Processor and graphics card specifications are live

Glossary/GPU/Shader throughput

Shader throughput

Shader throughput is how much arithmetic a GPU's programmable cores complete per second, and it limits performance in shading-heavy modern rendering.

Also known as compute throughput, ALU throughput, shader performance

Measured by SystemCheck

Part of the four-stage run: a 60-second sustained CPU load reporting throughput, decay and consistency; a memory stage reporting bandwidth and latency; a calibrated GPU raster load reporting frame time, pacing and decay; and a path tracer reporting samples per second. The paragraphs below say which metric carries this term.

A GPU's shader cores (CUDA cores, stream processors, EUs) execute the per-vertex and per-pixel programs. Throughput depends on core count, clock, and how well the work keeps those cores occupied. Occupancy is often the real limiter: a shader that stalls on memory or diverges across a wavefront leaves silicon idle regardless of theoretical FLOPS.

Peak FLOPS from a spec sheet is a ceiling reached only by hand-tuned code. Real shaders hit a fraction of it. This is why two cards with similar theoretical numbers can differ substantially in games, and why measured throughput beats calculated throughput.

A browser cannot read a GPU's clock, core count, or native FLOPS. SystemCheck measures the outcome instead: the raster stage measures how fast real shader work completes through WebGL2 or WebGPU, and the path-tracing stage measures ray-shading throughput as samples per second. Both include browser and driver overhead, so they run somewhat below a native-API measurement of the same card, a consistent offset that still compares fairly between machines.

Related terms