[Bf-cycles] cycles, some questions about GPU

Jordan Miller jrdnmlr at gmail.com
Wed Nov 16 16:12:51 CET 2011

is there anywhere to demo Nvidia graphics cards? I'm imagining
installing Blender with cycles on demo machines and checking out
performance on test scenes. I did this a couple weeks ago at the apple
store and it was NOT significantly faster than my laptop from a couple
years ago. I'll definitely be steering clear of Mac desktops (even 8
core) for Blender Cycles for the time being.

for GPUs, GFLOPS is calculated with a very specialized highly parallel
algorithm written from scratch and directly for GPUs. so it is NOT
akin to clock speed on a CPU in terms of a ballpark performance
metric. it does not always achieve that rate, and IIRC, it can never
achieve that rate for things like CUDA or OpenCL etc.

I think you want the fastest one with the most memory and the most CUDA cores.


On Nov 16, 2011, at 10:03 AM, Sebastian Röthlisberger <lalli at gmx.net> wrote:

> On 16.11.2011, at 15:35, Brecht Van Lommel wrote:
>>> 2011/11/16 Sebastian Röthlisberger <lalli at gmx.net>:
>>> "will be limited by the max amount memory of 1 GPU" until the "bigger scene
>>> than memory" feature is added?
>> Right, multi GPU wouldn't help in increasing the memory available.
>>> about 2. GFLOPS.
>>> afaik GFLOPS is not a single feature but (wiki) Floating Point Operations
>>> Per Second.
>> What I mean is, it's not possible to say, look for X or Y in the GPU
>> specs, this will ensure good rendering performance. It's much more
>> complicated than that, but more flops is of course better.
> Ok, i thought GFLOPS are something like the "result" of the GPU specs.
>> Brecht.
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles

More information about the Bf-cycles mailing list