GPU acting erratically, possibly bad power delivery.

[Tadgh]

Recently my GPUs been acting up pretty badly - starting with blue screening whenever I alt-tabbed out of a game with the error: ERROR_MEMORY_MANAGEMENT, although it was an issue with alt tabbing out I was really optimistic that this might have been a RAM issue - as the error was fairly ambiguous, and RAM is much cheaper to replace than a GPU. So I swapped my RAM out for some modules I'm fairly sure definitely work - the BSODs didn't stop - so I took down the slight OC on my 280s boost - and the BSODs disappeared. Yay! The problems kept coming and keep coming and I'm really gonna need a hand from some folk smarter than I. I think it's a power delivery issue:

 

1. Games AND regular programs are inexplicably crashing to desktop, examples of this:

 

         - Shogun 2 crashed immediately in sight of buildings, but only in battles. So long as there are no buildings, everything runs fine.

 

         - Rome 2 crashes, because ???, I was in a multiplayer campaign, it crashed, BSODed on starting it up again, and hasn't launched since

 

         - Foobar 2000 crashes scrolling through playlists. It just hangs, and without an error message, stops responding, and needs to be ended in task manager

 

         - Waterfox crashes completely arbitrarily. It just.. It just, decides.

 

         - Windows explorer, rarley, but sometimes crashes - it's totally arbitrary, no indication or reproduction

 

          - The native Windows Photo app, has crashed a few times now.

 

           Funnily enough I can play some games just fine without any errors, typically older DX9 games  like RTW

 

2.  BSODs have started popping up with the effort CRITICAL_PROCESS_ERROR

 

3. The refresh rate on the second screen seems to have fallen, waving a window about on the second screen there's noticable lag. This stands up in screen-to-screen comparison.

 

 

 

Here's where I start to not know what I'm on about, please explain the specifics of power delivery if you can!

 

 

4. Opening up GPU-Z is what brought on my suspisions about the power delivery. My 280 IS drooping, with some strain on the PCI-e slot. Here's why:

 

          - VCDC Current is erratically jumping from 3.8-5.3A in irregular intervals. Again, IDK if that's a problem, but go for it.

 

          - VCDC Voltage is jumping, but a bit less erratically, from V0.0803 to V0.0805

 

          - Resource monitors go dead randomly, and after a while GPU-Z starts to report clock, and memory clock of 0MHz, oddly enough the thermometers keep working (possibly). The card is still on and connected, fans spinning, and screens on. When the card starts reporting that it's off, the crashes and lag get noticably worse. This could be placebo though.

 

 

-------------------------------------------------------------------------------------------------------------------------------

 

 

Here's what I've done to try and remedy this situation:

 

1. Replace my RAM

 

2. Reinstall my drivers

 

3. Replace Windows 10 (TP, Build 10074) to Windows 10 (!?! The released one)

 

4.Underclock the boost on my 280 from 1145MHz Core and 1550MHz Memory to 350MHz Core and 150MHz Memory with a -20% power limit and with +20% power limit, all 3 configurations have no effect.

 

 

None of these things have worked, and I'm kind of freaking out a little - today I'm gonna go home and change the PCIe socket the GPUs in, just in case it's an issue with a bent PCIe slot or dying PCIe bus (as unlikely as that might be, a man can dream right?). If anybody could confirm that this is an unresolvable hardware fault, or that this is a resolvable fault and how to do so - I'd really appreciate it! I don't wanna give up on my 280

[KarateHottie93]

I wish my GPU acted erotically too

[Godlygamer23]

What power supply are you using? I wouldn't necessarily be surprised if Windows 10 would be an issue as well. If the system crashes when the second card reports that it's off, Windows 10 may not be responding favorably. Did you remove your drivers using Display Driver Uninstaller and then reinstall them? Or did you just install over the previous installation?

[Tadgh]

I wish my GPU acted erotically too

You just gotta get nice and comfortable with it, is all

 

What power supply are you using? I wouldn't necessarily be surprised if Windows 10 would be an issue as well. If the system crashes when the second card reports that it's off, Windows 10 may not be responding favorably. Did you remove your drivers using Display Driver Uninstaller and then reinstall them? Or did you just install over the previous installation?

I've got an XFX 550w 80+Bronze, so efficient enough, never had problems with it before (but thanks for the suggestion I didn't think of that, might swap it out for another I have locally and see)

 

Just to be clear, it's a single card setup! But I'd believe that it's a Windows 10 issue I've been on the TP for almost a year now and haven't experienced this before yet though.

 

I replaced windows before doing a driver install, so I'm assuming that it's a fresh driver install - but then again, both screens were configured correctly at the right resolution from the install. When I go home I'll try using the Display Driver Uninstaller and installing the drivers again. Thanks for your help, this is a dark, dark day

[Godlygamer23]

You just gotta get nice and comfortable with it, is all

 

I've got an XFX 550w 80+Bronze, so efficient enough, never had problems with it before (but thanks for the suggestion I didn't think of that, might swap it out for another I have locally and see)

 

Just to be clear, it's a single card setup! But I'd believe that it's a Windows 10 issue I've been on the TP for almost a year now and haven't experienced this before yet though.

 

I replaced windows before doing a driver install, so I'm assuming that it's a fresh driver install - but then again, both screens were configured correctly at the right resolution from the install. When I go home I'll try using the Display Driver Uninstaller and installing the drivers again. Thanks for your help, this is a dark, dark day

Power supply efficiency and its ability to deliver power aren't related, FYI. Did you always have an R9 280 with W10?

[Tadgh]

Power supply efficiency and its ability to deliver power aren't related, FYI. Did you always have an R9 280 with W10?

I was thinking more along the lines of possibly over-volting the PSU, which sounds a little stupid on reflection, but I've done it before and goddamn it I'm paranoid about doing it again.

 

Yep, it's been working.. Poorly.. But working, the whole time.

[Tadgh]

This has gotten weirder yet:

 

I've deleted the GPU drivers that were present, and re-installed them, things are back to how they were pre-uninstall, laggy and glitchy. But I ran Valley Benchmark just to see, and it seems like I'm getting better performance than before.

 

The fact that GPU load sits at a constant 0% when navigating the OS/Web would indicate that it's indeed a problem with Windows. Can't help but be frustrated as hell right now, looks like it's either a rollback to 8.1 or a weekend troubleshooting again.. 

[Godlygamer23]

The fact that GPU load sits at a constant 0% when navigating the OS/Web would indicate that it's indeed a problem with Windows. Can't help but be frustrated as hell right now, looks like it's either a rollback to 8.1 or a weekend troubleshooting again.. 

And your video cable is plugged into the graphics card, yes? Is hardware acceleration activated in your browser?

[PlayStation 2]

Maybe faulty RAM on the GPU?

[Godlygamer23]

I was thinking more along the lines of possibly over-volting the PSU, which sounds a little stupid on reflection, but I've done it before and goddamn it I'm paranoid about doing it again.

 

Yep, it's been working.. Poorly.. But working, the whole time.

How did you overvolt it? Power supplies have rails that run at 12V, 5V, and 3.3V.

[Tadgh]

And your video cable is plugged into the graphics card, yes? Is hardware acceleration activated in your browser?

Yessum!

 

Maybe faulty RAM on the GPU?

I was thinking along those lines yeah, but really hoping otherwise. Having said that windows really shouldn't be taking up enough VRAM while navigating Explorer to make that an apparent issue

 

How did you overvolt it? Power supplies have rails that run at 12V, 5V, and 3.3V.

Well it was a 25 euro '600 watt' PSU, meeting no standards, made in china. So it's safe to assume that that was it's peak output, not rated continuous output. I proceeded to OC the GPU (confident in my '600 watt' cheap-as-chips PSUs ability to handle a slightly higher load). One summer night playing Shogun 2 with a mate it caught fire. Riperoni, a hard lesson to learn.

[Godlygamer23]

Well it was a 25 euro '600 watt' PSU, meeting no standards, made in china. So it's safe to assume that that was it's peak output, not rated continuous output. I proceeded to OC the GPU (confident in my '600 watt' cheap-as-chips PSUs ability to handle a slightly higher load). One summer night playing Shogun 2 with a mate it caught fire. Riperoni, a hard lesson to learn.

You probably didn't overvolt it. You were simply forcing more current through the +12V rail than it could handle. That being said, if it was a terrible unit, it likely wasn't delivering consistent voltage anyways.

[Tadgh]

You probably didn't overvolt it. You were simply forcing more current through the +12V rail than it could handle. That being said, if it was a terrible unit, it likely wasn't delivering consistent voltage anyways.

Ah, in my kinda shitty electrical knowledge I thought that's what overvolting a PSU referred to

 

It was 25 euro, so, I suppose it did all I could ask it to do.