NVIDIA Kepler speculation thread

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

Anyone who uses the AC benchmark for a review will be getting very inaccurate results as both NVIDIA and AMD have built optimizations into their driver stacks JUST for benchmark. In-game results in exterior scenes (most of the game) differ wildly from the benchmark.

http://www.xtremesystems.org/forums...orce-GTX-780&p=5071088&viewfull=1#post5071088
 
That's the same for Civ 5 too - the "late game benchmark" (favoured by anandtech) which used to favour Nvidia by such a large margin wasn't indicative of actual gaming performance, so they didn't bench it on Hardware Canucks.

SKYMTL might be sympathetic to the green side but he is certainly honest when it comes to this nonsense.
 
Min Fps

Could the driver be configured to only power boost when it detected low frame rates?

i.e. instead of running at 200 fps could the clock rates be lowered to give say 100 fps. Then when the frame rate drops (say 30 fps), boost the clocks (and power consumption) to compensate (until the chip heated up).
 
everyone seems fixated on comparing GK104 to Tahiti but if you stuck 6gbps memory on pitcairn and upped its clock, GK104 and Tahiti both wouldn't look that great as a GPU with a die of 210mm would be in spitting distance.

reality is Tahiti is underclocked, it will be interesting to see GK104 head room assuming equal or + a few percent preformance @around 1000/6000 .
 
Ninjaprime said:
So its 20% slower than 7970 unless you choke the 7970 with 64x tessellation?
No. You could only draw that conclusion if the 680 was also run with the tessellation factor limited. I can't see any reason one wouldn't expect the 680's performance to also increase with a lower tessellation factor.
 
No. You could only draw that conclusion if the 680 was also run with the tessellation factor limited. I can't see any reason one wouldn't expect the 680's performance to also increase with a lower tessellation factor.

on a 6970 alteast, performance only really increases once you limit it to 8x the difference between 16 and 32 isn't very much and 32 to 64 there isn't any. It was ages ago that i tested so things might be different on newer drivers.
 
Well, the performance clearly increased in the tests which were linked. Perhaps AMD "optimized" is lower than 16 or the 7970 just performs differently?

7970 Off -> On
Min: 25.8 -> 42.9
Avg: 61.6 -> 98.7
Max: 160.2 -> 195.4


While I think the "optimized" setting has merit for actual gameplay, it should obviously be disabled for benchmarking purposes as the whole point is to compare the amount of time it takes to do X amount of work.
 
Last edited by a moderator:
Back
Top