WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 10/29/2026, 12:56:56 AM
WebGPU device
Vendor—
Device—
Architecture—
Description—
WebGL device
VendorGoogle Inc. (NVIDIA)
RendererANGLE (NVIDIA, NVIDIA GeForce 8800 GTX Direct3D11 vs_5_0 ps_5_0), or similar
Browser
NameFirefox
Version159.0
Operating system
NameWindows
Version10
Benchmark Results
Throughput from this run, in bytes or operations per second. Higher is faster.
| Benchmark | Throughput |
|---|---|
atomic direct | 1.83 Gupdate/s |
atomic sharded | 2.78 Gupdate/s |
atomic workgroup | 2.91 Gupdate/s |
branch coherent | 23.1 TFLOP/s |
branch divergent | 15.6 TFLOP/s |
branch none | 37.4 TFLOP/s |
branch uniform | 27.2 TFLOP/s |
branch vec4 if | 1.37 Tchoice/s |
branch vec4 select | 1.37 Tchoice/s |
f32<->f16 convert | 19.3 TOP/s |
fp16 div | 6.44 TFLOP/s |
fp16 ln | 6.57 TFLOP/s |
fp16 mat4 FMA | 13.9 TFLOP/s |
fp16 matvec FMA | 21 TFLOP/s |
fp16 pow | 3.25 TFLOP/s |
fp16 rsqrt | 6.58 TFLOP/s |
fp16 scalar FMA | 37.9 TFLOP/s |
fp16 sin/cos | 3.13 TFLOP/s |
fp16 sqrt | 6.59 TFLOP/s |
fp16 vec4 FMA | 46 TFLOP/s |
fp32 clamp | 18.4 TOP/s |
fp32 div | 6.6 TFLOP/s |
fp32 fma() builtin | 37.4 TFLOP/s |
fp32 ln | 6.45 TFLOP/s |
fp32 mat4 FMA | 21.2 TFLOP/s |
fp32 matvec FMA | 21.9 TFLOP/s |
fp32 min/max | 37.4 TOP/s |
fp32 pow | 3.26 TFLOP/s |
fp32 rsqrt | 6.48 TFLOP/s |
fp32 scalar FMA | 36.8 TFLOP/s |
fp32 select | 6.02 TOP/s |
fp32 sin/cos | 3.27 TFLOP/s |
fp32 sqrt | 6.45 TFLOP/s |
fp32 vec4 FMA | 34.7 TFLOP/s |
i32 div | 1.74 TOP/s |
i32 mat4 multiply-add | 18.5 TOP/s |
i32 matvec multiply-add | 22.3 TOP/s |
i32 scalar multiply-add | 24.2 TOP/s |
i32 vec4 multiply-add | 26.6 TOP/s |
i32<->f32 convert | 6.4 TOP/s |
int8 dp4a | 38.3 TOP/s |
int8 dp4a matvec | 43.1 TOP/s |
layout aos | 5.82 Gparticle/s |
layout aosoa | 8.03 Gparticle/s |
layout soa | 9.14 Gparticle/s |
read dependent chain | 4.45 Mhop/s |
read gather 16kb | 2.05 TB/s |
read gather 4mb | 780 GB/s |
read gather 64mb | 171 GB/s |
read linear | 671 MB/s |
reduction serial | 2.13 Gelement/s |
reduction subgroup | — |
reduction workgroup | 4.57 Gelement/s |
texture interp (built-in) | 6.52 TB/s |
texture interp (manual) | 8.18 TB/s |
tile direct | 4.27 Goutput/s |
tile shared | 10.7 Goutput/s |
tile shared padded | 10.7 Goutput/s |
u32 byte pack/unpack | 12.3 TOP/s |
u32 countOneBits | 3.24 TOP/s |
u32 firstLeadingBit | 3.07 TOP/s |
u32 reverseBits | 61.7 TOP/s |
u32 variable shift | 5.68 TOP/s |
uint8 dp4a | 38.3 TOP/s |
workgroup 128 | 4.27 Gelement/s |
workgroup 256 | 10.7 Gelement/s |
workgroup 64 | 10.7 Gelement/s |
write linear | 780 GB/s |
write scatter 16kb | 3.2 GB/s |
write scatter 4mb | 195 GB/s |
write scatter 64mb | 182 GB/s |