GPU (CUDA CORES) FOR DSP processing?

Serious? of course I’m being serious, I’ve been doing this professionally for 30 years.

How many convolution reverbs do you need in your project? convolution reverbs and FFT uses aren’t the things that cause major CPU usage in my projects, algorithmic reverbs DO take a huge CPU hit and they can’t be parellized successfully.

with regards to IIR filters I was just giving you an example of another important aspect of Audio processing that can’t be parellized either, what isn’t serious about pointing this out?

Also as soon as you off load convolution to your GPU you loose Zero latency. Nebula for example uses convolution and as such can be run on the GPU but the latency issues make it unworkable in most situations.




MC