Michael S (already5chosen.delete@this.yahoo.com) on November 27, 2020 3:34 am wrote:

> IMHO, the correct question is "Why force zeros to main DRAM at all?".
> What's wrong with zeroing page in cache and relaying on normal cache evictions for taking care of the rest?
> Combined L3 bandwidth on Ryzen 3950X (4 L3s, dual-channel DDR4) could easily be 10x of bandwidth of main
> memory. And the chance that the page will be read by the same CCD at least once before evicted is high.

Doesn't it depend on when and what's doing the zeroing?

If a program wants to zero memory for its own use, I'm sure you're right.

If the OS is zeroing memory before it hands it out to processes, it may well want to keep a pool of pre-zeroed memory. In this case, the latency is less important, and it might be longer before the page is used.

A quick overview came up with this LWN article, suggesting that the Linux kernel community have at least looked into this -- I'm not sure what the kernel does these days.