Modders unlock hidden AI compute on RTX 20/30/40 GPUs: RTX 4090 hits 331.6 TFLOPS FP16
mariolefebvre · x · 2026-10-26
Modders have removed software-level limits on certain Tensor Core operations across NVIDIA's RTX 20, 30 and 40-series GPUs — no hardware changes or VBIOS flash required. Measured gains: RTX 4090 FP16 Tensor Core throughput jumps from 165.6 to 331.6 TFLOPS (2x), RTX 3090 Ti BF16 from 87.5 to 157.4 TFLOPS (1.8x), RTX 2080 Ti FP16→FP32 from 59.9 to 111 TFLOPS (1.9x), and RTX 3090 PyTorch GEMM from 73.0 to 122.8 TFLOPS (1.7x). The gains mainly benefit AI and compute workloads, not gaming.
Related event: Community Unlocks Hidden Tensor Performance on RTX 20/30/40 GPUs(3 posts)→