Article index:
- 1 – EVGA GTX 780 Gallery
- 2 – EVGA GTX 780 GPU Data
- 3 – EVGA GTX 780 Benchmarks
- 4 – EVGA GTX 780 Burn-in Test
3 – EVGA GeForce GpuTest Battle: GTX 780 vs GTX 680
This battle has two fighters from EVGA: the GTX 680 and the GTX 780. And for this battle, I used GpuTest 0.5.0 Both cards have been benchmarked with stock clock speeds and default power/temperature targets. The temperature target is a feature of the GTX 780 only.
Settings: 1920×1080 fullscreen, no AA, duration: 60 seconds.
3.1 – FurMark test
- GTX 680: 3021 points, 50 FPS
- GTX 780: 5711 points, 95 FPS (+89% boost)
3.2 – TessMark test
The TessMark test with a tessellation level of 32 have been used:
- GTX 680: 26600 points, 443 FPS
- GTX 780: 34686 points, 578 FPS (+30% boost)
3.3 – GiMark test
- GTX 680: 5554 points, 92 FPS
- GTX 780: 7255 points, 120 FPS (+30% boost)
3.4 – Piano test
- GTX 680: 850 points, 14 FPS
- GTX 780: 1091 points, 18 FPS (+28% boost)
3.5 – Volpolosion test
- GTX 680: 1982 points, 33 FPS
- GTX 780: 2490 points, 41 FPS (+25% boost)
3.6 – Plot3D test
- GTX 680: 64130 points, 1069 FPS
- GTX 780: 63646 points, 1059 FPS (+0% boost)
3.7 – Triangle test
- GTX 680: 11200 FPS
- GTX 780: 10800 FPS (-4%)
Analysis:
There are 3 groups of test:
– group 1: FurMark
– group 2: TessMark, GiMark, Piano and volplosion
– group 3: Plot3D and Triangle
FurMark is a well-known TDP application that puts a heavy stress on the GPU. But another less known side is that FurMark is a texture intensive application: the rendering of the fur requires a lot of texture fetches. The GTX 680 has a 256-bit memory bus while the GTX 780 has a 384-bit memory bus and a wider memory bus leads to faster memory access. Both memory bus width and the larger number of CUDA cores can explain that big jump in FurMark score.
In the group 2, the GTX 780 offers a performance boost of around 30% which is conform to NVIDIA marketing slide:
The group 3 contains some very light 3D tests (the triangle is a bit like the hello world of 3D programming) and it looks like the GK110 is too powerful for those light tests. But a likely better reason is that the graphics driver is not enough mature for the GK110 and NVIDIA need to bring some minor optimizations for the GTX 780.
Most of the graphics applications belongs to the group 2 and then the GTX 780 is around 30% faster than the GTX 680.
And to close this article, I quickly tested the GTX 780 under Linux and as usual, the GeForce is perfectly supported. I installed the R320.08 under Mint 15 and launched a GpuTest demo:
The performances under Linux are more or less similar to Windows ones.
Article index:
- 1 – EVGA GTX 780 Gallery
- 2 – EVGA GTX 780 GPU Data
- 3 – EVGA GTX 780 Benchmarks
- 4 – EVGA GTX 780 Burn-in Test
Still love the OpenCL performance of a truly GPGPU enabled 7970 😉
Yep, Radeon cards are real beasts in GPU computing. I’ll try to start an OpenCL benchmark asap because I’d really want to compare all those cards together!
http://www.anandtech.com/show/7166/nvidia-announces-quadro-k6000
First fully enabled GK110 2880 cores, clocked higher at 900MHz and 12GB of GDDR5 memory.