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.
fyodor, I wasn't aware the Opteron's cache was lower density than XP's. Where did you get that?
Paul
PS. Welcome to the board, by the way!
Tenchu, I was actually referring to the physical differences between the caches that allow Intel's to be smaller per unit of memory, and whatever other physical differences. I was already aware of the inclusive/exclusive difference.
Paul
Guys, at the risk of flaming arrows coming my way, couldn't yb perhaps have a point about Dell? Granted his depiction is simplistic beyond reason, but isn't it true that when it comes to development capabilities and the "brain trust" factor, that Dell is more like an efficient box builder and very capable at contracting out for designs, than a world class firm having the "from the ground up" engineering design capabilities of of an IBM, HP or even Sun or maybe Intel? I'd include AMD, but though they've got great folks, they're more narrowly "expertised" and focused than the rest. I, for one, have never been real excited about the possibility of Dell for AMD. They'd have to be a lot bigger to be anything more than solely Dell's slave, and the margins would likely be less than what the rest of the market will provide.
Paul
Tenchu, what can you tell us about the differences between Intel's and AMD's implementation of L2 cache?
New stepping? TbredA-B type repeat? Higher Freq/lower voltage.
http://www.theinquirer.net/?article=9152
subzero, you still didn't address my question, so I'll give up after this. In addition to what I've said before, the Opteron also has a memory controller working at full processor speed. If you want to find out whether or not there are power gains from SOI, try comparing apples to apples. You'd need to know either what an Athlon on SOI would do, or an Opteron on bulk, which we don't. Perhaps someone knows the gains IBM got with bulk vs. SOI for a similar clocked/architecture chip?
Paul
subzero, 5% accounted for the 64-bit part only. It had nothing to do with the extra 768K of cache compared to the Athlon of the time!
Paul
EP, OTOH, since the less dense cache is a disadvantage they've been used to dealing with, and the integrated stuff eases board design for manufacturers, and taking into account the increased margins Opteron will allow, I suspect AMD will do just quite well, thank you, even with the perceived disadvantages. Like you've said, now it's just a matter of whether or not they can make enough of them. And whether they can sufficiently boost the frequency to make it the unquestioned leader.
Paul
wbmw, that's why i asked here, I had trouble finding any info elsewhere; I knew y'all would know.
chipguy, 1) are we talking processor power or system power?
2)according to geek.com, Banias die size is 100mm^2. Do you have an Intel doc that says otherwise? I couldn't find it. In any case we already knew that Intel's cache is denser than AMD's, didn't we? Add that factor to the integrated memory controller & aHT circuitry and its size looks a lot more reasonable.
http://www.geek.com/procspec/intel/banias.htm
The only valid point you made here is speed. Why would you expect necessarily lower power for a chip with 1Meg of cache as opposed to 256K? Why is 193mm^2 such an abomination for a chip with 1Meg of cache? Get real.
Paul
sgolds, rumor had it that the tier 1 OEM, which wound up being IBM, didn't announce actual shipping product at the launch because they hadn't completed work on their own motherboard yet.
Paul
Customer sources said that AMD has told them to expect a top-tier OEM vendor to be on hand at the launch to talk about the Opteron's advantages. However, the potential customer will not announce a design win, as it has not finished its own board design.
http://www.extremetech.com/article2/0,3973,1034704,00.asp
I suspect a lot of companies will find this attractive.
Paul
http://racksaver.com/benchmarks/TPC-info.asp
Well, if they can ramp up the frequency, this is going to be KILLER! Multiprocessor systems already look great, more speed will allow killer workstations and gaming machines. From the rumors going around about it being a killer gaming setup, I'm hoping the AGP issues have been worked out.
I haven't had time to sit and digest Johan's review yet, but things looked great at a glance.
Paul
WOW, wbmw, are you impressed?!
aceshardware's review is up.
Translate fully for us, BUGGI1000? I don't want to get anything wrong.
IBM will have 2P and 4P units. Software and Hardware partner.
http://www.heise.de/newsticker/data/ciw-22.04.03-001/
Short SuSE 8.0 / Opteron review. Sounds smooth and "wicked fast".
Paul
http://newsforge.com/newsforge/03/04/21/1914216.shtml?tid=7
Simple. They didn't. This page shows from the System Properties that it was Trident Blade 3D. You don't think that could have any relevance, do you? <:0
Nah! ;p
Paul
http://www.tomshardware.com/cpu/20030422/opteron-11.html
Hans DeVries is convinced of Yamhill's existence.
http://www.chip-architect.com/news/2003_04_20_Looking_at_Intels_Prescott_part2.html
Brookwood(Insight64)likes Opteron after previewing it, says it outperforms Xeon and is close to Itanium for much less.
"One investment bank already ordered about 1,000 Opteron-based servers for its bond trading desks, to replace older Sun Microsystems and Compaq servers," Wong said. Wong is with Einux.
http://www.sfgate.com/cgi-bin/article.cgi?f=/c/a/2003/04/21/BU102975.DTL
Tenchu, I was referring to situations where system memory is not accessed, which is what I meant by (core).
Elmer, you can also tell from this image that it is socket754, therefore A64, therefore it can't ever have 2-channel memory.
http://www.xbitlabs.com/articles/cpu/display/athlon64_4.html
sgolds, are you sure about that? I thought that things had changed and the "Paris" variant was the one that would have 256K, but the earlier released ones would have 1Meg.
Paul
But would floating gates' speed be helped by floating loans? ;)
One of the glaring issues is that SOI hasn't bought AMD anything...
...yet.
Nah, CJ, it'll be 1.4GHz in December, just in time to miss Christmas! ;P
I don't get what all the negativity is about. @ 5% improvement(core) over equal clocked XP was expected, right? The memory controller seems to do quite well, although there are some substantial inconsistencies in the review re: memory, so let's see what Tuesday brings.
The PR rating is questionable, but that's not likely to be the final form anyhow. Seems to me Intel had a similar situation when the P4 first came out. Anyone think a 1.4GHz Willamette would hold a candle to a 1.4GHz PIII? The overall architectures are different and comparisons can be tricky. I'm confident it'll be worked out adequately.
My only "concern" is whether they'll be able to ramp up speed adequately. If they do, the "sky's the limit"! If they don't, look out below! Frankly, they have too much riding on this not to figure it out and make it work one way or another.
Paul
wbmw, not if they're planning rapid frequency ramp-up. ;)
But Fujitsu-Siemens and NEC aren't considered tier 1, are they? Which takes us back to whom? IBM, Dell, HPaQ? One would think that IBM would use the Newisys design, since they have such a heavy personnel influence there(I didn't realize that some came in from Dell as well, until I read a recent article). Sun, maybe? The Newisys board wouldn't be suitable for blade servers, would it? How exactly is a blade server defined, anyhow?
Things are clear as mud right now.
A few notes on Opteron from extremetech.com article
1) Though TDP is @80W, Hester says reality is closer to 40W, leaving lots of headroom for higher clocked CPUs.
2) Avnet says stock of the processors is sufficient to support launch and increased demand through the quarter.
3) Though it appears there won't be a tier1 at launch, it appears to be because they haven't completed their proprietary board design yet. Brookwood expects that landing a tier1 will add about one percentage point to the 2 points he expects in increased market share(annual increase) that the tier2s and white boxers will bring.
4) Einux says demand for Opteron is 8-10 times greater than for Athon MP.
5) nVidia Crush K8 should be shipping in volume in June. Workstations, anyone?
Tuesday looks to be an interesting day.
Paul
http://www.extremetech.com/article2/0,3973,1034626,00.asp
A couple of comments about the A64 preview.
1) VIA's first chipset implementations always stink. The "A" versions usually address this.
2) AGP is still handled by the "chipset".
3) Boot screen and BIOS state Mem clock as 200MHz. If you look at the Sandra screen on page 6, something doesn't seem to jibe. If it were DDR400(200MHz clock), shouldn't the throughput be @ 3200MB/s, especially since they mention the incredible efficiency? 1600MB/s sounds like PC1600, which is DDR200(100MHz clock). Something ain't right here.
4) I wouldn't expect it to do so well on benchmarks that fit in all the processors L1/L2 caches. After all, there clock rate would be a huge factor, just like the higher clocked Tbreds beating the lower clocked Bartons with the same PR.
I believe I'll reserve judgement until the real reviews come out.
I am encouraged that they claimed that DDR266/333/400 were supported, but there are too many inconsistencies in this preview for my taste.
Paul
sgolds, precede the graphic URL with "[ chart ]" and follow it with "[ / chart ]" (without quotes or spaces, of course). Oh, I believe you have to lop off the "http://" as well.
Paul
Don't gloat too hard, guys. It'll come back to bite you. Can you say ES? Things will change substantially.
Paul
AMD has definitely crerated closer ties in their relationship with Avnet. I first noticed this yesterday.
Paul
https://aac.avnet.com/AMDstore/
Contrary to what a lot of the hand-wringing the analysts are doing would lead one to expect, AMD seems to be holding its own quite well today, thank you kindly.
Paul
wbmw, doesn't sound like it's necessarily quite as far out as you think, since 193nm may be extended to use as low as 45nm according to this article.
Paul
http://www.siliconstrategies.com/story/OEG20030227S0068
neye, I think it's pretty clear further on down the thread that he did the benchmarking(between that and the earlier thread below)
http://www.aceshardware.com/forum?read=95028880
Great line re: Itanium ;)
Frome the Register http://www.theregister.co.uk/content/61/30274.html
"So far, the roll-out of Intel's Itanium appears to resemble the US airline industry - lots of pilots but little real business being done."
Thanks, Tenchu!