Vray rendeing issue: why 88 threads rending more slower than 24 threads?

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

REPLY TO THREAD
 
Thread Tools Search this Thread Display Modes
  12 December 2017
Vray rendeing issue: why 88 threads rending more slower than 24 threads?

Hi, all
I have a scene with same frame, same settings rendered on two computers:

A computer:
system: win2012
Max: 2017
Vray: 3.60.03
CPU: E5 2620 x2= 24 threads
Mem: 64GB
Total render time: 1min30s.
first GI Passes time: ~20s.
second GI Pass time: ~17s.
final image render time: ~42s.
B computer:
system: win2016
Max: 2017
Vray: 3.60.03
CPU: E5 2696 V4 x2= 88 threads
Mem: 128GB
Total render time: 2min 25s.
first GI Passes time: ~1min 30s.
second GI Pass time: ~26s.
final image render time: ~18s.
I also tested Vray 3.5 has the same result: B computer rending more slower than A computer.
Why B computer rending more slower than A computer?

Thanks.
 
  12 December 2017
That's definitely a question for the Chaosgroup forums, you'll get faster and more professional support over there.
 
  12 December 2017
This sounds like a memory access issue, the more threads you have, the worst the problem becomes.  XEONs are designed for running a large number of small parallel processes accessing their own separate data, not a single large process where all cores are accessing the same data.

Try running 2 or more separate renderings instead of rendering frames sequentially, but as been said, asking them directly is your best bet.

Last edited by KaaF : 12 December 2017 at 07:19 PM.
 
  12 December 2017
It will probably shoot ahead if you render something that lets the buckets run for longer. That's my experience with a 56 threads machine vs 8 threaded machines at least.
 
  12 December 2017
Windows 2012 and 2016? I take those are Windows server editions?
Isnt the amount of cpu's/cores you can use on windows server quite limited (depending of which edition you use)?
 
  12 December 2017
this is purely just for laughs...but I'll round down Usain Bolt's 100m record to an even 10secs..ok...(yeah in terms of speed, that's round down)

So Usain Bolt, with 2 arms, 2 legs, and it's taken him all his life to hone his biomechanics to jedi level so he can run 100m in 10s flat.

Now you tell Usain, hey U....sain...I'm gonna give you FOUR arms and FOUR legs......
Were you expecting Usain Bolt to simply run 100m in 5 secs now ?

01) he now has to carry more mass
02) all his jedi level muscle memory, biomechanics are thrown out the window
03) bigger body, more surface area, more drag...
04) needs more oxygen supply, he's still got only one heart
blah blah blah....

so simply whacking more stuff in doesn't automatically mean better...

Last edited by vusta : 12 December 2017 at 02:45 AM.
 
  12 December 2017
If you can talk with crowds and keep your virtue,       Or walk with Kings—nor lose the common touch,Please be nice to one another! 

Last edited by Travis : 12 December 2017 at 02:59 PM.
 
  12 December 2017
Originally Posted by KaaF: This sounds like a memory access issue, the more threads you have, the worst the problem becomes.  XEONs are designed for running a large number of small parallel processes accessing their own separate data, not a single large process where all cores are accessing the same data.

Try running 2 or more separate renderings instead of rendering frames sequentially, but as been said, asking them directly is your best bet.

Thank you. I just render one frame....
May be it's a memory access issue....

Last edited by alonezhby : 12 December 2017 at 05:22 AM.
 
  12 December 2017
pffffffffffffffft !!

Last edited by vusta : 12 December 2017 at 06:57 PM.
 
  12 December 2017
have you checked the bios? There could be something switched off in there regarding your CPUs
__________________
James Vella
3D Visualization
Portfolio
 
  12 December 2017
Originally Posted by NorthernDoubt: have you checked the bios? There could be something switched off in there regarding your CPUs
Thank you for your reply.
I have checked. Did not found special option. 
You can see only calculating GI is more Slower, final rendering is more fast.
 
  12 December 2017
My friend had exactly same problem in his studio. 
He bought few Xeons, and on some scenes they rendered slower then regular i7. 
I just asked him on Skype how he solved the problem, and he said that the Xeons ha had were ES (engineering sample), and these was early test versions of these processors. After few months testing, and not fixing the problem, he sold them out all, and bought a OEM versions (that is official released version). 
Now, he says that they work like charm. 

Hope this helps. 
__________________
Now it's the time to be extreme!
Demoreel
website
 
  12 December 2017
Originally Posted by d4rk3lf: My friend had exactly same problem in his studio. 
He bought few Xeons, and on some scenes they rendered slower then regular i7. 
I just asked him on Skype how he solved the problem, and he said that the Xeons ha had were ES (engineering sample), and these was early test versions of these processors. After few months testing, and not fixing the problem, he sold them out all, and bought a OEM versions (that is official released version). 
Now, he says that they work like charm. 

Hope this helps. 
Thank you.
Is his problem is: calculating GI is more Slower, final rendering is more fast ?
And my cpu is not ES version.
 
  12 December 2017
out of curiosity what happens when you try brute force/ brute force for first/secondary cache?
__________________
James Vella
3D Visualization
Portfolio
 
  12 December 2017
Originally Posted by NorthernDoubt: out of curiosity what happens when you try brute force/ brute force for first/secondary cache?
1st Render test:
24 threads: thinking ~20s start rendering, total render time is ~3min.
88 threads: thinking ~1min start rendering, total render time is ~2min 42s.
2nd render test:
24 threads: thinking ~20s start rendering, total render time is ~3min 4s.
88 threads: thinking ~1min start rendering, total render time is ~1min 54s.
3rd render test:
24 threads: thinking ~16s start rendering, total render time is ~3min 3s.
88 threads: thinking ~1min start rendering, total render time is ~2min 11s.
 
reply share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright ©2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 02:00 AM.


Powered by vBulletin
Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.