Independent telemetry verification. A permanent trail of evidence.

NodeRAYA reads the telemetry your GPUs already report and raises a rule finding only when a defined pattern persists. Separately, each GPU receives its own tamper-evident daily receipt after the day closes, backed by deterministic analysis and telemetry from the GPU itself.

Deterministic rules, inspectable logic, byte-level evidence. Fully read-only. It sits alongside your stack (Grafana, New Relic, Datadog) and runs fully automated.

Read-only Config-hashed receipts Physics-derived thresholds

Compute Inefficiency

Detects GPUs drawing excess power relative to their clock-aware mathematical physics baseline for 10+ minutes.

Thermal Throttling

Reads the GPU's own firmware throttle bitmask directly. Real physics from the silicon, not an inference from room temperature.

Daily Receipt

At day-end, each GPU receives a daily record with a SHA-256 integrity hash. Receipts are retained permanently for later verification.

How it works

Four simple steps. Every verdict traceable.

NodeRAYA reads telemetry your GPUs already emit, matches it against deterministic rule patterns, and raises a finding only when the pattern holds through the persistence window.

01
Read

Sits alongside your stack

Reads metrics your GPUs already export: utilisation, power, SM clock, temperature, and the firmware throttle bitmask. Zero deployment on the silicon. Fully read-only.

02
Match

Applies deterministic rule patterns

Each rule is a deterministic predicate over a window. RULE_01 for compute inefficiency. RULE_02 for thermal throttling. PEER for cohort deviation. The rules are inspectable and reproducible.

03
Verdict

Raises only on persistence

A finding fires only when the pattern holds through the full persistence window. No single-sample triggers.

04
Receipt

Opens into evidence

At the end of each day, NodeRAYA finalises one receipt for each GPU with recorded telemetry. It captures the daily telemetry summary, data-quality result, hardware status and SHA-256 integrity hash. Receipts are retained permanently in the daily archive.

Fleet

Live fleet telemetry.

Fleet overview
London-01 · Production · 10 GPUs · 2 models · 3 EPR deviations · 1 RULE_01 finding · 1 RULE_02 finding · 60m window
Fleet size
10
GPUs monitored
Models
2
B200 · H100 SXM5
EPR deviations
3
NODE 20, NODE 11 and NODE 10
Rule findings
2
1 RULE_01 · 1 RULE_02
NVIDIA B2005 GPUs · 1 EPR deviation
EPR deviation
NODE 20
0.55EPR ×
Util44.7%
Power764 W
Clock1924 MHz
Temp74.8 °C
Stable
NODE 16
73.5% util
Power728 W
Clock1857 MHz
Temp72.9 °C
EPR0.00×
Stable
NODE 18
75.6% util
Power631 W
Clock1414 MHz
Temp76.9 °C
EPR0.07×
Stable
NODE 17
73.8% util
Power726 W
Clock1856 MHz
Temp73.0 °C
EPR0.00×
Stable
NODE 14
73.8% util
Power728 W
Clock1856 MHz
Temp72.9 °C
EPR0.00×
NVIDIA H100 SXM55 GPUs · 2 EPR deviations · 1 RULE_01 · 1 RULE_02
Excess power · RULE_01
NODE 11
3.74EPR ×
Util6.7%
Power485 W
Clock1951 MHz
Temp68.5 °C
EPR deviation
NODE 10
0.52EPR ×
Util49.7%
Power479 W
Clock1587 MHz
Temp72.0 °C
Hardware slowdown · RULE_02
NODE 12
92.2°C
Util98.4%
Power537 W
Clock1283 MHz
StateHwSlowdown
Stable
NODE 02
68.8% util
Power434 W
Clock1641 MHz
Temp67.8 °C
EPR0.02×
Stable
NODE 01
88.0% util
Power521 W
Clock1781 MHz
Temp79.9 °C
EPR0.00×
Receipts

Every GPU gets a daily receipt.

Each receipt is finalised after the day ends, records one GPU’s daily telemetry and hardware status, carries an integrity hash, and is retained permanently.

Daily archive · one receipt per GPU

At the close of each day, NodeRAYA finalises a receipt for every GPU with recorded telemetry. Each receipt preserves that GPU's daily record, data-quality assessment and hardware status, together with its integrity hash. Receipts are retained permanently and remain available for later review.

Closed at day-end · Individually recorded · Permanently retained
Hardware status clear
GPU-HBM3-000 · NVIDIA H100 80GB HBM3
No RULE_02 hardware slowdown reported · 100.0% trusted telemetry
View receipt

GPU Receipt · GPU-HBM3-000

NVIDIA H100 80GB HBM3 · 2026 · Daily receipt · version 1

Data quality

100.0%
of the period had trusted data
Trusted Gap or untrusted Not reporting
Gaps0
Untrusted minutes0
Readings received1,440 of 1,440

Scorecard

Available and reporting100.0%
Average utilisation66%
GPU profile TDP700 W
Throttling events0
Peak temperature65 °C

Hardware status

Hardware slowdown RULE 02
None found in trusted data
Hardware status clear
GPU-HBM3-001 · NVIDIA H100 80GB HBM3
No RULE_02 hardware slowdown reported · 100.0% trusted telemetry
View receipt

GPU Receipt · GPU-HBM3-001

NVIDIA H100 80GB HBM3 · 2026 · Daily receipt · version 1

Data quality

100.0%
of the period had trusted data
Trusted Gap or untrusted Not reporting
Gaps0
Untrusted minutes0
Readings received1,440 of 1,440

Scorecard

Available and reporting100.0%
Average utilisation76%
GPU profile TDP700 W
Throttling events0
Peak temperature60 °C

Hardware status

Hardware slowdown RULE 02
None found in trusted data
Evidence

Every rule finding opens to its evidence record.

The pattern that fired. The thresholds that were in force. The exact telemetry rows behind it. Nothing is reconstructed after the fact.

Evidence drill-down · example finding

Onboard Thermal Throttling

Red RULE_02

NODE 12 · NVIDIA H100 SXM5 · HwThermalSlowdown held 60.0 min · 16:54 → 17:54 · config hash 1e39f4725cd2

Current die temperature
92.2°C
Throttled · investigate
HwThermalSlowdown bit set Throttle state confirmed
Timeline · full 60-minute event window GPU Temperature °C Power Draw W
16:5417:1417:3417:54
Every finding opens into a full evidence trail: triggering datapoints, thresholds in force, data confidence, and a hash of the configuration used.
Compute Inefficiency · RULE_01
01 · The verdict
Compute Inefficiency Red RULE_01

Low utilisation sustained through the persistence window while observed power exceeded the clock-aware baseline for this GPU profile.

Util below threshold Excess over baseline
3.74×
Excess power ratio
02 · Rule evidence
Thresholds in force
Utilisation thresholdconfigured
Excess power marginconfigured
Persistence windowconfigured
Observed in window
GPU modelNVIDIA H100 SXM5
Avg util6.7 %
Avg power485 W
Expected baseline102 W
Duration60.0 min
Data quality
gpu_utilization_pctTrusted · 98%
power_draw_wattsTrusted · 98%
sm_clock_mhzTrusted · 98%
03 · The proof
Triggering window · sampled every 5 minutes
TimeUtil %Power W
20:006.4483
20:056.6484
20:106.9486
20:156.7485
20:206.8487
20:256.5484
20:306.9485
20:356.6483
20:406.7485
20:456.8486
20:506.5484
20:556.9485
21:006.7484