| 08 December 2009
I believe that the lower minimum frames on the 5970 are caused by drivers. Sometimes when the scene changes, it takes awhile for it to load it.. resulting in a 1 second loading chop (I've verified this with another 5970 user). You also have to take into account that tessellation is done on the CUDA cores for the GTX 480. These tests are ran at 1x AF and 0x AA and AA/AF is also done on the CUDA cores. There is much speculation that the GTX 480 will have a very hard time with tessellation and AA/AF at the same time. The 5XXX series has it's own dedicated tessellation core, so it doesn't suffer from this. We will not have a true 5970 vs GTX 480 vs 5870 comparison until we get some 1920x1200, 4x AA, 16x AF results.
The GTX 480 & 5870 scores are from NVIDIA, the 5970 scores are from me, and the graph was created by IEATFISH. I used the same Heaven Benchmark settings as them. They were running at 3.33GHz so I lowered my CPU to those clocks for the stock 5970 run. I'm using the Catalyst 10.3 beta drivers on Windows 7 64-bit.
Yes, the GTX 480 is single-GPU but it's the best NVIDIA has at the moment. The GTX 485 should beat the 5970 but there is no release date and it could be months before it is released. I don't believe it's even been officially announced.
We will have to wait and see how much they want to charge for the GTX 480 before considering the card a hit or miss. It would need to be greatly cheaper than the 5970.
Edit #1: Removed the old benchmarks, and merged into 1 chart. The chart is messy, but if you know the source of the original chart (Youtube stream) you probably understand why it looks so messy. The information is accurate though.
Edit #2: Apparently there is a Heaven Benchmark v1.1 in the works that increases FPS by 30% when tessellation is enabled. No clue on if this actually exist, or if it is more geared for ATI or NVIDIA.



