I haven’t ran the CPU_BENCH file since I had my socket A 2500+ which was barely able to play it at most
rates & buffersizes, but I’m pleased to see how much things have improved
System spec:
Opteron 170 @ 2.75GHz on 1.45v
DFI Lanparty NF4 SLI-DR Expert
2x512MB Corsair XMS3200XLPT @ 275MHz (2.5-3-3-6-1T)
M-Audio Delta 1010
No Dual core optimiser installed
test.rns:
Frequency: 96KHz
Buffersize: 64-128
Latency: ‘0’-1ms
CPU Load Very jumpy from 60-99%
Very weird ‘lofimat’ sound, less apparent with 128 buffer
Buffersize: 256,384,512,768
Latency: 2-8ms
CPU Load: 92%, less jumpy
Still pops, 768 buffer is more listenable
896 buffersize hangs on playback.
Buffersize: 1024
Latency: 10ms
CPU Load: 76%
Dropouts stop after 30 seconds or so.
Buffersize: 1280
Latency: 13ms
CPU Load: 76%
No dropouts.
Buffersize: 2048
Latency: 21ms
CPU Load: 75%
So, no major improvement on CPU load going beyond 1280 buffersize at 96KHz.
Frequency: 44.1KHz
Buffersize: 64
Latency: 2ms
CPU Load: 43%
Buffersize: 256
Latency: 6ms
CPU Load: 41%
Buffersize: 2048
Latency: 47ms
CPU Load: 37%
So, as expected, not much improvement between even 64-2048 buffersizes at 44.1KHz. Very glad to not notice any dropouts.
CPU_LOAD_BENCH.rns
Frequency: 96KHz
with buffers from 64-768
Latency: '0-8’ms
CPU Load 99.9%
It plays without stopping or grinding to a halt (probably due to core management), but a ‘timestretched’ effect occurs due to dropouts. This improves as the buffer size is increased up to 768, where the majority of it plays smoothly.
Buffersize: 896
Hangs on playback once again!
Buffersize: 1024
Latency: 10ms
CPU Load: 95.5%
Rare dropouts.
Buffersize: 1280-2048
Latency: 13ms
CPU Load: 92%
No dropouts
Frequency: 44.1KHz
No dropouts with this samplerate!
Buffersize: 64
Latency: 2ms
CPU Load: 52.2%
Buffersize: 128
Latecncy: 3ms
CPU Load: 48.2%
Buffersize: 256,384, 512 & 768
Latency: 6-47ms
CPU Load 46-42%
No improvement with larger buffers, even when going as far as 2048 buffersize I only managed to get to 42.3% load.
After a fresh install of everything, I’m suprised not to notice bigger drops in CPU load at higher buffersizes. I also noticed that Renoise’s CPU load meter only represents one core or 50% of the total, even though taskmanager’s performance graphs clearly show load being distributed amongst both cores. Setting renoise’s affinity to either single CPU caused CPU_LOAD_BENCH.rns to creep up to 90.2%, and did not rectify when setting affinity to both cores again.
Some interesting results, which makes me think I won’t need an upgrade for a while