Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Petz who wrote (215164)	10/27/2006 6:39:28 AM
From: j3pflynn	Respond to of 275872

Petz that GPU SP issue is about to change, though. At least for ATi. Can't remember where I read it yesterday, but Orton said they were planning DP capability.

To: Petz who wrote (215164)	10/27/2006 11:55:06 AM
From: UpNDown	Respond to of 275872

Petz, on single precision A lot of scientific work can be done roughly in single precision, then the result can be tuned up in double. Matrix inversion, root finding and eigenvalue solution come to mind. For instance, for matrix inversion, find the inverse in single precision, then do a Newton-Raphson iteration in double precision to tune it up. Single precision is also good for many randomized algorithms, Monte Carlo simulation, etc. AMD already has their math library out; that can be extended to make calls to GPU or GPU-like functionality without requiring change from the application programmer. Just buy and install the care (and a new copy of the library) and many things magically speed up. Of course, specialized calculations could also be customized for the GPU calculations.