Chips & computingResearch findings
Measure HBM beyond the standard stress loop
Original published 2026.10.15 08:27 KSTAdded to Blackdeck 2026.10.18 11:38 KST
What has changed?
Physical-chip evidence identifies power conditions missed by a standard test pattern.
Key points
- Researchers measured operation- and data-dependent power on 36 real HBM2 chips.
- Their access pattern drew 23.6% more power on average than the highest-power JEDEC measurement loop.
- The study also develops a measurement-informed HBM power model.
Connect it to your work
Memory and server engineers can reconsider worst-case workload coverage.
What remains unproven
The percentage is specific to the tested HBM2 chips and comparison patterns.
Originals used for this note
Links open the original publisher in a new tab.