VOGONS


The XP SD Smasher!

Topic actions

Reply 20 of 29, by douglar

User metadata
Rank l33t
Rank
l33t
cyclone3d wrote on 2026-01-26, 07:20:

This is going to take a while... full write cycles is most likely somewhere between 8-10k

See here for a person that tested a massive number of SD cards:
https://www.reddit.com/r/homelab/s/VwIab3xD74

What a great link! So much good stuff to read.

I found a link to his website here: https://www.bahjeez.com/the-great-microsd-card-survey/

The same author had some comments on SD Trim support here:

https://www.reddit.com/r/raspberry_pi/comment … omment/nqg8e4z/

So here's the thing with trimming -- in the SD specs, it's referred to as "discarding", but it's basically the same thing. It was introduced in version 6.00 of the SD Physical Layer Specification, which came out in 2017. A lot of the cards I've tested don't support it, but some do -- including the PNY Pro Elite Prime's. But you also need a compatible reader. Most USB readers will simply present themselves as generic mass storage devices to the system -- they just take the incoming commands and translate them to the equivalent SD commands. The system can't send arbitrary commands to the card -- if the reader doesn't tell the host that it supports trimming, then you can't use it for trimming. And if the reader wasn't designed to be compatible with version 6.00 (or later) of the SD Physical Layer Specification, it will probably tell the host that it doesn't support trimming, even if the card does.

Most readers that are built into laptop computers, phones, and SBCs (like the Raspberry Pi), on the other hand, are handled by the mmc driver on Linux. This driver *can* send arbitrary commands to the card -- so it should be able to issue a "discard" command to the card regardless of whether the card reader was designed to support it. (In the SD protocol, the "discard" command is just an "erase" command with a different argument.)

So the answer is still "Good luck at getting trim support from a sinitechi", but maybe some day with a new firmware it could happen.

Reply 21 of 29, by wierd_w

User metadata
Rank Oldbie
Rank
Oldbie

I'd settle for 'lba48 that actually, truly works from a sintechi' 😁

But the quotation is very interesting. SDCard supporting DISCARD is news to me. (Then again, i've been treating them as trash tier flash, and doing gymnastics to make sure writes are atomic to flash pages, whenever I have the level of control needed for that. Windows does not give that to me.)

It's been my understanding that sdcard has a small RAM buffer of between 4 and 32kb in size, that's used to help it buffer writes while the microcontroller does its thing to read-erase-modify flash pages. As long as your writes are some even multiple of this buffer then the degree of write amplification will be mitigated, with maximum mitigation happening when full atomicity with flash pages is performed.

Incidentally, maximum read/write performance is *also* attained when full pages are transacted, so if you are abusing SDCard for user storage on Linux, 'it's just good medicine' to (ab)use EXT4's RAID features to force the desired atomicity.

The presence of DISCARD means the card *wants* to write on discarded pages before doing read-erase-modify though, but if the OS is not marking pages as discarded, then it will run out of them, and the wear levelling will cease operating as 'intended'.

Last edited by wierd_w on 2026-01-26, 15:05. Edited 1 time in total.

Reply 22 of 29, by douglar

User metadata
Rank l33t
Rank
l33t

So if I am writing 40MB every minute with the constant rebooting and I have 12GB free to write in, and if the card is good for 8000 write cycles, let's see, that's 4.5 Years.

Reply 23 of 29, by douglar

User metadata
Rank l33t
Rank
l33t
wierd_w wrote on 2026-01-26, 14:48:

I'd settle for 'lba48 that actually, truly works from a sintechi' 😁

Lol! (Except it hurts a little)

Maybe someday there will be an open source firmware for the Sinitechi devices, kind of like flash floppy is for Gotek.

Reply 24 of 29, by wierd_w

User metadata
Rank Oldbie
Rank
Oldbie

If we have that, getting TRIM from DISCARD becomes low hanging fruit.

It's my understanding though, that the interface spec (electrical signalling AND adv capabilities MMIO mode instructionset) is encumbered by an NDA by the SDCard Assn. 🙁

This is likely why we *DONT* have ATMel or other commodity microcontroller based SDCard adapters floating around. BitBash mode is public, but MMIO is not.

Reply 25 of 29, by douglar

User metadata
Rank l33t
Rank
l33t

Ohhh, he's got a patreon ... https://www.patreon.com/MattCole

He's like the apex predator in the SD world. So much SD cruelty there that he should be age restricted!

Edit - And a website--

https://www.bahjeez.com/
Why I’m Destroying a Bunch of MicroSD Cards

  • Details on Individual Cards
  • Browse My Data

Reply 26 of 29, by jmarsh

User metadata
Rank Oldbie
Rank
Oldbie
wierd_w wrote on 2026-01-26, 15:20:

BitBash mode is public, but MMIO is not.

That simply isn't true. SDIO mode is well and truly public (up to at least UHS-I / ~100MB/s speeds).

Reply 27 of 29, by douglar

User metadata
Rank l33t
Rank
l33t

12500 reboots and counting ......

Reply 28 of 29, by douglar

User metadata
Rank l33t
Rank
l33t

32768 reboots and counting ....

Reply 29 of 29, by RetroPCCupboard

User metadata
Rank Oldbie
Rank
Oldbie
douglar wrote on 2026-02-13, 12:25:

32768 reboots and counting ....

I like how your count is exactly the number that a signed 16-bit integer can hold. Lol.