Titan Black vs 2x970 rendering far too slow...

Was wondering why my renderings at work were so slow, so I decided to compare the same file with my machine at home and was struck by the difference. ie at work I have a TitanBlack and it took 4 minutes to render ( http://i.imgur.com/oMbMtgf.jpg ), while at home it took 40 seconds ( http://i.imgur.com/MPqobMB.jpg ). I know there should be a difference of about double the time going off these benchmarks http://blenchmark.com/gpu-benchmarks but 40 seconds compared to 4 minutes is a big leap especially when rendering for animations.

Any reason why this is happening? The specs for both machines are below (sorry I’m not much of a hardware man)…

Work (Win7): Titan Black

Home (Win10): 970x2

Titan blacks are from 2014 (a slightly higher quality version of the 780Ti)

970s are a much faster chip over the 780…

i would also suggest trying out a nightly build from builder.blender.org for your tests… they recently improved performance of cards with loads of cuda cores (which both the 970 and titan blacks have), especially under windows 10

I also suggest measuring the rendering time by commandline / console ( bpy.ops.render.render() )…as it is much more accurate in terms of maxing out gpu performance.

Ok thanks,

I’ll have to give that nightly build idea a go as I heard that too. And I’ll suss out the commandline too and get back to you.

I’m such an idiot!

I started recording all my rendering times with different tile sizes and was thinking, gee these are rendering really slow when the tiles are large just like when you render with the CPU… hmm maybe I haven’t turned on GPU rendering…


So now I’m rendering with GPU with these results:

128x128: (01:10)

240x270: (00:55)

256x256: (00:54)

300x200: (00:54)

600x400: (00:56)

1024x555: (01:25)

Thanks guys for putting up with my lunacy :slight_smile:

Should I still be worried though that the TitanBlack is still not faster than my 2 970’s?