from Hacker News

Sandy Bridge-era motherboard gets M.2 SSD boot support 12 years after launch

by mobilio on 5/7/25, 12:14 PM with 43 comments

by rfv6723 on 5/7/25, 2:05 PM
You can boot an OS on NVME drive from another bootloader on a USB drive.
I have been using CloverBootloader for years.
https://github.com/CloverHackyColor/CloverBootloader
by hedora on 5/7/25, 2:41 PM
It's cool they issued a new BIOS. I have an extremely similar (but older, I think -- it is the first gen of boards that support EFI) Gigabyte motherboard, and the community hacked in support for M.2 boot (with a PCIe adapter passthrough board) maybe 5 years ago.
I only looked into this because I had to upgrade the video card (NVidia's drivers go obsolete way before the hardware), and the new video card wouldn't post without a new BIOS. Gigabyte had issed one even though the motherboard was ten years old. I thought about going the NVME route, but didn't bother. The video card higher priority.
I even ran stable diffusion on a machine like that one, but a few years newer (with a covid-era video card).
by eqvinox on 5/7/25, 1:33 PM
Reminds me of editing NVMe support into my Supermicro X9DR7-LN4F server mainboard (Sandy/Ivy Bridge) I have at home. It involved "unpacking" the 'BIOS' (UEFI) image, adding an NVMe module that was published somewhere (I think it was actually generic, as in plain UEFI and not specific to the BIOS vendor, AMI in this case), repacking it, and then flashing it on the mainboard (with clenched buttcheeks — I'm not sure this board has a recovery mode…)
(It works perfectly to this day.)
It's a little sad that Supermicro didn't do this at their end, it's maybe an hour of work, two with testing…
> inclusion of the NVMe DXE trifecta (Nvme, NvmeSmm, NVMEINT13)
Now that I think about it, this sounds very much like what I threw into the X9DR7-LN4F BIOS image.
Ed.: apparently these are the files I used, though I cannot dig up how or where I did this :)
```
  -rw-r--r--   1      41917 May 28  2020 P-X9DR7-E-LN4F_BIOS_3_3_release_notes.pdf
  -rw-r--r--   1   30493975 May 28  2020 FD12LITE.zip
  -rw-r--r--   1   16293994 May 28  2020 Aptio_V_AMI_Firmware_Update_Utility.zip
  -rw-r--r--   1     818096 May 28  2020 UEFI_Shell_Spec_2_0.pdf
  -rw-r--r--   1    2104626 May 28  2020 AMIBCP 4.53.zip
  -rw-r--r--   1    5910147 May 28  2020 UEFITool_v0.27.0.rar
  -rw-r--r--   1      10665 May 28  2020 NvmExpressDxe_4.rar
  -rw-r--r--   1     404236 May 28  2020 MMTool 4.50.0.23.zip
  -rw-r--r--   1    6150536 May 28  2020 Aptio_4_AMI_Firmware_Update_Utility.zip
```
(…10kB compressed for the actual NVMe DXE…)
```
    ..A....     20832  2018-04-18 00:00  NvmExpressDxe_4.ffs
```
by AIoverlord on 5/7/25, 2:22 PM
Would be interesting to know what hardware generations are completly useless today due to other issues like this.
Linux packages no longer being available, sourcecode missing, CPU features missing, etc.
Windows 11 also kills quite a lot of hardware.
I had to upgrade my working laptop because my 4k display broke the GCP console when doing a screenshare (GCP Console became very slow when doing screensharing).
I had to upgrade my CPU when upgrading my camera because the RAW Fileformat of my new campare made my old high end CPU so much slower, it was crazy (i guess the new size broke cache line alignment or whatever, it wasn't slow with the older files).
My high end CPU from 6 years ago, i just replaced it with a new CPU (for less money than 6 years ago) and the CPU is double the performance.
I used a older system of mine on a hacker event for file sharing and other hacker experiments and redid the same thing with a better system. All the issues disappeared i had regarding performance.
Doing this now for 20 years, hardware is a lot more commodity than i thought.
by fithisux on 5/7/25, 12:25 PM
The issue here is the lack of open documentation of chipset, device providers.
The end user is hostage.
We talk about AI/smartphones whatever
but the devices are not open or smart.
by oso2k on 5/8/25, 3:38 AM
The Dell Optiplex community did this a few years ago and they were even able support older generations of Optiplex. The actual NVMe boot module is pretty generic and I’d suspect it would work with other machines.
https://tachytelic.net/2021/12/dell-optiplex-7010-pcie-nvme/
https://tachytelic.net/2022/02/dell-optiplex-790-990-nvme/?a...
by accrual on 5/7/25, 1:25 PM
Always cool to see old boards learn new tricks. I thought for a second "wait, my Thinkpad X230 from 2012 can boot from M.2" but then I remembered it has mSATA, not M.2.
by jeffbee on 5/7/25, 1:04 PM
It is weird how the article freely confuses M.2, a physical expansion card form factor, with NVMe, a logical peripheral device interface. The motherboard obviously did not gain M.2 slots after the fact, even as the article points out that it has always been possible to use adapters to put M.2 devices into any PCI slot, and there have always been NVMe devices in the larger card form factors. What the platform gained from its new software is NVMe support.
by Dwedit on 5/7/25, 2:35 PM
Way way back, there were hard disk size limitations that were really only software limitations rather than hardware limitations. So there were tools that could install themselves to the boot sector of the drive, and provide support for larger hard disks by replacing the BIOS's disk access functions.
There's the 504MB or 2GB limit that was removed by using EZ-Drive, and the 8GB limit that was removed by using EZ-BIOS.
by lotharcable on 5/7/25, 9:21 PM
Board firmwares are full of copy-paste and code lifted from other devices that is never intended to get used and is supposed to be disabled by default.
I wouldn't be surprised if this was 100% by accident.
by Calwestjobs on 5/7/25, 12:41 PM
Who cares, people still use bootloaders in era of UEFI (uefi can directly call your freaking kernel), so this is like throwing pearls to pigs.
But it is awesome that there is community of people who are interesting enough in this stuff to find these things out.