fp16 scalar FMA
Same eight independent, 4x-unrolled FMA chains as the fp32 scalar test, but every operand and accumulator is f16, so the chains run entirely in half precision.
Measured in FLOP/s. Higher values mean better performance. Benchmark your device.
Performance distribution
1.12 GFLOP/s208 GFLOP/s38.5 TFLOP/s
1.12 GFLOP/s208 GFLOP/s38.5 TFLOP/s
Performance thresholds
99% of results achieve at least 5.47 GFLOP/s
95% of results achieve at least 113 GFLOP/s
90% of results achieve at least 341 GFLOP/s
Median (50th percentile) 2.97 TFLOP/s
Best performer
Highest recorded performance: 38.5 TFLOP/s. View benchmark profile.
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