Va. Tech Cluster Gaining In Teraflops!
From The MacObserver
More here.
Quote:
In benchmark tests last week, the Virginia Tech supercomputer, powered by 2,200 IBM microprocessors, could compute at 7.41 trillion floating point operations (teraflops) a second. That would place it fourth in the list of fastest supercomputers in the world, according to The New York Times, which first reported the story.
But by Wednesday evening, when Varadarajan spoke to this correspondent, the machine had been powered up to 8.164 teraflops, pushing it up to a possible third place. "And it's only now operating at perhaps 50 per cent efficiency ", Varadarajan said in the interview.
In benchmark tests last week, the Virginia Tech supercomputer, powered by 2,200 IBM microprocessors, could compute at 7.41 trillion floating point operations (teraflops) a second. That would place it fourth in the list of fastest supercomputers in the world, according to The New York Times, which first reported the story.
But by Wednesday evening, when Varadarajan spoke to this correspondent, the machine had been powered up to 8.164 teraflops, pushing it up to a possible third place. "And it's only now operating at perhaps 50 per cent efficiency ", Varadarajan said in the interview.
More here.
Comments
i was thinking those initial numbers (17 Gflops) were a bit high, but any improvements are always welcome.
here's to hoping they keep chipping away at the ineffecient bits and pieces.
Project Web Site
BEN
http://www.netlib.org/benchmark/performance.pdf
Check page 53.
Varadarajan reported that "our latest numbers are 9.555 tera and we still have more tricks left. We are hoping for another 10 percent boost to become the first academic machine to cross 10 tera. The last ratings put us at number three worldwide." During the question-and-answer period at the end, an audience member from the Lawrence Livermore National Laboratory introduced himself as coming from the institution that had the Supercomputer that the Virginia Tech cluster had just passed. He asked whether the details of the Supercomputer would be published. The reply was that in addition to documentation and papers, the plans are to return the changes to MVAPICH to the open source project so that it would be freely available. There are also plans to open source the caching code and Varadarajan expects that Mellanox's code will be available.
Varadarajan said that they are getting requests for clones. "Expect to see a lot more G5 clusters."
Complete article here.
Wooo hoo. Go Apple. You go!
Originally posted by MacsRGood4U
Complete article here.
Wow. I had heard they were running a Beta release of Panther, but to find out they're getting this kind of performance running Jaguar is impressive. Can't wait to see how Panther boosts the numbers on this baby.
Varadarajan said Apple provided significant technical help and gave Virginia Tech some of the first G5s off the production line, but the college paid full price for the machines, which cost $3,000 apiece.
Anyway...If I could use that cluster to render my own reel....that would be awesome
Originally posted by alcimedes
lol, only Apple would have a customer ordering 1,100 machines pay full price for them.
Geez, not even an edu discount. Va. Tech would have to qualify as a edu customer, I'd think.
Who cares they went full price? It was cheaper than a Xeon cluster that would do the same performance. And the help from Cupertino was probably thrown in as a 'freebie'.
What's amazing is that a Mac cluster is 3rd. Who'd'ave thunk it three years ago? This is the begging of a great era.
Originally posted by bauman
What's amazing is that a Mac cluster is 3rd. Who'd'ave thunk it three years ago? This is the begging of a great era.
It'd be third on the June list. We don't know where it will rank on the next list. We also don't know which number they are going to use. If the October 1 date was a hard deadline for submitting results, then the 9555 GFLOPS number is not going to make the list...not even the ~8000 GFLOPS number.
Originally posted by MacsRGood4U
Where does it say October 1st was a hard deadline? The Report I posted above was dated October 25. I read elsewhere that the final list will be published in mid-November.
The date of the list's unveiling has no relation to when official results must be submitted. External articles written about benchmars outside potential deadlines are also of no significance.
I didn't say for sure that October 1 was a deadline for submitting results, but it could be. In such a case, the numbers posted here wouldn't be used for the rankings. Get it?