WebGPU Micro-Benchmark Result
Per-operation GPU timings for the profiled device. Benchmark details available in the webgpu-bench open source project.
Device
Recorded 9/12/2026, 5:33:03 PM
WebGPU device
Vendorapple
Deviceapple
Architectureapple
Descriptionapple
WebGL device
VendorApple Inc.
RendererApple GPU
Browser
NameSafari
Version27.0
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 | — |
atomic sharded | — |
atomic workgroup | — |
branch coherent | 1.71 TFLOP/s |
branch divergent | 948 GFLOP/s |
branch none | 2.69 TFLOP/s |
branch uniform | 1.75 TFLOP/s |
branch vec4 if | — |
branch vec4 select | — |
f32<->f16 convert | 3.55 TOP/s |
fp16 div | 1.07 TFLOP/s |
fp16 ln | 929 GFLOP/s |
fp16 mat4 FMA | 2.33 TFLOP/s |
fp16 matvec FMA | 2.14 TFLOP/s |
fp16 pow | 437 GFLOP/s |
fp16 rsqrt | 875 GFLOP/s |
fp16 scalar FMA | 2.8 TFLOP/s |
fp16 sin/cos | 223 GFLOP/s |
fp16 sqrt | 840 GFLOP/s |
fp16 vec4 FMA | 3.37 TFLOP/s |
fp32 clamp | 1.34 TOP/s |
fp32 div | 951 GFLOP/s |
fp32 fma() builtin | 2.72 TFLOP/s |
fp32 ln | 844 GFLOP/s |
fp32 mat4 FMA | 2.22 TFLOP/s |
fp32 matvec FMA | 1.76 TFLOP/s |
fp32 min/max | 2.69 TOP/s |
fp32 pow | 707 GFLOP/s |
fp32 rsqrt | 841 GFLOP/s |
fp32 scalar FMA | 2.69 TFLOP/s |
fp32 select | 976 GOP/s |
fp32 sin/cos | 306 GFLOP/s |
fp32 sqrt | 833 GFLOP/s |
fp32 vec4 FMA | 2.35 TFLOP/s |
i32 div | 155 GOP/s |
i32 mat4 multiply-add | 844 GOP/s |
i32 matvec multiply-add | 805 GOP/s |
i32 scalar multiply-add | 844 GOP/s |
i32 vec4 multiply-add | 1.2 TOP/s |
i32<->f32 convert | 630 GOP/s |
int8 dp4a | 536 GOP/s |
int8 dp4a matvec | 499 GOP/s |
layout aos | — |
layout aosoa | — |
layout soa | — |
read dependent chain | — |
read gather 16kb | 880 GB/s |
read gather 4mb | 102 GB/s |
read gather 64mb | 34.8 GB/s |
read linear | 89.5 GB/s |
reduction serial | — |
reduction subgroup | — |
reduction workgroup | — |
texture interp (built-in) | — |
texture interp (manual) | — |
tile direct | — |
tile shared | — |
tile shared padded | — |
u32 byte pack/unpack | 614 GOP/s |
u32 countOneBits | 422 GOP/s |
u32 firstLeadingBit | 421 GOP/s |
u32 reverseBits | 421 GOP/s |
u32 variable shift | 535 GOP/s |
uint8 dp4a | 846 GOP/s |
workgroup 128 | — |
workgroup 256 | — |
workgroup 64 | — |
write linear | 86.8 GB/s |
write scatter 16kb | 810 GB/s |
write scatter 4mb | 47.4 GB/s |
write scatter 64mb | 20.3 GB/s |