On Tue, Nov 27, 2012 at 04:14:06PM +0100, Bogdan Costescu wrote: > On Tue, Nov 27, 2012 at 3:40 PM, Eugen Leitl <eu...@leitl.org> wrote: > > Please do not top-post and please trim your replies. > > Thanks for the reminder ! It bugs me too... > > > I would be > > looking at least for an Exynos 5... > > ... to process code locally ... > > I'm still not convinced about computational capabilities of ARM cores.
Me neither, but there's always hope that http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone will deliver something useful early next year. 32 bit floats are quite enough for me. > So how about letting the ARM core deal with I/O and adding a dedicated > computing unit in form of FPGA ? Not suggesting a (GP)GPU to keep in That would be great, but no longer woud be COTS. > line with the low power envelope. Yes, this is basically not something most supercomputer users would need. It's a dedicated system to store and reprocess a very large voxelset. It's basically making a sea of smarter spindles. Their result would be processable on a classical cluster, though. _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf