Stereo Tool
https://forums.stereotool.com/

Stereo Tool 9.38 BETA
https://forums.stereotool.com/viewtopic.php?f=14&t=27324
Page 22 of 22

Author:  hvz [ Thu Mar 26, 2020 2:52 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
Quote:
Quote:
Something is really wrong with ST after v9.38 beta038. Cpu usage is not under control anymore. beta039 cannot work on 1 core anymore and beta038 works just fine.
Again, will "dynamic speeds" be optimized since that filter uses too much CPU?
I've compared the final release build (9.40) against 9.37 and I don't see any difference in CPU usage - well, _maybe_ 9.40 uses 1% more, but it's a bit hard to see the exact numbers. Also, I've loaded an old preset and turned all "Dynamic" speeds on, and that also gave about a 1% increase in CPU usage.

What differences are you seeing?

Note: If you've configured it to use a specific core, that is now (by default) a "preferrerd" instead of "forced" core. Meaning that Windows can, if it needs to, move it somewhere else. I've done tests and this turned out to work much better, sometimes Windows starts some background tasks which are also running on a specific core with high priority, and in that case if we force things to run on that same core it needs to wait until the Windows task has finished. You can still force it to run on a specific core though, see the new "Forced" checkboxes under CPU & Latency. Not recommended though!
1. Try run beta 039 and above on 1 core. If you watch Task manager You will see windows spreads it to few other cores. "Forced" does not help, it uses core0, then after few second core4, then 6 then 8, and i set it all to core 15. Beta038, and older betas, are fine. Sure i will not use 1 core, but it is easier to see what's happening.

2. When i use "dynamic speeds" on my 7 MB bands, cpu load rises from 44% to ~68%, on any beta/release. I just found "Start Attack Immidiately" Also uses much CPU.
Weird. "Forced" works fine here, I just verified it on the 9.40 release. If I select core 1 forced for multiple threads and increase the CPU load a lot, it all stays on core 1 and starts to buffer.

I also measured Dynamics on/off for both AGC and all MB1 bands, and it increased the CPU load from 51.5 to 52%, maybe (hard to see with such small numbers). So I don't know what's happening on your pc... Could be a cache issue - is this still an old Core2Duo or Core2Quad system?

Author:  Bojcha [ Thu Mar 26, 2020 3:12 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
Quote:
Quote:


I've compared the final release build (9.40) against 9.37 and I don't see any difference in CPU usage - well, _maybe_ 9.40 uses 1% more, but it's a bit hard to see the exact numbers. Also, I've loaded an old preset and turned all "Dynamic" speeds on, and that also gave about a 1% increase in CPU usage.

What differences are you seeing?

Note: If you've configured it to use a specific core, that is now (by default) a "preferrerd" instead of "forced" core. Meaning that Windows can, if it needs to, move it somewhere else. I've done tests and this turned out to work much better, sometimes Windows starts some background tasks which are also running on a specific core with high priority, and in that case if we force things to run on that same core it needs to wait until the Windows task has finished. You can still force it to run on a specific core though, see the new "Forced" checkboxes under CPU & Latency. Not recommended though!
1. Try run beta 039 and above on 1 core. If you watch Task manager You will see windows spreads it to few other cores. "Forced" does not help, it uses core0, then after few second core4, then 6 then 8, and i set it all to core 15. Beta038, and older betas, are fine. Sure i will not use 1 core, but it is easier to see what's happening.

2. When i use "dynamic speeds" on my 7 MB bands, cpu load rises from 44% to ~68%, on any beta/release. I just found "Start Attack Immidiately" Also uses much CPU.
Weird. "Forced" works fine here, I just verified it on the 9.40 release. If I select core 1 forced for multiple threads and increase the CPU load a lot, it all stays on core 1 and starts to buffer.

I also measured Dynamics on/off for both AGC and all MB1 bands, and it increased the CPU load from 51.5 to 52%, maybe (hard to see with such small numbers). So I don't know what's happening on your pc... Could be a cache issue - is this still an old Core2Duo or Core2Quad system?
No i am on i9 9900K at 5GHz.. and beta 038 works fine all after that does not work with cpu cores how i set.
Checked also on core2duo and i3 9100.. same thing. Also on all same HIGH cpu usage when using "dynamic speeds" on MB bands. One Pc is XP, one W7 and one is W10. Exactly same with this 2 issues.
Maybe i can send you preset i am using ?

edit
Ok Forced seems to work, i didnt check one checkbox. However it worked erlier without any "forced" checkmark.

Author:  hvz [ Thu Mar 26, 2020 3:32 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
Quote:
Quote:

1. Try run beta 039 and above on 1 core. If you watch Task manager You will see windows spreads it to few other cores. "Forced" does not help, it uses core0, then after few second core4, then 6 then 8, and i set it all to core 15. Beta038, and older betas, are fine. Sure i will not use 1 core, but it is easier to see what's happening.

2. When i use "dynamic speeds" on my 7 MB bands, cpu load rises from 44% to ~68%, on any beta/release. I just found "Start Attack Immidiately" Also uses much CPU.
Weird. "Forced" works fine here, I just verified it on the 9.40 release. If I select core 1 forced for multiple threads and increase the CPU load a lot, it all stays on core 1 and starts to buffer.

I also measured Dynamics on/off for both AGC and all MB1 bands, and it increased the CPU load from 51.5 to 52%, maybe (hard to see with such small numbers). So I don't know what's happening on your pc... Could be a cache issue - is this still an old Core2Duo or Core2Quad system?
No i am on i9 9900K at 5GHz.. and beta 038 works fine all after that does not work with cpu cores how i set.
Checked also on core2duo and i3 9100.. same thing. Also on all same HIGH cpu usage when using "dynamic speeds" on MB bands. One Pc is XP, one W7 and one is W10. Exactly same with this 2 issues.
Maybe i can send you preset i am using ?

edit
Ok Forced seems to work, i didnt check one checkbox. However it worked erlier without any "forced" checkmark.
Yes, please send me your preset.

About "Forced": Without "Forced" it works better. It uses the core that you specify if it's availalbe, if not, it will pick another one instead of just waiting for it to become available. That reduces the change to get dropouts if the pc starts to do other things. That's why we modified the old behavior, you can still force it if you really want to but usually that turns out to be a bad idea.

Author:  Bojcha [ Thu Mar 26, 2020 4:02 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
Quote:
Quote:

Weird. "Forced" works fine here, I just verified it on the 9.40 release. If I select core 1 forced for multiple threads and increase the CPU load a lot, it all stays on core 1 and starts to buffer.

I also measured Dynamics on/off for both AGC and all MB1 bands, and it increased the CPU load from 51.5 to 52%, maybe (hard to see with such small numbers). So I don't know what's happening on your pc... Could be a cache issue - is this still an old Core2Duo or Core2Quad system?
No i am on i9 9900K at 5GHz.. and beta 038 works fine all after that does not work with cpu cores how i set.
Checked also on core2duo and i3 9100.. same thing. Also on all same HIGH cpu usage when using "dynamic speeds" on MB bands. One Pc is XP, one W7 and one is W10. Exactly same with this 2 issues.
Maybe i can send you preset i am using ?

edit

Ok Forced seems to work, i didnt check one checkbox. However it worked erlier without any "forced" checkmark.
Yes, please send me your preset.

About "Forced": Without "Forced" it works better. It uses the core that you specify if it's availalbe, if not, it will pick another one instead of just waiting for it to become available. That reduces the change to get dropouts if the pc starts to do other things. That's why we modified the old behavior, you can still force it if you really want to but usually that turns out to be a bad idea.
Eh, you actually changed how that works, ok.

Issue2: Something in preset. Try load web preset '7wonders' and just turn all "dynamic speeds" in MB. Here difference in cpu usage is 59% vs 42%.

And here how it looks on 1 core with dynSpeeds On and off.
Image

Author:  hvz [ Thu Mar 26, 2020 6:32 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
Quote:
Quote:

No i am on i9 9900K at 5GHz.. and beta 038 works fine all after that does not work with cpu cores how i set.
Checked also on core2duo and i3 9100.. same thing. Also on all same HIGH cpu usage when using "dynamic speeds" on MB bands. One Pc is XP, one W7 and one is W10. Exactly same with this 2 issues.
Maybe i can send you preset i am using ?

edit

Ok Forced seems to work, i didnt check one checkbox. However it worked erlier without any "forced" checkmark.
Yes, please send me your preset.

About "Forced": Without "Forced" it works better. It uses the core that you specify if it's availalbe, if not, it will pick another one instead of just waiting for it to become available. That reduces the change to get dropouts if the pc starts to do other things. That's why we modified the old behavior, you can still force it if you really want to but usually that turns out to be a bad idea.
Eh, you actually changed how that works, ok.

Issue2: Something in preset. Try load web preset '7wonders' and just turn all "dynamic speeds" in MB. Here difference in cpu usage is 59% vs 42%.

And here how it looks on 1 core with dynSpeeds On and off.
Image
Ah! Found it. I was testing RMS mode, with "Heavy CPU (compatibility)" off. And then Dynamic speeds has nearly no effect. But in Peak mode, or with Heavy (CPU) enabled, the difference is big.

I have already replaced "No heavy CPU processing" by something with more different levels, and I'll change the behavior to jump to a step lower for Peak mode, in the "sane" and lower CPU usage settings.

Author:  Bojcha [ Thu Mar 26, 2020 11:37 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
Quote:
Quote:

Yes, please send me your preset.

About "Forced": Without "Forced" it works better. It uses the core that you specify if it's availalbe, if not, it will pick another one instead of just waiting for it to become available. That reduces the change to get dropouts if the pc starts to do other things. That's why we modified the old behavior, you can still force it if you really want to but usually that turns out to be a bad idea.
Eh, you actually changed how that works, ok.

Issue2: Something in preset. Try load web preset '7wonders' and just turn all "dynamic speeds" in MB. Here difference in cpu usage is 59% vs 42%.

And here how it looks on 1 core with dynSpeeds On and off.
Image
Ah! Found it. I was testing RMS mode, with "Heavy CPU (compatibility)" off. And then Dynamic speeds has nearly no effect. But in Peak mode, or with Heavy (CPU) enabled, the difference is big.

I have already replaced "No heavy CPU processing" by something with more different levels, and I'll change the behavior to jump to a step lower for Peak mode, in the "sane" and lower CPU usage settings.
Yes, peek mode (analog) However "Heavy cpu" does not make big change for peek here.

Author:  hvz [ Fri Apr 24, 2020 1:39 am ]
Post subject:  Re: Stereo Tool 9.38 BETA

Aand... continuing here: https://forums.stereotool.com/viewtopic ... 14&t=28247

CPU load should be a lot lower with Dynamic Speeds in Peak mode as long as you select "Sane" as the CPU usage. (Updated setting under CPU & Latency, Sane is the standard value if you weren't already using "No heavy processing" mode).

Author:  hvz [ Tue Jul 14, 2020 3:19 pm ]
Post subject:  Re: Stereo Tool 9.38 BETA

Quote:
been away for a while, Linux Jack 64 9.37 on Debian 9, stereotool shows once in main cable connector twice in the "patch pannel" is this intended behaviour? If so what is the difference?
Long time ago, but I have just setup a new 64 bit test environment and decided to fix this before releasing version 9.52. However, I don't seen anything wrong on my end. Can you show me what you mean, maybe with a screenshot? We haven't changed anything in the Jack code since 9.37 so the behavior should still be the same.

Page 22 of 22 All times are UTC+02:00
Powered by phpBB® Forum Software © phpBB Limited
https://www.phpbb.com/