A next gen Nvidia GeForce flagship graphics card might have just been spotted strutting its stuff in 3DMark Time Spy. Hardware Leaks, run by Twitter leakster RoGame, asserts that he unearthed an Nvidia Ampere gaming card in the 3DMark databases and its performance is pretty impressive - over 30 per cent faster in the DirectX 12 benchmark than a stock Nvidia GeForce RTX 2080 Ti Founders Edition.
As per usual for a pre-release graphics card, 3DMark's reporting of its specs are a bit skew-whiff. While the vendor (Nvidia) and GPU boost clock (1,935MHz) will likely be reliable, the memory clock of 6.0GHz, doesn't fit into line with any memory type currently out there. RoGame says that in his experience such odd numbers come from pre-release drivers and/or a new type of memory subsystem in use.
Above you can see a useful 3DMark Time Spy benchmark result comparison chart which puts the mystery Nvidia CPU into perspective. Given the relatively low-ish reported GPU boost speed the result is impressive. If this is a Founders Edition under test you can see an 30+ per cent uplift in DX12 gaming-type performance at this pre-release stage.
RoGame says he is "100 per cent" sure that the 3DMark Time Spy result he discovered and represented in the chart is authentic. It was a 'private result', a run completed by an Nvidia employee - but of course he doesn't want to disclose the employee name in case of repercussions.
Other recent Twitter leaks and gossip firm up the theory that Nvidia is testing three GA102 GPU-powered graphics card variants as follows:
- 2nd Gen TITAN RTX: GA102-400-A1, 5376 CUDA, 24GB GDDR6X, 17Gbps, 384bit, 816GB/s
- RTX 3090: GA102-300-A1, 5248 CUDA, 12GB GDDR6X, 21Gbps, 384bit, 1008GB/s
- RTX 3080: GA102-200-Kx-A1, 4352 CUDA, 10GB GDDR6X, 19Gbps, 320bit, 760GB/s
Remember to take the above rumours and leaks with a pinch of salt. I'm sure there will be plenty more spills on the path to the first Ampere-based GeForce official product reveal which could be as soon as 2 or 3 months from now.