Question A clichéd "Is my GPU going bad?" post

[King_V]

Ok, first, the full system specs:

Dell XPS 8700:

• Windows 10 Professional
• i5-4460
• Nvidia Founders Edition GTX 1080
• 4x4GB Patriot Viper 3 DDR3 1600 CL9
• 500GB Samsung 860 EVO SATA 2.5" SSD
• 500GB Crucial MX500 SATA 2.5" SSD
• 1TB HDD
• Seasonic Focus Plus SSR-650PX Fully Modular 650W 80+ Platinum PSU
• Acer XR382CQK 38" 3840x1600 IPS, 48-75 FreeSync

Also, I have GSync/FreeSync enabled in the drivers. My monitor confirms this, because there's a setting I've enabled where the LEDs at the bottom go from their selected color, to red, when FreeSync is active. Refresh rate in Windows is set to 60.

Now, I'm generally not on the latest and greatest games. I've been playing Pinball Arcade, and Don't Starve mostly. The latter, I've actually been playing on my Athlon 200GE based Micro Machine for the most part. The Dell I've specced out above, I do a lot of web browsing, watching videos, and keeping files (spreadsheets, banking and budgeting stuff, etc, as I still use Excel from Office 2013)

Two weeks ago, I fired up the Dell to play Pinball Arcade (the Micro Machine is Linux, and Pinball Arcade doesn't seem to work through Steam Proton). No problems, but I play that in windowed mode.

Then I played Don't Starve. It got weird. Every once in a while, the screen would go completely blank for about 2 seconds, and then come back. Everything else kept running, sound, action in the game, etc, just the screen would go blank, and come back. There was no regularity to it - sometimes it was as little as 2 minutes between blackouts, and sometimes as much as 15 minutes, and various durations in between.

This past weekend, I tried again. No problems with web-browsing, or playing Pinball Arcade windowed. Played some Don't Starve, and same issues as two weeks earlier. So, I downloaded GPU-Z for the speed/temp/fan/gpu graphics, and tried again. Same symptoms

Wondering if maybe there was some bug in a recent update for the game, I decided to try another non-demanding game: Portal. Settings were high, but not max, and I played.

The symptoms were the same - the screen would go blank at irregular intervals, but only for about 2 seconds. No freezing, no crashing, everything else going as normal. Also, there was no loss of signal, as had there been, the monitor would've tried cycling through the other inputs, which didn't happen.

After about an hour or so of playing Portal, things got a bit odder. I noticed that the game was jittering/shuddering at points, and Steam's frame counter was showing it fluctuating at various values between 44 and 48 (EDIT: actually, in hindsight, I'm fairly certain between only 47-48, without dipping lower than 47). There's no way the GPU should be struggling.

When I'd check the GPU-Z graphs, there were no spikes in utilization, temperature, fan speeds, or anything, so there was no thermal throttling going on. I checked the Windows logs in the Event Viewer, and there were no issues logged at the times that the blanking occurred, or even close to those times.

I further tried the following:

• Updated to the latest drivers (I was maybe 2 or 3 versions behind originally)
• Switching to a different DP cable
• Switching to different DP outputs on the same card

Nothing made a difference.

A friend at work suggested re-checking the PCIe connection (I'll give that a go tonight), as well as maybe replacing the PCIe modular cable with another one, though I'm dubious on those as I assume the system would probably crash if a brief loss of PCIe cable power was at issue.

It only happens in full screen, never when I'm doing anything on the Windows desktop. I haven't tried Portal or Don't Starve windowed, nor have I tried Pinball Arcade at full screen, though.

Thoughts or suggestions? Are there any steps I've missed? Any further diagnostics I can do? Or is the card actually going?

 

[logainofhades]

Have you looked at your CPU and GPU temps any?

 

[JaAntonio]

It seems the problems occurs when you get the gpu into load, look at temps for the first step. The power supply may be damaged or even your graphics card. You could also try different power supply, and conect the graphics card to another PC. You know what power supply you have?.

 

[King_V]

I didn't look at the CPU temps, but the GPU never even got up to 70 according to GPU-Z. I'm dubious as to the CPU temperatures being the issue given how little demand the games put on the system, and that I've played these (and more intensive) games in the past on the same system without issue, though I can give it a look next time I try (hopefully tonight).

PSU is Seasonic Focus Plus SSR-650PX (650W, Platinum, fully modular)

 

[JaAntonio]

Try to get other system to try the graphics card. One from your friends or family. If this still happening, your gpu is broken. If not, the most reasonable part to fail was that motherboard. That power supply is very good quality.

 

[King_V]

Update, been playing for 5 minutes, full screen, native resolution, Portal. I had GPU-Z tracking PEAK numbers, which, at the time of blanking:

Stat	Peak Value (not necessarily at time of blanking)
GPU Clock	1607.0MHz
Memory Clock	1251.5MHz
GPU Temp	31.0 C
Fan Speed	1108 RPM
GPU Load	87%
Memory Controller Load	67%
Video Engine Load	0%
Power Consumption (W)	53.4W
Power Consumption (%)	29.7% TDP
CPU Temperature	38.0 C

Doesn't seem like it was struggling much.


EDIT: Also tried Portal in Windowed mode, running 2560x1080 - and the blanking also occurs then, with the entire screen blanking out, not just the window the game was in. Again, only about 2 seconds.

 

[King_V]

Further updates:

• I tried switching which of the two end-plugs on the PCIe cable were plugged into the video card
• I tried switching which PCIe socket on the PSU that the PCIe cable was plugged into

Also:

• Played Portal full screen, and the problems continued
  • I also had a period well before an hour where the screen would get jittery, and the FPS would fluctuate between 47 and 48. After a few minutes, however, it went back to a rock-steady 60.
• Played Portal windowed - the problem exhibited itself. The entire screen would go blank, and sometimes, in windowed mode, the blanking would occur, the screen would come back for one second or so, then another blank would occur.
• Played Pinball Arcade full screen - the problem exhibited itself
• Played Pinball Arcade windowed - the problem exihibited itself. This had never happened on Pinball Arcade windowed before.

The problem is also happening with very noticeably increased frequency since the weekend. Peak values of things (temps, etc.) were about the same, but the GPU temp after extended playing had peaked up to about 50 degrees at one point.


My last attempt at diagnosis is to install the card on a different PC, which, fortunately, has a fresh install of Windows 10 Pro on it. It's running only 8GB RAM, and an i5-2320 CPU. I'm almost HOPING it has the same problems on that system, because if it doesn't, I'll be at a loss to know what to do.

 

[King_V]

Following up. My main computer that has the GTX 1080, and the Seasonic PSU, is a Dell. The older machine I'm going to try it out with (we'll call Old Dell) has the factory Dell PSU, and was purchased new in 2012, and is a Sandy Bridge i5. For the past few years, it's been running routinely with a 150W card in it (GTX 660Ti).

I tried it - but the PSU only had dual 6-pin PCIe connectors, and the GPU requires an 8-pin. So, I ran the PCIe cable from the main machine's power supply to the video card when it was in the Old Dell.

Same problems occurred. Clearly, it CANNOT be the main machine's motherboard or PCIe slot that's at issue.

I finally got a dual-6-pin-to-single-8-pin adapter, and it arrived yesterday. Thus, I finally got to try out running entirely on the Old Dell, including Old Dell's PSU. This thing is a 460W unit that specifically states "max 385W" for the 12V rails.

Everything was fine. Gamed for over an hour with no issues.

The only culprit left: The Seasonic power supply.

I'm starting to wonder if I missed something, because of how wildly improbable this seems. Focus Platinum, 650W, 10 year warranty, and yet, somehow, at just past the two year mark, it's having issues? Possible, sure... but the whole "ancient Dell PSU handling things better than the modern Seasonic is bizarre.


EDIT (Mar 19, 2020): well, I'd concluded it was because everything behaved fine when I tried another PSU for several hours. So I got a replacement Seasonic under warranty. Same issues. Then I went back to the non-Seasonic that behaved for several hours, and now it's doing the same thing! Insert image of me gnashing my teeth and smacking my forehead into my desk here...