Testing..testing....HD_process vs HD_accel vs Dual opteron

[guerillamixer]

I quick test non-scientific for some of the plug-ins I generally use in
PT 7.1cs4.
I don't have an Accel card, so the numbers are based on the instance
per chip given by dutchmuzik on the DUC

http://duc.digidesign.com/showflat.p...ev=#Post860557

or in the product manuals.


I based it on the plug-in count per card (not chip) vs RTAS running
@80-90% . To make sure the RTAS load wasn't crippling the system I tried
recording a bit of audio, then playback. The playback engine was 96 voices,
using 2 cpus and cpu usage at 90%.All tracks are mono.



Sony oxford eq +filters

process card 18 /accel card 46 /dual 246 180 @ 80%cpu

MCDSP filterbank E6

process card 81/ accel card 144/ dual 246 355 @ 80%cpu

McDSP Compreessor Bank CB3

process card 27/ accel card 69/ dual 246 220 @ 90%cpu

Analog Ch 2

process card 81/ accel card 144/ dual 246 355 @ 80%cpu

D-verb

process card 36/ accel card 38/ dual 246 60 @ 90%cpu

Digidesign Dynamics III comp/limit

process card 63/accel card 162/ dual 246 140@90%cpu

Mod delay II extra long

process card 18/ accel card 36/ dual 246 380@80%cpu


Some Observations.

There are some real gains in efficiency when using multishell
TDM plug-ins, as PT allocates then very effectively. For example
the Sony Oxford eq/filters uses about 80% of a chip for two
instances, leaving 20% that could be shared with other plugs.

RTAS plug-ins instantiate much quicker than TDM after the first
instance.

[justicetones]

very interesting.........