Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

The article literally says this very thing?

> Lion Cove suffers harder with backend memory latency, but far less from frontend latency. Part of this can be explained by Zen 4’s stronger data-side memory subsystem. The AMD Ryzen 9 7950X3D I previously tested on has 96 MB of L3 cache on the first die, and has lower L3 latency than Lion Cove in Intel’s Arrow Lake platform. Beyond L3, AMD achieves better load-to-use latency even with slower DDR5-5600 36-36-36-89 memory. Intel’s interconnect became more complex when they shifted to a chiplet setup, and there’s clearly some work to be done.

They only compared three games of roughly the same genre in terms of mechanics which is not exactly a representative benchmark either, even for games. The average shooter with a fixed map and say, Minecraft, have vastly different memory access demands.



Yeah, latency. Strix has higher bandwidth but it pays for it in higher (i.e. worse) latency.


They literally say AMD gets better latency on top of better bandwidth, are we reading the same sentences? Better = lower I would expect.


They're referring to the ordinary DDR5-on-a-stick Ryzen implementation, not the LPDDR5X Ryzen AI implementation.

ETA: The only non-laptop results I have seen for the Ryzen AI place it worse than the 9950X, which has 2-channel plain old DDR5. The 4-channel LPDDR5X only wins when it can exploit the bandwidth and hide the latency, such as physics simulations. But for most applications the 9950X beats it. Hence my original point now well upthread: still waiting to see if the Strix Halo architecture is good for anything.


Strix Halo has a big GPU (for a laptop chip). GPUs love memory bandwidth and can tolerate latency, so... Strix Halo should be (and is) good at graphics and GPGPU.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: