WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 9/23/2026, 4:10:52 AM
WebGPU device
Vendornvidia
Device—
Architectureblackwell
Description—
WebGL device
VendorGoogle Inc. (NVIDIA)
RendererANGLE (NVIDIA, NVIDIA GeForce RTX 5070 Ti (0x00002C05) Direct3D11 vs_5_0 ps_5_0, D3D11)
Browser
NameChrome
Version153.0.0.0
Operating system
NameWindows
Version10
Benchmark Results
Throughput from this run, in bytes or operations per second. Higher is faster.
| Benchmark | Throughput |
|---|---|
atomic direct | 3.37 Gupdate/s |
atomic sharded | 5.16 Gupdate/s |
atomic workgroup | 17.8 Gupdate/s |
branch coherent | 14.5 TFLOP/s |
branch divergent | 9.14 TFLOP/s |
branch none | 22.1 TFLOP/s |
branch uniform | 16.6 TFLOP/s |
branch vec4 if | 2.15 Tchoice/s |
branch vec4 select | 2.15 Tchoice/s |
f32<->f16 convert | 11.6 TOP/s |
fp16 div | 5.99 TFLOP/s |
fp16 ln | 6.4 TFLOP/s |
fp16 mat4 FMA | 26 TFLOP/s |
fp16 matvec FMA | 15.4 TFLOP/s |
fp16 pow | 3.22 TFLOP/s |
fp16 rsqrt | 6.35 TFLOP/s |
fp16 scalar FMA | 25.7 TFLOP/s |
fp16 sin/cos | 3.07 TFLOP/s |
fp16 sqrt | 6.1 TFLOP/s |
fp16 vec4 FMA | 24.6 TFLOP/s |
fp32 clamp | 10.7 TOP/s |
fp32 div | 6.47 TFLOP/s |
fp32 fma() builtin | 38.3 TFLOP/s |
fp32 ln | 6.52 TFLOP/s |
fp32 mat4 FMA | 21.6 TFLOP/s |
fp32 matvec FMA | 17.6 TFLOP/s |
fp32 min/max | 21.7 TOP/s |
fp32 pow | 3.32 TFLOP/s |
fp32 rsqrt | 6.62 TFLOP/s |
fp32 scalar FMA | 22 TFLOP/s |
fp32 select | 7.36 TOP/s |
fp32 sin/cos | 3.14 TFLOP/s |
fp32 sqrt | 6.62 TFLOP/s |
fp32 vec4 FMA | 22.7 TFLOP/s |
i32 div | 1.8 TOP/s |
i32 mat4 multiply-add | 23.1 TOP/s |
i32 matvec multiply-add | 18 TOP/s |
i32 scalar multiply-add | 25.8 TOP/s |
i32 vec4 multiply-add | 25.7 TOP/s |
i32<->f32 convert | 5.68 TOP/s |
int8 dp4a | 44.3 TOP/s |
int8 dp4a matvec | 46.7 TOP/s |
layout aos | 35.6 Gparticle/s |
layout aosoa | 53.3 Gparticle/s |
layout soa | 53.4 Gparticle/s |
read dependent chain | 8.02 Mhop/s |
read gather 16kb | 2.32 TB/s |
read gather 4mb | 846 GB/s |
read gather 64mb | 370 GB/s |
read linear | 794 GB/s |
reduction serial | 7.25 Gelement/s |
reduction subgroup | 64.1 Gelement/s |
reduction workgroup | 45.8 Gelement/s |
texture interp (built-in) | 6.54 TB/s |
texture interp (manual) | 8.21 TB/s |
tile direct | 63.9 Goutput/s |
tile shared | 64 Goutput/s |
tile shared padded | 63.9 Goutput/s |
u32 byte pack/unpack | 10.3 TOP/s |
u32 countOneBits | 3.29 TOP/s |
u32 firstLeadingBit | 904 GOP/s |
u32 reverseBits | 59.4 TOP/s |
u32 variable shift | 6.27 TOP/s |
uint8 dp4a | 44.4 TOP/s |
workgroup 128 | 43.7 Gelement/s |
workgroup 256 | 45.7 Gelement/s |
workgroup 64 | 37.6 Gelement/s |
write linear | 800 GB/s |
write scatter 16kb | 10.5 GB/s |
write scatter 4mb | 804 GB/s |
write scatter 64mb | 66 GB/s |