WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 9/14/2026, 2:52:24 PM
WebGPU device
Vendornvidia
Device—
Architectureblackwell
Description—
WebGL device
VendorGoogle Inc. (NVIDIA)
RendererANGLE (NVIDIA, NVIDIA GeForce RTX 5090 (0x00002B85) 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 | 2.94 Gupdate/s |
atomic sharded | 4.73 Gupdate/s |
atomic workgroup | 16.6 Gupdate/s |
branch coherent | 39.1 TFLOP/s |
branch divergent | 21.6 TFLOP/s |
branch none | 49.5 TFLOP/s |
branch uniform | 44.8 TFLOP/s |
branch vec4 if | 3.3 Tchoice/s |
branch vec4 select | 3.28 Tchoice/s |
f32<->f16 convert | 27.5 TOP/s |
fp16 div | 13.6 TFLOP/s |
fp16 ln | 14.6 TFLOP/s |
fp16 mat4 FMA | 39 TFLOP/s |
fp16 matvec FMA | 44.4 TFLOP/s |
fp16 pow | 7.4 TFLOP/s |
fp16 rsqrt | 14.5 TFLOP/s |
fp16 scalar FMA | 28.8 TFLOP/s |
fp16 sin/cos | 7.27 TFLOP/s |
fp16 sqrt | 15 TFLOP/s |
fp16 vec4 FMA | 52.3 TFLOP/s |
fp32 clamp | 25 TOP/s |
fp32 div | 13.4 TFLOP/s |
fp32 fma() builtin | 83.8 TFLOP/s |
fp32 ln | 14.8 TFLOP/s |
fp32 mat4 FMA | 43.2 TFLOP/s |
fp32 matvec FMA | 40.2 TFLOP/s |
fp32 min/max | 49.8 TOP/s |
fp32 pow | 7.48 TFLOP/s |
fp32 rsqrt | 13.9 TFLOP/s |
fp32 scalar FMA | 49.5 TFLOP/s |
fp32 select | 16.8 TOP/s |
fp32 sin/cos | 7.5 TFLOP/s |
fp32 sqrt | 13.4 TFLOP/s |
fp32 vec4 FMA | 50.9 TFLOP/s |
i32 div | 3.79 TOP/s |
i32 mat4 multiply-add | 50.3 TOP/s |
i32 matvec multiply-add | 40.2 TOP/s |
i32 scalar multiply-add | 46.3 TOP/s |
i32 vec4 multiply-add | 50.5 TOP/s |
i32<->f32 convert | 13 TOP/s |
int8 dp4a | 86.5 TOP/s |
int8 dp4a matvec | 91.8 TOP/s |
layout aos | 45 Gparticle/s |
layout aosoa | 53 Gparticle/s |
layout soa | 53 Gparticle/s |
read dependent chain | 7.4 Mhop/s |
read gather 16kb | 4.28 TB/s |
read gather 4mb | 1.45 TB/s |
read gather 64mb | 1.39 TB/s |
read linear | 4.84 TB/s |
reduction serial | 6.95 Gelement/s |
reduction subgroup | 63.5 Gelement/s |
reduction workgroup | 53.1 Gelement/s |
texture interp (built-in) | 13 TB/s |
texture interp (manual) | 15 TB/s |
tile direct | 63.8 Goutput/s |
tile shared | 63.4 Goutput/s |
tile shared padded | 63 Goutput/s |
u32 byte pack/unpack | 21.8 TOP/s |
u32 countOneBits | 7.45 TOP/s |
u32 firstLeadingBit | 2.09 TOP/s |
u32 reverseBits | 139 TOP/s |
u32 variable shift | 14.4 TOP/s |
uint8 dp4a | 91.3 TOP/s |
workgroup 128 | 45.5 Gelement/s |
workgroup 256 | 45.3 Gelement/s |
workgroup 64 | 39.9 Gelement/s |
write linear | 3.35 TB/s |
write scatter 16kb | 15.9 GB/s |
write scatter 4mb | 1.22 TB/s |
write scatter 64mb | 1.46 TB/s |