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.
John -
Yes, Ireland was the fab that was halted then restarted. Remember 12 inch wafers have about 2.4X the capacity of 8 inch wafers. Intel has 2 12-inch fabs on line now, F11X and D1C. 2 more in construction, D1D and F24 (I think) plus a 3rd is being converted from 8 to 12, F12 in Ocatillo Az. That will bring it to 5, 12 inch fabs probably by the end of 2004. Are 5, 12-inch fabs equal to 12, 8-inch fabs? Pretty close I would imagine. After that Intel may continue converting 8-inch to 12-inch which capacity wise, would be like getting an extra new fab each time they convert an old one but at a much cheaper cost.
NAS -
E-mails criticizing former employer(Intel) is protected speech, court rules.
While he is free to send the emails, Intel is still free to block them.
New SPEC scores released:
http://www.spec.org/osg/cpu2000/results/res2003q3/
New SPEC scores released:
http://www.spec.org/osg/cpu2000/results/res2003q3/
Sgolds -
It would seem to me that because there is one die, that every Opteron is a potential 1/2/4/8 way processor. If all the aHT ports work then how can there be any difference?
As for MBs, a prototype doth not a validation make. (For our non English Mother tongue friends, that was a little reference to Shakespeare).
Paul -
I think Hector's done all right. Now, if things don't look considerably better a year from now...
The Hammer architecture and process selection was cast in stone when Hector took over. SOI has serious problems and may never yield well. It's an old core and the integrated memory controller may prove to be more of a limitation than it's worth. How's Hector going to change that in a year with no money while he records record losses?
Don't blame Hector for the mess he inherited from Jerry.
John -
It seems like a long time since Intel has announced a new fab (maybe I missed it). Are the larger wafers providing that much of a productivity boost? If so, it should help the profits after depreciation some time in the future.
Fab11X opened earlier this year. It's a huge 12 incher. Another fab is in the final stages in Oregon, D1D. Another is going up in Ireland. And yes they are providing that much of a productivity boost.
Sgolds -
Although AMD is shipping a processor meant for 4-way and 8-way systems, I have not seen an 8-way Opteron motherboard yet. Hard to benchmark it if it does not yet exist.
Huh? How did they validate an 8-way if there's no motherboards? I guess it's just a coincidence that AMD intro'd a product that hasn't been validated and nobody can use on the same day that Intel introduced a product with astonishing performance and systems shipping today? Just a pure coincidence...
Hurricane -
I'm not so sure Intel has very good yields, at least on their 0.13 process. Sure it's bins well, and the cache is very dense, but I've this nagging doubt about the yield on that cache. The only reason (apart from market segmentation) for introducing so many different versions is to stop them being scrapped, due to cache failure. And most of this die is cache.
I didn't claim Intel has very good yields, I just showed what would happen if they did. You are entitled to your opinion. One this is for sure, they produce much larger cache that anyone else. Personally I don't think they would do that if they had problems making those sram arrays.
NEC Solutions America Sees Market Acceptance for 64-bit Servers Based on New Intel Itanium 2 Processor with 6M L3 Cache
SAN JOSE, Calif., Jun 30, 2003 (BUSINESS WIRE) --
NEC Solutions (America), Inc., a premier provider of integrated solutions for the "Connected Enterprise" in North America, reports strong and growing interest from corporate customers for its 64-bit servers, which utilize the powerful new Intel(R) Itanium(R) 2 processor with 6M L3 cache (code-named "Madison") announced today by Intel Corporation.
Positive customer response to NEC Solutions' 64-bit, 1000 Series servers has been seen across industry boundaries in wide-ranging applications, including datacenters, financial services, government and more. In most datacenter applications strategies are being implemented to move away from proprietary RISC-based systems in favor of industry standards-based systems. The new 1000 Series servers combine Intel and Windows compatibility with the tremendous performance gains received from Intel's new 64-bit processors running 64-bit operating systems and applications. Performance is further enhanced by NEC Solutions' high performance computing (HPC) technologies such as ccNUMA and crossbar switching, which dramatically improve memory and throughput speed, respectively.
During TPC benchmark testing, the NEC Solutions 1000 Series servers demonstrated performance and cost parity with many of the world's fastest, proprietary RISC-based systems. The TPC benchmark scores posted in April 2003 at www.tpc.org clearly showed that the NEC Solutions Itanium 2-based systems perform among the fastest in the world, particularly in applications where the proprietary systems previously held performance advantages. TPC-C is an industry-standard benchmark for measuring a system's processing performance based on an enterprise transaction model for handling orders.
"The Intel(R) Itanium(R) 2 processor with 6M L3 cache provides a significant performance boost for customers while allowing our 64-bit servers to come in at a price point well below that of proprietary RISC systems," said Larry Sheffield, senior vice president, NEC Solutions America. "Given the cost advantages of Intel's manufacturing process, and the fact that this is the first mass-produced 64-bit chip from the company, we expect the price/performance benefits to continue driving the market toward the EPIC architecture. Customer response makes it clear to us that even the highest end of the marketplace is ready to move away from the economic and technical traps of proprietary technologies."
Record TPC-C Benchmark with Intel(R) Itanium(R) 2 Processor with 6M L3 Cache
In April 2003 NEC Solutions America announced that a 32-processor server from NEC, based on the Intel(R) Itanium(R) 2 processor (6M), running Microsoft Windows Server 2003 Datacenter Edition for 64-bit and Microsoft SQL Server 2000 Enterprise Edition (64-bit), achieved the world's best TPC-C(1) benchmark result for a non-clustered, 32-processor SMP server. These groundbreaking results demonstrate that the most demanding scalability and throughput requirements can be achieved on Windows and Intel based servers from NEC Solutions America.
The 514,034.72 tpmC transactions per minute (tpmC) benchmark score achieved by the NEC Express5800/1320Xc system was accomplished while using the Intel Itanium 2 processor (6M) (code named "Madison") running Microsoft SQL Server 2000 Enterprise Edition (64-bit) on Microsoft Windows Server 2003 Datacenter Edition for 64-bit.
The benefits of 64-bit computing for commercial applications include 64-bit addressing of large-scale data sets. Systems using 32-bit technology today can address far fewer memory addresses than 64-bit systems, which can address hundreds of gigabytes of memory addresses. This provides outstanding system responsiveness for mission-critical enterprise applications. In addition to providing 64-bit memory addressing, NEC's design with the Intel Itanium 2 processor (6M) provides an expanded system bus, large number of registers and nearly unlimited virtual memory. The NEC Express5800/1000 server series has also been designed to incorporate future members of the Itanium processor family without requiring additional hardware upgrades, thus sustaining their leading edge technology edge at minimal costs. All attributes are key infrastructure benefits designed to take full advantage of Microsoft's 64-bit operating system and enterprise database application.
"Given the dramatic performance improvements associated with 64-bit computing, and the fact that 32- to 64-bit migration is made far easier with popular applications like Microsoft SQL Server, we're going to see more and more customers migrate to the 64 bit platform," said Rahim Kapadia, product manager, 64-bit systems, NEC Solutions America.
The high-performance NEC Express5800/1320Xc server, with 32 Intel(R) Itanium(R) 2 processors, is currently available for order North America. Limited product availability for systems configured as tested for the TPC benchmark tests is scheduled for August 2003. The NEC Express5800/1000 server series also includes 8-way and 16-way systems, which are available for order today. Further information about the server is available at www.nec64.com or by calling 866-632-3226. A white paper is available at http://servers.necsam.com/leads/WhitepaperInfo.cfm.
Intel(R) Itanium(R) 2 Processor with 6M L3 Cache (code named "Madison")
The Intel Itanium 2 processor with 6M L3 cache is the third in the Itanium processor family, a line of enterprise-class processors from Intel that brings the performance and volume economics of Intel(R) architecture to the most data-intensive business-critical and technical computing applications. Systems based on the Intel Itanium 2 processor (6M) are expected to deliver 30 to 50 percent more performance than today's Itanium-based systems.
K -
No I don't agree. AMD announced benchmarks with their 1-4 way configurations last April. Now, on the same day Intel announces performance numbers that set many industry records, AMD announces an 8-way with no benchmarks whatsoever. That says to me that AMD doesn't want to be compared to Madison.
YB -
I guess Intel has the excess capacity even if demand is going to be very high.
I would not make that assumption if I were you.
ChipGuy -
374 mm2
Thanks, so pulling out my handy yield calculator I see that if Intel has very good yields then it would place their die cost at around $75 assuming $1800 for raw wafer cost. Using your estimate of $25 for packaging and test, which I agree with, would give us $100 manufacturing cost. Not too bad for a product with high ASPs.
ChipGuy -
Have you seen a die size published for Madison?
State Supreme Court puts crimp in policing internal e-mail
By Howard Mintz
Mercury News
Putting a crimp in corporate efforts to police internal e-mail systems, a deeply divided California Supreme Court today refused to allow the Intel to block a disgruntled ex-worker from bombarding the Santa Clara chip giant with thousands of electronic messages critical of its labor practices.
Breaking new ground in Internet law, the Supreme Court, by a 4-3 vote, refused to extend traditional property rights to company e-mail systems, thus siding with ex-Intel worker Ken Hamidi in his long-running legal feud with the Santa Clara-based chip giant. Intel five years ago obtained an injunction blocking Hamidi from sending the company's employees an avalanche of e-mails regarding his firing and the chipmaker's labor practices, but the Supreme Court ruled that the order violated Hamidi's rights.
Specifically, the Supreme Court determined that Intel had failed to prove that the e-mails harmed its property rights, saying Hamidi ``no more invaded Intel's property than does a protester holding a sign or shouting through a bullhorn outside corporate headquarters.''
In dissent, three justices warned that the ruling ignores the reality of the computer age and a company's right to control its property.
The e-mail sent to Intel, the dissenters wrote, ``is more like intruding into a private office mailroom, commandeering the mail cart and dropping off unwanted broadsides on 30,000 desks.''
Hamidi, who was still reviewing the decision, expressed relief at the decision. Intel was also still reviewing the four separate opinions totaling 78 pages.
YB -
I know you addressed this to ChipGuy but I'm going to offer an opinion too. I would guess that the die cost is under $100 so there is still good profit there to be made. Much more important is it's impact on capacity should demand be high.
sgolds -
Deerfield and its larger brothers may share the same core, but different dice since the die consists of the core plus peripheral logic (like cache).
How do you know they're different dice?
K -
Could you help out with a link for Gallatin-SPEC-numbers @2,8? Or @2,5 at least?
Typically Intel does not publish SPEC scores for their Xeon MP line.
sgolds -
This is a great day! New Opteron models for 1x, 4x and 8x platforms!
A cynic might say that AMD had to do something today in light of the incredible performance of Madison. Looks to me like AMD did a Vaporon release. Where are the 8-way benchmarks?
wbmw -
I didn't see a benchmark link. Are there new benchmarks?
No but their website has an 8-way product announcement and a link to the old benchmarks. No 8-way. Obviously AMD had to do something today and it looks like the 8-way is pure Vaporon.
wbmw -
but you can get a $2149 model 844 Opteron (1.8GHz), which will probably have lower performance than the $1,338 1.3GHz Madison
A cynic (not me of course) might say that AMD couldn't allow today's Madison announcement to go unanswered. A quick look at their benchmark link that was attached to the 8-way announcement shows no 8-way benchmarks! Add that to the promised 2GHz that never showed up and the TPC-C score that was apparently pulled from the TPC-C page and one really has to wonder if AMD is sinking faster than we thought.
wbmw -
Incredible scores! Simply amazing.
You can see why IBM has an internal fight going on as to which processor to use! This has to be very humbling for them...
Meanwhile no 2GHz Lateron as promised.
sgolds -
So I don't give weight to what Jerry said for public consumption at that time.
I wouldn't either except it was in keeping with the pattern.
Sell PLD division to raise funds to do battle with Intel.
Sell the Headquarters building to raise funds to do battle with Intel.
Kill the 29K to have more funds to do battle with Intel.
Do you see a pattern there?
sgolds -
AMD dropped it as an embedded processor because they were not successful in getting it designed into products.
That's not what they said when Jerry canceled the project. He said AMD (meaning Jerry) couldn't afford to fund both x86 and 29K. Something had to go and it was 29K. Another shining example of how Jerry's hatred for Intel destroyed a successful product line.
ChipGuy -
and higher test costs
I still think you're wrong on this one butI admit, it's based on what AMD should do, not on what they did do.
With half the cache and lower clock frequency than Barton it remains to be seen if AMD can get the performance from A64
I don't think they can.
drjohn -
The Xeon MP has 512K L2 and either 1 or 2 Meg of L3 cache.
We are starting to get indications that on Monday, in addition to Madison, Intel will be releasing 2.8GHz versions of the Xeom MP processor. This will be a major performance boost over the current 2.0GHz MP.
http://www.theregister.co.uk/content/61/31469.html
Monday is also expected to see the launch of 2.8GHz Xeon MP processors.
sgolds -
I should also point out that money does not lie, but it may lay.
I think you meant ly.
SZ -
I lightened my position significantly today. Monday will not be good for AMD.
YB -
I keep trying to tell you but you don't listen.
There are no 3.2Ghz P4s available with anything other than an 800MHz FSB. i845 doesn't support an 800MHz FSB.
YB -
how many people use P4 3.2 with 845 motherboard? I claim much more than zero.
P4 3.2 requires a 800Mhz FSB. i845 doesn't support it.
This sort of helps to balance all those new Opteron design wins... Oops! I forgot... There aren't any....
SGI boasts new Altix supercluster wins
Jun 27, 2003 (Datamonitor via COMTEX) --
Silicon Graphics Inc continues to win new customers for its Altix 3000 Intel Corp Itanium 2-based Linux supercluster server, even if the increasingly popular server cannot prevent the company's current financial slide.
The Mountain View, California-based server and workstation vendor announced two new Altix customers this week: the Center for Biological Sequence Analysis (CBS) at the Technical University of Denmark, and the French oil and gas company Total.
The majority of Altix customers announced so far have been educational establishments, so the deal with Total is particularly important to reviving SGI's fortunes. The deployment sees the installation of a 256-way 900MHz Itanium 2 Altix 3000 supercluster with 1TB of memory and 16TB of disk storage at Total's technical center in Pau, France.
The supercluster will run seismic processing applications, which will be used by Total's 44 exploration and production subsidiaries to identify and develop onshore and offshore oil and gas prospects.
The CBS supercomputer, meanwhile, will be used for the virtual analysis of proteomic- and genomic-based cell models. The deployment includes a 64 processor Altix 3000 with 196GB of memory and a 10TB SGI TP9400 Fibre Channel RAID array, and also includes MySQL AB's eponymous open source database management system.
Since its launch in January, the Altix 3000 has been purchased by universities in Australia, Japan, Austria, Poland, Spain and Germany for such projects as bio-informatics, marine ecology research, earthquake research, and chemical, scientific, engineering and physics computation.
While the product has proved popular, it has not done much so far to boost SGI's flagging financial fortunes. In its recently completed third quarter the company made a net loss of $34.9m, down from a profit of $10.3m on revenue 30.7% lower at $217.1m.
K -
You're right. I hadn't considered the possibility of impacting your other sales. See, that's why I'm not a finance guy!
K -
Seems to me that once the wafers are done the money is already spent. If they can sell them for more than the incremental packaging and final test cost then it beats throwing them away. Anything over about $10 bucks cuts their losses.
ChipGuy -
I think the disagreement is from the POWER4/5 guys fearful for their ricebowl. :-P
BTW, this article describes the x450. It should be pretty competitive. But later this year IBM will bring out the x455. It reportedly uses a completely revamped chipset. From what I hear IBM is going to give HP and SGI real competition for the IPF performance crown at a given CPU count. Not bad for a company with "enormous ambivalence". ;^)
I think IBM knows Madison is going to clean their clock. No doubt they've had samples for many months and they can see for themselves. Either they get on the bandwagon or SGI and HP will do it for them. I think IBM also realizes they can't go head to head with Intel in process technology much longer either. Intel can produce Madison in volume and cheaply, relatively speaking. The handwriting is on the wall. After seeing the TPCC scores posted recently by HP and the SGI scores posted here today, I'm really anxious to see what IBM systems do. Monday is going to be very interesting...
IBM's Intel servers to reach new high
By Stephen Shankland
Staff Writer, CNET News.com
June 26, 2003, 5:40 PM PT
IBM is expected to announce on Monday several new servers based on Intel chips, including two Itanium systems and a top-end Xeon system with 32 processors.
The servers are the high water mark thus far in IBM's years-long "X Architecture" plan to endow its Intel servers with some of the capabilities of its higher-end server lines, notably its venerable but expensive mainframes. Intel's Itanium processor is geared for higher-end "big iron" systems that handle demanding data-processing chores where IBM has long been top dog.
IBM's x450 has four of the 1.5GHz Itanium 2 6M processors--code-named Madison--that Intel is expected to announce Monday. IBM had already begun selling its x450 system in April with the earlier 1GHz "McKinley" version of Itanium, but Big Blue put that system on hold after an NEC server customer found a glitch in the chip.
IBM also sells its pServer and iServer lines based on its own Power chips as well as its zSeries mainframes. The company has been ambivalent about Itanium, but it eventually came around, said Illuminata analyst Gordon Haff.
"Whatever spin IBM wants to put on it, there was obviously enormous ambivalence and disagreement within the company on whether to do that system. But the decision does seem to have been made that...IBM's going to go for it," Haff said. "They may not push it as aggressively as the pSeries for high-end commercial customers, but they do at least want to be able to cover that big iron window of opportunity."
While Itanium relations haven't always been smooth sailing, another agreement between the two companies is expected to bear fruit. IBM and Intel are working together to design thin "blade" servers that fit into IBM's BladeCenter chassis.
Intel will begin shipping the blades in July, Bill Zeitler, head of IBM's server group, said in an interview. "What we want to do is create a standard, like the IBM PC, that will accelerate adoption," Zeitler said.
IBM is freely licensing the design, Zeitler said, and networking companies Cisco, Nortel and Brocade also are planning to support it. An Intel representative confirmed that the blade products are scheduled to begin arriving in the third quarter.
The x450 uses IBM's Enterprise X Architecture (EXA) chipset, code-named Summit, to join the four processors to memory. A relative of the x450 expected later this year will permit several four-processor building blocks to be stacked together to form a 16-processor system.
The EXA chipset also has close relatives already in use for servers with Intel's Xeon processor. Xeon is a 32-bit design intended for lower-end and midrange servers, whereas Itanium is a 64-bit design that more easily supports large amounts of memory, has better features for multiprocessor systems and has mechanisms to protect against data corruption.
The first IBM Xeon servers using the EXA chipset were the four-processor x360 and the x440, which could grow from four to 16 processors. On Monday, IBM will reveal the successor to the x440, the x445.
As reported, the x445 will accommodate as many as 32 Xeon processors, though when the system begins shipping on July 1, it will come with up to only 16. The 32-processor version is due later this year.
While the 32-processor machine isn't likely to appeal to a large number of customers, it is likely to make the lesser versions more appealing, Haff said.
"What you often see with big systems is a lot of people would prefer not to buy the very largest. By having a 32-way, you make the 16-way that much more attractive," Haff said.
For customers building high-performance calculation machines out of clusters of lower-end systems, IBM also will introduce a rack-mountable two-processor Itanium 2 system called the x382 that's 3.5 inches thick. [Edit: Sounds like a direct competitor for Opteron with much higher performance]
The x445 has a starting price of $18,599. The x450, available July 18, will start at $25,999. The x382 will be available Aug. 20 starting at $26,589, IBM said. All the systems support Windows and Linux.
IBM also will upgrade several existing Xeon servers--the x360, x205, x225 and x255--with the new processors, the company said.
Spokeshave -
I'm forced to place you on ignore for a while.
I'm returning the favor.
Spokeshave -
It is worth noting, though, that Intel is equally evasive about P4 power.
You're getting a little carried away here Spokeshave. Try and calm down a little and keep some credibility.
Spokeshave -
I am willing to accept that number in order to avoid rehashing the same old tired argument that you are trying to goad me into.
You can't even accept a compliment.
Why are you in such a bad mood today?