Search papers, labs, and topics across Lattice.
SK hynix
2
0
3
0
NELSSA achieves a staggering 5.5x increase in decode throughput for mixed-length LLM workloads by intelligently integrating GPUs with PNM accelerators.
ITME achieves a remarkable 35.7% throughput improvement by leveraging CXL-hybrid memory to expand remote memory for LLMs, tackling the challenges of disaggregated shared storage.
