OTOY Forums

does the 690 spread the 4gb ram between 2 chips like the 590 did? - thus leaving the user with 2gb actual ram limit?
and does the 680 function like the 580 and leave us with 4gb of usable ram limit?

thanks in advance for your advice

Hi jmfowler,

Short of grabbing a couple just to physically check myself yes - all reviews and specs on these cards show the same design as the 500 series.
the 680 can have 4GB available ram, the 690 is still 4GB shared.

Cheers
Chris.

thanks, one more question -

I understande the 600 series is much more power efficient BUT

with the 680's having 1500 CUDA cores and the 580's with 500, will the 680's eventually be 3 times faster when CUDA 5 is implemented into octane ? or is there a trade off with the 600 series that will always limit their speed with octane?

trying to understand why all the extra CUDA cores in the 600 series doesn't boost octane render times?

cheers,

You can't really say that 1 500 series cuda core is equal to 1 600 series core. So you can't compare them - I very much doubt that a 680 will ever run as fast as 3 580's in octane render. Currently the 680 runs at 60-70% the speed of a 580 in octane render.
Although we are hopeful to improve performance by tweaking/optimizing for kepler we have no concrete info on exactly how much this will help and if their performance in octane will equal or surpass the 500 series. It's definitely not as simple as compiling with cuda 5 and everything runs super fast on kepler. In fact our initial test are producing results slower than cuda 4.2 and 4.0.

Sorry, there is no concrete info i can give you on where these cards will settle in terms of render speed.

If you want more info about the differences between fermi and kepler then i'm not a hardware architecture expert but there are articles out there on the design of kepler.
Things like this:
http://www.nvidia.com/content/PDF/keple ... epaper.pdf
http://www.nvidia.com/content/PDF/keple ... 012_LR.pdf

Cheers
Chris.

Thank you Fooze, I see more 580's in my future me thinks...