Corvex reports HGX B200 retains roughly 2× training and 2.5× inference advantage over HGX H200 with CC enabled
~2x / ~2.5xestimate
| Value kind | estimate — Interpret this value according to its displayed kind, scope, source, and as-of date. |
|---|---|
| Scope | Provider-reported cross-generation advantage; no paired same-host CC-on/off overhead or universal workload result is established. |
| As of | 2026-01-27 |
| Source | Corvex — Rodolfo Leffa and Troy Benjegerdes, 27 January 2026. · Confidential Computing Meets NVIDIA HGX B200; How NVIDIA HGX B200 Secures Workloads without Performance Penalties; cross-generation comparison. |
| Review | checking…review by 2026-09-05 · standard cadence |
| Recorded changes | last 2026-06-29 |
| Claim id | training-inference-advantage-hgx-b200-retains |
Where the guide uses it
Not quoted in a chapter yet; it is kept in the curated register.
← Full numbers register — every date-stamped figure in the guide, with revision history.