interesting benchmark results
Posted: Sun Sep 18, 2011 7:43 pm
Hi all!
i ran some 5 minute testrenders on my workstation with the new indigo 3 trail version today.
tested pathtrace vs pathtrace with GPU on the Caterpillar and Pintest scenes.
Got some very interesting results!
Caterpillar:
pathtrace = 5m 56 s/px, 138900 s/sec
pathtrace GPU = 5m 309 s/px, 720424 s/sec (fastest)
Pintest:
pathtrace = 5m 1428 s/px, 2282538 s/sec (fastest)
pathtrace GPU = 5m 534 s/px, 853952 s/sec
pathtrace MTL = 5m 958 s/px 1530646 s/sec
my specs are:
2 x Xeon X5670 @ 2.93GHz
(total 24 logical processors)
24gb RAM
Quadro 4000
running Win 7 Pro SP1
I did not expect GPU to speed up Caterpillar and slow down Pintest. also MTL was a little slower on Pintest.
I figured maybe the CPU's sen't to many rayhits to my GPU so it could not handle them all?
and because Caterpilar is a simple scene my GPU was more able to speed thing up?
But this does not explain to me why MTL also was a little slower then just pure pathtrace.
Of course one must choose the right metod on a scene basis. I'm just curious why the results came out this way on my system.
Simply trying to get my head around this
Hopefully some of you guys hacan explain this to me?
Thanks!
/Stefan
i ran some 5 minute testrenders on my workstation with the new indigo 3 trail version today.
tested pathtrace vs pathtrace with GPU on the Caterpillar and Pintest scenes.
Got some very interesting results!
Caterpillar:
pathtrace = 5m 56 s/px, 138900 s/sec
pathtrace GPU = 5m 309 s/px, 720424 s/sec (fastest)
Pintest:
pathtrace = 5m 1428 s/px, 2282538 s/sec (fastest)
pathtrace GPU = 5m 534 s/px, 853952 s/sec
pathtrace MTL = 5m 958 s/px 1530646 s/sec
my specs are:
2 x Xeon X5670 @ 2.93GHz
(total 24 logical processors)
24gb RAM
Quadro 4000
running Win 7 Pro SP1
I did not expect GPU to speed up Caterpillar and slow down Pintest. also MTL was a little slower on Pintest.
I figured maybe the CPU's sen't to many rayhits to my GPU so it could not handle them all?
and because Caterpilar is a simple scene my GPU was more able to speed thing up?
But this does not explain to me why MTL also was a little slower then just pure pathtrace.
Of course one must choose the right metod on a scene basis. I'm just curious why the results came out this way on my system.
Simply trying to get my head around this
Hopefully some of you guys hacan explain this to me?
Thanks!
/Stefan