Bifrost Benchmark

[morteza.cg]

woudn't it be cool to know how different processors perform with bifrost?!

it can help us build better pc and workstations for bifrost simulations.

I have never created a benchmark before so it may not be that good but i hope that even in it's current state it come in handy.

currently it's only for Aero, let's check it and if it was working as expected i will add mpm and instancing as well.

when you open the scene file you will see this window.

• at the very top you have two check box one for pausing viewport, it can help you if you don't have a powerful gpu. it's checked by default.

when you run any of the aero sims it will write a text file for you with your spec and simulation time

• if the second checkbox is enable it will remove that result file and create a new file for you everytime you run the simulaton and if you leave it off it will keep adding to same file.

after that we have 3 button each one will run a simulation.

• Explosion: a combustion test for an explosion like simulaton, it will run for 30 frame but source is only active for first 5 frame.
• Smoke: a basic smoke with collision, simulation will run for 100 frame
• rolling flame: a continous combustion simulation, it will run for 30 frame

after you run each simulation it will show the simulation time in the fields on the right.

----

Moderator's note: Removed the scene due to it containing the scriptJob described here. Please use the security tools for Maya. An updated version has been posted in the thread below. 

[morteza.cg]

i only have some low end cpu around to test this with. i really like to know how long it takes for more power processors to simulate these.

Asus fx502Surface pro 2017

[mjcg91]

This is awesome, thanks for making something like this. Everyone loves benchmarking simulation. I've shared your post on the Bifrost Addicts Slack channel. If you'd like to join, I believe people will share their benchmark results: https://join.slack.com/t/bifrostaddicts/shared_invite/zt-jq7gwf53-AtAgPGWF_NPan_Tu7cfJTw

In the meantime, here are my results:

[sachin.shrestha]

This is a really nice benchmark tool Morteza! My personal PC config is pretty similar to yours so I guess my test results  should be close to yours. Thanks for sharing this!

[morteza.cg]

Thanks, I'm glad you guys liked it.

I didn't know about the Bifrost slack channel, I will join now.

[hamed_msv]

Xeon 2630 x 2

[tig5]

What a fun tool. Thank you @morteza.cg for making this!

[cescocm · Accepted solution]

Hello, here my results:

Also, keep in mind that the scene has this this scriptJob.
I cleaned it from the .ma before opening the scene because last week we had the same issue in my workplace. I'm also attaching the cleaned .ma just in case.
Cheers!

[morteza.cg]

Thank you for the fix!

I can't seem to edit my post, I've notified the moderator hopefully they will change my attached scene with yours.

[sachin.shrestha]

@cescocm  - Holy aero smokes! That's one fast machine!

[jonah.friedman]

Very cool tool (and it's very interesting to see the results too!)

I have removed the attached scene that contains the bad scriptJob though - I didn't post the updated one though as that might be taken as a verification that it's clean, which we haven't done. Feel free to accept @cescocm's post as a solution though to draw more attention to that post. Thanks! 

[mihirsbcc]

I tried your scene. When I opened the scene, it gave me the famous Maya's error........ An error occurred while reading file. There will be data lost..... I ignored it. selected the Explosion tab and play the time line of 200 frames. It crashed after completing 1oo frames. I will try it again and keep an eye on it.

[mihirsbcc]

Thanks for the fix

[mihirsbcc]

[tig5]

Interesting.. after running Yuri's (@1usmus) CTR diognostic, I got stable figures for an undervolt/overclock and created a test profile in ryzen master. Looks like it lowered my cpu temps while shaving a few seconds off this test. Still no threadripper lol but cool if it remains stable with day long renders.

[michael_nielsen]

Great initiative.

Here are the results for two other machine configurations:

I also just want to point out that there are a few ways to speed up the benchmarks and simulations in general as shown in this image:

[morteza.cg]

Thank you for the tips Michael specially the Viewport bypass processing.

[Anonymous]

Thank you for this, I'm gonna try running this on my 3990X. I do have a question though, when I open the file, the benchmark window doesn't pop up, how can I get it to open?

[Anonymous]

I have a 3990X and ran the sims, the times are above. Is there any way to script this? I'd like to run it on some other machines.

[Anonymous]

Hi @michael_nielsen when I try connecting that has_property node like you show, I get an error. I think it's because the updated scene doesn't have the "found" property on the output node, but I'm probably missing something here.