blades) running a commodity microprocessor and OS at 2-3X the speed of current x86 systems thanks to the non-commodity architecture built around the parts," the analyst firm wrote in the report. "Running Solaris rather than Red Hat would probably provide an additional performance boost. Sun's value added would be in the design of the board, the I/O, and raw performance. If 3X performance is achieved, then a data center could save significantly on A/C, electricity, and real estate."
These performance projections may sound nuts, but Sun really believes that its new Opteron gear along with its Thumper project will give it a big competitive advantage. If the performance is all Sun claims, customers may well pick Solaris x86 over Linux. And, as has been discussed before, Solaris sales are worth much more than Linux sales to Sun.
What a moron!
"A: No, it doesn't. Moreover, I believe we have more experience than AMD in designing memory controllers. Actually, we'd like to see AMD stop putting the memory controller in the CPU. I approached AMD about this several times, but they have their own reasons for integrating the memory controller with the CPU."
I'll say. No wonder SiS is a third-rate player.
keeping the prices of its Pentium 4 CPU lineup unchanged
I actually believe that to be the more important information, if correct.
With 2 more weeks in the year I don't imagine any great effect on 2004 Itanium units.
Optimistic about Itanium sales next year? Don't you think they've Osbourned themselves through all of 2005 with all the hype around Montecito, which won't be generally available until Q1 2006?
The Inquirer suggests that one possible impact has been an abandonment of current Tukwila design work in favor of a new approach.
Forbes has an odd habit of reposting old news stories anywhere from a day to several weeks after they are first published.
To your point, it isn't a new chip and it is most likely a Sempron of one sort or another.
kpf was more specific on the SI board about the rumor. The December 28th event is rumored to be the 90nm Opteron systems launch, and kpf's source is from one of the other companies with a product launching at that event.
Well, there is that. How bad is this year likely to be? One of the articles noted that Intel will miss the 200K goal this year, but it was not more specific. 150K? 175K? 125K?
A true believer, eh? :)
The move is a net boon for IPF
Boon for a boondoggle? :)
Seriously, if this is good news for IPF, what would bad IPF news look like?
The reporter has a huge error in his market share comments. 13.5% is the correct overall share for Spansion in Q304, but this was a decline from 15.7% in Q204, not 30% !!
EDIT: I see this was discussed earlier, and that the 30% figure was for NOR share only. Very sloppy reporting.
kpf, does your source continue to confirm a December 28 event involving AMD and partners?
thank you.
My feeling is that Intel was as optimistic as possible with the schedules presented yesterday, to the point of being misleading in some areas. They are fighting recent very negative PR.
Paul Otellini presented prototypes of the 65nm Cedar Mill core
If you mean holding up a wafer and claiming it contained Cedar Mill parts, then yes, he did. Were they functional? Probably not.
Yonah is due to ship in 2005, but will not likely be available in products until 2006. There is a slide that shows the Napa platform for Yonah, and it says Q1 06.
D-step was the "dumb shrink" from 130nm although there were some minor bug fixes and optimizations. E-step is a relayout, together with SSE3 and additional optimizations including additional write-combining buffers. It seems likely that the E-step is the one that incorporates an extra two metal layers as part of the revised layout. AMD has also spoken of a second generation of strain, but I do not know if this is part of the E-step. It would be reasonable.
I would expect frequency increases and lower power consumption, depending on the segment.
IBM Sells PC Unit to China's Lenovo
BEIJING/SAN FRANCISCO (Reuters) - China's largest personal computer maker, Lenovo Group Ltd. (0992.HK: Quote, Profile, Research) , said on Wednesday it is buying control of IBM's PC-making business for $1.25 billion, capping the U.S. tech giant's gradual withdrawal from the business it helped pioneer in 1981.
The agreement, which forms the world's third largest PC business, calls for Lenovo to pay IBM $650 million in cash, $600 million in Lenovo Group common stock and for Lenovo to assume $500 million in net balance sheet liabilities from IBM.
IBM (IBM.N: Quote, Profile, Research) will hold an 18.9 percent stake in Lenovo.
The deal closes an era for the world's largest computer company and kicks off a new age in which China's top PC maker Lenovo steps onto the world stage as a major PC brand and IBM partner.
The sale of IBM's PC desktop and notebook computer lines frees the company to focus on higher-margin businesses such as computer services, software, more powerful server computers, and storage as well as computer chips, analysts have said.
For Lenovo, which is battling intense competition in its home market, the deal with the world's largest computer company marks a breakthrough in its efforts to build its business overseas. It would also make the company part of a small but growing group of Chinese manufacturers buying overseas brands.
Lenovo will take ownership of IBM "Think" trademark family, including its ThinkPad notebook brand and its ThinkCenter desktop line. Lenovo will also buy out IBM's interest in its joint venture with Lenovo rival Great Wall Technology (0074.HK: Quote, Profile, Research) , China's No. 2 PC maker.
Lenovo will hire 10,000 IBM PC employees -- including about 2,300 in the United States -- mostly product designers, marketers and sales specialists -- and some 7,700 elsewhere, principally in China, where IBM operates a manufacturing joint venture.
http://www.reuters.com/newsArticle.jhtml?type=businessNews&storyID=7022858
Meanwhile, AMD has started shipping 90nm Opterons.
http://www.siliconinvestor.com/readmsg.aspx?msgid=20836095
Intel slides say Xeon will not be dual core until 2006. They are talking about Itanium when they say "server products next year"
The key word is "introduced".
Did any of their dual core demonstrations include AMD64 capable parts?
My understanding is that Yonah (dual core P-M, 65nm) is 32 bit only, which strikes me as a strange product from a marketing perspective in Q1 06.
How about the two-die "dual" Smithfield? Will that run in AMD64 mode or not?
Lenovo Confirms Acquisition Talks
http://biz.yahoo.com/rb/041206/tech_hongkong_lenovo_1.html
IBM-Lenovo Pact On Tap To Be Announced Tuesday
http://www.crn.com/sections/breakingnews/dailyarchives.jhtml?articleId=54800686
This section is most important:
Barring any last-minute hiccup, IBM is expected to announce an agreement in principle to sell its PC business, including the desktop and notebook business, to Lenovo, sources said. If the deal is consummated after a due diligence phase, which could take four months, Lenovo will end up manufacturing and marketing IBM-branded PCs and notebooks through the current IBM channels including the IBM direct sales force, sources said. The pact does not include IBM servers. IBM and LeNovo were unavailable for comment Monday.
December 28 90nm Opteron launch
http://www.siliconinvestor.com/readmsg.aspx?msgid=20832850
Over on that board kpf provides more information, saying that his one source claims the Dec 28 event is the 90nm Opteron (and related systems) launch.
3 HP Opteron blade servers by early January
http://www.siliconinvestor.com/readmsg.aspx?msgid=20833171
I read 90nm A64s as "backordered" and "new listing". Could be demand driven, given the very positive reviews the 90nm parts have received.
But would raise serious question about WTF problems AMD
was having with it. Four full mask level respins of a shrink job?
Revisions A, B, C, CG were all 130nm.
D0 is the shrink job (of CG), with minor bug fixes and improvements resulting in 1-3% IPC improvement by some measures.
E0 is the second generation part, probably having the additional metal layers and 2nd generation strain, in addition to adding SSE3, etc.
But Dec 28th is the rumored date for the 90nm Opteron launch. I don't understand your "too close to Christmas" argument with respect to a server/workstation part. Also realize that AMD's Q1 05 begins on December 27.
I think you should re-evaluate your strategy.
The reference to not being "turned on" would appear to be limited to DDR2. These chips will likely perform about 5% faster than their equivalently clocked revision CG brethren due to the accumulated bug fixes and small optimizations, which is something to be happy about, beyond the addition of SSE3. Further power reductions should also be possible. What is there not to like?
I find it odd that you or anyone would reference base scores instead of peak scores when quoting SPEC int/fp 2K results. Base scores are subject to manipulation by compilers that identify benchmark source code and apply the best optimizations.
SPECint/fp_peak2k:
FX-55 1854/1878
P4/3.8 1671/1842
Notice that the P4/3.8 gains almost nothing going from base (1666/1839) to peak (1671/1842).
What does "NOSIG" mean? Thank you for the update.
Greetings, kpf.
Would you have any further details you could share regarding the December 28 product launch rumor you posted about a few days ago? If this is the launch of 90 nm Opteron systems do you know at what speed and core revision the parts will be?
Thank you.