Superhypersam brought this up,( I am beginning to think I missed something) and of course being a curious little monkey that I am, did a quick test...
single grid, object source emitter, 20 frames:
All Available checked (4 cores): 01:40
4 threads: 01:41
3 threads: 01:22
2 threads: 01:12
1 thread: 00:44
interesting... Why are my sim times so much faster when only using one core? How is the processing done? Kind of strange that the sim calculates in almost a minute faster using only one thread.
Could someone please enlighten me?
Obviously I haven't tried it one a more complex scene yet to see if there is any difference (would take a few nights me thinks) but anyway any explanations would be greatly appreciated.
-John
Interesting !
I can not explain it directly, but a´ve had a similar expirience with Backburner-Rendering (2 Dualcore-Machines).
I was wondering, because i tendered 3 minute on one machine (no BB in use). The two machines needed 51min50sec.
But it may be the thing: on short sequences the preparing of data is as time consuming, that the total processing is slower than single-threaded.
(The data has to be splitted up for each thread, then it has to be sent to each core, and each core wants to be managed all the time by the application).
Would be interesting to hear, what happens on longer sequences.
Best wishes
Dominik
Hello,
We are aware of this problem and we're looking into the cause.
As soon as we fix it, update will be available from our members area.
Best regards,
Kresimir
Hi Dominik,
What you are saying makes sense but three more seconds per frame? Seems high to me considering the speed of the processor/ram. Maybe more if it were caching to disk but it isn't.
Thanks Kresimir, I figured maybe I had missed something, somewhere, somehow.
Ups, sorry, i wrote an error  i meant:
Single machine: 3 minutes (no BB in use)
2 machines via BB needed 5min50sec (and not 51 minutes).
But again, i think with BB thishappened because of preparing data, sending,starting Max... for serious results this scene was too small.
What Kesimir said, VERY nice to hear  i´m exited to see, what performance-boosts we will get...
Hello,
I want to update you on this issue.
We have found what is causing slow down.
Fix will be included in FumeFX 1.0a and you can expect FumeFX to be around 30% faster on dual core (or 2 CPU) machines.
Regards,
Kresimir
Wonderful!
Thank you 
-John
sitnisati Wrote:Hello,
I want to update you on this issue.
We have found what is causing slow down.
Fix will be included in FumeFX 1.0a and you can expect FumeFX to be around 30% faster on dual core (or 2 CPU) machines.
Regards,
Kresimir
Hi Kesimir,
since you gave some information about the Afterburn4-release-date, i want to ask: Can you give similar information about a FFX-update - especially this performance-issue and/or about moving FFX-Grids ?
Thank you
Dominik
Hello Dominik,
FumeFX 1.0a is released today.
We're waiting TurboSquid for details how existing users will obtain their two free Simulation licenses. We'll keep you informed.
Regards,
Kresimir
Super! Thanks Kresimir
Two sim licenses too... sweeeet!
What was the info on a Afterburn 4 release date?
Chad
cnaeger Wrote:What was the info on a Afterburn 4 release date?
Chad
We're looking for September, but we'll know more when when enter beta testing.
Regards,
Kresimir
Retested and whatever you guys did... WOW! that's more like it
single grid, object source emitter, 20 frames:
All Available checked (4 cores): 00:11
4 threads: 00:11
3 threads: 00:14
2 threads: 00:16
1 thread: 00:27
That makes a lot more sense. Thanks again for your guys hard work :wink:
BTW excited to see afterburn run under the fusion renderer.
jrand Wrote:BTW excited to see afterburn run under the fusion renderer.
Is this planned for the next afterburn release ?
as far as I know/have heard, it has been mentioned.
|
