Unbricking a BIOS-Bricked Motherboard | Hacker News Reader

Unbricking a BIOS-Bricked Motherboard

davidesnotes.com

144 pointsbelter51 comments

When I was 13 or so I bricked my A7A-266. This had me beyond tears and in that petrified “my world is ending” state.

After hours of a sunken stomach and reading online, I cobbled together a floppy disk with a bunch of special files that I didn’t understand. I think I even had to type one out in notepad. I put it in the disk drive. It tells you to wait five minutes and then reboot but I waited ten. There were no beeps or signals. I rebooted and the thing came right back to life and lived for years until an A7V8X promised me a good time with faster RAM.

Godspeed my first private computer motherboard. You recovered from a teenager-induced coma and got me through most of high school.

That feeling of absolute dread, thinking 'my parents are gonna kill me when they realize I killed the family computer' did a lot to prepare me for the white-knuckle experience of troubleshooting problems in Production. :)

Bless the talented engineers that put together last-ditch recovery methods and bless the folks that documented things well enough to let us make use of them.

My first family computer was a Tandy 386-sx that came with DOS 5.0.

Not soon after, we got an update to DOS 6.22, which also came with Doublespace, an on the fly disk compression system, which upgraded our 80mb hard drive to 116 glorious megabytes.

A few months later, I did something stupid and decided to reinstall everything from scratch. Used a boot disk and formatted the hard drive, tried to reinstall DOS 5.0 and the hard drive reported that it only had 2 Megabytes of storage available.

Instant panic mode.

I had no internet, I had the manual for DOS 5.0 and 6.22 and about 6 hours before my mom got home from her night job before I got murdered.

After frantically reading page after page of documentation, I found that once Doublespace is activated, the hard drive only has enough normal readable room to launch Doublespace, not to launch DOS.

A bit more reading, and I found that I could make a DOS 6.22 bootable floppy disk with the Doublespace system on it, so that is what I did.

Breathlessly I booted the PC with the floppy and it worked! Full disk file size visible! I hurriedly opened Doublespace, de-doublespaced the drive, and rebooted.

From there it was a race against time to reinstall DOS 5.0 (from 5 disks), install the DOS 6.22 update (3 disks), install Windows 3.11 (8 MORE disks), install all of the other software (12-15 disks in total) and then make sure everything was signed in and set up before the garage door opened.

I was literally putting the finishing touches on everything when my mom surprised me by walking in the front door. She had parked in the driveway instead of the garage. She looked at me, asked me what I was doing to the computer, and I told her that I had messed something up and her solitaire history had been deleted (which was true).

She said it was ok and went to bed and I collapsed after a marathon 8 hour blind computer repair.

Oh man. I bricked a board doing a bios update while I had time time set funny (post 2038). Well at least that's what I think the problem was. Floppy drive recovery didn't work out for me. But I ended up getting a duplicate board on a job later, and ruined that one by spilling water on it in a caseless setup. But I was able to pull the BIOS chip and swap it to the board I had messed up earlier. I guess I probably could have contacted support and gotten a new BIOS chip before, but that didn't occur to me until now, 25 years later.

This was a Shuttle super7 board with the AMD 640 chipset that got disappeared. I was trying to flash the bios related to getting the software for a distance learning macroeconomics course to work. The course was still in development, there were a bunch of videos on cd-rs... Ended up getting a new computer instead.

I feel you, around the same age I constantly bricked my BIOS. I forgot what motherboard it was. I had to find the same model motherboard from somebody, boot it with their BIOS, swap in when booted and use some rescue floppy to clone to BIOS.

It happened quite a lot. Later I found out that it always happened when shutting down Windows 95. Just turning the power off did not result in a bricked BIOS.

It was also the first time I felt ripped off. I went into a computer repair store with it. Told them what was wrong and how they could fix it. Two days later they told me to collect my PC. They did not fix it because the cause was a bricked BIOS, had to pay all my pocket money for the cost of their 'research'.

Still wonder why shutting down Windows 95 made this happen, never found out, maybe somebody here knows it?

This only works for certain motherboards that have support for flashing the BIOS from USB without a even a cpu or memory populated.

I'm not sure, but think that functionality requires at least some part of the previous BIOS to be intact.

If this solution doesn't exist/work, you will need to go the route of finding the (often undocumented, sometimes unpopulated) SPI flashing header, and using an external device to directly write a working BIOS onto the SPI flash chip.

Flashback functionality without a functional CPU generally is accomplished using a separate microcontroller on the board which runs its own code separate from the BIOS. It shouldn't need any of the original BIOS code in-tact, since there's nothing to execute it anyway if there's no CPU installed. :)

I have no idea how it's actually implemented.

It would be smart to implement it as seperate chip, so it can work even with an empty SPI flash. Would also explain why it's only available on some motherboards.

But it wouldn't entirely surprise me if they managed to implement it as pure software inside an extra microcontroller core inside the chipset (or another chip). There are dozens of microcontrollers cores spread out over the various chips on the motherboard which might be abused for such purposes.

Long ago, I botched a mobo BIOS update (Abit BH6), and for the reasons you mentioned/my incompetence, couldn't do another write through the floppy method.

Sold it on eBay (disclosed as such).

Later sent a message to the buyer asking if/how they fixed it. They shared that they did: with a similar mobo, booted the good one up, hot-swapped the EEPROMs and then ran the BIOS update on the bad chip.

Smart.

I managed to do that once after flashing the BIOS for the wrong board, back when the EEPROMs were in DIP sockets.

Another time I was less lucky and broke an edge off the CPU die while installing the heat sink. This was during the time when the CPUs had exposed silicon with no metal lid. Instant dead CPU, expensive mistake for a college kid.

These were PLCC32 or PLCC40 or something like that. Probably an 29f400 or something like that. But thankfully socketed!

Yes, it needs an intact bootblock to do this. I'm actually slightly surprised that BIOSes now have USB support, or at least just mass storage and FAT, in the bootblock. The floppy controller is much easier to program directly to do this.

> you will need to go the route of finding the (often undocumented, sometimes unpopulated) SPI flashing header, and using an external device to directly write a working BIOS onto the SPI flash chip

Y'all these days are spoiled with your headers, test clips, howto videos and cheap EPROM programmers.

Had to reflash an i2c 24-series eeprom on a satellite receiver. You don't want to power the whole board, so gotta lift the VCC pin off the SOIC chip (not hard if you warm it up and pick at it with a soldering iron), solder kynar to leads/nearby lines for VCC, data and clk lines, and aligator clip for GND. Connected to some self-made janky 24 series parallel port interface.

I think the article in question actually is for older BIOSes that don't have flashback or similar. Instead this relies on the "BIOS Boot Block" still being intact in the flash, so you do need a partially working BIOS flash, CPU and memory for this method.

My favourite, most reliable method is to use a TL866 family rom-burner, they are also dirt cheap.

There's conveniently an open source "minipro" programming tool for TL866 family. (linked by another poster)

That + a clip for SOIC, WSON or whatever applies to the motherboard, if you're not lucky enough to have a socketed DIP EEPROM.

Else, something supported by "flashrom", which has a load of supported programmers, including a range of microcontrollers and even bitbanging with dirt cheap FTDI uart usb dongles.

That or an even cheaper CH341A (which is what I use).

Careful the most common cheapo such board uses 5v for io[0] even though it has a 3.3v regulator onboard. Shit design.

0. https://www.eevblog.com/forum/repair/ch341a-serial-memory-pr...

Years ago, on an old (~2000/2001, AMD K series maybe?) system, I was doing some patches for a friend of mine and noticed there was an update for his BIOS. Went through the normal process of backing up to a floppy (luckily, I didn't always do this on my own upgrades before this point!), began the flash process and it all appeared to work fine. Rebooted and the machine didn't come up at all. Blank screen, nothing. I totally panicked. He came to check in and I must've been profusely sweating by this point and we chatted it through, but I noticed, on reboot, the floppy drive gave a tiny click, not like the normal exaggerated one it would do when normally booting. I had a glimmer of hope that it might still read a boot record from the disk, which it did. When the floppy stopped grinding, I carefully typed in the commands (even though there was no output to screen), I knew were on the disk (after running back home to build a bootable disk with a flasher and the BIOS dump) managed to restore the BIOS back to running perfectly with the old version!

Was absolutely terrifying, but very rewarding to get back a machine that we were both sure was bricked!

I remember having to flash the spi chip with a raspberry pi, flashrom and janky wiring on an oldish Thinkpad. Not so good times :)

coreboot/libreboot? I still keep the 8-pin chip clip I used, even thought I'm probably never going to flash another chip using it again (terrible contacts), but at least a good memory!

Or just kill the wifi blacklist and make it hackintosh ready. ;]

Close, it was to kill the stupid wifi blacklist but just to upgrade the wireless n card with an ac one I had lying around form a broken machine.

What is this referring to? What wifi blacklist?

Some laptops (notably Thinkpads) enforce a whitelist of PCI IDs for wireless cards - an "unapproved" card will display a warning in the firmware and won't be visible by the OS.

The excuse if I remember right is FCC compliance, but I call BS given that most other manufacturers successfully get away without it.

What about the vast majority of us who don't care what the atavistic FCC say?

That is interesting... I never knew that until now...

Are there any other defining features of the cards that are accepted or the cards that are rejected other than PCI ID? (i.e., Manufacturer, Chipset, MAC, Country Of Origin, Year Manufactured, ?, ???)

Also (and I say this not to you, but to the Internet at large) --

If every PCI ID belongs to a Manufacturer (a company -- and they do!), and some companies cards are accepted and other companies cards are rejected, then...

Wouldn't rejecting some manufacturers Wi-Fi cards -- be in conflict with Federal Trade Commission Antitrust regulations?

https://en.wikipedia.org/wiki/Federal_Trade_Commission

https://en.wikipedia.org/wiki/United_States_antitrust_law

...In other words, while IBM or Lenovo, or whoever makes the Thinkpad these days -- while they would be complying wtih FCC regulations...

Wouldn't they also simultaneously be in conflict with FTC Antitrust regulations?

I managed to brick my Asrock X370 Taichi a while ago... I had hoped that a BIOS update would fix some issues that I had, but did not notice that they removed a few things, presumably to free space and add support for newer CPUs. The update (from 6.40 to 7.xx) itself went fine, but unfortunately broke IOMMU groups and maybe more. Downgrading was not an option, I assume to prevent people from installing an older version that is incompatible with their CPU.

I managed to flash an older image with a tool from the BIOS vendor, but apparently messed up some settings and the PC would not even boot anymore. Fortunately, there was a pin header on the board right next to a serial flash chip and after some probing and fiddling with a cheap stm32 board I managed to get it running again.

I might upgrade to a newer CPU soon-ish which will again require the never BIOS version, let's see how that goes...

There are dozens of us! (It's nice to know I'm not alone)

I bricked an Alienware gaming PC back in 2014 and was facing having to junk the whole PC because replacing the motherboard would be too expensive. Found a community online that stepped me through using my RPI to reflash the BIOS. Used the PC for another 5 years after and still have it in my closet now its something I shouldn't be so proud of but still am ha

Thanks for posting this. I would never have though to check if this was possible even though I know rPi are being used as pseudo mod chips in gaming consoles.

I bricked a gpd pocket at one point and one of my proudest accomplishments was unbricking it.

I was trying to run openbsd on it and openbsd was not happy. so I install some firmware(bios) with all the options unlocked. I then proceeded to disable usb3, the idea here was that I knew obsd had some usb3 problems and disableing it would force usb2. however it had the effect of disabling all usb on the device. On the gpd pocket everything is usb, so the keyboard did not work and no way to get back in the bios and reenable it. At this point I would go for a cmos reset. but the gpd pocket has no cmos reset capability. Completely stuck, even tried to drain battery to see if that would do something. I saw a forum post you could pull the mainboard and reflash the winbond cmos chip, so I bought a cheap chip flasher and connector kit but was not having much luck. But did not know why so it sat on the healing bench for 2 years. At which point I was cleaning up and saw it and decided to give it another try. I bought a nice sioc8 clip(pelican) this time, hooked it up to the same chip flasher/3volt level converter setup used flashrom and the infernal thing flashed successfully. the cmos had reset and the battery even had a charge after all that time. the system came right up.

The moral of the story is that it pays to have a nice mechanical interface. especially if you do not know what you are doing.

One would think this situation would be better documented by vendors, instead of relying on obscure incantations. But one would probably be told by Customer Services that this defect is not covered by warranty (if you still have it) and you're out of luck.

Some do, e.g.: https://support.hp.com/us-en/document/ish_3932413-2337994-16

Ah, if only I knew that years ago.

A botched BIOS upgrade bricked my mobo. I had a friend with the identical same model, had to remove the chip from mine, go at his house, and with his system turned on remove his chip, put on mine, flash the BIOS and put his back again.

Scary but a learning moment.

I remember when I broke our first ADSL USB (not RJ45 modem, the model was Netopia Cayman something). I was trying to port forward for NAT, but that damn thing had an old firmware that didn't have that feature. I was naive then, didn't quite understand what a firwmare was. So, I read on the official website that the new firmware version had port fortwarding. I downloaded the firmware and started the firmware update process through the web-interface. I waited 30min or so, I realized something was off, it should not take that much to flash a couple of mb. But stupid me, didn't know that unplugging the power cord while it was flashing was a bad idea.

I was scared to death. Once broke, I didn't know where to look for informations to unbrick it. Spent 2 weeks with the ISP support center to get a replacement. Till today, I have anxiety when it comes to flashing firmwares.

If I recall correctly (many years have passed) that was used to recover from the effects of Chernobyl virus:

https://en.wikipedia.org/wiki/CIH_(computer_virus)

but (for whatever reasons) it didn't work on all motherboards, and when it didn't work the hot swap method was used/needed.

I remember fixing a couple motherboards with the latter, being lucky enough to find a similar BIOS chip from a stack of broken motherboards at a local computer shops.

I bricked a Nintendo ds while trying to flash a bios (? Not sure if right term) that would skip the health and safety screen on boot.

You had to open up the back and hold some aluminum foil in a hole where it would bridge a couple contacts on the main board to make the bios writable. I must have slipped at some point in the process and it got bricked.

The bios on those things was stored on a daughter board together with the wifi. Ended up having to hotswap that daughter board from a working ds to fix it.

I wonder if this also should work with AMI APTIO UEFI capsules.

Currently dealing with a similar issue where for no apparent reason on a regular reboot motherboard fails to even start to POST. Without RAM modules it emits the manual-defined two beeps, but with a RAM module in but no GPU it doesn't emit the defined no-GPU three beeps. Not clear if those beeps require a functioning CPU or not - but knowing that would be very helpful in knowing which component(s) might have failed!

Working on the supposition the NOR Flash chip had somehow been corrupted I connected my long-unused TL866A EPROM programmer to the provided SPI header on the mobo via a mash-up of a PATA IDE 2.5" to 3.5" adapter (the only component in my collection that fits the mobo header) and fly leads to the TL866A.

Repeatedly failed to read the chip ID until finally realising the chip takes 1.8V not 3.3V! (Winbond W25Q128FW SPI NOR). Ordered a 1.8V interposer for the TL866A. Tried again with the same result but suspect reading using 3.3V might have killed an otherwise good chip!

Ordered a couple of replacement chips early December that finally arrived early January (no, not from China, from 40 miles away - thanks UK Postal workers on strike!) only to discover they are 3.3V chips - I'd misread the listing and ordered W25Q128FV not "FW" (V = 3.3V, W = 1.8V).

Finally took the plunge and desoldered the chip from the mobo whilst holding my breath in case the hot air gun blew the miniscule 1mm resistors close to the chip off the board! Tried reading the chip whilst disconnected but same failed result so assume device is trashed.

Proved I can program one of the 3.3V chips so - assuming the 1.8V interposer isn't faulty - just need the correct chips and resolder. Am now awaiting delivery of the correct chips.

At that point I'll likely discover the fault is either the motherboard or the CPU at which point it gets more expensive as have to obtain and swap in alternatives to deduce the failed part!

Don't you just love the right (and ability) to repair!? ironic laugh

On the positive side I discovered the completely open-source TL866A minipro software [0] and the TL866 firmware updater [1] tools that have made the software side of the process much easier than wrestling with MS Windows applications and WINE. Also UEFITool for inspecting and extracting binary ROM files from UEFI Capsules [2].

[0] https://gitlab.com/DavidGriffith/minipro

[1] https://github.com/radiomanV/TL866.git

[2] https://github.com/LongSoft/UEFITool (also available in Debian/Ubuntu archives)

If you a connected a 3.3v reader to a 1.8v chip while it was in-circuit, there is a good chance you have blown up everything else on the same 1.8v power supply on the motherboard...

There is also a small chance that the circuits survived. Some devices break at a higher voltage that the one specified in the datasheet.

Given the symptoms you most likely have a problem with PSU, not with BIOS.

Well, at first. Now you have more problems..

PSU and VRMs all check out. Went through the basics first.

How do you flash your BIOS on Apple hardware? Did they make it more convenient?

It is part of the system update packages. Users are not aware, that any firmware is being flashed; only that the system reboots multiple times.

Ok, how robust is it? Can you power-off the machine in the process?

Sorry, I have no idea about robustness.

If you can fix it then it isn't bricked.

If you can't fix the situation via conventional means, I'd say it's bricked.

An SPI programmer is unconventional?

I'd suspect the use of SPI would probably go beyond the capability of an average IT hardware support person, much less -- an end-user.

AFAIK that's the difference between a "soft" and a "hard" brick.

Reply on news.ycombinator.com