Nvidia GeForce RTX 4080: High-end Ada Lovelace graphics card is up to 47% faster than Ampere on Geekbench and Blender ↺
The GeForce RTX 4080 16 GB will be the only one of its kind to hit shelves on November 16 after Nvidia "Unlaunched" its 12 GB sibling. Initial benchmarks of the penultimate Ada Lovelace graphics card pointed at a ~50% performance uplift over the GeForce RTX 3080 10 GB in gaming and synthetic workloads. It has been tested again, this time on Geekbench, and the results are more or less consistent.
Benchleaks spotted three instances of the Nvidia GeForce RTX 4080 on Geekbench. It scored 248,932 in the OpenCL benchmark, 300,728 in CUDA and 148,838 in Vulkan, making it 25%, 37% and 47% faster than its Ampere-based predecessor. This suggests a ~40% generation-over-generation improvement in rasterization performance, which falls short of pre-launch rumours about the GeForce RTX 4080. It is about 15% faster than the GeForce RTX 3090 Ti based on Geekbench metrics alone. However, that figure is a lot higher in Blender, where it scores 9478.64 (H/T @TUM_APISAK), a full 50% faster than the Ampere flagship.
AMD's Frank Azor said that the Radeon RX 7900 XTX was targeting the GeForce RTX 4080 16 GB and not its beefier sibling. It'll be interesting to see how both cards fare against each other in the real world. Extrapolated performance estimates peg the RDNA 3 flagship to be about 10% slower than the GeForce RTX 4090. One can assume that figure to be closer to 20-30%, given that it was extracted from AMD's first-party data, but that would still be perilous for the GeForce RTX 4080, which is rumoured to cost upwards of US$1,600 in some region. Low availability will almost certainly drive that price even higher.