Stenfox

/ Stenfox GPU · Standalone product

New

Stress test and benchmark any GPU.

A dedicated product for graphics processors, separate from computer diagnostics. From laptop chips to data-centre AI accelerators, every GPU runs the same protocol and leaves with a score, a grade and a signed report.

NVIDIA H100 PCIe, 80 GB

Illustrative
GradeA
Compute (FP32)
50.8 TFLOPS
Memory bandwidth
1.94 TB/s
VRAM tested
80 / 80 GB
Peak temperature
71 °C

0 errors

Every class of GPU.

Integrated

Built into Windows laptops and desktops

Discrete cards

Consumer and workstation PCIe cards

External

Plug-in eGPU enclosures

Data centre and AI

Server accelerators for AI and compute

Two modes, side by side.

Stress test

Sustained full load until weaknesses show, inside safe limits. The temperature curve is recorded through warm-up, sustained load and cool-down.

Benchmark

Compute throughput, memory bandwidth, VRAM and peak temperature are measured, then turned into a score and a grade for resale.

NVIDIA H100 PCIe, 80 GB

Illustrative
Core temperature
Safety limit
Warm-upSustained loadCool-down
GradeA
Compute (FP32)
50.8 TFLOPS
Memory bandwidth
1.94 TB/s
VRAM tested
80 / 80 GB
Peak temperature
71 °C

0 errors

The protocol, in six stages.

  1. 01

    Compute stress

    Sustained load, every result checked for correctness.

  2. 02

    VRAM integrity

    Patterns written, read back and compared.

  3. 03

    Thermal behaviour

    Hotspot and throttling under load.

  4. 04

    Load transitions

    Idle and peak alternate to expose instability.

  5. 05

    Fault detection

    Driver errors, disconnects and resets.

  6. 06

    Safety supervision

    Slows or stops at conservative limits.

What the report records.

  • Model and unique ID
  • Driver
  • Benchmark scores
  • Grade
  • Temperatures
  • Power draw and limit
  • Clocks and throttling
  • VRAM results
  • PCIe link width
  • Fault events
  • Pass or fail per stage

The report always states the depth that ran.

  • Full testStress protocol and benchmark, end to end.
  • LimitedMonitoring only, when a full load is not possible.
  • Identification onlyModel, unique ID and driver recorded.

Stenfox Portal

Workflows, grading forms, reports
  1. Intake

  2. Diagnose

  3. Erase

  4. Report

  5. Route

/ Honest comparison

From a boot test to a graded, signed result

Most GPUs are resold after a visual check or a quick boot. Stenfox GPU runs a repeatable protocol and records what it found.

Criterion
Typical check
Stenfox GPU
Test
Visual check or boot test
Sustained load with verified results
Performance
No performance figure
Benchmark score and grade
Evidence
No record
Signed report per GPU

/ FAQ

Frequently asked questions

Which GPUs does Stenfox GPU support?

Integrated GPUs in Windows laptops and desktops, discrete PCIe cards (consumer and workstation), external eGPU enclosures, and data-centre / AI accelerators.

Isn't this already in Stenfox Computing?

Stenfox Computing includes a basic GPU check among its hardware tests (26 on Mac, 23 on Windows laptops). Stenfox GPU is a separate, dedicated product for full stress testing and benchmarking.

What if a GPU can't run a full test?

The report always states the depth that ran: full test, limited (monitoring only) or identification only.

Is it a data wipe?

No. A GPU is not storage media. Stenfox GPU is a health, performance and grading test.

Next step

See it run on your own devices.

  • Live demo on your own devices
  • Pilot in your workshop before you commit
  • Direct line to our engineers
  • Onboarding and training for your team

Tell us what you process and at what volume. We reply within one business day.