WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 9/25/2026, 4:19:30 PM
WebGPU device
Vendorapple
Device—
Architecturemetal-3
Description—
WebGL device
VendorGoogle Inc. (Apple)
RendererANGLE (Apple, ANGLE Metal Renderer: Apple M5 Pro, Unspecified Version)
Browser
NameChrome
Version153.0.0.0
Operating system
NameMac OS
Version10.15.7
Benchmark Results
Throughput from this run, in bytes or operations per second. Higher is faster.
| Benchmark | Throughput |
|---|---|
atomic direct | 12.3 Gupdate/s |
atomic sharded | 10.1 Gupdate/s |
atomic workgroup | 12 Gupdate/s |
branch coherent | 3.15 TFLOP/s |
branch divergent | 3.16 TFLOP/s |
branch none | 7.81 TFLOP/s |
branch uniform | 5.29 TFLOP/s |
branch vec4 if | 957 Gchoice/s |
branch vec4 select | 921 Gchoice/s |
f32<->f16 convert | 6.48 TOP/s |
fp16 div | 3.21 TFLOP/s |
fp16 ln | 3.25 TFLOP/s |
fp16 mat4 FMA | 6.72 TFLOP/s |
fp16 matvec FMA | 7.5 TFLOP/s |
fp16 pow | 2.05 TFLOP/s |
fp16 rsqrt | 4.06 TFLOP/s |
fp16 scalar FMA | 10.5 TFLOP/s |
fp16 sin/cos | 632 GFLOP/s |
fp16 sqrt | 3.11 TFLOP/s |
fp16 vec4 FMA | 13.9 TFLOP/s |
fp32 clamp | 3.95 TOP/s |
fp32 div | 3.96 TFLOP/s |
fp32 fma() builtin | 7.82 TFLOP/s |
fp32 ln | 3.92 TFLOP/s |
fp32 mat4 FMA | 4.95 TFLOP/s |
fp32 matvec FMA | 4.8 TFLOP/s |
fp32 min/max | 7.9 TOP/s |
fp32 pow | 2.04 TFLOP/s |
fp32 rsqrt | 3.91 TFLOP/s |
fp32 scalar FMA | 7.82 TFLOP/s |
fp32 select | 3.06 TOP/s |
fp32 sin/cos | 527 GFLOP/s |
fp32 sqrt | 3.96 TFLOP/s |
fp32 vec4 FMA | 7.81 TFLOP/s |
i32 div | 525 GOP/s |
i32 mat4 multiply-add | 4.05 TOP/s |
i32 matvec multiply-add | 3.69 TOP/s |
i32 scalar multiply-add | 4.09 TOP/s |
i32 vec4 multiply-add | 4.09 TOP/s |
i32<->f32 convert | 1.97 TOP/s |
int8 dp4a | 2.67 TOP/s |
int8 dp4a matvec | 3.83 TOP/s |
layout aos | 14.6 Gparticle/s |
layout aosoa | 20.1 Gparticle/s |
layout soa | 23.6 Gparticle/s |
read dependent chain | 9.81 Mhop/s |
read gather 16kb | 2.8 TB/s |
read gather 4mb | 380 GB/s |
read gather 64mb | 69.9 GB/s |
read linear | 288 GB/s |
reduction serial | 5.13 Gelement/s |
reduction subgroup | 27.4 Gelement/s |
reduction workgroup | 11.6 Gelement/s |
texture interp (built-in) | 1.36 TB/s |
texture interp (manual) | 1.71 TB/s |
tile direct | 19.6 Goutput/s |
tile shared | 17.8 Goutput/s |
tile shared padded | 16.9 Goutput/s |
u32 byte pack/unpack | 1.47 TOP/s |
u32 countOneBits | 1.98 TOP/s |
u32 firstLeadingBit | 960 GOP/s |
u32 reverseBits | 1.97 TOP/s |
u32 variable shift | 1.94 TOP/s |
uint8 dp4a | 3.51 TOP/s |
workgroup 128 | 17.8 Gelement/s |
workgroup 256 | 17.7 Gelement/s |
workgroup 64 | 17.8 Gelement/s |
write linear | 304 GB/s |
write scatter 16kb | 15.6 GB/s |
write scatter 4mb | 260 GB/s |
write scatter 64mb | 50.7 GB/s |