What about Cuda performance with GTX680? Are there any tests?
3x more CUDA cores, better core clock speeds and faster memory, yet when it comes to computing power, it is actually worse than GTX580.
That is what initial tests on SmallLuxGPU show so far. It has yet to be tested in iRay, but if the result is same, then you can see how dirty cheap tricks would nVidia be willing to use to force people buy GPU computing dedicated cards (Teslas)
So far it seems like there is intentionally crippled computing performance.
Actually, most performance benchmarks have it outperforming all other cards currently on the market.
It's also being sold for the same price as the current 580 ($499) which means they are skipping the usual "premium" for first adopters.
What you refer to with the LuxMark is intentional, like you mention, to differentiate markets. Different chip yields will dictate different products. Quadro and Tesla will have uncapped performance for compute tasks (32 and 64 float operations)
Shit! I was hoping that "1500" will triple core performance, but nvidia seems to go on the road "is more marketing of the cores" "people buy digits in the brochures"
miracles do not happen
I'm very disappointed
I don't see that observation. In fact, all current observations are that it outperforms everything on the market. The compute performance is explained below.
"According to Gupta, the second Kepler implementation will include a lot of capability not present in these first gaming-oriented products. In particular, it will have a lot more double-precision capability (which is not required for most graphics applications) and include new compute-specific features. And of course the raw power of these chips will be quite a bit higher than the mid-range graphics version introduced this week.
Although the company is not yet giving any of the speeds and feeds on the second Kepler, one would expect the core count and peak double precision performance to be two to three times higher, and memory bandwidth to get at least a 50 percent bump. Clock speed will almost certainly be whittled down from the current 1.3 GHz on the Tesla M2090, but perhaps not so aggressively as in these first Kepler gaming parts."
Last edited by Remydrh; March 23rd, 2012 at 22:26. Reason: typo
load/store and floatingpoint units, mainly the only stuff you should care about when looking for GPU in rendering, run in kepler at core speed, before there was the so called two half warps at 2x (shader) speed so that the hardware was able to provide for 2 clocks a 4 halfwarps. now that shader clock is gone you have to double the hardware. so, to say nothing about the hardware scheduler removal, do a general kepler cores /2 do get something in line with fermi.
I wonder, will 3dsmax 2013 bring support for Kepler GPUs?
or possibility to load and unload geometry on the fly, so memory limit will no longer be a problem for large scenes...
I am wondering when Kepler support will happen for Iray. I am hoping it is packaged with the upcoming release of Max 2013.
Anyone have anymore insight or news on this?