{"name":"gpu-unlock","repo":"Zagitta/gpu-unlock","url":"https://midorreal.com/project/zagitta-gpu-unlock","repository":"https://github.com/Zagitta/gpu-unlock","description":"Your GeForce runs AI at half speed. This fixes it for 20/30/40 series cards. Up to 2× FP16 / BF16 / TF32 tensor throughput","facts":{"language":{"value":"C","source":"https://github.com/Zagitta/gpu-unlock","checked":"22 Nov 2026"},"last_release":{"value":"unraid, 27 Oct 2026","source":"https://github.com/Zagitta/gpu-unlock/releases","checked":"22 Nov 2026"},"releases":{"value":"1","source":"https://github.com/Zagitta/gpu-unlock/releases","checked":"21 Nov 2026"},"contributors":{"value":"9","source":"https://github.com/Zagitta/gpu-unlock/graphs/contributors","checked":"21 Nov 2026"},"archived":{"value":"no","source":"https://github.com/Zagitta/gpu-unlock","checked":"22 Nov 2026"}},"stars":20700.0,"writeup":{"what_it_is":"gpu-unlock is a Linux kernel module that removes a firmware cap on tensor core throughput for FP16, BF16 and TF32 math in supported GeForce RTX 20, 30 and 40 series cards. It runs at boot before the stock NVIDIA driver loads, and does not patch the driver or the VBIOS.","audience":"Linux users with supported GeForce RTX cards who run AI training or inference workloads","claims":[{"kind":"specific","claim":"Doubles FP16, BF16 and TF32 tensor throughput on supported cards","excerpt":"**Up to 2× FP16 / BF16 / TF32 tensor throughput**","status":"not_checked"},{"kind":"specific","claim":"Gives a 25–55% speedup in LLM prefill","excerpt":"**+25–55% LLM prefill**","status":"not_checked"},{"kind":"specific","claim":"Gives a 25–50% speedup in fine-tuning","excerpt":"**+25–50% fine-tuning**","status":"not_checked"},{"kind":"specific","claim":"Shows an RTX 3060 BF16 rate going from 27 TFLOPS to 51 TFLOPS","excerpt":"| RTX 3060 12 GB | 27 TFLOPS | 51 TFLOPS | +87% |","status":"not_checked"},{"kind":"specific","claim":"Outputs are bit-identical before and after the unlock","excerpt":"Outputs are bit-identical before and after; only the speed changes.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-07"},"why_now":null,"owner_supplied":null,"score":{"verdict":"overhyped","hype":100,"reality":54,"gap":46,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-10-27T23:09:01+00:00","text":"Latest release unraid","source":"https://github.com/Zagitta/gpu-unlock/releases"},{"date":"2026-10-18T01:28:31+00:00","text":"Repository created","source":"https://github.com/Zagitta/gpu-unlock"}],"cite":"gpu-unlock is a GitHub project at https://github.com/Zagitta/gpu-unlock written mainly in C, described there as \"Your GeForce runs AI at half speed. This fixes it for 20/30/40 series cards. Up to 2× FP16 / BF16 / TF32 tensor throughput\".","attribution":"Data from Mid or Real, https://midorreal.com/project/zagitta-gpu-unlock, checked 22 November 2026. Attribution with a link is required.","still_real":null}