SAW DAW Performance Puzzle

[DennisC]

Greetings,
I have had SAW for over a year but am just now in a major project with it. I am having an unexpected result on my PC's performance on all .EDLs and one in particular. Computer specs will be listed at the end of this post.

I have about ten .EDLs with simple "scratch tracks" of click, acoustic gtr, and voice on each. A couple have a few more tracks.

General Performance Puzzle:
With three tracks playing and using channel EQ and Compression on two of them I get a MT LOAD of 6% with the SRC Load at 0%. Each instance of a SAW Reverb adds 3% to the MT LOAD. Each instance of a Timeworks DX Pluggin adds 5% to the MT LOAD. This seems very high.

One Particular Puzzle:
5 tracks playing, 4 of them using channel EQ and Comp, 2 tracks using three pluggins, 2 tracks using 1 pluggin --- MT LOAD 68%! When I change the out buffers from 256 to 128 it goes to 78% MT LOAD!

These MT LOADS seem high. The system config and steps taken so far:

Asus P4P800SE mobo
2 Gig RAM
Pentium M 1.6 (using Asus socket 479 adapter on mobo)
Drives - SAMSUNG Spinpoints 80 and 120 GB
Windows Home SP1
RME FireFace800 - configured with buffers at 128 on FF
SAW configured using MME out buffers 4 x 256
Matrox G550 - dual function disabled

Bob's Windows XP Tweaks done
PassMark testing software used and:
Hard drives perform excellent
CPU performs OK -- ~same as a P4 1.8-2.0

Ideas?

Thanks,
Dennis

[Bruce Callaway]

Hi, those figures definitely look high to me. Probably needs Bob's input. One question, has this been the trend all the time or has it only just occurred.

[DennisC]

Bruce,
I can not say with confidence.

One more piece of data on the One PArticular Puzzle - with only a single track selected for playback - the MT LOAD is at 28% - each additional track selected adds ~2% to the MT LOAD. Each use of EQ/Comp adds another 1-2%. All while the SRC LOAD stays a 0%.

Thanks,
Dennis

[SoundSuite]

Is the system getting hot and the CPU slowing down, by chance?

[Bob L]

The first numbers seem fine... it seems that on the one puzzle edl, you might have some plugin patched on the output track... there is no reason why 1 track should give you 28% unless that's going thru a cpu hog of a plugin.

Bob L

[mOjO Fet]

Are your drives defragmented?

I once had a high load because of that.

Michael

[jarvissound]

Bob, isn't there something about processing blank tracks with plugins that eats CPU power. I seem to remember having a similar strangeness awhile back (a few years ago) where I was eating up my system resources and, based on your input, I tracked it down to plugins that I had on tracks that had basically nothing on them. The fix was to bypass the plugin in automation until I needed it and then popping it in when I did need it. That made a huge difference in my MT loads. If I remember, SAW native plugins didn't have this problem but Direct X ones did or something.

FWIW,
Brett

[Bob L]

Some plugins that use floating point processing can bog down in an endless loop when the floating point procesor sees data that approaches zero... you have to program specifically to get around that issue... many don't.

Bob L

[DennisC]

Bob was correct with the possible plugin(s) on an output track. My buddy, whom I am trying to lure from the dark side of Nuendo (just kidding, please no flames), and get him to look at SAW, has been teaming with me on this new project. Actually, he is a much more experienced an engineer than me.

Anyway, we have been sharing the engineering duties. He also has brought in some of his favorite Timeworks plugins which he knows well. He had patched two across the output buss and I didn't know it. When I took them off the 28% dropped down to 3%.

These Timeworks plugins seem to eat a LOT of CPU. Anyone else notice that?

Thank you Bob for your help. Thanks to all who helped as well.

Dennis