EXP-058REJECTED
Canonical GPU engineering / SHA-256d

Nsight Compute profiling permission block

Three independent Nsight Compute profiles can expose the bottlenecks of the frozen T512-N32 baseline.

Claim statusNO END-TO-END MINING ADVANTAGE DEMONSTRATED

Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.

Date2026-08-14
HardwareNVIDIA GeForce RTX 5080 · exact SHA-256d verifier
Run scopePreregistered campaign with sealed result and audit
ReproducibilityRun the frozen command on a system where NVIDIA performance counters are explicitly authorized, without changing the binary or metrics.
01 / Question & hypothesis

What was tested?

Three independent Nsight Compute profiles can expose the bottlenecks of the frozen T512-N32 baseline.

02 / Scientific basis

Why the test is meaningful

Hardware-counter profiling is useful only when the counters are actually collected; a successfully executed kernel with unavailable counters cannot support a profiling conclusion.

valid profile ⇔ kernel completes ∧ counters availableERR_NVGPUCTRPERM ⇒ environment-invalidno counters ⇒ no performance inference
Nsight Compute profiling permission blockVisual reading of the published metrics and gates for EXP-058; it summarizes the registered result, not a mining advantage.EXP-058 / NSIGHT COMPUTE PROFILING PERMISSION BLOCKPROFILES COLLECTED0 / 3DRIVER RESULTERR_NVGPUCTRPERMSCIENTIFIC VERDICTINVALID ENVIRONMENT
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-058; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Attempt the preregistered three-profile campaign against the frozen binary and stop immediately if the driver denies GPU counter access.

04 / Observed result

What happened

Profiles collected0 / 3
Driver resultERR_NVGPUCTRPERM
Scientific verdictINVALID ENVIRONMENT

The kernel completed, but the driver returned ERR_NVGPUCTRPERM and Nsight exposed zero counters. The offered elevated-session path was cancelled and global permissions were not changed.

05 / Validation

Exactness and statistical controls

The attempt is classified as environment-invalid, not positive or negative evidence about the kernel.

06 / Interpretation

What the result means

No profiling conclusion can be drawn. Preserving the blocked run prevents permission failure from being misreported as a scientific result.

Limitations

  • No counter data were collected.
  • The profile was not repeated elevated.
  • Static SASS analysis is a different experiment.
07 / Reproduction

Evidence trail

Run the frozen command on a system where NVIDIA performance counters are explicitly authorized, without changing the binary or metrics.

Canonical variants

CANONICAL-EXP-134PREREGISTERED-CAMPAIGNSEALED-AUDIT

Source: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.