Bengaluru Systems Meetup
Bengaluru Systems Meetup

Speak at Bengaluru Systems meet-up

Call for talks & demos for monthly meet-ups

All submissions

Pranav M

@prawns

The Interface Tax: How Much of Near-Memory Compute's Speedup Survives in LLM Decode?

Submitted Oct 27, 2026

LLM decode is memory-bound, so near-memory compute (PIM/PNM) looks like an easy win. But how much of the ideal speedup survives the interface?

This talk looks at where the speedup goes, using a case study from my own small near-memory engine for binary similarity search, HADES, and a few recent PIM papers.

Length: 12–13 minutes. No PIM background needed.

About me: Pranav M, undergrad ECE at PESU, working on computer architecture, processing near memory and RTL.

All submissions

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Hosted by

Bengaluru Systems Meetup
Bengaluru Systems Meetup
Bengaluru Systems Meetup

Supported by

Venue host

Lossfunk
Lossfunk
a lab based in Bangalore, India focused on foundational questions on artificial and biological intelligences.