Jeremy Blosser wrote : > Subject: [Prime] Double Precision finally coming to the GPUs? > > I was reading a review of the GeForce 200 series GPUs coming > soon and saw there was support for the IEEE-754R double-precision > floating-point standard. So is this something where we can > actually get a port of prime95 or mprime to the CUDA architecture? > > I haven't looked at the details of how they are doing the double > precision (is it baked into the hardware or emulated using single > precision for example), but if the performance increase for > double-precision operations scales linearly with the FLOPs numbers, > then we are looking at the GTX280 GPU being ~10x faster than a > quad-core Penryn processor. > > Anyone else following the GPGPU applications care to chime in? > > -Jeremy
Martin wrote : > Why is it not useful at single precision? > > A GPU should give a tremendous speedup! > > Cheers, > Martin All this is discussed extensively at http://www.mersenneforum.org/showthread.php?t=10275 Jacob Visser _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
