FreeBSD killed my GPU

I have installed FreeBSD on a spare disk. Everything worked fine until I tried to get X.org running with my AMD 6600 RX. I eventually gave up when I couldn't get pass the black screen when loading the amdgpu module.

I rebooted the computer and tried booting my regular OS. At first I couldn't even get pass Grub. The image just turned black. I forced reboot the computer by holding the power button for a few seconds. I eventually got the OS to boot, only to be greeted with a black screen before it was done loading.

I then tried to boot my second OS, which booted fine after a few tries. It didn't take more than a few minutes of usage before the screen turned black.

And right now I'm not able to get any picture at all, not even for BIOS...

So... It looks like FreeBSD killed my GPU... Great! I'm really looking forward to spend another few hundred euros on a new GPU that I barely even use to begin with! 🤬
People who posted here earlier have pointed out how to deal with the hardware - so I won't rehash THAT.

I can, however, help with the software side of things after hardware issues are eliminated.

So I will start by asking: Did you read the Handbook on how to install the correct driver for your RX 6600? If you skipped steps, then of course the driver won't load and you'll get a blank screen.

If you look for Codename of your GPU on TechPowerup.com, you'll see that it's Dimgrey Cavefish. And that is definitely present in https://cgit.freebsd.org/ports/tree/graphics/gpu-firmware-amd-kmod/pkg-plist.

Because of that, all you need to do is follow directions in the FreeBSD User Handbook (Section 5.3, graphics cards), don't skip steps or do them out of order.

Just install graphics/drm-kmod, then add the amdgpu driver to /etc/rc.conf, and then reboot.

This can be done even if you have to SSH in from the side.

Please, tell me what you ate. 😂
I ate a bowl of thick split pea soup with some garlic, onion, and Anaheim pepper! ;)
 
People are going to have questions. They're going to bring scenarios and often include information that may not be germane. It happens. No one here was born an expert in anything. We ALL started somewhere, we all learned, and we all received help from someone else.

If you can't be helpful, be quiet. Some things just do not need to be said.
I 100% agree with you. No, I really do. You are absolutely right.
The point is:
In this case the OP did not asked a question.
He simply blamed FreeBSD to be responsible for killing his GPU.
Everything then was trying to help him,
by transforming his initial post in some kind of a question which can be answered - to help him,
which includes the primary requirement to explain to the OP, the cause needs to be searched elsewhere,
which - I agree again - in some cases by the one or the other may haven been put more gentle, more helpful, or one may say less rude. Okay.
But I also understand 'our side':
Some new one just rumbles into the forums and simply unloads his frustration, without even asking for any help.

Being kind, and helpful needs more words than keeping it short, which also can be, and often is misunderstood as being rude - especially when like in this case some were pissed off in the first place, inlcuding the OP, and the understanding of the OP being pissed.

Plus we're talking about how to talk - which also does not shorten this thread, but hopefully increases the way we talk to each other, which - I like to underline - by all my experiences I had in many other forums, here is at a exemplary high level of understanding, patience, professionalism, and respect, even when disagreeing - at least most of the times by the majority.

But back to topic:
I hope the OP tested his GPU carefully again (as several posts here suggested how to.)
I hope his GPU did not die, and I'm pretty sure everybody else here feels the same, including eternal_noob.
And anyway furthermore I hope this will not make the OP to turn his back on FreeBSD, but see this wasn't FreeBSD's fault, the community is really helpful, and useful, actually, and he'd miss a valueable experience if he decides against it.

We will see if he comes back, or not.
 
When checking the graphics card also check the motherboard connector, pins & local chips on the motherboard. I have come up against motherboards going faults & damaging gpu cards by voltage regulator faults that most people find by installing a new card & killing that as well. The best test for that possible faulty card is trying it in another system if it is possible & if the motherboard has an igpu output select that in the bios & try booting any usb or iso image with the internal drives disconnected.
 
Not long ago, OpenZFS with a mirror for two SSD drives burned my motherboard. How it happened: the board was from 2013, it worked, worked and died under Firefox, when I was peacefully watching something on YouTube@FullHD.
I did not bother with additional and high-quality heat dissipation. I saw how
my modest PC (with a built-in video card) turned on the fan at maximum speed, breathing heat, like an album of the "Morbid Angel" group. I ignored this too.
Who fried the chip? Me or OpenZFS?
Answer: me.
 
I don't want to believe such a product exists in the market, but I can think of a quite unbelieavable case that usually-sane driver can burn a graphics card.
  • Not using (and considering) reference design by GPU manufacturer by graphics cart manufacturer,
  • Fan speed of the graphics card is controllable and defaulted to 0rpm (off),
  • The firmware on the graphics card doesn't handle the fan speed and rely on the specially crafted genuine (by graphics card vendor) driver, only provided for supported versions of Windoze,
  • Attempting to run the GPU on the graphics card with extremely high load.
In this quite specific and unlikely case, the generic driver (including of FreeBSD, but not limited with FreeBSD) does not know how to activate cooling fan of the graphics card, and extremely high load should make GPU on the graphics card to be overheated, thus, kills the card, even with generic (provided via Microsoft) driver on Windoze for the GPU (not mean the specific cart, but cards with reference design).

Of course, such a silly graphics card can easily cause fire hazard, thus, should not be allowed to ship to anywhere.

Moreover, if the graphics card doesn't show anything at the start (not at all working), how FreeBSD driver make it running on heavy load? Shoud be nonsense. Just a silly technical possibility.
 
I can think of a quite unbelieavable case that usually-sane driver can burn a graphics card.
ATi, now AMD, has had famously bad drivers. I wouldn't be surprised, either, if the card just broke down, and it would have done that with FreeBSD's help or not. It's a thing that happens.
 
Software damaging graphics cards can absolutely happen, sometimes you don't even need a bad driver.

When Amazon's new MMORPG started ("New World" or something) it damaged quite a few GPUs in a 2D startup screen. Was just sitting there with like 600 fps and some NVidia cards didn't like it.
 
Software damaging graphics cards can absolutely happen, sometimes you don't even need a bad driver.

When Amazon's new MMORPG started ("New World" or something) it damaged quite a few GPUs in a 2D startup screen. Was just sitting there with like 600 fps and some NVidia cards didn't like it.
Not to mention that many years ago you could just fry your CRT monitor using wrong refresh rates in XF86Config .
 
When Amazon's new MMORPG started ("New World" or something) it damaged quite a few GPUs in a 2D startup screen. Was just sitting there with like 600 fps and some NVidia cards didn't like it.
So what I take from this is that running FreeBSD is *safer* because there is less 3D software available for it than other popular platforms. ;)
 
When Amazon's new MMORPG started ("New World" or something) it damaged quite a few GPUs in a 2D startup screen. Was just sitting there with like 600 fps and some NVidia cards didn't like it.
Turned out to be bad quality cards.

Instead, it reiterates and confirms that it's a problem with GPU manufacturers and poor graphics card build quality.
EVGA has already confirmed that its first batch of RTX 3090 cards built a year ago had faulty solder joints, which caused the cards to die under the strain of New World — and if it wasn't New World, it would have been some other game or application.

And with any product, even good high quality ones, a small percentage will die within one or two years. That's why there's such a thing as warranty.
 
I don't want to believe such a product exists in the market, but I can think of a quite unbelieavable case that usually-sane driver can burn a graphics card.
  • Not using (and considering) reference design by GPU manufacturer by graphics cart manufacturer,
  • Fan speed of the graphics card is controllable and defaulted to 0rpm (off),
  • The firmware on the graphics card doesn't handle the fan speed and rely on the specially crafted genuine (by graphics card vendor) driver, only provided for supported versions of Windoze,
  • Attempting to run the GPU on the graphics card with extremely high load.
In this quite specific and unlikely case, the generic driver (including of FreeBSD, but not limited with FreeBSD) does not know how to activate cooling fan of the graphics card, and extremely high load should make GPU on the graphics card to be overheated, thus, kills the card, even with generic (provided via Microsoft) driver on Windoze for the GPU (not mean the specific cart, but cards with reference design).

Of course, such a silly graphics card can easily cause fire hazard, thus, should not be allowed to ship to anywhere (except Naz* Rus*?).

Moreover, if the graphics card doesn't show anything at the start (not at all working), how FreeBSD driver make it running on heavy load? Shoud be nonsense. Just a silly technical possibility.
The latest NVidia GPU (RTX 5090) did get so hot it melted: https://www.theverge.com/news/609207/nvidia-rtx-5090-power-connector-melting-burning-issues

Even the 4090 would melt if pushed by a good benchmark.

But be careful about putting another country on blast, T-Aoki ... some Forums members are from there.
 
I'm not sure if FreeBSD does anything with AMDGPU sys stuff for voltage control, but generally speaking I don't believe an OS-side driver to be able to cause hardware damage unless that driver is touching power or voltage stuff.

I would hard-reset CMOS (battery pull, jumper, NVRAM erase key combo on laptops). I vaguely recall having something with VBIOS act unexpectedly with an AMD GPU until I reset the mobo CMOS. I'm not sure if it was with a RX 580 or 6600 XT, but I had both those GPUs with Linux and Windows. I think it was switching between Legacy and UEFI boot and something like a VBIOS (GOP vs Legacy) not being loaded by BIOS because of the previous boot state (maybe min/quick boot and not re-initializing?)

As for New World, as I understand that was basically bad hardware that couldn't or wasn't tested to handle a unique load (if it wasn't NW then anything 3D 1000+ FPS likely would have done it, which "nobody" realistically does). StarCraft II before HotS had a patch to fix something similar to limit main menu FPS :p

Although, what DE was used? What about running Xfce or not GNOME/Plasma without compositor; would that theoretically cause an unlimited FPS scenario? I'm guessing anything 3D would respect Mesa's default vblank, but what about 2D?
 
This is my field of expertise though I don't think of it much anymore. (My degree is in EE)

Any hardware that lets itself catch fire via software deserves to burn in hell.

Way back when, there was supposed to be an assembly instruction HCF which stood for Halt and Catch Fire. I don't recall if it was a joke or misinterpretation but you can Google for it.
 
I'm not sure if FreeBSD does anything with AMDGPU sys stuff for voltage control, but generally speaking I don't believe an OS-side driver to be able to cause hardware damage unless that driver is touching power or voltage stuff.

I would hard-reset CMOS (battery pull, jumper, NVRAM erase key combo on laptops). I vaguely recall having something with VBIOS act unexpectedly with an AMD GPU until I reset the mobo CMOS. I'm not sure if it was with a RX 580 or 6600 XT, but I had both those GPUs with Linux and Windows. I think it was switching between Legacy and UEFI boot and something like a VBIOS (GOP vs Legacy) not being loaded by BIOS because of the previous boot state (maybe min/quick boot and not re-initializing?)

As for New World, as I understand that was basically bad hardware that couldn't or wasn't tested to handle a unique load (if it wasn't NW then anything 3D 1000+ FPS likely would have done it, which "nobody" realistically does). StarCraft II before HotS had a patch to fix something similar to limit main menu FPS :p

Although, what DE was used? What about running Xfce or not GNOME/Plasma without compositor; would that theoretically cause an unlimited FPS scenario? I'm guessing anything 3D would respect Mesa's default vblank, but what about 2D?
There are crappy board partners for GPUs. Asus and Gigabyte are somewhat pricier than others, but they make good boards for AMD GPUs. I'd stay away from the cheapest ones. But Gigabyte/Aorus GPUs that have the Radeon RX 6600 are pretty decent quality stuff. You generally get what you pay for.
 
This is the only instance I can recall of an operating system destroying hardware. I used to use this OS but luckily did not have an LG drive.

That stuff is bullshit, it was probably a dud unit from LG to begin with. I used Madrake Linux back then, and yes, my CD-ROMs were from LG back then. The OS ran fine, hardware was not destroyed on my watch. I got tired of the OS design and features, true, but if hardware fries, that's generally because one either has a dud unit or a bad PSU.

You can always try to intentionally run something that would melt the hardware (like the Furmark GPU benchmark), but otherwise, it's just impossible for an OS to fry the hardware, even if the hardware is not supported by the OS.
 
Of course software can destroy hardware. You've heard of Stuxnet? That was done by intention.
Stuxnet does have a very interesting story behind it. Do some research and try to understand how things are normally supposed to work if you want to be taken seriously.
 
Software damaging graphics cards can absolutely happen, sometimes you don't even need a bad driver.

When Amazon's new MMORPG started ("New World" or something) it damaged quite a few GPUs in a 2D startup screen. Was just sitting there with like 600 fps and some NVidia cards didn't like it.
I exactly wanted to write that, but I forgot the game name 😅

Besides...
You can in windows force in the GPU driver options a specific frame refresh rate so the GPU does not cross that line and render like 600+ FPS, and suddenly dies.
But, I heard about that story, although hard to believe without experiencing it myself.
Something like the RTX 40 series suffer from bad cables where the main problem was the user behind it not plugging the power supply cable properly in the GPU.
Gamers Nexus in youtube stated something similar among that lines.
 
Back
Top