WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 9/24/2026, 5:43:55 PM
WebGPU device
Vendor—
Device—
Architecture—
Description—
WebGL device
VendorBrave
RendererBrave
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 | 3.4 Gupdate/s |
atomic sharded | 5.16 Gupdate/s |
atomic workgroup | 7.65 Gupdate/s |
branch coherent | 11.8 TFLOP/s |
branch divergent | 6.83 TFLOP/s |
branch none | 15 TFLOP/s |
branch uniform | 13.5 TFLOP/s |
branch vec4 if | 1.59 Tchoice/s |
branch vec4 select | 1.6 Tchoice/s |
f32<->f16 convert | 8.29 TOP/s |
fp16 div | 4.36 TFLOP/s |
fp16 ln | 4.37 TFLOP/s |
fp16 mat4 FMA | 13.2 TFLOP/s |
fp16 matvec FMA | 10.3 TFLOP/s |
fp16 pow | 2.23 TFLOP/s |
fp16 rsqrt | 4.47 TFLOP/s |
fp16 scalar FMA | 8.83 TFLOP/s |
fp16 sin/cos | 2.19 TFLOP/s |
fp16 sqrt | 4.47 TFLOP/s |
fp16 vec4 FMA | 8.54 TFLOP/s |
fp32 clamp | 7.49 TOP/s |
fp32 div | 4.36 TFLOP/s |
fp32 fma() builtin | 25.7 TFLOP/s |
fp32 ln | 4.35 TFLOP/s |
fp32 mat4 FMA | 14.5 TFLOP/s |
fp32 matvec FMA | 13 TFLOP/s |
fp32 min/max | 15 TOP/s |
fp32 pow | 2.23 TFLOP/s |
fp32 rsqrt | 4.47 TFLOP/s |
fp32 scalar FMA | 15 TFLOP/s |
fp32 select | 5.07 TOP/s |
fp32 sin/cos | 2.23 TFLOP/s |
fp32 sqrt | 4.46 TFLOP/s |
fp32 vec4 FMA | 15.5 TFLOP/s |
i32 div | 1.18 TOP/s |
i32 mat4 multiply-add | 15.7 TOP/s |
i32 matvec multiply-add | 12.8 TOP/s |
i32 scalar multiply-add | 14.9 TOP/s |
i32 vec4 multiply-add | 15.5 TOP/s |
i32<->f32 convert | 3.92 TOP/s |
int8 dp4a | 29 TOP/s |
int8 dp4a matvec | 32.3 TOP/s |
layout aos | 35.3 Gparticle/s |
layout aosoa | 25.6 Gparticle/s |
layout soa | 24.8 Gparticle/s |
read dependent chain | 8.14 Mhop/s |
read gather 16kb | 1.57 TB/s |
read gather 4mb | 841 GB/s |
read gather 64mb | 289 GB/s |
read linear | 625 GB/s |
reduction serial | 7.59 Gelement/s |
reduction subgroup | 27.6 Gelement/s |
reduction workgroup | 39.6 Gelement/s |
texture interp (built-in) | 4.26 TB/s |
texture interp (manual) | 4.81 TB/s |
tile direct | 27.7 Goutput/s |
tile shared | 26.4 Goutput/s |
tile shared padded | 25.6 Goutput/s |
u32 byte pack/unpack | 7.07 TOP/s |
u32 countOneBits | 2.22 TOP/s |
u32 firstLeadingBit | 630 GOP/s |
u32 reverseBits | 42.8 TOP/s |
u32 variable shift | 4.31 TOP/s |
uint8 dp4a | 29 TOP/s |
workgroup 128 | 37.3 Gelement/s |
workgroup 256 | 35.9 Gelement/s |
workgroup 64 | 37.3 Gelement/s |
write linear | 636 GB/s |
write scatter 16kb | 8.54 GB/s |
write scatter 4mb | 651 GB/s |
write scatter 64mb | 49 GB/s |