Could My RX 480 Be Killing PSUs?

zombieno7

Honorable
Aug 3, 2012
33
0
10,530
I bought an RX 480 at launch. Since then, a series of strange issues surfaced, which don't seem to have anything to do with the RX 480. I should also probably mention that I'm running Linux.

So, I got an RX 480 at launch. Not long after, the power supply that I was running, an old Ultra X3 1600w, started giving incorrect voltages, running hot, and eventually smelling like burning electronics. I immediately replaced it with a Corsar AX 860i

The day after I installed the AX 860i, the computer started restarting randomly. After about of month of running memtest and messing with drivers, I figured out that the PSU was bad, and returned it for another AX860i.

For about three weeks, the new PSU worked great. Last night, the computer restarted again. Since the computer is configured to boot into the command line, it did. When I tried to start the GUI(XServer), it crashed immediately and restarted. It then looped several times, crashing when the kernel loaded the AMD firmware during the boot process. I powered down the machine and removed the storage hard drives to check if they were damaged. When I was done with that, I went back to test if it would boot again, and it ran fine.

Since it booted fine, I began to run benchmarks to stress test it. I ran both Prime95 and Unigine Heaven both separately and at the same time for several hours to stress it. Nothing happened.

I have no idea what's causing this. Could I really have just had 3 bad power supplied? Is it possible that the RX 480 is killing them? Any input would be greatly appreciated.
 
Solution
Then I'm not sure. You have enough power, the fix is implemented. So unless it fried your board before the fix was posted it shouldn't be a power problem. You can try taking voltmeter readings but I'm not sure anything would turn up. Any chance these are used PSUs? Beat up before you got them?
I don't think it ever was an issue. I've been using the open source Linux drivers and have been updating them regularly. Even still, it has never happened under a heavy load.
 
? It was an issue and if the Linux drivers have never got the fix the motherboard you have is probably toast/going. You failed to give us system specs so we have no idea if your board can handle the higher load the 480 is putting on it. Assuming you bought new, the 860i has more then enough power for a 480.
 
Specs: AMD FX 8350 OC'd @ 4.6GHz, ASUS Crosshair V Formula Z(Switched from ASRock Fatality 970 because it was hot), XFX RX 480 8GB Stock, Corsair AX860i, 16GB(2x8GB) Kingston HyperX Savage 1866MHz

I'm running the open source Linux drivers on Gentoo with kernel 4.7.2. I've been updating the drivers regularly since I installed the card. I hope that gives a clearer picture of what's going on.
 
That's just the thing. It isn't being logged properly. I eventually figured out that the first AX860i was bad because the generic "crash" messages being logged started exactly a day after I installed the PSU. This last time wasn't logged at all, and the logs still showed the machine running from the time before the crash.

I also read that the AX860i has a problem with fan regulation. It does so by wattage being drawn instead of temperature. It seems like the crashes tend to happen at night, when the machine is just acting as a server. Could the PSU just be overheating because the fans aren't coming on?
 
Not sure, case temps would help. The 860i is a very good PSU and it shouldn't be making a lot of heat. More so if your CPU and GPU are in (mostly) idle mode. Temps shouldn't get high enough to cause the 860i to have issues. Unless you have your case in a box or other such problem.

If it's the 860i, isn't there a way to use the digital interface to check temps, etc? I don't remember if there was a way to change the fan speed/curve on it. I would look into accessing the digital part of that PSU and see if you can change the fan speeds.
 
Corsair Link has no Linux support, so I don't have any way to check anything from the PSU in the OS. The case is a Fractal Arc XL. I have it sitting on my desk in an air conditioned room. The CPU is being cooled by Phanteks PH-TC14PE and tops out at 57C on Prime95. The GPU is stock with a stock cooler, but even running Unigine or Furmark, it only gets up to 85C.
 
Then as I said I don't see how the 860i can overheat while it's idle(ish). Yes the fan wouldn't be spinning because it's not putting out a lot of wattage, but it was designed to handle heat so at low loads it wouldn't be getting that hot anyways. It's possible the board got fried/is getting fried and it's not designed to handle the voltage the 480 is calling for. Just because it's a 990FX chipset with good coolers on the board doesn't mean it's not failing in some way. I would look into seeing if the Linux drivers fixed that problem or not like Invalid suggested.
 
I'll look into whether or not the Linux drivers ever got fixed. Honestly, the ASrock board was okay until the RX 480.

BTW: I've been running Linux for over 7 years. I know how to use it. I run Gentoo both on my home systems and servers. If something isn't supported, it just isn't. The RX 480 has had poor support since day one. I've had to compile the drivers from source and use bleeding edge kernels in order to make it work.
 
In Kernel 4.7? Or drivers?

https://linuxconfig.org/getting-the-rx-480-running-with-amdgpu-on-linux

Update : As of now, Kernel 4.7-rc7 supports AMDGPU as well. You can use the source from Kernel.org, and use that instead.

Did a fast search online and 4.7 supports the 480, but that doesn't mean it fixed the "pull too many Amps from the PCIe slot" problem. That was fixed with a driver in windows and I'm not sure if that happened on Linux. This was the second search result, but I don't know if it was ever implemented.

http://www.phoronix.com/scan.php?page=news_item&px=AMD-RX-480-Power-Issue
 
The fix was implemented a few days after in the kernel. The last comment on that article actually talks about it. The AMD Linux drivers work much differently than Windows. There are both closed source firmware and open source kernel modules in the Linux kernel. Then, there are other open source drivers for the display server that interact with the kernel. The power fix was pushed into the kernel. The Git commit is here: https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/gpu/drm/amd/amdgpu?id=eeade25ad029cb1f31f27f8e0ddc9bb9c22b5537
 
Then I'm not sure. You have enough power, the fix is implemented. So unless it fried your board before the fix was posted it shouldn't be a power problem. You can try taking voltmeter readings but I'm not sure anything would turn up. Any chance these are used PSUs? Beat up before you got them?
 
Solution
It can't be that the board was fried before. I replaced the ASRock board I had after the fix was implemented, thinking that it may have been the problem. These PSUs aren't used, but they are refurbished. I've been reading around, and it looks like the AX760i and the AX860i are plagued with similar issues. I think I may just keep getting garbage units.
 


No.

90W load wouldn't kill your 850W PSU.

The issue was centered around the motherboard having potential issues supplying enough power through the PCIe slot and the motherboard frying as a result. Only a few claimed instances of this issue ever materialized. It was mostly all speculation based.

Your RX 480 isn't frying your PSU.
 
Okay. I just wanted to make sure, seeing as I've gone through 3 PSUs in less than three months. It looks like I've just been getting poorly refurbished garbage PSUs. I'm going to go return this last one(within the 30 days) and get an EVGA. I hope that works.