Geekbench OpenCL database reveals the first performance tests of NVIDIA’s GeForce RTX 4070 Non-Ti graphics card.
The NVIDIA GeForce RTX 4070 boasts a bandwidth of 504 GB/s, thanks to its 12 GB GDDR6X VRAM with a 21 Gbps speed. This is a 12.5% increase over the RTX 3070’s bandwidth. The GPU core is the AD104-250/251, which has 5888 CUDA cores and a huge 36 MB L2 cache. The L2 cache is nine times the size of the GA104 GPU that powers the RTX 3070.
The RTX 3070 has an 8% higher TGP than the graphics card, which is expected to have a 200W TGP. The average gaming power of the graphics card is 186 Watts, which is 13.5% and 22.5% lower than the RTX 3070 and RTX 3070 Ti, respectively. It has the same power consumption as the RTX 3060, but it should perform as well as or better than the RTX 3080 12 GB, which uses 300-320W. This shows a huge improvement in efficiency.
The NVIDIA GeForce RTX 4070 graphics card got 177,594 points in the Geekbench 5 OpenCL benchmark, which is similar to the RTX 3080 but 16% slower than the Ti version in this benchmark. This is just one “Invalid” result, but the score should be around the same range.
NVIDIA GeForce RTX 4090 OpenCL (Geekbench 5) Benchmark
The NVIDIA GeForce RTX 4070 has lower cores and clock speeds for the AD104 GPU than the Ti model, so it performs lower as expected. The VRAM specs and configuration are the same as the Ti model. It seems that NVIDIA aims for GeForce RTX 3080 12 GB performance with native rasterization & faster than 3080 Ti / 3090 with DLSS 3.
FP32 Compute Horsepower Comparisons (Higher is Better)
The launch date for the NVIDIA GeForce RTX 4070 graphics card is the 13th of April. Reviews for the FE variant will come out a day before and reviews for custom AIC models will come out on the same day. The graphics card will cost $599 US officially but some custom models with better cooling and overclocks will be slightly more expensive.