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.
Haddock,
Time to go long SGI? Altix is pretty impressive. But then so is having a negative book value
I guess these "commitments" to Itanium were not cheap.
Joe
wbmw,
It seems pretty clear, then, that Opteron has very limited headroom.
Yes, current stepping has a limited headroom. It has been pretty clear from availability of parts.
But a new revision of the core should be available within a month or 2, with improved clock speeds.
Joe
wbmw,
If HP gets the same percentages, they will have 1275 SPECint and 2076 SPECfp, higher than any other SPEC scores to date.
That score would surprise a lot of people. Let's wait for the official submission.
Joe
thanks for the reminder. That would make a viable system, but would make sense only if prices of Opteron 2xx and Athlon64 were vastly different.
On the memory types, I am hoping that one of the hardware sites will give write an article explaining what's possible. Is it possible to have unbuffered memory on Opteron, registered on Athlon64?
Joe
Elmer,
Sure Joe. They're announcing before Intel announces because Intel asked them too. Brilliant!
Hmm... Then why announce now?
Come Monday Performance will be redefined.
Has Intel released the SPEC scores of Madison (and others in the family)?
Joe
Petz,
Why aren't they announcing when it will be ready right now?
Because they don't know how long it will take to get it to work.
They are announcing now because Intel asked them to.
Joe
UpNDown,
chipguy posted this sometime ago:
http://www.investorshub.com/boards/read_msg.asp?message_id=966713
Are you saying that none of this is implemented in any OS?
That's a significant omission on part of chipguy.
Joe
yb,
I don't understand your interest in dual A64. FIrst of all, there may never be such a thing, second of all, why would you not buy dual Opteron instead?
Price is one issue, I understand, but the difference maynot be huge. There will always be lower clocked Opteron CPU selling for less than high end, and I am sure someone will release motherboard in $150 - $250 range.
Joe
yb,
So, is IBM server Newisys-based or not? That's interesting...
One of the articles said that the IBM system 's memory capacity will be 12GB, which I guess means 6 DIMMs. Newisys has 8 DIMMs.
Joe
Some interesting quotes in the articles published. First C/Net: http://news.com.com/2100-1010-1020636.html
"For those segments that have already embraced AMD, the enthusiasm for Opteron is very, very high," said Dave Turek, leader of IBM's new "Deep Computing" team, pointing to government laboratories in particular. "We have substantial pull from customers."
...
Turek said IBM's Opteron systems will arrive in early fall. They'll cost somewhat more than Intel Xeon systems but much less than Itanium systems, he said.
...
IBM expects interest in the Opteron systems from customers in government agencies and universities, as well as in the life sciences, petroleum and automotive design industries, Turek said. To accommodate some of the demand, IBM this year will put some Opteron servers in its supercomputing center in Poughkeepsie, N.Y., which customers will be able to tap into as needed instead of buying systems on their own.
The same guy at Infoworld: http://www.infoworld.com/article/03/06/24/HNibmcluster_1.html
"The simple fact of the matter is that the uptake on Itanium has been slow," said Turek. He blamed Itanium's slower-than-expected adoption in part on the fact that ISVs and customers have been reluctant to port their applications to Itanium's 64-bit instruction set. "It is a bad thing to do when you introduce a new technology to tell customers, 'by the way, everything's new,'" he said.
Different IBM guy, this time Linux guy at ExtremeTech: http://www.extremetech.com/article2/0,3973,1136374,00.asp
IBM's dual-processor 1U Opteron server has not yet been branded, Jursik said. IBM intends to work on a 1U server first, then explore alternatives, such as 2U products and blade servers, he said.
Joe
belgiangenius,
You've simply got to be short on AMD going into earnings. It's free money.
Are you now going to claim that your alleged position was actually a short position?
Well, your position seems to be evolving. Have you considered a career in politics?
Joe
belgiangenius,
Oh I have a small position in AMD; not enough for me to care much. But, it amazes me how much one management entity can screw things up so consistently and royally.
LOL. Whatever you say.
Just for the record, my opinion is that you don't have a position in AMD, and you care a lot, enough to play a silly game of creating an alias etc.
Just an opinion, WTFDIK?
Joe
belgiangenius,
AND THERE IT IS!!
The usual feeling of what it is to be an AMD investor.
On a day like this one, I wish my AMD position was fictional, like yours.
Joe
Haddock,
How about Opteron score with gcc 3.3 compiler that is available, in SpecInt, Opteron has score 1045 vs. 800 for Apple.
Joe
Haddock,
That doesn't necessarily mean there will be 8-way systems. The 8xx chips are needed for 4-way systems too. If 8-way systems were that close you would expect to see SPEC and database benchmarks popping up by now.
Yes, that's true. Still, I hope there will be some announcements from third parties that will go along with the 8xx and 1xx lines announcements. A really nice one would be an official announcement by IBM for a full featured server (hopefully including 4-way). Another could be from MSFT, announcing beta of AMD64 Windows. June 30 happens to be exactly mid-year.
Joe
chipguy,
You mean like how Northwood co-existed in "very low volume for a limited time" with 180 nm Athlons?
No, like Athlon 250nm and Coppermine 180nm. <g>
Joe
Elmer,
if you read the CNET story and follow the link you will find this in the April link:
Big Blue on Tuesday announced its intention to use the AMD server chip in future server hardware, including a #1 server platform and #2 high-performance computing clusters.
http://news.com.com/2100-1006-997827.html?tag=nl
They are going to be showing #2:
The system, to be unveiled at the ClusterWorld conference in San Jose, Calif., is a dual-processor, rack-mounted server measuring 1.75 inches thick, an IBM representative said. It is expected to ship in the second half of 2003. The machine, which IBM first discussed in April, is designed to be clustered in large numbers to form a single, powerful computational engine.
http://news.com.com/2100-1010_3-1020033.html
Joe
DARBES,
re: "I am wondering if anybody will build a beast like this"
My suspicion is that we will not have to wait very long and that for many applications, it will be amazing.
Speaking of which, deep in this article, I found this quote:
AMD, meanwhile, is expected to announce a version of its Opteron chip for eight-processor servers by the end of the quarter, which ends June 30, said sources.
http://famulus.msnbc.com/famulusgen/cnet06-23-083501a.asp?t=CNTEK
Joe
Elmer,
So IBM will introduce a HPC Opteron. Is this really such good news? It seems it fulfills IBM's promise to offer an Opteron system but does the HPC market really offer AMD much in the way of volume?
The article didn't say that IBM would limit their Opteron systems to this market.
Joe
Elmer,
A simple google search of "prescott die size" returns a bunch of hits. Size estimates range fro 80 - 109mm2. I'd be inclined to believe the high side.
I didn't ask about estimate, I asked about the actual. Intel has not yet released the actual die size just as AMD has not yet released the die size. No need to get into Osama Bin Laden conspiracies.
If this is correct and if Intel has world class yields, their Prescott die cost will be a little over $5 before any shrinks. Meanwhile AMD could have lower performance and a die cost above $50. Not a good competitive position.
I don't know where you are getting these numbers, but they are most likely wrong. Besides, you are comparing 90nm and 130nm CPUs which will co-exist in very low volume for a limited time.
Joe
Paul,
You just did. I was a few 100s posts behind on iHub, so I am not sure if it was posted here or not.
Joe
ChipGuy,
For the record, AMD still has not released Opteron 1xx processors. They used 1 in a submission (Hardware available in July 2003), but they used it in a multiprocessor system, with registered PC-2700 ECC memory. When single processor Opteron (and systems) are released, and installed in more appropriate system / configuration, we may see another submission with higher score.
Joe
wbmw,
aHT doesn't really help the Opteron once it scales past 4 CPUs. You then have to use strange, high latency cube or chain topologies, or go with an aHT switch, which is the same thing that Madison CPUs would have to go through in a large scale system. Either way, Opteron has little advantage after 4-way. If it doesn't outperform in that space, the advantage could go either way.
Just a speculation on my part, but my guess would be that Opteron's strongest relative performance (vs. Itanium) would be in 8-way server. The reason I think that is that Opteron still stays glue-less, and even if HT links start to get saturated, it remains very low latency, much lower than the hit of connecting multiple high latency 4-way modules. And you get 16 channels of memory for an unbelievable bandwidth. I am wondering if anybody will build a beast like this.
Above 8 way for Opteron or 4 way for Itanium, I guess it depends on how these modules are connected.
Joe
Elmer,
Same goes for A64. Why don't they just say what the die size is? What are they trying to hide?
What is the die size of Prescott?
Joe
8-/,
I don't think microprocessors are the same as OSs. There is some inertia (once a model is introduced by OEM, it stays that way), but new models are introduced on quarterly bases. Servers less often.
Joe
spokeshave,
It depends on what kind of volume AMD is planning on for 1 MB Athlon 64. If the volume they are planning on is modest (less than 1M per quarter) they can save resources by just using Opteron die.
The real savings in die space will happen when AMD introduces 256K Athlon 64 for low end. And, 1 MB Athlon 64 will move to 90nm, probably (or hopefully) in Q1 2004.
Joe
sgolds,
replacing a CPU with a faster one will give you the cheapest way to extend a life of a PC. So it is a good thing if the socket is forward compabible (as opposed to backward compatible). It is not always straight forward, but with Socket A, and multiplier unlocked (or unlockable) CPUs, it is even possible to get all the performance out of the new CPU.
This is something you would do on a tight budget. Suppose you feel like you need a faster PC a new PC would cost you somewhere between $500 and $1000, you can get a CPU for $50 to $100 and be content with your system for another year.
Joe
wbmw,
I think the site you linked is wrong about the first Socket 370 introduction. The difference between Celeron 300 and 300A was addition of L2. It was still Slot 1 CPU. Later on, with introduction of Celeron 366, Intel introduced Socket 370, and I think they released the slower speed grades (333 and 300) as well later on. Here are some links from January 4 and January 9 1999 (as opposed to November 1998):
http://www.anandtech.com/cpu/showdoc.html?i=525
http://www.hwupgrade.com/cpu/celeron_333_370pin/
Joe
NAS,
The first socket 370 was a Celeron. The first ones released were 366 and 400 MHz parts (but I think they later released slower and faster versions).
Joe
wbmw,
An oversight on my part. I agree that Intel should have made more of an effort to keep the Tualatin socket 370 processors compatible with the Coppermine versions.
The one I really minded was Celeron (100 MHz FSB) to Coppermine (100 MHz FSB), which to me was completely unnecessary.
I believe there were some changes in voltages from Coppermine to Tualatin, so I could see it as a valid reason, but there could have been 2 specs instead of 4 (not 3 I mentioned):
- 100 MHz for Celeron (pre-Coppermine) + CopperMine
- 133 MHz for Coppermine + Tualatin (able to handle different voltages).
I ended up holding the bag with the pre-Coppermine Celeron speced mobo. There were a lot of hints that the Celeron socket would accept Coppermine, which in the end it didn't. I had a dual Celeron machine, and I hoped to upgrade to dual Coppermine when the prices of Coppermine dropped, which in the end I could not do (so I went with a single Athlon Tbird instead).
Joe
wbmw,
Agreed about 333 MHz. The problem may be qualifying memory, but I what I would like to see is even if the CPU maker releases 333 (or 533 in case of Intel), the mobo specs should make a leap to 400 / 800, so that when the next step is upon us, the mobos don't have to change.
BTW, you couldn't help yourself from showing more changes on AMD side, and do so by skipping 3 versions of Socket 370 in the same time-span.
Joe
spokenshave,
I don't know about Elmer's current position, but in the past he said he had covered calls or sold naked puts, which are neutral to bullish positions.
Joe
sgolds,
wbmw's link said there is no retail package of WIndows XP IA-64. It would be nice if there was one for AMD 64 version...
Joe
wbmw,
I was actually looking for this info - how to buy Windows XP 64 for Itanium. I am disappointed that MSFT chose this route. I would like to see a CD (DVD) with all the binaries (when ready) for i386, IA-64 and AMD64 all on one disk for the same price.
Joe
wbmw,
Joe, I hadn't made a QuantiSpeed post for weeks until the one you saw. Just because you happened to stop by and see one
Well, that is a news to me and you could have said it right away, but from where I started reading, it seemed like 75 of the 150 posts you did your best to keep the flame burning.
The only reason why we are having this conversation is because you can't seem to let it go.
_I_ can't seem to let go? LOL. The conversation is about the fact that _you_ can't seem to let go of the subject that you have beat to death, and as a conclusion you say that _I_ can't seem to let go?
I visit this thread a lot less because I just don't have the time to go through all your QS posts, and I can't seem to let go?
I have to give you a compliment for your attempt to twist words and meanings around, but I am really tired of these games. I just want the info / opinions that will hopefully help me make money, not word games.
Joe
wbmw,
I have to say that I am dissapointed in both AMD and Intel dispensing the faster FSBs in a nickle and dime fashion, rather than infrequent leaps.
Joe
Keith,
Interesting info from Tech Tour. I would really love for 2 GHz A64 to exceed my expectations and beat P4 3.2 soundly.
But, I think the most important info has to do with Windows XP AMD64 desktop. Microsoft hasn't said much about it officially. Windows XP in 2H 2003 would be huge. It would certainly more than make up for the info coming out yesterday, that AMD64 version of Server Windows will be available only in Entrprise version, not Standard.
Joe
Keith,
Yes, you are right. I don't think I posted about that subject much or at all on IHub. Mainly on SI.
Anyway, Nforce3 will give us a good opportunity to compare how much performance gets left on the table by choosing the single channel route.
Or more precisely, motherboards based on nForce3, because the memory support is between the CPU, mobo traces, DIMM chips and BIOS. The chipset doesn't get involved.
It would be nice to get support for unbuffered memory in Opteron 1xx. I don't know if it is possible at all with Opteron. Maybe I should read the BIOS writers guide to find out ...
As far as comparison with single channel, for that we will have to wait 3 more months until the releas of A64. Pete posted on SI that Opteron can be used with DIMM filled in only 1 memory, but I am not sure if it is the case. I have not seen any other comments on that either on BBSs or in reviews.
Joe
Keith,
That's an excellent news!. Wow, I am impresed with the machine. I wonder what you get in computer stores in Germany, if the PC sold in supermarket is at the top or near the top on performance.
I would certainly trade it for my current PC. I am not quite sure about the price, since they mention 1000 Euro and 1899 Euro.
Since they use 3200, it means that AMD is getting a good volume of it. Also, I am surprised that nVidia moved fairly quickly to supply the 5900 card. I believe this one has doubles the width of the memory path over the first GeForceFX cards. nVidia is certainly getting good revenue from this machine, since it also has the nForce-2 400, and Asus A7N8X has the high end southbridge of nForce-2, with excellent sound.
Joe
wbmw,
I think people are very interested to hear about prospects of Athlon 64, but you can't compare the volume of my posts about my pet-peeve of A64 having only 1 x PC-3200 rathere than 2 x 3200 unbuffered with the volume of your posts about QS. The ratio is probably 10:1.
Joe