Best way of stress testing a GPU?

[Falkentyne]

In the past it could be useful because both AMD and Nvidia used to leave things behind that could cause conflicts. That stopped happening years ago. DDU runs the risk of removing things you don't want to remove so can just cause extra problems. Nowadays there are no benefits at all from using it. It's one of those programs which are a relic from the past where the benefits used to out way the cons, but now it's the other way around

DDU is absolutely NOT a relic from the past!

It's constantly updated. Perhaps you're mixing it up with Driver Sweeper or something.
Not only is it updated, Wagnard is also very active on his forums. Yes it sometimes causes issues for some people, but post a log and he fixes them.

Some people have had issues with inconsistent strange 3dmark results. Ignoring the latest 3dmark "Ampere buggy" update which caused a 500 point loss in Port Royale, as well as a loss of FPS in Heaven right after running Port Royal (even though you closed 3dmark)--which apparently UL has been able to reproduce and will hotfix or something--, one way people suggest to fix driver bugs is the following.

1) Uninstall MSI Afterburner. Reboot.
2) Run the driver install package for the current driver you already have installed. Tick advanced options and choose "Clean install" (even though the driver is already installed). Reboot.
3) Download the latest DDU. Reboot to safe mode (hold left shift while choosing restart windows). Unplug internet, run DDU in safe mode with internet unplugged.
4) Reboot, reinstall current drivers with internet unplugged. Reboot again.
5) Install MSI Afterburner. Replug internet.

 

[no_1_dave]

DDU is absolutely NOT a relic from the past!

It's constantly updated. Perhaps you're mixing it up with Driver Sweeper or something.
Not only is it updated, Wagnard is also very active on his forums. Yes it sometimes causes issues for some people, but post a log and he fixes them.

It's 100% a relic from the past... Both AMD and Nvidia drivers uninstall correctly now. It doesn't matter how often DDU is updated because it is no longer needed

 

[The Old School Gamer]

DDU a relic from the past lol... funny.

A 3090 and 5950X on a 850w PSU, yeah I'd be measuring total power draw at the wall as it could be close.
I can tell you first hand a 850w THOR couldn't cope with a 10900k & 3090 STRIX, we banged in a 1200w problem solved.

 

[xPETEZx]

DDU a relic from the past lol... funny.

A 3090 and 5950X on a 850w PSU, yeah I'd be measuring total power draw at the wall as it could be close.
I can tell you first hand a 850w THOR couldn't cope with a 10900k & 3090 STRIX, we banged in a 1200w problem solved.

Those must have been really high OC'ed??

My 5900x + 3090 barely draw over 550W at the wall. And that was with a mild OC on both. Measured using a watt-o-metter thingy. So the PSU being Gold rated, probably means actually draw DC was closer to 500-510W.

Granted the FE isnt as power hungry as a Strix, and a 10900k draws more than a 5900x, but I wouldnt have thought 300W+ ?

 

[OldBoy75]

I had similar issues with my 3090 last year and so did some other forum members. Theres a few threads on the forum somewhere. From my experience id be looking towards the PSU. mine also was a 850w gold rated ( good brand - seasonic ). changed it for a 1000w and the problems were solved

 

[LtMatt]

I had similar issues with my 3090 last year and so did some other forum members. Theres a few threads on the forum somewhere. From my experience id be looking towards the PSU. mine also was a 850w gold rated ( good brand - seasonic ). changed it for a 1000w and the problems were solved

This was my initial thought too, power hungry GPUs need a solid PSU and given the high wattage this game draws, it can easily sniff out any instabilities in a system.

 

[JohnG7]

Transient Spikes.

 

[Benass]

FurMark is like the Prime95 of GPU stress testing, it places a lot of stress on the GPU and is known to make GPUs beg for mercy before the eye of Sauron or the Donut of Doom. This is a GPU stress test known to strike fear in lesser GPUs cores and VRMs. This is one bad boy of a stress test that makes sure your GPU can handle everything else if it survives an hour of FurMark. It will wreak havoc on bad GPU cooler designs and is to be used with utmost care. This is my go-to GPU stress test for absolute stability. The UI is simple and effective.

3DMark is generally regarded as the most popular benchmarking tool, used widely and used to rank CPUs and GPUs. The tests are closer to gaming than stress tests and offer workloads that stress your components in a realistic manner. An overall great tool for the overclocker or stability tester to give an accurate picture of how the PC will perform in gaming-focused applications. Your CPU, GPU and RAM all have an impact on your score which lets you see if your rig is performing correctly.

For monitoring software there are a few great choices like HWiNFO64, HWMONITOR and Open Hardware Monitor (OHM) are all great pieces of software with different functions. Use any of these and you will see what your system is doing at any time and how your temperatures are. I must mention that HWMONITOR and OHM are more starter friendly with a simpler UI design, but HWiNFO64 offers more functionality and customizability.

 

[The Old School Gamer]

Yeah if the power limiter was dialled right back it was perfectly stable, no hard lock ups or reboots, but the whole point is to crank such a card under water to the max.
We did measure a peak of over 800w at the wall so it was certainly up there in the 'oh **** would you look at that' zone.

I've also seen a MSI 3080 Trio drawing a unreasonable amount of power as well, that one may of been faulty though as it was returned by the customer because you could fry your breakfast on the back of it.
Also had Gigabyte 3080 Vision go faulty, that one was squealing like a distressed pig under load (I mean it was really loud like someone was whistling directly in your ear).

Check out this build my work buddy J just completed for the boss man.
I didn't get that involved with this build as I'm not really into water cooling.
But credit where credit is due, he did do a damn fine job.

Spoiler: ...

 

[Nitefly]

Thanks all - still at work so unlikely I’m going to be able to make much progress today but I will get round to it and say how I get on I promise, really appreciate the input thank you.

DDU a relic from the past lol... funny.

A 3090 and 5950X on a 850w PSU, yeah I'd be measuring total power draw at the wall as it could be close.
I can tell you first hand a 850w THOR couldn't cope with a 10900k & 3090 STRIX, we banged in a 1200w problem solved.

Ah man, my psu is a 850w ROG Thor.... but as said no overclocks.

 

[LtMatt]

Ah man, my psu is a 850w ROG Thor.... but as said no overclocks.

Should be okay assuming you have two PSU cables.

 

[nlel1975]

I’ve recently upgraded my PSU from a 550W to a 750W in order to run a 6800XT when they are sensibly priced again, rest of my system contains a Ryzen 5 2600, 2x 16GB RAM sticks, 1x NVme drive, 2x SATA HDD and a few fans.

I’ve also got an old 1275W PSU that was given to me for free which makes a horrible squeaking noise after being on for a awhile but rather then chuck it I’ll keep it for testing in future.

 

[Nitefly]

So after about an hour of trying to get the afterburner overlay to run (I didn’t install rivatuner at first instance, derp, we may be getting somewhere).

Look at this power draw on the start menu.



Seems pretty high.....!

 

[Nitefly]

Further to my play testing yesterday, the game seemed to crank the GPU at 100% usage, temperature of 65 degrees centigrade, and it would then die within 5-10 minutes.

My psu - which displays total power consumed and is 850watts - never peaked above 550w, according to the display.

Any thoughts on next steps? I’m guessing the settings just rinse the GPU too much!

 

[TNA]

Further to my play testing yesterday, the game seemed to crank the GPU at 100% usage, temperature of 65 degrees centigrade, and it would then die within 5-10 minutes.

My psu - which displays total power consumed and is 850watts - never peaked above 550w, according to the display.

Any thoughts on next steps? I’m guessing the settings just rinse the GPU too much!

Did you do any of the steps mentioned above?

 

[Nitefly]

Did you do any of the steps mentioned above?

No well not all of them - I basically haven’t had time beyond my initial observation exercise last night to do anything, so was trying to get as much info as possible from what I had before embarking on anything else, but you’re right there’s plenty more I could be doing. I only have about an hour free on any day

I have run stress tests on the ram and I’m happy that the sticks themselves are not faulty (no errors on the testing) but I could dial back the speed.

To make it clear I’m not intending to take the ****, I hereby pledge £30 to charity for the help thus far, regardless of the outcome, and will pledge further as the process goes on!

 

[SpartaP]

Further to my play testing yesterday, the game seemed to crank the GPU at 100% usage, temperature of 65 degrees centigrade, and it would then die within 5-10 minutes.

My psu - which displays total power consumed and is 850watts - never peaked above 550w, according to the display.

Any thoughts on next steps? I’m guessing the settings just rinse the GPU too much!

Keep in mind that wattage displayed is just for the GPU, not the whole system. The gpu temp is fine. What dies, the game or does the whole PC lockup/switch off?

 

[TNA]

No well not all of them - I basically haven’t had time beyond my initial observation exercise last night to do anything, so was trying to get as much info as possible from what I had before embarking on anything else, but you’re right there’s plenty more I could be doing. I only have about an hour free on any day

I have run stress tests on the ram and I’m happy that the sticks themselves are not faulty (no errors on the testing) but I could dial back the speed.

To make it clear I’m not intending to take the ****, I hereby pledge £30 to charity for the help thus far, regardless of the outcome, and will pledge further as the process goes on!

From experience, having no errors on ram does not mean it is stable. I have run 16 hours plus mem stability tests that passed, then get a blue screen when playing a game. Go back to xmp and it stops crashing.

First thing you need to make sure is remove all overclocks. If that does not solve it, try lower clocks on gpu.