GeForce RTX 4090 performs up to 67% better than the RTX 3090 Ti in Geekbench 5's CUDA and OpenCL benchmarks

Nvidia launched the RTX 4090 on September 20 alongside two RTX 4080 SKUs. The RTX 4090 packs 16,384 CUDA cores, base/boost clocks of 2.23/2.52 GHz, and 24 GB of GDDR6X memory. Nvidia claimed at the Lovelace reveal event that the GPU performs 2 to 4 times better than the RTX 3090 Ti. Since the card isn't publicly available yet and Nvidia has yet to lift the review embargo, we can’t say for certain whether Team Green’s claims hold up. That said, thanks to recent Geekbench entries for the RTX 4090, we now have an idea about the board’s performance at least in synthetic benchmarks. The information comes to us courtesy of Benchleaks.

The RTX 4090 in question visited Geekbench inside a system configured with a Ryzen 9 7950X, an Asus ROG CROSSHAIR X670E HERO motherboard, and 32 GB of DDR5 RAM running at 2998 MHz. Finally, the GPU ran at a maximum frequency of 2.61 GHz, prompting us to assume that it was overclocked.

In Geekbench 5’s CUDA benchmark, the RTX 4090 scored an insane 433,619 points. As a comparison, the RTX 3090 Ti clocks in at 260,346 points, making the Lovelace flagship almost 67% faster than the Ampere powerhouse. While not quite the 2x that Nvidia promised, the result is still spectacular.

Moving on to the OpenCL Geekbench 5 benchmark, the RTX 4090 produced a score of 366,286. Similar to the CUDA result, the Lovelace board is 59% faster than the RTX 3090 Ti (229,738 points).

To sum it all up, the RTX 4090 is looking like a solid performer compared to the Ampere flagship. Unfortunately, we don’t have definite gaming performance data, but we do know that the board is up to 2.4x faster in Cyberpunk 2077 with DLSS 3 Frame Generation at 4K compared to the RTX 3090 Ti with DLSS 2 Performance mode. Keeping this in mind, we can expect the RTX 4090 vs RTX 3090 Ti battle to be a landslide win for the Lovelace board.

Buy the PNY GeForce RTX 3090 24GB XLR8 on Amazon