The two Geekbench 4.0 benchmarks come from regular leaker @TUM_APISAK (via Tom’s Hardware). They both use the same system: Intel Core i9-10900K, ROG MAXIMUS XII EXTREME motherboard, 16GB DDR4, and the 8GB RTX 3070.
One of the tests brought an OpenCL score of 359,349, while the other result is 350,093.
As noted by Twitter user @Sebasti66855537, most of the RTX 2080 Ti results on Geekbench 4.0 lie somewhere between the 345K and 355K mark. Tom’s confirmed this is the case, though it pointed out that there are outliers significantly higher and lower the RTX 3070 scores.
Based on Nvidia’s chart comparing the RTX 3070, RTX 2070, and RTX 2080 Ti, the upcoming card is, on average, around 8 percent faster than $1000+ last-gen product and 68 percent faster than the 2070.
Nvidia will have cherrypicked the games and apps that favor the RTX 3070, of course, and two Geekbench results don’t tell the whole picture.
Responding to a question on whether the RTX 3070 is “better” than the RTX 2080 Ti, APISAK said they believe the latter card is the better of the two, except when it comes to ray tracing.
The RTX 2080 Ti has 4,352 GPU cores, 272 texture units, and 88 ROPs, while the RTX 3070 has 5,888 cores, 184 TMUs, and 96 ROPs. Other early benchmarks have also suggested the new card will have the edge, though if the launch is anything like the previous RTX 3000-series, the worst element of the RTX 3070 will be getting hold of one—a scenario AMD is trying to avoid with its Radeon RX 6000 series.
We’ve also heard that Nvidia is working on a 16GB version of the RTX 3070, but new rumors claim the alleged December launch of both this card and the 20GB RTX 3080 have been canceled.