News for hackers. Stuff that matters.
Re:Apple’s M1 Positioning Mocks the Entire x86 Business Model by PartiallyTyped on Friday April 23, 2021 @06:55AM (#26913342)
Re:Apple’s M1 Positioning Mocks the Entire x86 Business Model by necovek on Friday April 23, 2021 @08:00AM (#26913801)
What makes you think it's random RW?

Fastest SSDs barely reach 200MiB/s on 4k block random writes (reads are usually slower, likely because they can't be cached if they are truly random):

- 124MB/s for Samsung 980 Pro (https://www.servethehome.com/samsung-980-pro-500gb-pcie-gen4...) - 211MB/s for Sabrent Rocket 4 Plus (https://www.servethehome.com/sabrent-rocket-4-plus-2tb-revie...) - 171MB/s for Intel 670p (https://www.servethehome.com/intel-670p-2tb-m-2-nvme-ssd-rev...)

32/64 queue depths generally do not apply to desktop computing since you are unlikely to read that many file streams simultaneously (or rather, your one program that you care about right now might be loading only a few files at once, which is what you'd perceive as the SSD speed: the fact that OS services might be accessing other stuff in the background does not help much there).

Re:Apple’s M1 Positioning Mocks the Entire x86 Business Model by danieldk on Friday April 23, 2021 @07:37AM (#26913630)
BlackMagic tests video workloads using sequential reads/swrites. 2.2/2.7 GiB/s is pretty average for a modern NVMe SSD. Even consumer-level Samsung NVMe SSDs are faster.