FRESH

Hacker News

Three types of LLM workloads and how to serve them

74 points by charles_irl

by rippeltippel

1 subcomments

> Gallia est omnis divisor in partes tres.
OCD-driven fix: The correct Latin quote is "Gallia est omnis divisa in partes tres".

by ZsoltT

1 subcomments

> we recommend using SGLang with excess tensor parallelism and EAGLE-3 speculative decoding on live edge Hopper/Blackwell GPUs accessed via low-overhead, prefix-aware HTTP proxies
lord

by omneity

0 subcomment

Very cool insights, thanks for sharing!
Do you have benchmarks for the SGLang vs vLLM latency and throughput question? Not to challenge your point, but I’d like to reproduce these results and fiddle with the configs a bit, also on different models & hardware combos.
(happy modal user btw)