[cl3misch]

> By default, many modern SSDs, including the one in this laptop, report a smaller sector size than their optimal one. This is done for compatibility with Windows XP and older systems…

I had never heard of this. Of course, checking on my new NVME SSD I also get the suboptimal 512 bytes. I won't do a reinstall for this, but I wonder how large the performance impact actually is?

[wtallis]

Not much actually does 512-byte IOs, at least not in any performance-critical way. Your OS and filesystem are probably doing most stuff in increments of 4kB, and that's the granularity the SSD is tracking allocations with, even when it is presenting the illusion of 512-byte sectors.

The flash itself probably has a page size of at least 16kB, and erase block size of several MB, but trying to align and chunk your IO to those boundaries has marginal benefits.

As long as you don't have partitions that are misaligned relative to 4kB boundaries, you're fine.

[cl3misch]

Thanks, I was expecting something along those lines. I assume misaligned partitions are kinda hard to produce these days if you are using the default partitioning from your GUI Linux distro installer? Last time I heard this being really relevant was over ten years ago iirc.

[jdecourval]

The truth is, I realized this too late, and now I would also need to, maybe not reinstall, but at least reformat. So I have no idea what sort of impact this has, but clearly, this is not huge.

What supposedly really kills performance is if your filesystem's blocks are unaligned with the real physical sector size (4k), but most FS correctly handles this: https://wiki.archlinux.org/title/Advanced_Format#File_system...

[teo_zero]

I think wrong block sizes, or, to be more precise, misaligned blocks, affect writes more than reads, so the performance impact will be limited in all common cases. But they will worsen the write amplification and thus shorten the life of your SSD.