Understanding DAW/ CPU performance.

Configure and optimize you computer for Audio.
Shawnobi
KVRer
3 posts since 6 Oct, 2021

Post Sat Dec 04, 2021 7:49 am

Hope this is the right place to ask this..

Having decided to upgrade my CPU after hitting the old S1 performance meter a bit hard, I replaced my 3600X with a 5800X.

Before removing I opened a new song and threw as many SGA1566 plugins on the insert on a single track until the meter hit 95% as a reference. I managed 12.

Then, following the upgrade I reopened the project, the dial had dropped to 85%. I managed an extra 2 instances of the plugin before I was back at 95%. Got to admit I was hoping for more. However maybe this is to do with single core speed and hence no real difference? So have I wasted my money or should there have been another test I should have done?

The interesting thing was that with the 3600 I could see that CPU2 was maxxed out at 100%. But with the 5800 my pc seems to be barely breaking a sweat (8% usage according to windows!). So there is definitely more power there. I'm just not sure how S1 is making use of that if it is able.

Scotty
KVRAF
2399 posts since 23 Dec, 2002

Post Sat Dec 04, 2021 7:53 am

It seems like you aren't getting the performance you should be getting. Is your ram speed set-up properly in the BIOS and did you do the Windows optimizations steps? Did you lock all cores or did you let them float? They used to recommend locking all cores but now letting them float seems to give better performance. I'm on a Ryzen 5950x for a few months now. I had similar "concerns" but I am getting the performance I paid for now after some tweaking.

ShawnG
KVRian
987 posts since 27 Apr, 2005

Post Sat Dec 04, 2021 8:08 am

Usually daws run a track and all it’s inserts on a single core. So throwing successive inserts on a single track will only test your single core performance. If you run the test wide with inserts on multiple tracks that will give you a better idea.

Also make sure your bios is set up correctly for your ram, and set the system up for good dpc latency (there are a few things to do here, google is your friend, but the biggies are disabling usb device power settings, and doing a clean install of your graphics card drivers)

Shawnobi
KVRer

Topic Starter

3 posts since 6 Oct, 2021

Post Sat Dec 04, 2021 9:23 am

RAM seems to be running at the correct speed and and I've gone through the old optimisation stuff. Not heard of core locking before, though?

ShawnG - Yeah the single track/ core thing was what I was wondering. That my quick and dirty test isn't a good one. I guess I'm mostly looking for a tangible result that will validate my purchase. Will have a look into DPC latency. Is that the same as 'regular latency'. I've never had an issue with that and LatencyMon results seem good.

ShawnG
KVRian
987 posts since 27 Apr, 2005

Post Sat Dec 04, 2021 3:25 pm

Dpc latency is the thing that LatencyMon exists to check, so if that’s good, you’re good.

ShawnG
KVRian
987 posts since 27 Apr, 2005

Post Sat Dec 04, 2021 3:28 pm

It’s generally better to lock your cores to a speed in bios rather than letting the system determine what your speed should be based on how hard it thinks the system is working, even if you aren’t overclocking.

User avatar
Pictus
KVRian
694 posts since 21 Aug, 2017 from Brasil

Post Sun Dec 05, 2021 12:22 am

To check the DPC latency https://www.resplendence.com/latencymon
System temps/voltages/stuff https://www.hwinfo.com/download/
Memory related settings https://zentimings.protonrom.com/

Some tweaks you should check viewtopic.php?p=7624448#p7624448

The 5800x is a hot CPU and needs a good cooler
https://youtu.be/yW1x101pDfE

But the Curve Optimizer with Scalar = 1 and Max Boost = Zero
can keep the CPU more cool&quite.
https://youtu.be/dU5qLJqTSAc
https://youtu.be/gDLuwNp06-Q

CoreCycler - tool for testing Curve Optimizer settings
https://www.overclock.net/threads/corec ... s.1777398/

Shawnobi
KVRer

Topic Starter

3 posts since 6 Oct, 2021

Post Sun Dec 05, 2021 2:38 am

Thanks for that.
Seems like my cooler and case are doing a job. with S1 at 100% load the CPU is sat at 40 degrees. So lots better than the 80+ gamers have been seeing. But my CPU is only showing 17% utilisation. Understanding how DAW's use CPU seems a bit of a black art that's harder to understand which processor you need compared to the gaming world.

EDIT: Just downloaded Reaper and threw a bunch of SGA1566's on a track. My CPU's are barely above 2% compared to getting one of them maxxed out with S1. Is S1 really that bad at utilizing CPU? Using Reaper was kinda horrid, but it made me think!

rardier
KVRist
190 posts since 22 Feb, 2014

Post Thu Dec 09, 2021 9:48 am

Yes and no , if you are using default setting on reaper it seems to add latency..so you should compare at the same amount of latency..but i also found that on reaper i can add more latency than on other daw (cubase and s1) so at the end yes you can add more plugin and use more your cpu but if you go to far your whole system become laggy

jdnz
KVRian
1001 posts since 3 Oct, 2011 from Christchurch, New Zealand

Post Thu Dec 09, 2021 11:53 am

rardier wrote:
Thu Dec 09, 2021 9:48 am
Yes and no , if you are using default setting on reaper it seems to add latency..so you should compare at the same amount of latency..but i also found that on reaper i can add more latency than on other daw (cubase and s1) so at the end yes you can add more plugin and use more your cpu but if you go to far your whole system become laggy
reaper does latency compensation using the block size of your audio device - which is why it tends to report higher latency ( say if a plugin reports latency of 13 samples but your audio block size is 256 samples then reaper will round up to 256 )

it's a two edged sword - that approach is one of the reasons reaper is so cpu efficient, but it also leads to seriously huge latency on chains.

Return to “Computer Setup and System Configuration”