Here it's fine, results on timedemo on a short demo of spam in a pub with a default config on my client:
(pollofeliz, /timedemo should ignore com_maxfps)
/*
Win64:
Bumpy:
dmaHD: 188.5-189.4
dmaEX: 195.9-195.7
(oops, I was accidentally running the Win64 client for this, but let's go with it too anyway)
Ikalizer: Not Applicable (Not available in Win64)
Non-bumpy:
dmaHD: 172.1-171.8
dmaEX: 178.0-177.0
Ikalizer: Not Applicable
Win32:
Bumpy:
dmaHD: 213.9-212.7
(differences in Win64 and Win32 numbers are normal since Win64 is in more need of x86_64 code related optimizations)
dmaEX: 221.2-220.9
Ikalizer: 202.2-202.8
Non-bumpy:
dmaHD: 194.4-194.8
dmaEX: 201.4-200.3
Ikalizer: 185.7-185.5
(two results are there for two /timedemo runs)
*/
The system is CPU bottlenecked (in this setup and nvidia effects). This is important, since if you try to benchmark a sound system on a GPU bottlenecked system it's at least partly pointless since the CPU may be able to cope up to a point but for at least some range of FPS the GPU keeps it down.
e.g. you may have on ALL systems 50 FPS because the GPU can't go above it but the CPU can handle all three systems fine.
pollofeliz, since you tested on 2 different clients, it may had been the optimizations of building.
If I get on a dual core lower FPS on Ikalizer (which uses SMP) on its default settings, it means not only it's heavier, it's a lot heavier.
--
After profiling dmaHD_PaintChannelFrom16() ranks very high on CPU usage. I assume it's normal but pointing it out here in case it's not.
It takes 14-15% of CPU time and ranks 1st but it's on a 200FPS timedemo (spam of sounds).
This post has been edited by mitsubishi: 13 November 2010 - 05:33 AM