NVIDIA Kepler speculation thread

592x528px-LL-bb90b16a_2rp4r6e.jpeg


Control application for the turbo clocks in Kepler.
 

It seems the GTX 680 is a whopping 88% faster than my stock GTX 570 with extreme tess. :oops: Same settings except the windowed option. I don't think the windowed option would make any difference anyway. If anything else it should impose a handicap on the GTX 680.

heaven2012-03-1923-01-08-41.png


Here's hoping we will see some great improvements, in real life, non ridiculously tesselated synthetic benchmarks as well.

Unfortunately the steep prices quickly makes me shift focus to the 670. If it could give me 50% more performance compared to my 570, at a reasonable price (<400€) I am game, otherwise I'll see this gen out I guess.
 
There's a distinction to be made between what GCN as an overarching initiative promises is possible and what a specific implementation can do.
 
Min FPS are very inconsistend in Unigine Heaven due to loading stutters between the scenes. You have to press enter and cycle through all the scenes at least once for the Min FPS to become correct values during the run.
 
It seems the GTX 680 is a whopping 88% faster than my stock GTX 570 with extreme tess. :oops: Same settings except the windowed option. I don't think the windowed option would make any difference anyway. If anything else it should impose a handicap on the GTX 680.

heaven2012-03-1923-01-08-41.png


Here's hoping we will see some great improvements, in real life, non ridiculously tesselated synthetic benchmarks as well.

Unfortunately the steep prices quickly makes me shift focus to the 670. If it could give me 50% more performance compared to my 570, at a reasonable price (<400€) I am game, otherwise I'll see this gen out I guess.

Nice, was wondering 570 # on those settings offhand, I sold my 570 SLI to go for GTX 680 or SLI of that even a bit ago. The GTX 680 is shaping up well!
 
Here's something I don't get about "boost." In theory if your GPU is operating below whatever power/heat threshold they have, then the GPU isn't getting pushed hard and hence it's unlikely you'll need the boost in power. Well, except for benchmarking I suppose. So you'll get more rendering power but if it's already at some ridiculous number that what's the point? Overclocking? Wasting electricity just for the sake of wasting electricity?

On the other hand when you actually do NEED the boost, you're going to be limited by the power/heat threshold since that's when your card is likely to be getting pushed the hardest. But since you're at the threshold you won't be getting the boost.

Powertune I could always understand. It limited power to prevent you from exceeding the safe operating parameters of the chip and card. On the other hand it didn't articially increase power consumption in situations where it wasn't needed.

Perhaps the forthcoming reviews will better explain why boost is there in the first place. It isn't like the CPU world where singled threaded apps are still common and hence you can boost the clocks due to 3/4 of the cores going virtually unused.

Regards,
SB
 
It depends on what's limiting the performance, doesn't it? If you're hitting the perfoamce limit due to a bottleneck in block A, but block A is far away from being the most power hungry of the chip, them increasing clocks is a useful thing to do.
 
Yeah SB, you're making some really sweeping assumptions there about the correlation between fps and power consumption.
 
There's a distinction to be made between what GCN as an overarching initiative promises is possible and what a specific implementation can do.

That's it: architecture vs. implementation. Although you could technically call it half-rate for DPADD I might add for completeness' sake.
 
Conclusion: The OC'd gtx680 is faster using 4x MSAA in Batman AC than the OC'd hd7970 is using FXAA with all other settings equal!
I don't follow. 4xAA > FXAA, but 112 < 130 and very high < max (according to SimBy). The 680 appears to be slower than the 7970 in that game at those settings, which aren't at all equal. How did you reach the opposite conclusion? :???:

(I have no idea which card is faster, how CPU affects Batman AC, or if Batman AC even has a built-in standard benchmark, so I'm just going by the data in your post.)
 
I don't follow. 4xAA > FXAA, but 112 < 130 and very high < max (according to SimBy). The 680 appears to be slower than the 7970 in that game at those settings, which aren't at all equal. How did you reach the opposite conclusion? :???:

(I have no idea which card is faster, how CPU affects Batman AC, or if Batman AC even has a built-in standard benchmark, so I'm just going by the data in your post.)

The particular hd7970 benchmarked Batman AC screenshot you are referring to completely differs from what review sites get. The guy with the gtx680 might be trolling, but the particular benchmark you are referring to is a DEFINITE troll.

Batman AC HD7970 DX11 Max settings, all different AA modes
1920x1200: 81fps http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/20
1920x1200: 68.6 fps http://www.techpowerup.com/reviews/AMD/HD_7970/8.html
1920x1200: 94 fps http://www.hardwareheaven.com/revie...-graphics-card-review-batman-arkham-city.html

hd7970 @ 1265mhz / 1505mhz: 103fps http://www.hardwareheaven.com/revie...-graphics-card-review-batman-arkham-city.html
 
On the other hand when you actually do NEED the boost, you're going to be limited by the power/heat threshold since that's when your card is likely to be getting pushed the hardest. But since you're at the threshold you won't be getting the boost.

Or maybe the program running on the GPU isn't able to keep all the cores in the GPU active at all times, so even if it's fully utilizing the GPU, there is enough of it inactive that it's not at it's power threshold and can therefore increase clocks.
 
Back
Top