WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 11/3/2026, 3:34:55 AM
WebGPU device
Vendornvidia
Device—
Architecturelovelace
Description—
WebGL device
VendorGoogle Inc. (NVIDIA)
RendererANGLE (NVIDIA, NVIDIA GeForce RTX 4090 (0x00002684) Direct3D11 vs_5_0 ps_5_0, D3D11)
Browser
NameChrome
Version156.0.0.0
Operating system
NameWindows
Version10
Benchmark Results
Throughput from this run, in bytes or operations per second. Higher is faster.
| Benchmark | Throughput |
|---|---|
atomic direct | 2.34 Gupdate/s |
atomic sharded | 3.95 Gupdate/s |
atomic workgroup | 13.6 Gupdate/s |
branch coherent | 26.8 TFLOP/s |
branch divergent | 16.4 TFLOP/s |
branch none | 36.8 TFLOP/s |
branch uniform | 30 TFLOP/s |
branch vec4 if | 3.7 Tchoice/s |
branch vec4 select | 3.71 Tchoice/s |
f32<->f16 convert | 23.5 TOP/s |
fp16 div | 10.6 TFLOP/s |
fp16 ln | 10.8 TFLOP/s |
fp16 mat4 FMA | 27.3 TFLOP/s |
fp16 matvec FMA | 36.9 TFLOP/s |
fp16 pow | 5.38 TFLOP/s |
fp16 rsqrt | 10.8 TFLOP/s |
fp16 scalar FMA | 38.5 TFLOP/s |
fp16 sin/cos | 5.13 TFLOP/s |
fp16 sqrt | 10.8 TFLOP/s |
fp16 vec4 FMA | 41.7 TFLOP/s |
fp32 clamp | 18.8 TOP/s |
fp32 div | 10.9 TFLOP/s |
fp32 fma() builtin | 64.8 TFLOP/s |
fp32 ln | 10.8 TFLOP/s |
fp32 mat4 FMA | 37.3 TFLOP/s |
fp32 matvec FMA | 37.7 TFLOP/s |
fp32 min/max | 37.3 TOP/s |
fp32 pow | 5.38 TFLOP/s |
fp32 rsqrt | 10.7 TFLOP/s |
fp32 scalar FMA | 37.2 TFLOP/s |
fp32 select | 9.68 TOP/s |
fp32 sin/cos | 5.43 TFLOP/s |
fp32 sqrt | 10.7 TFLOP/s |
fp32 vec4 FMA | 39 TFLOP/s |
i32 div | 3.06 TOP/s |
i32 mat4 multiply-add | 33.8 TOP/s |
i32 matvec multiply-add | 33.9 TOP/s |
i32 scalar multiply-add | 41.7 TOP/s |
i32 vec4 multiply-add | 44 TOP/s |
i32<->f32 convert | 10.1 TOP/s |
int8 dp4a | 65.3 TOP/s |
int8 dp4a matvec | 83.1 TOP/s |
layout aos | 29.4 Gparticle/s |
layout aosoa | 42 Gparticle/s |
layout soa | 42.2 Gparticle/s |
read dependent chain | 9.47 Mhop/s |
read gather 16kb | 3.64 TB/s |
read gather 4mb | 1.22 TB/s |
read gather 64mb | 545 GB/s |
read linear | 3.44 TB/s |
reduction serial | 3.37 Gelement/s |
reduction subgroup | 48 Gelement/s |
reduction workgroup | 37.5 Gelement/s |
texture interp (built-in) | 9.95 TB/s |
texture interp (manual) | 12.6 TB/s |
tile direct | 28 Goutput/s |
tile shared | 48.4 Goutput/s |
tile shared padded | 47.1 Goutput/s |
u32 byte pack/unpack | 19.9 TOP/s |
u32 countOneBits | 5.33 TOP/s |
u32 firstLeadingBit | 1.16 TOP/s |
u32 reverseBits | 101 TOP/s |
u32 variable shift | 9.19 TOP/s |
uint8 dp4a | 65.3 TOP/s |
workgroup 128 | 36.7 Gelement/s |
workgroup 256 | 39.1 Gelement/s |
workgroup 64 | 29.7 Gelement/s |
write linear | 987 GB/s |
write scatter 16kb | 11 GB/s |
write scatter 4mb | 604 GB/s |
write scatter 64mb | 338 GB/s |