HPCC 2008: Steve Wheat's talk

I’m at the 22nd annual HPCC Conference in Rhode Island over the next two days. I’m chairing the session today so I can’t spend a lot of time posting or even post a nugget for each speaker, but as particular items strike me I’ll post them here.

Intel logoIntel’s Stephen Wheat just got done talking. He referenced Pat Geslinger’s upcoming HPC strategy brief in Shanghai, and Intel’s “rolling thunder” strategy of discussions and peeks at the strategy leading up to what one supposes will be a significant announcement at SC08.

At the end of his talk he touched on something near and dear to my heart: acquisitions, and the degree to which the federal government’s current acquisition model doesn’t lend long term support to HPC. My take on this is that this lack of support means lack of innovation directed at our needs; if the small/mid end HPC market ever does coalesce and start buying equipment in droves, then they’ll determine what products the high end has available to cobble together into large machines.

Anyway, Steve closed with some good recommendations on how we might fix this. I didn’t take great notes on this slide (wish I had), but I remember two solutions: offer tax credits for R&D in HPC as has been done in other industries, and adopting a runner-up model.

The runner-up model is intriguing: rather than creating winner-take-all acquisitions, add provisions for a portion of the award to go to the second (or even third) place winners. In the raw this idea is a little to socialist for my tastes, but expanding on this idea, you could even have a large acquisition that puts says 80% of the money in a typical low cost procurement and reserve 20% for a high risk, uniquely innovative, solution that has a prayer of preserving innovation in the market.

Whatever path is adopted, I do think that change is needed, and is in fact long overdue.

Resource Links: