16 Oct
2004
16 Oct
'04
1:39 a.m.
On Oct 15, 2004, at 3:49 PM, siggmuss wrote:
Is this possible with pd?
Anyone what know the latency of a PCI-Express CPU->GPU->CPU transaction is? I know that it's highly unlikely to do with reasonable block sizes over AGP because the read is so damn slow. Offline rendering would be a no-brainer for things like huge convolutions, FFT and FIR operations, but what about real-time low latency situations?
cgc