TU104 - The Same GPU as RTX 2080
Just like the GeForce RTX 2060 SUPER has the same TU106 as the RTX 2060 had, the new RTX 2070 SUPER uses the TU104 GPU which is the same as the RTX 2080 -- up from the TU106 used on the normal RTX 2070. And again, like its RTX 2060 SUPER sibling the new RTX 2070 SUPER has an upgraded GPU with more SMs and CUDA cores, as well as higher GPU clocks.
NVIDIA's new GeForce RTX 2070 SUPER and its TU104 GPU has 2560 CUDA cores over the 2304 CUDA cores in TU106 that powers the RTX 2060, while there's now 40 SMs over the 36 SMs in the RTX 2060. The GPU clocks increase a fair amount as well with the new RTX 2070 SUPER with a base GPU clock of 1605MHz versus the 1410MHz on the RTX 2070 while the boost GPU clocks go up to 1770MHz versus the 1620MHz on the normal RTX 2070.
We have 9.1 TFLOPs of FP32 compute performance on the RTX 2070 SUPER against the 7.5 TFLOPs on the RTX 2070, with 72 TFLOPs of Tensor FLOPs on the RTX 2070 SUPER versus the 60 TFLOPs on the regular RTX 2070.
No GDDR6 Upgrades Here
Unlike the GeForce RTX 2060 SUPER, the new RTX 2070 SUPER has no upgrades to its 8GB of GDDR6. We have 448GB/sec of memory bandwidth on the RTX 2070 SUPER and even when writing this review, the 448GB/sec on the RTX 2060 SUPER feels so ridiculously high that I genuinely double checked it as I was writing this part of the review. Crazy stuff. But sorry, no upgrades in the RTX 2070 SUPER and its 8GB of GDDR6.
- Page 1 [Introduction & Pricing]
- Page 2 [Specs: Turing GPU & No GDDR6 Upgrade]
- Page 3 [TWIN FROZR 7 Cooling]
- Page 4 [Detailed Look]
- Page 5 [Test System Specs]
- Page 6 [Benchmarks - Synthetic]
- Page 7 [Benchmarks - 1080p]
- Page 8 [Benchmarks - 1440p]
- Page 9 [Benchmarks - 4K]
- Page 10 [Thermals & Power]
- Page 11 [Final Thoughts]