The Geekbench Compute scores of the GeForce RTX 4070 Super have leaked out ahead of the CES 2024 announcement. We’ve got the CUDA and the OpenCL scores of the first Super SKU, demonstrating a healthy improvement in GPGPU performance. Remember that Geekbench compute (or any compute benchmark) isn’t a good representation of gaming performance that relies more heavily on the memory and cache subsystem.
In the CUDA-accelerated OpenCL benchmark, the GeForce RTX 4070 Super scores 219,237 points, up from 186-190K points on the RTX 4070. The RTX 4070 Ti nets ~230K points in the same test, while the 4080 yields over 300K. These numbers line up with the 15% uplift expected over the vanilla 4070.
In the (non-CUDA) OpenCL benchmark, the RTX 4070 Super nets 195K points, up from 176K-179K on the vanilla 4070 and slightly lower than the 206.6K achieved by the RTX 4070 Ti. The RTX 4080 scores 239.9K points in this benchmark, leaving a sufficient gap to the 70-class GPUs.
The RTX 40 Super graphics cards will be announced during NVIDIA’s CES 2024 “Special Broadcast.” The RTX 4070 Super will be the first to launch, followed by the 4070 Ti Super and the 4080 Super. The exact dates are unknown, but all three launches should be done by the end of the month.