OTOY Forums

I'd like to know what is more important for a better performance in octane, more cuda cores or less cores but with higher clock speed.
The later is true for gaming (in fact an overclocked GTX460 performs quite close to the GTX470 despite having just 336 cores)
but I'm not sure about cuda applications like octane

Thanks

my way for having an idea is multiplying # of cores by clock to get an "overall clock sum" of the GPU.

Hi,

Overclocking GPUs for use with octane is a very bad idea as octane taxes the GPU %100 and you have a big chance of damaging your GPU with overclocking it.

Radiance

Speaking of GPU load. I was watching GPU-Z while rendering a simple scene.
In Pathtracing mode my GPU load was 95-97% and the GPU temp was 90C.
Swithing to Directlighting the GPU load dropped to 82-85% and the GPU temp dropped to 87C.
Is there a reason why directlighting doesn't run on full throttle like Pathtracing does ?

directlighting is quicker/easier to compute.
on small resolutions you might reach your vsync or tonemapping limit.

Radiance

Many GTX460 have been released with factory OC so If I decided to overclock myself I can assume it's OK as long as I can keep the temperature within safe limits, right?
Anyway this all was about which card is better for octane: an OC'ed GTX460 2GB or a GTX470 as both are similary priced.

As long as at full load for about 10-15 mins your within manufaturer recommended limits then an OC shouldn't be a problem.
Remember, OC'ing too far can cause instability and graphics glitching so be careful if your really pushing it!

I plan on upgrading to a 460 2GB and will probably apply a mild overclock. (I'll show results when I manage to get one!)
Here a 470 is about £40-50 more than the 460 2GB and only has 1280MB so to me its a no brainer (I can always wait a little longer for the render but can never increase the amount of available vram)