The official AMI System Firmware hate thread

[Ericloewe]

After having spent a few hours messing around with my workstation, trying to get it to boot, I decided it's time for an AMI BIOS hate thread.

These things are insanely popular - most of you will know them as the !§"%@?!$ that Supermicro uses on their motherboards (where they're also a pain in the ass).

I'll start by airing my grievances:

• Random boot loops. On both my ASRock X99 WS and my X11SSM-F, these things often hang in the middle of the boot process. Sometimes they even reboot. When it does get to the OS, everything works normally, most of the time...
• Trying to fix the above on my ASRock board, I tried to update to the latest version (I hadn't updated in over a year, I think). Big mistake. With the latest BIOS, Windows 10 would not boot. Under any circumstances. At all. It would just hang after a few seconds. I tried everything short of removing internal hardware (except for a small USB Bluetooth adapter I forgot in one of the USB ports) and it would not work. Not in UEFI mode, not in BIOS compatibility mode. Same goes for the boot recovery assistant, same exact symptoms. In the end, I had to flash an older BIOS, so I went from 1.6 to 3.2 and back to 1.7. It didn't seem to work at first, but then magically worked. POST hanging issues still abound, though...
• Their legacy compatibility stuff is absolute trash. Back in the 9.3.1 days, I tried to install it on my X11SSM-F - which didn't work at all, as documented elsewhere. Then I tried my ASRock X99 WS - no luck either. Finally, an Asus Haswell board (exact model escapes me at the moment), with the same AMI scourge, gets the job done - after some hassle getting the boot override functionality to work.
• Their boot override option doesn't work consistently! How could they mess that up? It works badly on my Supermicro boards, it works badly on the Asus board and it's not even present in the 3.2 BIOS for my X99 WS! It's so bad they removed it!

Phew, that's all I can remember for now.

I eagerly await horror tales of AMI BIOSes.

 

[MrToddsFriends]

Under certain circumstances my A1SAi-2750F (AMI BIOS Revision R 1.1, A1SAi5_109) doesn't reliably boot from the devices configured in Boot -> Boot Option Priorities.

With USB boot sticks (a mirrored pair of SanDisk Ultra Fit 16GB USB3 devices, both of them configured as boot devices) plugged into USB2 ports of the board, roughly 5 out of 10 boot attempts result in the famous "This is a NAS data disk" message. I don't know if the BIOS refuses to obey it's settings or if it isn't able to find the configured boot devices when this happens.

With the same devices plugged into USB3 ports of the board (usable since FreeNAS 9.10) booting from the configured devices seems to operate reliably.

 

[Ericloewe]

Two days after my last adventure, they released ASRock 3.3 BIOS. Still nothing, except this time I tried everything. Different graphics card, no peripherals except USB keyboard, another USB keyboard in case the first one was causing the issue... And the piece of crap refuses to boot Windows or Ubuntu, at which point I just gave up and I'm going to make their customer support hear about this.

 

[hateredblast]

AMI bios is total trash. It was trash back in the 90s, when I got my first PC. It is total trash now. I hate it with a passion!

 

[hateredblast]

I registered on this forum and replied to a 7 year old thread just to tell you all how much I hate that piece of shit AMI!

 

[Ericloewe]

7 year old

Are you trying to give me a heart attack? Work has not yet had that much of an impact on my perception of time...

In any case, the system firmware update with the updated microcode with the spectre and meltdown mitigations magically solved my "can't boot with BIOSes later than 1.x" problem. It's the craziest thing.

These days, running with CSM disabled, the system firmware almost feels like a decent piece of software. Until the built-in boot manager fails to correctly display graphics, but I'll blame that on ASRock instead of AMI and just install rEFInd somehow when I next mess around with boot devices. If I feel really adventurous, I might try to patch into the ROM instead of ASRock's crappy boot manager.

 

[Ericloewe]

It's been another couple of years. AMI continues to piss me off.

So, on X10SDV boards, and likely many others, the setup application does not actually allow you to create boot options, i.e. to define a device, FAT32 partition and path to boot. But they do have a mystery option that says "create boot option". Unfortunately, it creates a driver rather than a boot option. Apparently, a driver entrypoint is indistinguishable from a boot payload entrypoint, because the firmware will happily execute said "driver" and load the payload...

...before it completes hardware initialization. So, loading ZFS Boot Menu with this weird option will bring it up, it will be able to do OS things and load a Linux kernel and userland... But much of the hardware will be messed up. USB controllers go crazy for a while, though Linux does successfully intimidate them into being controlled by Linux. PCI devices all seem fine, though. The real surprise is that the system will have booted with a single CPU and 2 GB of DRAM (out of 8 threads and 128 GB, in the case of this machine).

On a positive note, if anyone wants to load an NVMe driver from USB on X10SL* boards, this option should do the trick nicely. If you can get a suitable NVMe driver.

So thanks for another hour of my life wasted figuring this one out, AMI.

 

[Ericloewe]

Random boot loops. On both my ASRock X99 WS and my X11SSM-F, these things often hang in the middle of the boot process. Sometimes they even reboot. When it does get to the OS, everything works normally, most of the time...

With seven-and-a-half more years of grey hairs, a few things to look out for on this topic:

• On nominal systems, a lot of this is due to CSM. Disable all that crap yesterday, CSM is a major contributor to dragging system firmware down in terms of functionality, quantity of bugs and general unpleasantness. It's 2024, if your OS does not boot reliably from UEFI, change your OS. If you absolutely cannot, then it should have been in a VM years ago.
• This can also be caused by a dying clock battery, corrupting settings (notably memory controller DIMM training caches often cause trouble) and generally creating lots of unpredictability. If your system "suddenly" starts doing weird things when booting, try replacing the battery. Afterwards, clear the settings and reconfigure your system (and let the firmware retrain the memory controllers).

 

[Davvo]

Still can't believe it has been 8 years since 2016.

 

[homer27081990]

After having spent a few hours messing around with my workstation, trying to get it to boot, I decided it's time for an AMI BIOS hate thread.

These things are insanely popular - most of you will know them as the !§"%@?!$ that Supermicro uses on their motherboards (where they're also a pain in the ass).

I'll start by airing my grievances:

• Random boot loops. On both my ASRock X99 WS and my X11SSM-F, these things often hang in the middle of the boot process. Sometimes they even reboot. When it does get to the OS, everything works normally, most of the time...
• Trying to fix the above on my ASRock board, I tried to update to the latest version (I hadn't updated in over a year, I think). Big mistake. With the latest BIOS, Windows 10 would not boot. Under any circumstances. At all. It would just hang after a few seconds. I tried everything short of removing internal hardware (except for a small USB Bluetooth adapter I forgot in one of the USB ports) and it would not work. Not in UEFI mode, not in BIOS compatibility mode. Same goes for the boot recovery assistant, same exact symptoms. In the end, I had to flash an older BIOS, so I went from 1.6 to 3.2 and back to 1.7. It didn't seem to work at first, but then magically worked. POST hanging issues still abound, though...
• Their legacy compatibility stuff is absolute trash. Back in the 9.3.1 days, I tried to install it on my X11SSM-F - which didn't work at all, as documented elsewhere. Then I tried my ASRock X99 WS - no luck either. Finally, an Asus Haswell board (exact model escapes me at the moment), with the same AMI scourge, gets the job done - after some hassle getting the boot override functionality to work.
• Their boot override option doesn't work consistently! How could they mess that up? It works badly on my Supermicro boards, it works badly on the Asus board and it's not even present in the 3.2 BIOS for my X99 WS! It's so bad they removed it!

Phew, that's all I can remember for now.

I eagerly await horror tales of AMI BIOSes.

I currently have a "serverized" desktop in a rackmount box (living with my parents, they couldn't stand the noise of my DL380P Gen 8, so I sold it, anyway...) with a Ryzen 9 5900, 4X16GB DDR4 4200MHz RAM running at 3600MHz and a Gigabyte mobo with AMI UEFI...

It always needs to be rebooted in order to boot, both after a reboot or a shutdown.

• No errors,
• Memory OK,
• I have a UPS (EATON Ellipse ECO 1000KVA),
• PSU OK (seasonic),
• No GPU, so no GPU problems,
• BIOS updated,
• All temps normal and below

System also has:

• One NVME (mini PCIe) drive,
• One SATA HDD (exos 14TB),
• An Intel 2-port gigabit NIC (actually, one NIC per port) for my PfSense...

I have already considered and excluded the possibility of a defective mobo (all caps check out, no over-current spikes (meaning no undervoltage, sustained or sipkey), tested with different RAM, different timings and different CPU (from my desktop, so, confirmed working).

Also tested my Ryzen 9 on my desktop mobo for good measure, no problem there...

In all, 68 hours of testing down the drain.

I will end up just spending another day, switching the mobos of my desktop and server... Just too lazy to get around to it.
Also not a huge fan of rebooting my desktop every time I want to boot or reboot...

And, I understand that reading this, most people would jump up to say "sshhhh man, defective mobo, RMA it", but trust me, it is not the hardware that's defective.

Once you reboot (and to add here, doesn't matter when you reboot, doesn't matter if you drain all caps with the power off and then reboot, reseat the CPU, etc, just that you do reboot) the system is fine.

Currently having an uptime of 38 days (consecutive).

Sort of a poltergeist or witchcraft, it has got to be the BIOS.

Kind of a long rant, sorry, but that quirk means that I cannot even update and reboot without being next to the machine physically.

EDIT: Forgot to say, of course, even if there is no overcurrent happening, I mounted the whole system on my wooden kitchen table (grounded wrist with 1MΩ resistor to mains ground) and checked the case for points of metal-to-metal contact, even on the IO shield, but got zilch.

 

[homer27081990]

Also, while I know this should/could be a thread on another forum, seeing as that would solve most of my problems with the board and be a nice project-learning opportunity, I would love to see a guide for a dirt-cheap way to make your own PiKVM... While I can make my own (pretty basic) THD and large SMD boards (both the design with KiCAD and the assembly with my soldering station), I don't trust me enough to make something and mount it on my 1700 euro server that has irreplaceable and business related data on it...

 

[homer27081990]

Still can't believe it has been 8 years since 2016.

True...

 

[Ericloewe]

I would love to see a guide for a dirt-cheap way to make your own PiKVM

The hardware I've typically seen used has been rather expensive. With the long unavailability of Pis of all sorts, I never looked into it long enough to care much about it. By now, all the machines I care about have IPMI that works...

 

[homer27081990]

The hardware I've typically seen used has been rather expensive. With the long unavailability of Pis of all sorts, I never looked into it long enough to care much about it. By now, all the machines I care about have IPMI that works...

My problem is mostly a geographical and geopolitical one... All IPMI enabled boards I could find have a 150-250€ markup in Greece... Couple that with the fact that I cannot use refurbished enterprise grade HW (because most of the time, the sound like an Airbus A380 and I don't live alone AND am prone to migraines...) and you get why I think this is a good idea

 

[homer27081990]

The hardware I've typically seen used has been rather expensive. With the long unavailability of Pis of all sorts, I never looked into it long enough to care much about it. By now, all the machines I care about have IPMI that works...

Also, ha, I have a RPi 4 laying around with a HaOS that does almost nothing...

 

[Davvo]

Couple that with the fact that I cannot use refurbished enterprise grade HW (because most of the time, the sound like an Airbus A380 and I don't live alone AND am prone to migraines...) and you get why I think this is a good idea

I use a second hand server-grade motherboard and it's wisper quiet... it depends on how you tune i. Also, in Europe you can find such MB for a decent price... example:

Supermicro Motherboard MBD-X11SCL-LN4F-O E-2100 LGA1151 C242 | eBay

Le migliori offerte per Supermicro Motherboard MBD-X11SCL-LN4F-O E-2100 LGA1151 C242 sono su eBay ✓ Confronta prezzi e caratteristiche di prodotti nuovi e usati ✓ Molti articoli con consegna gratis!

www.ebay.it

 

[homer27081990]

I use a second hand server-grade motherboard and it's wisper quiet... it depends on how you tune i. Also, in Europe you can find such MB for a decent price... example:

Supermicro Motherboard MBD-X11SCL-LN4F-O E-2100 LGA1151 C242 | eBay

Le migliori offerte per Supermicro Motherboard MBD-X11SCL-LN4F-O E-2100 LGA1151 C242 sono su eBay ✓ Confronta prezzi e caratteristiche di prodotti nuovi e usati ✓ Molti articoli con consegna gratis!

www.ebay.it

The problem is that I also want as low as possible of an idle power draw and nice performance under load, and on a reasonably recent platform, so I cannot use those. I found new AM4 ones (supermicro with IPMI) for ~450€, but... I cannot produce the amount, hahahaha. To tell the truth, it was either my RTX4060Ti, or the supermicro, so I played it heads or tails... Tails won.

 

[Davvo]

Then it's not a noise issue

 

[homer27081990]

Then it's not a noise issue

Touche... Hahahahaha...

 

[Etorix]

I found new AM4 ones (supermicro with IPMI) for ~450€

Don't restrict yourself to Supermicro and have a look at this thread.
AM4 with IPMI: Gigabyte MC12-EC0, 55 E + shipping, with cooler.
Even cheaper, as it already includes the CPU and uses RDIMM, but with limited expandability: Gigabyte MJ11-EC1, should come out just under 100 E with a SFF-8654 4i to 4*SATA breakout cable from eBay (China) and a quiet 60 mm fan.