Based on those numbers the Intel one actually looks faster... look at the amount of cores the intel has vs the amd.... .. I wish AMD would stop hyping people keep there trap shut and release stuff on time. Because intel knows where amd is at.

Now I know what computers they use in the CSI shows labs.

While Intel does perform better per core, about 9.5 Gflops per core compared to AMD's 8.5 Gflops the real kicker is the performance per core.

AMD comes out to 19.16 Wh per core while Intel sucks up 40.82 Wh per core. So, for a 10.5% improvement per core in performance Intel uses up over 100% more energy. Honestly I'd say that's pretty impressive of both the AMD and Cell architecture.

Rich,

Before harping on about AMD’s issues and the fact that the Intel systems are more powerful, I think you should really take a look at the design of the processors and the super computing systems to see how Intel and AMD differ. From reading an article here a number of months ago, if my memory serves me correctly the AMD chips are providing the interconnecting communication backbone, rather than the sheer processing power – it is rather the PowerXCell processors that are actually crunching the numbers.
This super computing system harnesses the strengths of both the different processing packages (the Opertons and the Cell Broadband Engine) and combines them to provide a hybrid system that has proven is superior in many ways to Intels current offerings (power consumption, inter process package communication bandwidth to name only a couple).
The hyper transport interconnects offered by the Opertons are being exploited to provide suitable bandwidth for inter processor communications and the PowerXCell processes and their specialist processing cores are providing the number crunching power.
Rather than sit there and grumble about a company who experienced an issue with TLB, please just try to look at the engineering prowess of the super computer itself, the engineers who envisioned it and the skills for making it a reality.
Yes this system is a hybrid, but a very powerful one which may turn out to be a template for future systems, if this is the case, I personally feel AMD currently has the lead on Intel with its Hyper Transport communication system, in comparison to Intels new offerings.

This looks rosy for AMD now, but supercomputers take years to design and deploy. In one year I bet AMD has four in the top 10 and in two years they have one or two.

The way to go

I love that AMD hybrid is kicking butt and I agree with John, but it should be interesting to see the effect of the new i7 arch and then compare. Intel was getting its butt handed to it as the prior posters have alluded to due to some the FB-DIMM and other arch inefficiencies. I still think AMD is rockin, but the next few years over all (AMD 45nm with new hybrid C/Gpus) and the new TICK from Intel (don't know what they are really changing) should create very new and very powerful computers.

UT didn't get their results in by the cut off date. So the results for Number 6 are nearly a year old. Ranger has switched from 2.0 Barcelona's to 2.3 Barcelona's and is now 504 Tflops. http://www.utexas.edu/features/2008/ranger/ That would put it at Number 3. The more interesting rankings will be the DARPA HPC Awards which have replaced Linpack since it isn't difficult enough any more, its 32 bit. I would not be surprised to see AMD win all twelve awards. If you are serious about computing those are the awards that matter. Blue GeneL is the only non AMD machine that has a chance. http://www.hpcchallenge.org/ Right now Intel is 0 for 84 in HPC awards, AMD will probably be 21 of 84 with the rest held by IBM. The power ratings are a little misleading since all you really care about is how long it takes. On the HPC benchmarks where performance is really tested a run takes about a week. You can see the real difference. You have to weigh the time it takes as part of the equation. Ames uses about 30% more kwh per run than Ranger. On these machines that cam be $10 million per year per installation. The power ratings is kind of a squirrelly number since some include the AC and some don't.

Sweet.

Its not just put 100 processors together to get 100 more power, your have many issues on this kind of architecture...

Its not so easy (memory management, hazards, etc).

So, I completely disagree your comments that Intel is more processor performance efficient because you have to look not only to processor but to solution.

rise the prices per processor make the enterprise market less look for it

3 SGI Altix ICE 8200EX51,200487 2090 Intel
It has only one third of the cores of the second place bus just half the performance. Has one quarter of the cores of 4th place but has about 10% more power and consume lees energy. So those intels really kick asses, is it? Can same one has a explanation for that?