Page 1 of 1

High Cpu use in BD in Beta 0.9.5b for PC

Posted: Wed Dec 13, 2006 9:49 pm
by Motiatronix
Hi D16

Sorry but problem with high cpu use with just BD playing has returned...if I set just 4 hits for eg (dosen't seem to matter what instrument settings). It is not near as bad as the denormal problem in an earlier beta but get it swinging between 20%- 70% cpu use,usually settling if it does around 70% ...but it rarely spikes.

It's reproducible.

Everything else no problem.

This was not a problem in 0.9.4b.

BTW:

Thanks for all the new features:shuffle,flam,total accent per pattern are working superb,so many nice syncopations!!! to be had from pattern to pattern if wanted/needed.....mute/solo behaviour with automation is a hoot! and gives some nice gated stuff etc.Even switching internal sync on and off Drumazon syncs in/out with host no problem.Not to mention that great new instrument preset browser and having saved/updated knob states.Great stuff!


Regards,

Thanks.


P4 1700 Mobile
512 Meg Ram
Gforce 420 Go 32 meg Ram
Internal Sound Card
Asio for all driver at around 5ms
Win Xp Home
Tracktion 2

Posted: Wed Dec 13, 2006 10:30 pm
by Jacek@d16
It's fixed, so wait until next beta :)

Posted: Wed Dec 13, 2006 10:58 pm
by Motiatronix
Fantastico! :D

regards

Posted: Wed Dec 13, 2006 11:30 pm
by usrfobiwan
Thats weird no high cpu usage here just 3 bars in SX3 monitor. My system:

AMD A64 X2 4800
4GB mem
Asus A8N5X mobo
Sata drives
OS: XP Stripped for DAW version

Soundcard: GadgetLabs 824 (6ms latency)

Posted: Thu Dec 14, 2006 12:49 am
by Jacek@d16
It's because AMD has a better (faster) coprocessor handling for numbers denormalisation than Intel

Posted: Thu Dec 14, 2006 1:15 am
by Motiatronix
Hi usrfobiwan

It's possible although (I can't be sure) it is still a denormal related problem.I'm no expert but what I know is the math goes crazy in the Intel cpu in some way with the calcs but Amds resolve the problem very efficiently (again not sure if denormals is still a prob with the Amd FX2 architecture with respect to earlier models ),so may not be noticeable.

If it is related to this I had similar thought to you the 1st time it happened.....I was using my main core 2 duo which I had no problem with then was using same beta on this "scatchpad" laptop.After D16 explained what it was I realised Core 2 is not P4 architecture and I believe denormals is not a problem for these new chips anymore.

Perhaps somebody else can explain properly and way better than my rudimentary understanding of what happens in this scenario and has the knowledge.

I reckon I'd keep testing on this little system as more than likely a lot of people still using the P4 to show maybe if more problems with denormals.D16 also mentioned that they were also aware of others and or on their case to squash or have already squashed.It's a bit of a dinosaur compared to whats around these days :wink:

Posted: Thu Dec 14, 2006 1:18 am
by Motiatronix
Just writing my last post and posted and saw Jaceks reply.

Posted: Thu Dec 14, 2006 1:30 am
by Jacek@d16
Denormalsation occur when after some calculations You get an equal less than 10e-24. Don't know what exaclty P4 is trying to do with this equal (denormalisation ofcourse, but it's simplier to put 0.0 :D), but i know that it takes much more than a lot of FPP's cycles :)