Multi-Core/Processor/LP Proformance  

no avatar
ThinKVR
New member
 
Topic author
Posts: 2
Likes: 0 post
Liked in: 0 post
Joined: Wed Jul 20, 2016 9:08 am
Info

Multi-Core/Processor/LP Proformance

by ThinKVR » Wed Jul 20, 2016 9:39 am

Hello all,

Dell Precision T7810
Windows 10 x64
Single Xeon E5-2695 v3, 14c (28 Logical Processors) (dual processor capable)
64GB Ram
EVGA GTX 980 SC
Samsung EVO 850s SSD's all round

Video input H.264 from 3x Kodak PixPro SP360 4k Cameras

===============
When computing "Compute RMS curve" a 3m46s video took almost 21minutes to compute.

When looking at the task manager, I see that there is barely any load on the CPU's and where there is load, its hap-hazard across the logical processors. (see attached screenshot)

I played around with changing Processor Affinity, from 2, 4, 6, 8, 10, 14 then 28 cores (choosing those numbers as most common processors use these amount of cores) but found no significant performance increases (ie - time it took to process the RMS) curve on changed affinity settings.

I also monitored the GPU using the EVGA Precision X16 monitoring and overclocking software.

Rendering the same 3m46s video (Apple ProRes MOV, 4096x2048, 4444XQ) seems to be 'somewhat' more efficient as the CPU average across all cores/LPs ran at 49%, and took 9m41s @ 12.12 fps average.

Is there a optimum setting that can be changed somewhere to allow APG to utilise all hardware available?
Does AVP perform well with SLI configurations if we were to add an additional GTX 980?
Attachments
Render ProRes.png
Render to ProRes screenshot
Compute RMS Score.png
This is the RMS computing screenshot

User avatar
Emeric
Kolor Team
 
Posts: 156
Likes: 3 posts
Liked in: 14 posts
Joined: Tue Apr 12, 2016 9:45 am
Info

Re: Multi-Core/Processor/LP Proformance

by Emeric » Wed Jul 20, 2016 10:45 am

Hi,
First I guess you are using the Autopano video Pro right ?

Autopano PRO use the CPU to decode input videos and encode the stitched output. The GPU is used to do the stitching and we do not support SLI.
It's always pretty hard to tell what part of the rendering is slowing the whole process. Encoding 4k videos is always very slow and the hardest part to do on a CPU with many thread.
Using 50% of a CPU with 14 cores is already avery good result (and 12fps for 4k rendering is alright), but what you want is a faster CPU with less core, so if you can freeze a few cores so the others have more room to get more frequency that could be a good idea.

no avatar
ThinKVR
New member
 
Topic author
Posts: 2
Likes: 0 post
Liked in: 0 post
Joined: Wed Jul 20, 2016 9:08 am
Info

Re: Multi-Core/Processor/LP Proformance

by ThinKVR » Wed Jul 20, 2016 10:54 am

Ok. some good insights there.

Is there any system tweaking that can be done that you're aware of that could assist in a performance boost?

Just an FYI as well, perhaps Kolor should look at working with Dell. Dell have a program called "Dell Precision Optimizer" (http://www.dell.com/learn/us/en/04/camp ... =bsd&cs=04) that when we use it with programs such as Adobe AE and Premiere Pro, 3DS Max etc, notice a moderate increase in overall system performance within these programs.

You can watch more about the program from this youtube video: https://www.youtube.com/watch?v=Jt3XprWEfFM


Who is online

Users browsing this forum: No registered users and 0 guests

cron