Version 2.0, 16 September 2026. Extend the previous 58 configurations from eight documents with six IBM Deskstar 75GXP configurations from the manufacturer's Revision 2.0 manual, fifth edition with June and November 2000 dates. The manufacturer document is preserved on a third-party mirror. Scanned pages were OCR-transcribed; the raw PDF and extracted page text are retained locally. PDF page 21 gives exact logical bytes, page 22 rotational speed, latency and maximum density, page 24 seek specifications, and page 27 sustained transfer rates. Use 37.7 MB/s outer-zone sustained rate, not 444 Mbit/s media rate or 100 MB/s interface rate. The manual's typical read seek is 8.2 ms; a contemporary review's rounded 8.5 ms is not substituted.
All drives are 7,200-rpm desktop configurations. IBM capacities use the exact byte counts; other source sheets provide decimal nominal capacities, a precision difference recorded in notes. Full-read reference hours = capacity_GB × 1000 / sustained_MBps / 3600. This is an idealized reference at the chosen rated speed, not an independently measured minimum or a prediction of actual whole-drive throughput. IBM labels its outer-zone rate typical; most later rates are maximum outer-diameter sustained rates. These qualifiers are retained.
Recover density for the same model IDs wherever the selected source provides it. Convert 91,560 Mbit/in² to 91.56 Gbit/in². Preserve maximum versus average density; do not combine them into one indexed series. Four aligned average-density examples use the largest capacities in the selected 2006, 2007, March 2011 and 2017 Seagate manuals. They are illustrative model comparisons, not annual market-frontier estimates. Higher capacity also depends on surface count, so no causal density-to-capacity model is fitted.
Seek-plus-rotation, where available, excludes command overhead and data transfer. A printed less-than qualifier survives into the derived field. Equal 7,200 rpm implies a theoretical half-rotation of 60,000/(2×7,200) = 4.1667 ms; minor 4.16/4.17 differences in sheets are rounding, not evidence of a latency trend. Missing modern seek figures remain missing. Do not infer an end-to-end access-time improvement from this incomplete series.
The package preserves observations, source URLs, hashes, locators, qualifiers and calculations. Existing excluded sources remain excluded where only interface or raw-media rates were found. The contribution is the model-level reconciliation and derived read-time comparison, not a hardware benchmark.
Version 3.0 adds a complete-case 2006/2011 comparison using the 400 GB ST3400832AS and 1 TB ST31000524AS. The 2011 detailed Section 2.6 table gives typical average read seek 8.5 ms, while the summary says <8.5; preserve both locators/definitions rather than silently treating the summary bound as a measurement. Add 4.16 ms average rotation to the typical seek values for a reference mechanical access time. This excludes controller, command and data-transfer overhead. Average areal density, rated sustained speed and capacity are available at both endpoints. The 2017 extension does not have seek data.