Weird GPU performance issue

[SHayden]

By trying different PSU's in your PC.
Try using different of the same power and of higher, in my case the PSU I had was too weak for the new GPU(I upgraded from some old radeon rx something 1 gb to gtx 1050 Ti gaming 4 gb and power consumption was considerably higher)
But even if you didn't change anything the cables may also wear off (from what I've heard) especially if the PSU isn't of the best quality and if it's strained for extended periods of time.

edit: or alternatively trying your own PSU in other configurations with same or lower power consumption
Also don't take my advice for 100% when I said check if it's not hardware issue I meant generally hardware issue, I just mentioned my case of PSU because often people would check everything but would completely forget about PSU and in my case I literally turned my whole PC over both software and hardware and didn't find solution until I figured PSU was the issue

[Marisa the Magician]

Oh yeah, that could be the case. The GPU might be getting throttled due to not being able to draw enough power.

[Graf Zahl]

There's just one (small) problem here: I do not have a second computer to check against (everybody I know uses laptops) so I'd have to buy a new PSU just to try.

So, @SHayden, what problems did you experience? Just some throttled performance or also some other effects like system instability?

[SHayden]

There's just one (small) problem here: I do not have a second computer to check against (everybody I know uses laptops) so I'd have to buy a new PSU just to try.

So, @SHayden, what problems did you experience? Just some throttled performance or also some other effects like system instability?

In my case my PC would freeze and/or crash randomly, often when I performed demanding task such as gaming sometimes even watching movies in full HD, like Marisa said, probably because of the lack of power. And somewhat similar, let's say I opened the game, it would work for first minute or two, then it would begin to stutter and finally just dead often with those funny graphical glitches and stuff...Like I said I checked (almost) everything you could think of and found out PSU was too weak.

[Darkcrafter]

It happened to me with used MSI Radeon HD 6770. I used MSI Afterburner to give it some oomph so I overclocked it a bit more than it could handle but kept the same non-overclocked fan speed. Install MSI Afterburner and try to decrease clocks just a bit, it could be a sign of throttling. If you still have a warranty it's time to take a visit to the store you bought it at.

In the end of it, take the card's cap off (untight and unscrew every little bolt on it first in otherwise you can damage the PCB) and see if a thermal solution melted across all the chip including little resistors. If it is take a Q-Tip and gently remove it. Even if it's not, replace the thermal solution, but wipe off every bit of it from the chip and radiator surface before you apply it again. Make sure it's important not to put too much of it so it doesn't melt across tiny resisors again. A thin layer of the thermal solution is enough.

Also try to find out if the radiator CLIPS to the chip uniformly, if it's not the reason could be that PCB got bend in this place. So in the first place install MSI Afterburner and monitor your GPU load and temperature. Some people also tend to test graphics cards with Furmark, but it's dangerous

[Rachael]

It happened to me with used MSI Radeon HD 6770. I used MSI Afterburner to give it some oomph so I overclocked it a bit more than it could handle but kept the same non-overclocked fan speed. Install MSI Afterburner and try to decrease clocks just a bit, it could be a sign of throttling. If you still have a warranty it's time to take a visit to the store you bought it at.

In the end of it, take the card's cap off (untight and unscrew every little bolt on it first in otherwise you can damage the PCB) and see if a thermal solution melted across all the chip including little resistors. If it is take a Q-Tip and gently remove it. Even if it's not, replace the thermal solution, but wipe off every bit of it from the chip and radiator surface before you apply it again. Make sure it's important not to put too much of it so it doesn't melt across tiny resisors again. A thin layer of the thermal solution is enough.

Also try to find out if the radiator CLIPS to the chip uniformly, if it's not the reason could be that PCB got bend in this place. So in the first place install MSI Afterburner and monitor your GPU load and temperature. Some people also tend to test graphics cards with Furmark, but it's dangerous

We established very early on that it's not a heat problem.

In fact, it seems to be a PSU power draw problem, and that's not going to be solved by replacing the thermal paste. This is bad advice to give for the wrong problem. While it is a good idea to replace the thermal paste occasionally, Graf *just* got this card sometime mid-to-late last year - it's going to be some time yet before he has to consider such things.

Graf does not break 30C with that card - that's extremely low for there to be a heat problem. I would not even consider it with that low of a temp.

[drfrag]

I don't think it's a hardware problem. If it were the PSU the system would be ustable and the game would not start running slower after a minute or so.

[Rachael]

Power draw is a tricky thing, drfrag - it's never explained by one single symptom. In fact, the only way to factually diagnose it is to do exactly what others have already suggested - running the system on another PSU.

[SHayden]

I don't think it's a hardware problem. If it were the PSU the system would be ustable and the game would not start running slower after a minute or so.

Well you can't know until you roll up sleeves and check it out first-hand
I was lucky because I have brother who is good with those computer thingies and knows people who are good at it as well so I didn't have to pay someone to do it, but even if I haven't I would probably find an expert.

[Darkcrafter]

Does it only happen with GZDoom? Did you try running other games? What's about DirectX?

[drfrag]

You seem to forget i've been/was? a computer technician for 20 years. I'm not sure about anything right now tough.
Next thing i'd do is watching cpu usage in task manager, something else could be eating the cpu in the background.

[Darkcrafter]

That may not be the root of the problem but still, I have encountered many systems with useless processes eating up a lot of CPU in the idle time slowing the computers down so much they were getting extremely hot for no obvious reason and stopping to work stable. In Windows 10 I completely disabled SysMain component in autoload section, on some other laptops I removed Intel's useless things that were really disabling systems to work at all. Delete such stuff like Intel's disk managment.

Upd: do you have an m2 SSD installed in your system? I heard on some motherboards it could cause conflicts with pci-e slot slowing down either an ssd or a graphics card.

[Apeirogon]

Double drfrag. Im not a computer hardware guy, but I suspect that computers works in the same way as radio-related machines.
If a power source really too weak to supply videocard it would just wont run computer at all. I have seen this few times, computer turn on, all lights blink and and fans chase the air, but monitor left black.
It CAN be run, in this case, using some electrical trick (plug computer in electrical socket, wait a little so coils in power supply...pfffff....rectifier, as as prompts google translate, store some energy in magnetic field and then turn off power supply by specific tumbler on it and in the RIGHT moment turn it back on (which can be noticed by watching on green light diode on motherboard, right when it flashes brightly after slowly fading away ITS TIME) you would run it, dont remember exact name of effect which cause this) But I wont recommend this, since it can damage motherboard/videocard/cpu/etc transistors.
To completely exclude power supply from possible reasons of this I advise to check that it supply around 2kW of power. That must be more than enough for desktop computer.

Also, maybe there are happen something before this performance drop? Cup of tea, thunderstorm, nuclear war, etc.
Or this is new videocard in relatively old computer?

[Graf Zahl]

Does it only happen with GZDoom? Did you try running other games? What's about DirectX?

My main problem here is that most engines lack decent profiling tools and that this only becomes apparent when really pushing the hardware with lots and lots of draw calls. The performance drop is not strong enough to register while just playing, more when benchmarking.

You seem to forget i've been/was? a computer technician for 20 years. I'm not sure about anything right now tough.
Next thing i'd do is watching cpu usage in task manager, something else could be eating the cpu in the background.

I can rule that one out - the drop is entirely on the GPU side (i.e. it's the call to glFinish that blocks longer.)

Or this is new videocard in relatively old computer?

It's a one year old graphics card in a 7 year old computer. The power supply should be fine, though - my old Geforce 550ti was drawing more power than the 1060 is doing.

[Darkcrafter]

If your previous card worked fine while still consuming more power then it's not PSU. I would take your card to another computer with a modern motherboard (and if possible with modern graphics card too) to see if it behaves in the same fashion. If it does, there is something wrong with the card. You could either use warranty, repair it, or even sell.