AMD Finally Finds Fixes For Improving Inter-Core Latency On Strix Point APUs, Will Resolve The Issue With BIOS Updates

wccftech.com:

The Strix Point APUs had over 170ns of latency between Zen 5 and Zen 5c cores, which resulted in performance setbacks in multi-threaded operations.

ASUS and AMD announced Inter-Core latency fixation with the upcoming BIOS update for Strix Point APUs, now delivering around 60-70 ns of latency

The complex die design of AMD's Zen 5 chips has caused some issues with the core-to-core latency, which has been known to affect performance in specific workloads. This problem has affected both the desktop high-end Ryzen 9000 chips and AMD's Strix Point APUs, including Ryzen AI 9 365, Ryzen AI 9 HX 370, and Ryzen AI 9 HX 375 processors.

Related Story AMD Threadripper “Shimada Peak” 96-Core & 16-Core CPUs Based On Zen 5 Architecture Spotted

The high core-to-core or inter-core latency is particularly prevalent in high-core count chips, which involve CPUs with dual CCDs. This hasn't spared even the Strix Point APUs, which bring both Zen 5 and Zen 5c core clusters. Anandtech illustrated this with the review of the Ryzen 9 9950X as well as the Ryzen AI 9 HX 370 CPU.

While the problem was resolved for the desktop Ryzen 9000 processors through a BIOS patch, the high latency on the Strix Point APUs still exists. Thankfully, this is soon going to be history, as ASUS and AMD have managed to sort out this problem. Even though the information hasn't come officially from either of the companies, @9550pro has provided the early update, which shows that AMD has successfully fixed the inter-core latency drastically.

As you can see from the table, the usual inter-core latency usually stays between 160 and 180 ns, which is quite high compared to the inter-core latency in the Zen 5 core cluster. Going outside the Zen 5 cluster has affected the multi-core performance of Strix Point APUs noticeably in workloads that require more cores for faster processing. With an experimental update, AMD and ASUS have been able to reduce the latency down to 60-70 ns, which is almost 2x+ improvement.

While the core