# AMMA preprint reports lower attention latency with memory-centric chip design

_Published Monday, November 2, 2026 at 12:20 AM EST · Science, AI · Latest · Tier 2 — Notable_

Researchers report that AMMA, a proposed chip architecture for serving long-context AI models, achieves 15.5X lower attention latency and 6.9X lower energy consumption than NVIDIA's H100 in their evaluations. The preprint targets attention during token generation, when accessing memory can constrain performance.

AMMA replaces GPU compute dies with cubes that combine high-bandwidth memory and nearby processing, roughly doubling available memory bandwidth. Its design also reorganizes parallel work and communication between chiplets. The results concern attention workloads; they do not establish equivalent gains for an entire model-serving system.

## Sources

- [cs.LG updates on arXiv.org](https://arxiv.org/abs/2604.26103)

---
Canonical: https://techandbusiness.org/newswire/s4si3UprDcg1u-Ur7veXAW
Published: 2026-11-02T05:20:38.447Z
Story chronology: 2026-11-02T05:00:00.000Z
Retrieved: 2026-11-02T07:51:44.681Z
Publisher: Tech & Business (techandbusiness.org)