Chips & computingResearch findings
PICNIC: Silicon Photonic LLM Inference Accelerator
Original published 2025.11.06 13:15 KSTAdded to Blackdeck 2026.11.19 16:31 KST
What has changed?
Silicon photonic interconnection reduces communication latency between chiplets effectively.
Key points
- Silicon photonic interconnection reduces inter-chiplet communication latency.
- Develops optimized hardware scheduling for LLM workloads.
- Achieves 3.95x speedup and 30x efficiency improvement over Nvidia A100.
Connect it to your work
Proposes a 3D-stacked chiplet-based LLM inference accelerator integrating non-volatile in-memory computing and IPCN.
What remains unproven
The abstract reports simulation results; real-world performance is not established by this input.
Originals used for this note
Links open the original publisher in a new tab.