Yesterday, we received confirmation on some of the specifications for NVIDIA's upcoming GeForce RTX 50 Series SUPER refresh, which is currently rumored to arrive in early 2026. Outside of the GeForce RTX 5070 SUPER, which also increases the CUDA Core count slightly, the significant improvement this time around is the 50% increase in VRAM capacity for all SUPER GPUs.

This will allow the SUPER cards to better handle path tracing with max settings in titles like DOOM: The Dark Ages, while also expanding the cutting-edge AI capabilities of these GPUs. Rounding out the SUPER refresh, today we've got specs for the upcoming GeForce RTX 5080 SUPER graphics card courtesy of the same insider, @kopite7kimi.
Responding to a post about the potential specs for the GPU over on X, @kopite7kimi confirmed that the current specs indicate that the GeForce RTX 5080 SUPER will feature the exact SM and CUDA Core count as the baseline GeForce RTX 5080, albeit with a faster 3 GHz Boost Clock speed, and 24GB of fast GDDR7 clocked at a higher 32 Gbps speed.
Interestingly, the GeForce RTX 5080 SUPER will see its power rating increase by 15% from 360W on the baseline model to 415W. This is a notable increase in power draw, so it will be interesting to see how this translates to performance. It's unlikely that the GeForce RTX 5080 SUPER will match the GeForce RTX 4090, which is around 20% faster according to our benchmarks and testing; however, it should close the gap.
Ideally, we'd love to see NVIDIA increase the CUDA Core count on the GeForce RTX 5080 SUPER to achieve performance comparable to that of the RTX 4090, which would be fantastic when paired with DLSS 4 and Multi Frame Generation at 4K. As it stands, based on the specs of the RTX 5080 SUPER, gaming performance might only be around 5% faster than the baseline model.
| GPU Specs | GeForce RTX 5080 | GeForce RTX 5080 SUPER | GeForce RTX 5070 Ti | GeForce RTX 5070 Ti SUPER | GeForce RTX 5070 | GeForce RTX 5070 SUPER |
|---|---|---|---|---|---|---|
| Architecture | Blackwell | Blackwell | Blackwell | Blackwell | Blackwell | Blackwell |
| Process | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N |
| CUDA Cores | 10752 | 10752 | 8960 | 8960 | 6144 | 6400 |
| Tensor Cores (AI) | 330 (5th Gen) | 330 (5th Gen) | 280 (5th Gen) | 280 (5th Gen) | 192 (5th Gen) | 200 (5th Gen) |
| Ray Tracing Cores | 84 (4th Gen) | 84 (4th Gen) | 70 (4th Gen) | 70 (4th Gen) | 48 (4th Gen) | 50 (4th Gen) |
| GPU Boost Clock | 2617 MHz | TBC | 2452 MHz | TBC | 2512 MHz | TBC |
| Memory | 16GB GDDR7 | 24GB GDDR7 | 16GB GDDR7 | 24GB GDDR7 | 12GB GDDR7 | 18GB GDDR7 |
| Memory Interface | 256 Bit | 256 Bit | 256 Bit | 256 Bit | 192 Bit | 192 Bit |
| Bandwidth | 960 GB/sec | 1024 GB/sec | 896 GB/sec | 896 GB/sec | 672 GB/sec | 672 GB/sec |
| TGP | 360W | 415W | 300W | 350W | 250W | 275W |




