Hacker News new | past | comments | ask | show | jobs | submit login

Considering 100 GB HDD's came out 20+ years ago that's not very suprising. If people are comparing SD cards to HDD's they're probably comparing to HDD's made in the last 10 years.



OP specifically cited an old hdd in the pre-ssd days. New HDDs are still at best on par with a high end microsd used as internal storage (compare an A2 microsd 'rated' at 2000 but actually getting about 800 iops vs a high end hdd benchmarking at 200-300).

Mounting without a buffer on a badly chosen filesystem in what could well be the M1's abysmally slow sd reader isn't a representative test.

Hare braned formatting tools will also often offset the filesystem by 512 Bytes (necessitating two writes and two reads for one 4k write) or use blocks that aren't 4kiB


FWIW my MacBook Pro is the 2015 Intel version, and my card is HFS+ formatted (Android would use FAT32 or Ext4). I got similar performance with an external USB reader IIRC, but maybe it's time to test again.


Aye, there's your problem.

https://web.archive.org/web/20151222043711/https://support.a...

It's like connecting a QLC SSD through an IDE port and complaining it's slow. It /is/ slow but not that slow.


Is the issue the 480mbit/s USB limit? That's still a lot more than what I was getting with random writes.


USB 2 has very poor latency for this use case and the sd connector lacks the pins for newer sd cards.

A usb 3 UHS III reader would be able to talk to the card at full speed, then it's a matter of convincing the OS to trim it and buffer as if it's an internal hdd. Kinda expensive and pointless to go to that effort to get 2x hdd performance when there are NVME enclosures for cheaper SSDs that have 100x the IOPs though.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: