reanimator/bee
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

Refine NVIDIA benchmark phase timing

Browse Source
This commit is contained in:
Mikhail Chusavitin
2026-04-14 14:12:06 +03:00
parent b1a5035edd
commit 2be7ae6d28
6 changed files with 450 additions and 133 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
audit/internal/platform/benchmark_report.go
View File

@@ -88,10 +88,10 @@ func renderBenchmarkReportWithCharts(result NvidiaBenchmarkResult) string {
b.WriteString("- Thermal and power limits are inferred from NVIDIA clock-event counters plus sustained telemetry.\n")
b.WriteString("- `result.json` is the canonical machine-readable source for the run.\n\n")
b.WriteString("**Compute score** is derived from two phases:\n\n")
b.WriteString("- **Synthetic** — each precision type (fp8, fp16, fp32, fp64, fp4) runs alone for a dedicated window. ")
b.WriteString("- **Synthetic** — each precision type (int8, fp8, fp16, fp32, fp64, fp4) runs alone for a dedicated window. ")
b.WriteString("Measures peak throughput with the full GPU dedicated to one kernel type. ")
b.WriteString("Each result is normalised to fp32-equivalent TOPS using precision weights: ")
b.WriteString("fp64 ×2.0 · fp32 ×1.0 · fp16 ×0.5 · fp8 ×0.25 · fp4 ×0.125.\n")
b.WriteString("fp64 ×2.0 · fp32 ×1.0 · fp16 ×0.5 · int8 ×0.25 · fp8 ×0.25 · fp4 ×0.125.\n")
b.WriteString("- **Mixed** — all precision types run simultaneously (combined phase). ")
b.WriteString("Reflects real inference workloads where fp8 matrix ops, fp16 attention and fp32 accumulation compete for bandwidth and SM scheduler slots.\n\n")
b.WriteString("**Formula:** `Compute = Synthetic × (1 + MixedEfficiency × 0.3)`\n\n")