According to a 3DMark TimeSpy benchmark unearthed by _rogame, NVIDIA’s GeForce RTX 3080 will be around 30-35% faster than the present Turing flagship, the RTX 2080 Ti. Keep in mind that the leaked benchmark doesn’t explicitly mention the graphics card model, but I believe that it’s the RTX 3080. The gains are in line with what you’d expect from a new generation non-Ti card and on top of that, this is likely just an engineering sample, not the final product:
As you can see from the above graphics provided by the source, the supposed RTX 3080 is around 30% faster than the stock RTX 2080 Ti. Strangely, the reported core and memory clocks are quite low, even for an Engineering Sample. The boost clock of 1935MHz is nearly the same as existing Turing GPUs. This could either be due to the lack of a proper cooling solution or the final boost not being calibrated on this model just yet. Or it could simply mean that NVIDIA isn’t using TSMC’s 7nmn node for consumer Turing cards as expected, rather has opted for a lower density 8nm or 10nm process.
As far as the memory is concerned, you’re looking at a 12Gbps effective clock (1,500MHz= 12,000/8 actual). This is lower than even the RTX 2080 which has a 14Gbps memory clock while the 2080 Super has a clock of 15.5Gbps. This likely means that the GPU is still far from the final product and we could see an RTX 3080 beating the RTX 2080 Ti by as much as 40-50%. Then again, there’s a chance that it could be the 3080 Ti as well but I’ll bet my money on it being the former. Time will tell.