Revision History
NVIDIA H200 · 10 revisions
Sizes are character counts of the article source. The signed number is the change from the previous revision.
Recent edit summaries
Detailed summaries recorded by editors. Generic maintenance summaries are omitted here; every recorded revision remains below. Dates describe the edit, not necessarily the event it covers.
- Correction: Corrected Tensor Core throughput figures that presented NVIDIA's with-sparsity numbers as dense (and doubled them), the L2 cache size (50 MB, not 60 MB), the HBM stack layout and a wrong per-pin bandwidth calculation, the Google Cloud A3 Mega listing (an H100 machine), MLPerf v4.0/v5.1 details, a statement wrongly attributed to Ian Buck, claims that Stargate Abilene and JUPITER run H200-class hardware, and the outdated China export status (plus 22 minor fixes)
Version 10 · Nov 2, 2026, 08:31 AM
- Improve internal links (2026-10-13)
Version 8 · Oct 13, 2026, 10:19 AM
- Added 6 contextual internal links
Version 7 · Jul 23, 2026, 03:29 PM
- Pre-update snapshot before enhancement to 7283 words.
Version 3 · May 18, 2026, 05:10 AM