GPU benchmarks

[savage309]

I've update the results already. The test with the newer driver and M6000 is done on another machine running Windows. The Linux drivers seems to be lacking behind.
The times ratio can be a bit tricky - to get from 8 to 4 takes as much effort as from 4 to 2 .

[Booska]

I just ran it on my Quadro K5000 and got 6 min 32 secs.

[chris-x//]

EVGA GTX 980 Ti Superclocked / 11 min 18 secs / 353.06
Not Bad !

[3LP]

Awesome mate!
Better than a Titan X

Stan

[super gnu]

interesting one: bechmark 1.0, gtx 670 4gb, 3.20.02

cuda: 4m 39 secs.

opencl, with 2x 3930k using intel opencl driver

2m 3.8 secs.

i had to play with the RPP and ray bundle size to get my cpu usage up on the slave.

if youve got a mix of cpus and gpus (i.e. most of us) then i wouldnt write off opencl just yet.

[vlado]

Interesting Thanks for sharing.

Best regards,
Vlado

[3LP]

Ah ha, yes that's good to know.
Put aside the fact that's SP2, it seems that 2 x 3930k is equal to 1 Titan X.

Stan

[super gnu]

No actually 2x 3930k (at 4.5 ghz) are approximately equivalent to a gtx 670

Those three together equal one titan x. So it's 4x 3930k (at 4.5) equal one titan x. Gpu is obviously far more powerful. But.. If you already have a cpu render farm, and are testing / using a small number of gpus for rt (a common scenario I recon) then you don't have to leave the farm idle when using rt. My particular little setup, the gpus and CPU capacity I gave are evenly matched. So the only way I can get a boost from rt (without upgrading) is to use opencl and combine.

[super gnu]

Autocorrect messed a few words there and edit isn't working...

ok i edited! (and edited this) ive noticed that when i hit "edit" i often get a blank posting window (i.e. it forgets what ive already posted) so in order to edit rather than replace, i have to copy my text, and paste it into the edit window before editing.

im sure this never used to happen..

[voltron7]

Thanks super!

[savage309]

Just wanted to point that adding the V-Ray version with which the test is made is important. Even minor V-Ray versions can have different results (as is the case with 3.20.03 and 3.20.02).

[Guest]

Dual GTX 980 Ti 52.7 seconds in V-Ray 3.20.02 using driver version 353.30. This was in the original benchmark scene.

[3LP]

I'd be interested to see what would the 980 Ti do on the V2 benchmark if you have a chance to try

Thanks
Stan

[Guest]

I'd be interested to see what would the 980 Ti do on the V2 benchmark if you have a chance to try

Thanks
Stan

Can it be saved down at all? On 3ds Max 2012 at the moment with the dual 980's.

[3LP]

Sure!
here it is :
https://dl.dropboxusercontent.com/u/...scene_2012.rar

Jesus mate, 2012, that seems far away
I'm adding the download link to the benchmark spreadsheet just in case.

Thanks
Stan