I benchmarked my new GTX 980 against my GTX 780 Ti and wanted to share the results. Both cards are average overclockers.
1. EVGA GTX 780 Ti SC, stock boost to 1150. I was only able to reach 1175 on the core and 3700 on the memory
. Flashed with custom BIOS with 175% power limit, no throttling there. This card is one of the lowest clockers here, but still around 15% overclock over stock. Benched with 340.52 drivers.
2. EVGA GTX 980 SC, stock boosts to 1366. Game-stable clocks are 1480 core / 3954 memory
, which somewhat is low range for 980s from what I see here. This is with 125% power limit and +25mV voltage offset 1.243v. Exhibits little throttling due to power limit, but mostly runs at those clocks. Benched with 344.16Average FPS
Here, on average, 980 (1480/3954) is 9% faster
than 780 Ti (1175/3700).
980 performance scaling with overclock: going from stock SC clocks (1366/3506) to 1480/3954 gives 10% performance increase
on average.Minimum FPS
Here, we have some weird out-liners like Crysis 2 and Metro 2033 Redux, where min FPS actually went down. Not only that, but there was a large variance for min FPS reported between multiple runs. I think these 2 just don't report it accurately. On average, min FPS went up 6%
on 980 compared to 780 Ti.
Minimum FPS performance scaling with overclock: going from stock SC clocks (1366/3506) to 1480/3954 gives 17% performance increase
on average. This number is obviously padded by weird Crysis 2 score.
And lastly, Firestrike Extreme and Catzilla
Settings/benchmark results by game:Warning: Spoiler! (Click to show)
GTX 780 TiGTX 980 stockGTX 980 OC
My conclusion: the only game that showed true performance increase is Metro Redux. Other than that, it looks more like sidegrade from GK110. Of course, this is on raw release drivers against matured drivers for Kepler, so performance delta will increase with new drivers for Maxwell. But, for someone with GTX 680/770/HD 280X level of performance, this should be an awesome upgrade. ~Cheers