Im curious, how often is SSD / storage speed performance really useful? I feel like for many people it's akin to gigabyte wifi in that its nice to have, but not really particularly necessary

It's definitely important for editing video and doing VFX work, and lots of the editors/VFX folks I know are using Studios. Raw footage from cine cameras now can be 50-100 MB per second now that we have 12 or 18K cameras, and even the low-quality proxies people work with are still pretty heavy, so in-memory caches aren't really an option and stuff is constantly being read from (and cached to) disk.

Apple’s “LLM in a flash” paper [1] sheds some light on why this can be a big deal, especially if they are working on codesign of model and hardware in this dimension.

[1] https://arxiv.org/abs/2312.11514

When I was booting from a spinning disk, at that time it took over a minute to boot. I upgraded the system to SSD later, and that same system took less than 20s. That was 10 years ago.

Another example as a developer, one particular large project took an hour to build on a spinning disk, on a SSD it takes 2mins

So, yeah, it's really noticable improvement

Unrelated to disk, I've definitely been thinking about this more on the network side recently. I upgraded to 1.4/900 home fibre a while back and then got a modern Unifi UCG/U6 wifi setup, and while I do get 1.4GB down from my hardwired workstation in an explicit speed test I have yet to get anything close to that on real-world workflows. Torrents, container-pulls, backups, all of it seem to cap out around 20-50 MB/s.

So while have having this massive almost-symmetric fibre pipe is cool on paper, I haven't felt a huge need to install 2.5G gear all over the house.

It's very important. Every time you launch an app or open a file it's critical for performance. Also when memory runs out and the OS swaps to disk it makes a huge difference.

it's very very important, not at all something you can compare to WiFi.

data processing, LLMs, model loading, MoE loading, etc, etc relies on very fast storage to keep your GPU saturated.

It's nice when using large sample libraries so you can stream them off the drive without as much caching in memory. Local LLM workflows probably also benefit from being able to run models too big for RAM off the drive.

Boot time, loading game time, mostly.

[dead]