NVMe - make the most of your PCI-e Slots (How-to config Supermicro boards for AOC-SLG3-2E4T et al)

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

pipe2null

New Member
May 25, 2021
14
3
3
I have successfully booted various M.2 NVMe's on X10DRU-i+ motherboards, which is not supported with stock bios. These booting drives are NOT on any compatibility list I am aware of, but maybe they are?: Samsung 970Evo 2TB, Western Digital WD Black SN750 500GB, Western Digital SN730 256GB. I have tested with both single M.2 slot and 4x quad M.2 slot adapter boards (the Asus one, turning on x4x4x4x4 bifurcation). By adding one DXE UEFI driver to the bios image, I've had no problems booting M.2. It does not support M.2 RAID arrays, but single drives show up as boot options just like any SATA/SAS/USB drive. I use UEFI mode only, no legacy, and I have no idea if a similar solution would help with U.2 formfactor drives especially with HBAs and backplanes added to the HW mix...

The guides I followed are:
 

Eta_Carinae

New Member
Feb 24, 2020
12
2
3
40
Perth, Western Australia
I have successfully booted various M.2 NVMe's on X10DRU-i+ motherboards, which is not supported with stock bios. These booting drives are NOT on any compatibility list I am aware of, but maybe they are?: Samsung 970Evo 2TB, Western Digital WD Black SN750 500GB, Western Digital SN730 256GB. I have tested with both single M.2 slot and 4x quad M.2 slot adapter boards (the Asus one, turning on x4x4x4x4 bifurcation). By adding one DXE UEFI driver to the bios image, I've had no problems booting M.2. It does not support M.2 RAID arrays, but single drives show up as boot options just like any SATA/SAS/USB drive. I use UEFI mode only, no legacy, and I have no idea if a similar solution would help with U.2 formfactor drives especially with HBAs and backplanes added to the HW mix...
Both my intel 1.2TB 750 drives show up as being bootable, but I'm not booting off them (this is incorrect and you're right @pipe2null)
— they're storage only. I boot from satadom. Both have worked without issue for quite some time (not including them running only at PCIe 1.0). The BIOS modification you linked to discusses allowing the 750 to be bootable which isn't a problem for me. The p4610 also shows up as bootable, but causes intermittent hang during the boot process and then BSOD or reset upon reaching Windows.

I think I need to switch the AOC-SLG3-4E4T from the 1st PCIex16 slot (CPU2 Slot 5) over to the 2nd PCIex16 slot (CPU1 Slot 3) per the BPN-SAS3-826EL1-N4 backplane jumper configuration options:

screenshot-www.supermicro.com-2021.06.03-10_32_13.png

as there's no option to run all 4 NVMe from CPU2 like there is for CPU1.
 
Last edited:
  • Like
Reactions: Almighty

Eta_Carinae

New Member
Feb 24, 2020
12
2
3
40
Perth, Western Australia
You were correct, @pipe2null. None of the NVMe are bootable. I'd never tried and just asumed, but looking at the BIOS the NVMe are not presented as being bootable.

Also, switching the AOC-SLG3-4E4T HBA to the other x16 slot changed nothing nor did it resolve the 750s negotiating at PCIe 1.0.
Supermicro released a new bios 3.3 for my motherboard as it's currently on 3.2 but I doubt it will resolve this issue.

I think I'll have to look at adding the the DXE UEFI driver to the new 3.3 bios revision.
 

Eta_Carinae

New Member
Feb 24, 2020
12
2
3
40
Perth, Western Australia
Looks like I can rule out the BIOS (3.3 installed), AOC-SLG3-4E4T HBA & BPN-SAS3-826EL1-N4 backplane. Just tried the p4610 in a U.2 to PCIe card and same problem. Not too comfortable with BIOS modding so I think I'll give up.
 

pipe2null

New Member
May 25, 2021
14
3
3
My bios suggestion was just a shot in the dark, especially if you're booting your system off of a satadom. But if bios is not handling the drive correctly or only partially handling it instead of just ignoring it, I suppose it could cause problems handing the device off to the booting OS kernel.

I'm no expert on the subject, just speculating.

Modding bios might not have any positive effect on your problem, but before you give up on bios modding as an option, I would suggest going through the steps to test your comfort level first. You can start with a binary bios image straight from supermicro, the same flash image that you would normally use to flash/update your mobo. Flashing the resulting modded bios image is the very last step and everything up to that point is completely risk-free. If you are able to flash your bios over IPMI, your BMC can keep running even if your bios is corrupted since BMC has it's own flash and is kinda a separate machine running on the same mobo, so that is one recovery option if modded bios fails.

If you somehow end up with a corrupted bios on your mobo AND are unable to flash over ipmi AND system rescue USB procedure does not work: If you have a raspberry pi laying around, you can pick up a pomona SOIC-8 test lead clip (available on amazon) and directly re-flash the IC. On an X10DRU-i+ motherboard, this can be done completely in-circuit with no lifting of pins nor soldering nor external resistors, only clipping onto the bios flash IC with pomona clip and wiring to a raspberry pi. There are other supermicro mobo's that are designed the same way to allow "easy" no fuss in-circuit reprogramming. Grounding yourself for static is important, and completely unplugging power/etc is required. Also good idea to pop out the cmos battery as well, just remember to reinsert it before booting. NOTE: The process also cleared all previously installed Certs/settings/etc and reverted to a completely clean/default bios state.:

Pomona+Pi pinouts to use, and read/write commands:

Probably should ignore the "-l layout -i bios" part and ignore the hex hacking:

There are also various servethehome threads on using hardware programmer (or raspberry pi) to flash firmware

I recently bought (2x) X10DRU-i+ motherboards off ebay that were advertised as "Ready to Reuse!!!", but they arrived with both IPMI and BIOS passwords set, and the boot device configuration settings did not allow booting off the media I tried with. Clearing CMOS memory on one mobo left the system in a bad unbootable state. The second motherboard I immediately did the USB system recovery method but that failed and left it in a different unbootable state. Both boards were essentially bricked even though IPMI was still running but non-default password locked. But the raspberry pi + pomona clip flashing method worked to recover both motherboards bios with cleared passwords and once I had a booting OS I used the ipmitool (or equiv) to clear the ipmi password.


It is of course a "proceed at your own risk" kinda thing, but everything up to the point of flashing a modded bios is risk-free.
 

pipe2null

New Member
May 25, 2021
14
3
3
As an afterthought, if your drive setup is supposed to be hot swappable (for an NVMe), you might try booting windows and THEN inserting the drive to backplane to see if windows detects it and/or it is working correctly. If windows can use it just fine when inserted after system boot, it might indicate a bios or bios->kernel device handoff kinda problem.
 
  • Like
Reactions: gb00s

uldise

Active Member
Jul 2, 2020
209
72
28
I'd like to report that the Supermicro AOC-SLG3-4E4R card fitted in slot 6 of a Supermicro X11SPi-TF motherboard doesn't work in my setup.

Intel DC P4510 (1TB) NVMe drives connected via the BPN-SAS3-826EL1-N4 backplane of my Supermicro SC829HE1C4-R1K02LPB chassis to ports NVME0, NVME1 and NVME2 of the Supermicro AOC-SLG3-4E4R card with Supermicro CBL-SAST-0590 cables are recognised correctly and work fine.

From the moment I connect any of my 4 Intel DC P4510 (1TB) drives to port NVME3 of the Supermicro AOC-SLG3-4E4R card, the server remains stuck during the boot process at the screen DXE-IPMI Initialization.
sorry for hijacking old post, but looks like i just stepped on the same problem with AOC-SLG3-4E4R in X11DPi-NT. two ports works just fine (NVME0 and NVME2). when i connect additional cable to NVME1 and/or NVME3, system just hungs on boot with error code 92.
do i have options, or in this board only two ports will work?
have anyone tried this AOC-SLG3-4E4R on newer X/H12 boards?
 

uldise

Active Member
Jul 2, 2020
209
72
28
@RolloZ170
so i tested two more combinations:
1. i connected two ports to AOC-SLG3-4E4R, two ports to motherboard Oculink NVME slots - boot hangs at the same place with error code 92.
2. two ports connected to motherboard Oculink ports, and then i added 4 port x16 M2 NVME card to the same PCIe slot2 with 3 M.2 NVME drives populated - this card needs bifurcation too. and all is working - i see total 5 nvme drives in Linux.

so that mean - either my new AOC-SLG3-4E4R is broken or is incompatible with this board..
 

RolloZ170

Well-Known Member
Apr 24, 2016
5,316
1,604
113
so that mean - either my new AOC-SLG3-4E4R is broken or is incompatible with this board..
M.2 drives direct is not like long NVMe cable with unknown device at the end.
0x92 PCI Bus initialization is started
 

borja

New Member
Sep 14, 2022
1
0
1
Hi all!
I dont know if this thread is where i've to post that... if not just tell me.

i've got a supermicro server with X10SRL-F motherboard and i have to install 4 Supermicro AOC-SLG3-2E4R-0 cards.
By the moment i'm trying with only 1 pci card but it dont work...
so i follow all the steps that you explain here for my MB but it does not work.... any idea?

thx
 

sth

Active Member
Oct 29, 2015
379
91
28
Ive had issues with compatibility with those cards and some motherboards. Supermicro bounced back my support request a few years back but maybe things have changed since. Theres some posts from me on here for like 4 years ago that may have some details. I had success in one system with one card that would on warm reboots detect one of the drives, but on cold it would detect neither. It played havoc with my ZFS arrays having drives be randomly disconnected etc.

smcompat.jpg
 
Last edited: