PDA

View Full Version : ASIO buffer in relation to CPU



Mandarin Man
06-05-2006, 09:44 AM
Something I'm wondering..

The smaller the buffer, the lower the latency, the bigger the strain on the CPU.

Yet I noticed that when I set my audiocard to it's smallest buffer (64 samples) and then play a high-polyphony VI, it starts cracking at a certain polyphony count, eventhough my CPU is at 50%. Of course when increasing the ASIO buffer this problem goes away. Does this mean that I'm just running into the hardware/driver limitations of my audiocard when setting it to it's smallest buffer?

Thanks,

Roy

wja
06-05-2006, 12:00 PM
Something I'm wondering..

The smaller the buffer, the lower the latency, the bigger the strain on the CPU.

Yet I noticed that when I set my audiocard to it's smallest buffer (64 samples) and then play a high-polyphony VI, it starts cracking at a certain polyphony count, eventhough my CPU is at 50%. Of course when increasing the ASIO buffer this problem goes away. Does this mean that I'm just running into the hardware/driver limitations of my audiocard when setting it to it's smallest buffer?

Thanks,

Roy
Yes, pretty much. You should set it for about 11-12 milliseconds of latency, which is hardly noticeable and will reduce the strain on your soundcard and cpu. If you are not playing live and just doing midi playback of samples through your sequencer, you can set it even higher.

Mandarin Man
06-05-2006, 01:46 PM
Thanks wja.

Does anyone here ever change the buffer while working on a project? E.g. lowering it for tight recording and increasing it to playback high-polyphony sequences.. Or do you generally keep it at a certain buffer that works for most recording and playback situations, e.g. 512 samples?

andreas
06-05-2006, 03:37 PM
Thanks wja.

Does anyone here ever change the buffer while working on a project? E.g. lowering it for tight recording and increasing it to playback high-polyphony sequences.


I almost always work that way. I usually start out at 128 or 256 samples while I'm playing in parts, and then as I add plugins and processing, I increase the buffer size as needed. By the time I do my final mix, I might be at 1024 or even higher.
Of course, you need a DAW that has automatic plugin delay compensation to be able to work this way. Otherwise you'd have to keep delaying and shifting tracks around manually as you increased the buffer size. I use Nuendo, which has PDC, and also lets you freeze any audio/VSTi track. If I need to play in a part at a late stage in the project, when I'm at a higher buffer size, I'll start freezing tracks until I can lower it back down to something reasonable, like 256. If I'm unable to get that low, what I'll sometimes do is bounce out a 2-mix of the project, open it in an empty sequence (with a low buffer size), record my part to that, and then import that part/track back into my main sequence. A bit cumbersome, but it works in a pinch.