GPU Creating Horrible Smell Then PC Reboots

MAXimized

Prominent
May 30, 2017
3
0
510
Hello, everyone!

I've been noticing a bad smell coming from my PC recently and I couldn't seem to find the source. I thought it was probably dust build up in my heat sinks so I sprayed all of them with compressed air, but it didn't help.

Today, while playing X-Plane 11, my computer abruptly black screened. A few seconds later, I got the Windows startup screen. Upon closer inspection, I found that the smell was coming from my graphics card.

I tore the thing apart looking for melting plastic, dust buildup, bursting capacitors, but found nothing. I decided to do a little CPR and ran the computer without the graphics card and everything went fine.

Things to Note:
*According to Afterburner, Open Hardware Monitor, and FurMark, the temperature are never reaching dangerous levels. The card idles at around 40 to 50 degrees Celsius and under load never goes above 80 degrees.
*After just 5 to 8 seconds of running the 1080p benchmark on FurMark, the smell began coming from the card.
*At idle, it does not create the smell.

Pc Specs:
CPU: AMD FX 6350 @ 3.9
GPU: MSI RX470 w/ 8GB (GPU: 1245 MHz - MEM: 1600 MHz)
RAM: 16GB @ 1800 MHz
Motherboard: ASUS M5A78L-M/USB3
PSU: EVGA 500W 80+

Does anyone know what could be causing the smell or how to fix this issue?
Thank you so much in advance!!

-Max
 
Solution
Wait. So you went from molex to a 6pin, then a 6pin to 8pin then an 8pin to 2x 6pin? Did I read that right?

No wonder it burned up. I'd be surprised if there's not further damage. A molex connector tops out at @ 10-11A 12v. That's 120-130ish watts for that entire wire, if there's 2x molex on that wire, it's the wire that's the standard, so still 130w for both, not each.
A 6pin is rated at a combined wire total of just 75w. The 8pin connector is a combined 150w total. 2x 6pin would be 150w.

So what you did was ask a gpu that can pull as much as 145w to draw power from a 130w source, through a 75w connector before being split to a 150w demand.

In simple terms, since the psu is obviously substandard since it does not have the 2x 6pin...

MAXimized

Prominent
May 30, 2017
3
0
510


Yeah, it smelled like something was burning.
 

King_V

Illustrious
Ambassador
I'd say shut everything down, and see if you can track down the root of the smell. Are you absolutely, 100% certain it came from the GPU, and not, say, the power supply?

If it is from the GPU, I'd suggest getting an RMA and have it replaced.
 

Karadjgne

Titan
Ambassador
Gpu is toast. There's 2 important areas on a gpu, the processor and VRM's/VRam. Out of those 2,only the processor has a temp reading, so even if the gpu reports max temps of 80° with normal idle of 45ish, it's entirely possible/probable that the VRM's/VRam has reached well over 90° even as much as 100°C and is burning up, causing the blackscreens. Generally, the gpu should be within a few degrees of the cpu at idle, so with 40-50° idle temps, something is wrong from the get-go, more than likely hot running VRM's/VRam heating up the heatsink and you are seeing hotter temps at the processor as a result.
 

MAXimized

Prominent
May 30, 2017
3
0
510
Aha! I have found the problem!

It appears that the converter cable that I was using to power my graphics card was actually burning itself until it broke. I was using a 8 pin to 2x 6 pin connector hooked up to a 6 pin to 3 pin molex. The 6 pin connector (from the molex to 6 pin) was literally melting itself and the 6 pin connector on the other converter. I'm horrible at explaining all of this, so here are some images:

.eJwVzNsNwyAMAMBdGACDeSRkG0QQoQo1wu5X1d3b3gD3Vq91q0NdIpMPgLNzoXVqFlq5Vd2I2l3z7KwLDcgiuVyjPoUBdxu24IyzNqToPUZwaH4wJUzB-C3uBvr4N4_Z1OcLvS8hvQ.BvoKu45gdTNPUg_C50Rv5i152mg


.eJwVzNERwiAMANBdGIAAgajdhqMcxSuGI_Gr5-7qG-Bd5r1Os5lDdcoGsHcpvHYryiu3ahtzO2ueXWzhAVk1l2PUlwqEu0-3hA69Tw-KMRBgcD-IRBSQoicHffyb52zm8wW8GyGu._U40mL2taPjikF1-kfHyfRbdio8


.eJwVzNENgyAQANBdGICDO0FxG4IEMdIjcP0y3b3tG-A96j1utatTpM8d4Kgz8Tj0FB6xZF2Yy51jr1MnbhBFYjpbfskE3KxbHRmy1gW_LOiB0PyQx804whDQQG3_5upFfb68PyGx.XjqlfaGs0wa-guyEiy0jnnlq3zY
 

Karadjgne

Titan
Ambassador
Wait. So you went from molex to a 6pin, then a 6pin to 8pin then an 8pin to 2x 6pin? Did I read that right?

No wonder it burned up. I'd be surprised if there's not further damage. A molex connector tops out at @ 10-11A 12v. That's 120-130ish watts for that entire wire, if there's 2x molex on that wire, it's the wire that's the standard, so still 130w for both, not each.
A 6pin is rated at a combined wire total of just 75w. The 8pin connector is a combined 150w total. 2x 6pin would be 150w.

So what you did was ask a gpu that can pull as much as 145w to draw power from a 130w source, through a 75w connector before being split to a 150w demand.

In simple terms, since the psu is obviously substandard since it does not have the 2x 6pin or 6+2pin necessary to power that gpu, it's damned lucky you didn't burn down the entire pc. Go purchase the correct psu for the job, there's actually no reason why anyone should ever use molex-pcie adapters. Ever. Count yourself lucky, for now.

Honestly hope the gpu still works after this though, if it wanted 140w at 12v it's going to try and get that wattage however it can, if that means amperage goes up, oh well. No wonder the card was running hot, it was using way too much amperage to make up the wattage difference, which further heated up that 6pin connection. The board really isn't designed to handle that kind of current, so it's no wonder the VRM's finally shut you down when gaming. It's a fair bet if the processor was seeing 80°C, the VRM's were topping 90°or higher.

Even after getting the correct size psu, if you still have blacks creeps happen, it's the gpu is toast and can no longer support the wattage demand. You killed it with using that molex disaster of supply the first time around.

The psu is the single, unarguably most important component in any pc. It's the one component that powers everything and is the sole component responsible for the health of everything else. What you did would be like using the heart of a 5yr old in a Olympic athletes body. Something gotta give somewhere.
 
Solution

USAFRet

Titan
Moderator


As often happens with hacked together cables from a PSU that lacks the proper connections to run something.

The PSU lacked the proper connections and power to run your GPU.
Hacking some cables together does not change that fact.

Time for a real PSU at the very least.
 

logainofhades

Titan
Moderator