Global memory bandwidth

← All benchmarks · Chapter 5

![Peak bandwidth by operand size](/benchmarks/media/global-bandwidth.svg)

global.copy.peak (GB/s) -- memory/globalCopy.cu: peak device-memory copy bandwidth per operand size (max over unroll x block-size shmoo); series keyed by operand bytes

GPU 1 2 4 8 16 n
Ampere
GeForce RTX 3060 290.76 302.9 305.61 306.35 307.67 2

global.read.peak (GB/s) -- memory/globalRead.cu: peak device-memory read bandwidth per operand size (max over unroll x block-size shmoo); series keyed by operand bytes

GPU 1 2 4 8 16 n
Ampere
GeForce RTX 3060 300.37 311.13 315.52 317.42 318.49 2

global.write.peak (GB/s) -- memory/globalWrite.cu: peak device-memory write bandwidth per operand size (max over unroll x block-size shmoo); series keyed by operand bytes

GPU 1 2 4 8 16 n
Ampere
GeForce RTX 3060 304.65 311.19 306.71 308.38 307.46 2

Submissions

  • rtx-3060 · nwilt-2026-07-12-2.json -- Nicholas Wilt · driver 580.95.05 · CUDA 12.9
  • rtx-3060 · nwilt-2026-08-27.json -- Nicholas Wilt · driver 580.95.05 · CUDA 13.0