5950x 16 cores 95% use(high def), Will 32 cores drop use back to the 50% dvds level, TMPGEncVMWorks 7

[jmc111]

Processing dvd level video files (video denoise, contour, color) uses around 40-60% of 16 cores.

Which is fine, plenty of room for me to do simple video editing at the same time.

 

6Mbps 1080p files at 95% use does not leave room to do other things smoothly.

So looking at 32 core threadripper (Zen 4 next year) (64 cores is too expensive)

 

Keep hoping to read of 32 core Zen5 desktops but not holding breath.

 

Is expecting double the cores to drop the use back to 50%+/- reasonable?

 

Thanks.

 

[Sharkyx1]

5 minutes ago, jmc111 said:

Processing dvd level video files (video denoise, contour, color) uses around 40-60% of 16 cores.

Which is fine, plenty of room for me to do simple video editing at the same time.

 

6Mbps 1080p files at 95% use does not leave room to do other things smoothly.

So looking at 32 core threadripper (Zen 4 next year) (64 cores is too expensive)

 

Keep hoping to read of 32 core Zen5 desktops but not holding breath.

 

Is expecting double the cores to drop the use back to 50%+/- reasonable?

 

Thanks.

 

Would it be more wise to simply use software to lower the CPU usage and slow the process down?

It would be a reasonable expectation that doubling the core count should increase general windows performance, but alternative choices would be a dedicated rendering PC on side from your main PC for user tasks.

[GuiltySpark_]

It sounds like the software you're using (probably ffmpeg underneath) is scaling well to all the cores you have. Adding more cores may not do what you think its going to do. ffmpeg has the ability to limit thread usage assuming you can give it a specific command or the GUI you're using has that option available. 

 

Alternatively, just upgrading to a faster chip may allow you to get work done faster negating the need to do two things at once. 

 

Is hardware/GPU encoding not an option? Not familiar with the software you're using. 

[jaslion]

It will most likely just scale to all the cores again to get the work done faster. @GuiltySpark_ has the right idea here.

 

What software are you using? You can normally just say only use x cores and that solves it.

[OddOod]

Agree with everyone else: assuming it's using FF under the hood, it will likely just expand to all available cores. The best option here is to limit the cores it has access to. Also, when I used Handbrake to smash down some files, I was able to enable NV acceleration which dropped my 5950X to ~70% which is plenty enough headroom to operate smoothly

 

[jmc111]

I was testing files to see cpu use and realized that "Topaz AI video enhancer" is the one that really pushes up the high cpu use.

Topaz is run on the GPU and I take the 8Mbps mp4 file down to 2Mbps with high cpu use.

No idea why. I tried to create a same spec file with "TMPGEnc Video Mastering Works 7" and had much lower cpu use taking it down to 2Mbps.

So problem seems to be a special use case.

 

TMPGEnc Video Mastering Works 7 does have a cpu use limiter but that would bug me a lot.

This runs fast enough that I'm fine with the superior cpu output. (Topaz is just too slow for cpu use)

 

I believe that ffmpeg has limits per the VideoReDo  devs but don't remember what it was. 

 

Hate to have to slow things down. Tweak things to get faster. I've have run two pcs at once but really don't like that.

 

So I'll look forward to a 7000 level threadripper next year but hope AMD releases a Zen5 32 core desktop.

(and this is mainly to get a decent number of motherboard slots. Current ones are terrible, maybe 3 slots.)

 

Thanks to all for the thoughts!

 

 

 

[Bagzie]

8 hours ago, jmc111 said:

I was testing files to see cpu use and realized that "Topaz AI video enhancer" is the one that really pushes up the high cpu use.

Topaz is run on the GPU and I take the 8Mbps mp4 file down to 2Mbps with high cpu use.

No idea why. I tried to create a same spec file with "TMPGEnc Video Mastering Works 7" and had much lower cpu use taking it down to 2Mbps.

So problem seems to be a special use case.

 

TMPGEnc Video Mastering Works 7 does have a cpu use limiter but that would bug me a lot.

This runs fast enough that I'm fine with the superior cpu output. (Topaz is just too slow for cpu use)

 

I believe that ffmpeg has limits per the VideoReDo  devs but don't remember what it was. 

 

Hate to have to slow things down. Tweak things to get faster. I've have run two pcs at once but really don't like that.

 

So I'll look forward to a 7000 level threadripper next year but hope AMD releases a Zen5 32 core desktop.

(and this is mainly to get a decent number of motherboard slots. Current ones are terrible, maybe 3 slots.)

 

Thanks to all for the thoughts!

 

 

 

Couldn't you just set process affinity to leave 1 or 2 cores idle so you can still use the pc? 

[jmc111]

15 hours ago, Bagzie said:

Couldn't you just set process affinity to leave 1 or 2 cores idle so you can still use the pc? 

Yes, but that just tells me that I need a new faster computer.

Next year going 32 cores!

If that's not enough I'll just have to live with it.

[GuiltySpark_]

1 minute ago, jmc111 said:

Yes, but that just tells me that I need a new faster computer.

Next year going 32 cores!

If that's not enough I'll just have to live with it.

This tells me you don't quite understand how software like this is supposed to work. Most people would want an encoding application to go as fast as possible, meaning scale linearly across as many cores as possible. If you grabbed a Threadripper 7995WX when they launch soon, good software than can scale may still use 85-95% of the CPU (lot of variables here). 

 

If you want cores that aren't used no matter what, regardless of what you upgrade to, you may be forced to limit it manually using process affinity. 

[freeagent]

It’s just going to use more cores if you have them. You have to use task manager to manage it, or use different software. You could use your GPU for video stuff and keep your CPU free for other tasks if you have an Nvidia GPU.