If true, the reports that NVIDIA has lost the deal with Oak Ridge Labs to power one of its upcoming super computers would be a huge blow to the ground NVIDIA has won in the high performance computing market.  A story over at SemiAccurate repots as follows:

We said “would have been” because word has reached us that the win is now dead and gone. Stone cold dead. Actually, the blame is being put on Fermi and it’s power use, so it might not be stone cold, maybe burnt to death by hot stones. In any case, Dear Leader is now zero for two in the ‘wins’ he touted at GDC. Fermi is massively delayed, underpowered, hot, and the stopgap GPUs are so good that Nvidia won’t push them directly. Talk about opening a can of whoop-ass!

During this fall’s GPU Technology Conference hosted by NVIDIA, the company was proudly touting the fact that this world famous computing lab would be utilizing the upcoming Fermi GPU architecture for future systems.  From the NVIDIA press release on September 30th:

SANTA CLARA, Calif. —Sep. 30, 2009—Oak Ridge National Laboratory (ORNL) announced plans today for a new supercomputer that will use NVIDIA®’s next generation CUDA™ GPU architecture, codenamed “Fermi”. Used to pursue research in areas such as energy and climate change, ORNL’s supercomputer is expected to be 10-times more powerful than today’s fastest supercomputer.

Jeff Nichols, ORNL associate lab director for Computing and Computational Sciences, joined NVIDIA co-founder and CEO Jen-Hsun Huang on stage during his keynote at NVIDIA’s GPU Technology Conference. He told the audience of 1,400 researchers and developers that “Fermi” would enable substantial scientific breakthroughs that would be impossible without the new technology.

Well, whoops.  I hope those “scientific breakthroughs” weren’t that important to human society since they are once again going to be “impossible” without Fermi. 

Has NVIDIA lost its big HPC win with Oak Ridge Labs? - Graphics Cards 2
The Fermi Architecture

We do not know what Oak Ridge has decided to use in place of the Fermi architecture and it is possible that because of power considerations they might just be waiting for another evolution of the NVIDIA architecture to address some key issues before fully adopting it.
REMEMBER THE TRIUMPHANT WIN for Fermi at the Oak Ridge National Laboratory that Nvidia heavily touted at its GTC conference keynote? The supercomputer project was just killed for power reasons. Fermi power reasons. Whoops.

That keynote ‘win’, shortly followed up by the showing of faked boards, was the highlight of an otherwise dull show, but it was used to show the potential of Fermi. Faked boards aside, putting GPUs into HPC clusters and supercomputers is what Nvidia has staked much of it’s future on. The Oak Ridge National Laboratory (ORNL) win was a massive PR statement, even if it was unlikely to net Nvidia any money directly.

These ‘wins’ tend to be ‘sold’ at breakeven or a loss when all the numbers are added up, but they provide some very compelling selling points for smaller and more lucrative corporate clusters. It was a halo that would have been used to sell many more Fermi and Quadro boards over the next few years.