How we test
Nine stages. 24 to 72 hours. One signed result.
Every card we buy runs the same bench on our own test servers: Ubuntu LTS, NVIDIA’s datacenter driver, DCGM and CUDA. One orchestrator runs the stages in order, records time series, and hands the results to the signer.
01 — The bench
Most of the time is burn-in. That’s the point.
| No. | Stage | Tool | Duration | Pass criteria |
|---|---|---|---|---|
| 01 | Intake | Barcode and serial scan, photos of both sides, connectors and label, weight | 5 min | Serial matches the probe report. No physical damage. |
| 02 | Identity | nvidia-smi -q | 1 min | Board part number and VBIOS family match the SKU. |
| 03 | Health snapshot | ECC, row remapper, PCIe link | 1 min | No remap failure, no pending remaps, link at full generation and width. |
| 04 | NVIDIA diagnostics | dcgmi diag -r 3 (and -r 4 for Grade A) | 30–90 min | All tests pass. |
| 05 | Memory stress | DCGM memory tests and HBM stress | 1–2 h | Zero uncorrectable errors. No new remapped rows. |
| 06 | Performance | BF16/FP16 GEMM, HBM bandwidth, PCIe host↔device with nvbandwidth | 20 min | At least 95% of our reference for the SKU. |
| 07 | Burn-in | gpu-burn or DCGM targeted power at TDP | 24 h (B) · 72 h (A) | No Xid errors, no ECC growth, temperature under limit, no thermal throttling. |
| 08 | Re-snapshot | Repeat of stage 03 | 1 min | Nothing changed since stage 03. |
| 09 | Grade and sign | Platform signer | Instant | Grade A, B, C or Reject, signed with ed25519. |
Thresholds are where we start, and will be tuned against our own warranty outcomes. Performance is measured against a reference we record on known-good cards of the same SKU, not the spec sheet.
02 — Why these tests
Built on published failure data.
During a 54-day snapshot of Llama 3 pre-training on 16,384 H100 GPUs, Meta logged 268 GPU-related interruptions. Each one maps to a stage on our bench.
Source: The Llama 3 Herd of Models (Meta, 2024), Table 5. These were interruptions on new cards under full training load, and not all were permanent failures.
03 — Grades and warranty
What each grade promises.
| Grade | Requirements | Warranty |
|---|---|---|
| Grade A | 72 h burn-in, zero remapped rows, at least 98% of reference performance. | 6 months included. 12 months optional, paid. |
| Grade B | 24 h burn-in, a few correctable remapped rows that stayed stable, at least 95% performance. | 6 months. |
| Grade C | Works, but has a history: uncorrectable remapped rows that are now stable, or minor throttling. Sold at a discount. | 3 months, or sold as-is. |
| Reject | Remap failure, a failed diagnostic, uncontained ECC errors, or under 90% of reference performance. | Not sold. Returned to the seller or sent to an EPR-registered recycler, with the recycler certificate on file. |
Claims
Open an RMA from the buyer portal. Attach the card’s Sentinel snapshot if you run it, and we can ship a replacement before the faulty card comes back.
Renewal
When warranty ends, Sentinel data prices the renewal. A card that has run clean costs less to cover.
04 — Reference sheet
What we check each card against.
Power, bandwidth and link speed must match the SKU’s baseline. A relabelled or emulated card can’t fake them.
SXM modules are tested inside their HGX server, so for now we buy and sell SXM as whole servers. Loose-module testing comes once volume pays for a spare baseboard.
| Model | TDP | Memory | Bandwidth (spec) | PCIe |
|---|---|---|---|---|
| H100 PCIe 80GB | 350 W | 80 GB HBM2e | ≈ 2.0 TB/s | Gen5 |
| H100 SXM5 80GB | 700 W | 80 GB HBM3 | ≈ 3.35 TB/s | Gen5 |
| H200 SXM 141GB | 700 W | 141 GB HBM3e | ≈ 4.8 TB/s | Gen5 |
| A100 80GB PCIe | 300 W | 80 GB HBM2e | ≈ 1.9 TB/s | Gen4 |
| A100 80GB SXM4 | 400 W | 80 GB HBM2e | ≈ 2.0 TB/s | Gen4 |
| L40S 48GB | 350 W | 48 GB GDDR6 | ≈ 0.86 TB/s | Gen4 |