How much of each GPU works, beside whether anything does
CI / test (push) Failing after 6m42s

The busy figure is nvidia-smi's utilization.gpu: the share of time any
kernel ran, which a kernel on one SM fills. On cuda01 decode read 92% busy
while the SMs had work 69% of the time and issued on 18% of their cycles.

Busy is now said to be that, and beside it: each card's draw against its
power limit (efficiency table, fleet and node charts, the console's GPU
work chart), and where the profiling counters can be read, SMs active, SM
occupancy, tensor cores and DRAM (llmgw_gpu_*_active_percent), from NVML's
GPM on Hopper and later or a dcgm-exporter on the node
(discovery.dcgm_exporter). The GPU map colours by work by default and keeps
busy as a mode. A MIG instance is never published with its card's profile.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Ilker ManapandClaude Opus 5.5 committed 2026-10-31 20:11:26 +01:00
1 parent 7bfe70c8df
commit c866b3a8ee
34 files changed
+954 -52

No files matched your search

+12 -3
View File
@@ -13,8 +13,13 @@ type modelEfficiency struct {
// TokensPerSec is output tokens a second; GPUs the cards it holds now.
TokensPerSec float64
GPUs float64
// Watts is what those cards draw; Util their average utilisation.
Watts, Util float64
// Watts is what those cards draw; Util the average share of time a
// kernel ran on them, which says nothing of how much of each chip it ran
// on. PowerLoad (draw against the power limit) and SMActive (the SMs'
// share of time with work, where GPM or DCGM reads it) are how much, in
// percent; -1 is not known.
Watts, Util float64
PowerLoad, SMActive float64
// PerGPU and PerWh are the two ways of saying how much work the cards
// do: output tokens a second per card, and per watt-hour drawn.
PerGPU, PerWh float64
@@ -34,6 +39,10 @@ var efficiencyQueries = []struct {
func(e *modelEfficiency, v float64) { e.Watts = v }},
{`avg by (model) (llmgw_gpu_utilization_percent * on (node, gpu) group_left (model) llmgw_gpu_assigned)`,
func(e *modelEfficiency, v float64) { e.Util = v }},
{`avg by (model) ((llmgw_gpu_power_watts / llmgw_gpu_power_limit_watts) * on (node, gpu) group_left (model) llmgw_gpu_assigned)`,
func(e *modelEfficiency, v float64) { e.PowerLoad = 100 * v }},
{`avg by (model) (llmgw_gpu_sm_active_percent * on (node, gpu) group_left (model) llmgw_gpu_assigned)`,
func(e *modelEfficiency, v float64) { e.SMActive = v }},
}
// loadEfficiency is the table, best per card first; nil when the metrics
@@ -56,7 +65,7 @@ func loadEfficiency(ctx context.Context, h History) []modelEfficiency {
}
e := byModel[s.Name]
if e == nil {
e = &modelEfficiency{Model: s.Name}
e = &modelEfficiency{Model: s.Name, PowerLoad: -1, SMActive: -1}
byModel[s.Name] = e
}
q.into(e, s.Last())