Re:Audio processesing using the GPU
on
GPU Gems 2
·
· Score: 3, Informative
Some audio cards already use DSP chips developed for graphics processing (notably the UAD-1 DSP card,) but the main problem with using GPUs sitting on an video card to process audio is that the video card doesn't have any way of sending the processed data back to the rest of the system at a useful rate. So you can process audio at 40+ GFLOPS, but you can't send it back to your sequencer software and/or audio card. If/once video card manufacturers hack some sort of interface to send data back to the system, then things are going to get pretty interesting for sure.
And that's exactly why it's being used in several top SuperComputers/Clusters, including (but not limited to) NCSA's Tungsten, IBM BlueGene/L, LLNL's Thunder, BSC's MareNostrum and NASA/Ames' Columbia.
Count them, that's 4 out of the 5 fastest [publicly known] SuperComputing clusters.
Slashdot Mirror
Some audio cards already use DSP chips developed for graphics processing (notably the UAD-1 DSP card,) but the main problem with using GPUs sitting on an video card to process audio is that the video card doesn't have any way of sending the processed data back to the rest of the system at a useful rate. So you can process audio at 40+ GFLOPS, but you can't send it back to your sequencer software and/or audio card. If/once video card manufacturers hack some sort of interface to send data back to the system, then things are going to get pretty interesting for sure.
It's true, Linux isn't really scalable.
And that's exactly why it's being used in several top SuperComputers/Clusters, including (but not limited to) NCSA's Tungsten, IBM BlueGene/L, LLNL's Thunder, BSC's MareNostrum and NASA/Ames' Columbia.
Count them, that's 4 out of the 5 fastest [publicly known] SuperComputing clusters.
"Similarly" setup :P
It's still better than dumb people, who tend to choke all the time :P
Would you replace your (potentially hypothetical) Ferrari's motor with a Lada motor?