TLDR Tech / Practical
NVIDIA’s Olympus Core: Pushing Server Single Threaded Performance Boundaries
NVIDIA's Olympus core is a high-performance server processor designed to maximize per-clock performance and push single-threaded boundaries. Featuring a 10-wide out-of-order layout, massive register files, and a novel spatial multithreading implementation, Olympus delivers strong throughput that approaches desktop performance levels. However, testing shows that while its SMT approach performs well on frontend-bound integer workloads, it can encounter throughput losses on backend-bound floating-point workloads.
Read the full story