Hacker News new | ask | show | jobs
by jerf 2529 days ago
Is this even going to be noticeable at all out of benchmarks?

Nominally, my personal system has something that can do 2GB/s, but in normal operation, as programs and OSes do their thing, the software is incapable of making requests quickly enough to come even close to saturating the bandwidth. I'm not sure if you dropped a part in there that maxed out at 512MB/s that I could even notice.

Amdahl's law codifies the observation that as you get closer and closer to 0 time taken for a particular subtask, you'll rapidly stop gaining actual performance due to all the other subtasks that didn't speed up. It seems like 1GB/s is likely to be as close to infinite in practice as 2GB/s is.

1 comments

> I'm not sure if you dropped a part in there that maxed out at 512MB/s that I could even notice.

I've done exactly this. I tried to run an NVMe drive and a SATA drive back-to-back in the same laptop to see if I could notice any difference in my day to day workload. I couldn't, and because of that I bought a 2TB SATA SSD instead of 2TB NVMe. SATA usually still has better power consumption than most NVMe drives, and they're cheaper too.

Which OS did you test this on? Every time I have used a Windows device with an NVMe drive the difference is immediately obvious, but on my Hackintosh and Linux boxes I haven't noticed any difference at all.
Arch Linux and Windows 10 Enterprise LTSC. While it may be slower to use Windows, the difference hasn't been immediately obvious. I use Windows for gaming and browsing only and stay in Arch most of the time.
Yeah, NVMe is so fast, few people will ever need the speed. You'd have to, what, record raw video to fill it that fast? What other uses are there? Do OSes have optimisations to better use swap space?
I bought one as an upgrade disk because IMO they are easier to install. I don't need to worry about mounting, routing a cable, etc.
SATA SSDs also come in M.2 format.