6800 xt computer crashing on demanding games. psu problem?

[CindyH]

I had this exact same issue with a 6700xt. It sometimes worked via DX11 but anything like F1 or MSFS on DX12 the screen would go black then reboot. I thoroughly tested the memory and used a rail tester on the PSU. I also tested the graphics card in another system and got the same results. In the end I had to RMA the graphics card which resolved the problem.

 

[Noci]

to Thebadgamer98, as you can see in the comments of this thread there are multiple scenario's which could lead to the problems you are experiencing.

Before spending money on new hardware (unless of course you're dying to do so ), I would advise to do extensive testing with logging of the data. Than you can pinpoint trouble areas based on deviations in the data (towards comparible hardware and load conditions).
It will eliminate items of the list of possible culprits and you can do a targeted search on the remaining items.

Keep in mind that most spontanious crash problems are often a combination of issues that enhance each other. So maybe it is not as simple as to point at one thing only, it just could be the famous drop that lets the bucket over flow as a sort of domino effect.

 

[Thebadgamer98]

to Thebadgamer98, as you can see in the comments of this thread there are multiple scenario's which could lead to the problems you are experiencing.

Before spending money on new hardware (unless of course you're dying to do so ), I would advise to do extensive testing with logging of the data. Than you can pinpoint trouble areas based on deviations in the data (towards comparible hardware and load conditions).
It will eliminate items of the list of possible culprits and you can do a targeted search on the remaining items.

Keep in mind that most spontanious crash problems are often a combination of issues that enhance each other. So maybe it is not as simple as to point at one thing only, it just could be the famous drop that lets the bucket over flow as a sort of domino effect.

Yup. Glad I asked here before just buying a PSU.

Ran MemTest86 overnight and it came back clean, no errors.

Now to try CPU on Eco mode and see if that does anything to help.

Eco mode appears to have done it. Before, MW2 was crashing as I loaded into a match but now it's running fine for about 30min straight. Only more time and testing will tell but I'm feeling great about it so far. Thanks for all the advice, y'all!

 

[VuurVOS]

Eco mode appears to have done it. Before, MW2 was crashing as I loaded into a match but now it's running fine for about 30min straight. Only more time and testing will tell but I'm feeling great about it so far. Thanks for all the advice, y'all!

Do you have "3DMark Advanced Edition"? To crash my system, I preheated the VRM with OCCT for 20 minuten. Then with "CPU Profile" in 3DMark I could instandly crash my system!
This only happend in my case when the processor wasn't in eco mode on the old motherboard and the VRM was warmed up to about 60~65 degrees.

If your system stays stable when the processor is in eco mode, my advice is to replace the motherboard.

 

[Thebadgamer98]

Do you have "3DMark Advanced Edition"? To crash my system, I preheated the VRM with OCCT for 20 minuten. Then with "CPU Profile" in 3DMark I could instandly crash my system!

haha I do not, I think I'll leave that particular test to you my friend

 

[TheoneandonlyMrK]

You should never use 2 8-pin connectors daisy chained on the same cable. Your GPU doesn't know that those 2 8-pins are on the same cable so it might pull the maximum rated safe wattage from each as if each is it's own separate cable. In effect it could be pulling double the maximum wattage the cable is rated for. There's a decent chance that's the cause of your issue.

It wouldn't be split , legally if not capable of supplying two times 8pin power.

The issue is power spikes going out of spec.

That is why separate power cables are spoken of, they support the power with enough overhead usually to sustain the spike.

If it didn't support two times 8 pin power they could face a day in court Afaik.


IMHO your infinity fabric is stable only upto a point and those games go beyond.

Something I can replicate on my pc going from 1800 to 1866/1900 ?!?.

 

[Macro Device]

Undervolt your GPU. If it continues crashing in the same manner (id est after the same amount of time) then it's not your PSU. If it continues crashing but crashes are later than expected or it stops crashing then it's your PSU or GPU VRM issue. Most likely the former but can't hurt to test the latter.

After such testing, I'd get a couple outdated watt-greedy GPUs a-lá GTX 770 so it's not hurting my budget but still dunks on my PSU if I were in your shoes. The problem goes away despite wattage going higher, it's no more a PSU's fault. It's your time to RMA the GPU. The problem stays or worsens, it's time to invest in a beefier PSU.

And yes, optimising the rest of your system is a must. Sell your RAM to some Intel boy and get yourself something more AMD-friendly.

 

[evernessince]

It wouldn't be split , legally if not capable of supplying two times 8pin power.

The issue is power spikes going out of spec.

That is why separate power cables are spoken of, they support the power with enough overhead usually to sustain the spike.

If it didn't support two times 8 pin power they could face a day in court Afaik.


IMHO your infinity fabric is stable only upto a point and those games go beyond.

Something I can replicate on my pc going from 1800 to 1866/1900 ?!?.

There's no such things as legal limits to the ATX spec. Both Nvidia and AMD have gone outside spec before. That's considering that the ATX spec itself is pretty loose, hence why you get cheaper PSUs with thinner wires. Companies aren't going to get sued due to user error. If you run a cheap PSU and protections kick in because you plugged into 2 PCIe connectors from the same cable that's entirely on the user. Many PSUs are moving towards one PCIe 8 pin per cable to completely prevent user error entirely, like my Seasonic Prime Titanium.

Yup. Glad I asked here before just buying a PSU.

Ran MemTest86 overnight and it came back clean, no errors.

Now to try CPU on Eco mode and see if that does anything to help.

Eco mode appears to have done it. Before, MW2 was crashing as I loaded into a match but now it's running fine for about 30min straight. Only more time and testing will tell but I'm feeling great about it so far. Thanks for all the advice, y'all!

The problem is that we haven't really narrowed now the root cause. It could be the PSU or the motherboard's VRM, CPU overheatings, or a number of other things. Things working fine now doesn't guarantee into the future.

 

[Thebadgamer98]

The problem is that we haven't really narrowed now the root cause. It could be the PSU or the motherboard's VRM, CPU overheatings, or a number of other things. Things working fine now doesn't guarantee into the future.

Oh, I see. I thought this indicated the fault lay with the VRM, thanks for clarifying. Do you know of any methods to test some or all of those potential issues to, hopefully, find the root cause of the issue?

 

[oxrufiioxo]

Oh, I see. I thought this indicated the fault lay with the VRM, thanks for clarifying. Do you know of any methods to test some or all of those potential issues to, hopefully, find the root cause of the issue?

Try running a cpu only benchmark somthing like CB2024 and see if it crashes outside of eco mode. Maybe a couple hours.

 

[evernessince]

Oh, I see. I thought this indicated the fault lay with the VRM, thanks for clarifying. Do you know of any methods to test some or all of those potential issues to, hopefully, find the root cause of the issue?

I'll just point to what oxrufiioxo said as it's an excellent recommendation:

Try running a cpu only benchmark somthing like CB2024 and see if it crashes outside of eco mode. Maybe a couple hours.

With Eco mode disabled, if your system crashes in a CPU heavy workload it would inform us that your issue is CPU / Motherboard related. Given that you've run MemTest with no errors we can safely rule out memory as a potential cause.

I'd also recommend Y-Cruncher as a test given it's quite intensive on Zen CPUs. It was my go to for stress testing as for me it found bad overclocks when other programs did not. Make sure to be running HWInfo while doing this testing with logging enabled. In this instance you want to look at CPU temperatures, CPU Clock speeds, and VRM Temperatures. You can double click on values to bring up a chart so you can visually see change over time. CPU and VRM temps should be no higher than 90c preferably. CPU clock speeds should be 4.1 GHz to 4.2 GHz for all core workloads, any lower than that would indicate throttling, whether that be from the VRM or CPU temps.

 

[Thebadgamer98]

I'd also recommend Y-Cruncher as a test given it's quite intensive on Zen CPUs. It was my go to for stress testing as for me it found bad overclocks when other programs did not. Make sure to be running HWInfo while doing this testing with logging enabled. In this instance you want to look at CPU temperatures, CPU Clock speeds, and VRM Temperatures. You can double click on values to bring up a chart so you can visually see change over time. CPU and VRM temps should be no higher than 90c preferably. CPU clock speeds should be 4.1 GHz to 4.2 GHz for all core workloads, any lower than that would indicate throttling, whether that be from the VRM or CPU temps.

Took the CPU out of ECO mode and put it on default, then ran Y-Cruncher and logged with HWInfo. I can provide the logs but the pertinent bits are that the CPU stayed at 95.4 C throughout the entire half hour of testing and the VRM steadily increased from 45 C to 61 C throughout the half hour. CPU clocks hovered around 3.7 GHz (give or take 0.1) for the duration of testing.

Of course I don't really know how to interpret this information, but it sounds like the problems lie more with my CPU than with the Mobo? Please correct me if I'm wrong. Advice is appreciated!

 

[Toothless]

Or VRM can't handle that chip. 95.4c is pretty hot too.

 

[Thebadgamer98]

Or VRM can't handle that chip. 95.4c is pretty hot too.

Yeah, google searches agree that it's just not really compatible. Something about the b450m only able to safely handle a CPU that pulls 65w while the 3900x pulls a healthy 105w.

 

[Toothless]

Yeah, google searches agree that it's just not really compatible. Something about the b450m only able to safely handle a CPU that pulls 65w while the 3900x pulls a healthy 105w.

Mine pulls a nice 120w under load on a x570 Crosshair Hero. Maybe it's time for a proper board for you.

 

[Thebadgamer98]

Mine pulls a nice 120w under load on a x570 Crosshair Hero. Maybe it's time for a proper board for you.

Think you're right. I also considered sidegrading to a 5700x which has ~65w draw. Though I imagine the new board would be a bigger improvement in the long run. I'm nervous to wholesale swap my mobo! haha

 

[Toothless]

Think you're right. I also considered sidegrading to a 5700x which has ~65w draw. Though I imagine the new board would be a bigger improvement in the long run. I'm nervous to wholesale swap my mobo! haha

Look for board/cpu combos. Those typically have a good deal for both and when in doubt, ebay it out.

 

[evernessince]

Took the CPU out of ECO mode and put it on default, then ran Y-Cruncher and logged with HWInfo. I can provide the logs but the pertinent bits are that the CPU stayed at 95.4 C throughout the entire half hour of testing and the VRM steadily increased from 45 C to 61 C throughout the half hour. CPU clocks hovered around 3.7 GHz (give or take 0.1) for the duration of testing.

Seems to be that your CPU is overheating and throttling quite a bit. You should be seeing 4.1 to 4.2 GHz all core.

You didn't get any crashing? Given that we've established your CPU is overheating and throttling I have to wonder if the additional heat being dumped into your case by your GPU in something like a game could worsen the situation for the CPU and cause the crashes or if it's just the GPU / PSU.

Does a GPU heavy workload like furmark crash the system as well?

Of course I don't really know how to interpret this information, but it sounds like the problems lie more with my CPU than with the Mobo? Please correct me if I'm wrong. Advice is appreciated!

Correct. The VRM temps you reported of 41c to 61c are cool. VRM don't typically throttle until above 95c. Of course motherboards can cause all sorts of funky instability issues and crashes but those are near impossible to root out unless you swap the board. It's always possible that your CPU throttling issue could be separate from your crashing issue. Throttling in and of itself should not cause crashing unless the CPU cannot throttle enough or is defective. Sometimes heat can bring out crashing in a poor silicion quality chip that doesn't occur at lower temps.

 

[Noci]

Took the CPU out of ECO mode and put it on default, then ran Y-Cruncher and logged with HWInfo. I can provide the logs but the pertinent bits are that the CPU stayed at 95.4 C throughout the entire half hour of testing and the VRM steadily increased from 45 C to 61 C throughout the half hour. CPU clocks hovered around 3.7 GHz (give or take 0.1) for the duration of testing.

Of course I don't really know how to interpret this information, but it sounds like the problems lie more with my CPU than with the Mobo? Please correct me if I'm wrong. Advice is appreciated!

What CPU cooler are you using and which thermal paste? If you can get the CPU temps down during Y-Cruncher, it will be a whole other ballgame.

HwInfo also shows how much Watts your CPU uses, will be quite a difference between the CPU @ 3.7 or 4.2 GHz, let alone the ECO mode.

Temps of the VRM's look good when they peak @ 61 Celcius, despite the 65W CPU recommendation, they still have some overhead. So let's park that possibillity for now, because if they were at their limit, for sure the temps would hit the roof.

But you mentioned the crashes happend during gaming when the GPU also had to show some muscles. That combined with the CPU heating excessively, would tend me to think it indeed could be an insufficient or unstable power supply to the individual components. With the VRM's not in the picture (for now) let's focus on the stabillity of the power supply

If you can test with running folding@home, you can run the CPU & GPU simultaneously at 99% load (and you'll be working for the good cause at the same time). During that testing, logging with HwInfo could give some insight how your PSU is holding out on the 12V, 5V & 3.3V supply/stabillity. For example if your 12V drops significantly it could be an indication your PSU is not up to it's task.

Below the link:

https://foldingathome.org/start-folding/?lng=en

Make sure you give folding priority with the power slide fully to the right (see picture) and your system will be pushed to the limit.

 

[Thebadgamer98]

What CPU cooler are you using and which thermal paste?

Wraith Stealth Cooler - This is the stock cooler that came with my Ryzen 5 2600. My 3900x was used and did not come with cooler.
Corsair TM30 Performance Thermal Paste

But you mentioned the crashes happend during gaming when the GPU also had to show some muscles. That combined with the CPU heating excessively, would tend me to think it indeed could be an insufficient or unstable power supply to the individual components. With the VRM's not in the picture (for now) let's focus on the stabillity of the power supply

If you can test with running folding@home, you can run the CPU & GPU simultaneously at 99% load (and you'll be working for the good cause at the same time). During that testing, logging with HwInfo could give some insight how your PSU is holding out on the 12V, 5V & 3.3V supply/stabillity. For example if your 12V drops significantly it could be an indication your PSU is not up to it's task.

Folded for 20 minutes to get this snapshot. Despite 100% CPU and 90% GPU usage, it did not crash. The log can be found here, if there's any interest in it.

Some takeaways I noticed while folding:

• CPU averaged ~3.8 GHz during the folding, but increased to 4.5 GHz AFTER I stopped folding.
• CPU stuck on 95.6C during the folding process, with little fluctuation.
  • CPU fan was pumping at ~2600 RPM for the folding, significantly faster than the GPU fan (~1700). Normal?
• GPU hovered around 80C.
• 12/5/3.3V all stayed at their respective voltage.
• VRM stayed at 61-62C.

 

[Toothless]

I see a CPU problem. Poor thing is running way too hot and this is why AMD says to get a good cooler with the 3900x

 

[Thebadgamer98]

I see a CPU problem. Poor thing is running way too hot and this is why AMD says to get a good cooler with the 3900x

Yes, I've been eyeing upgrading to a Scythe Fuma 3 for that very reason. Also considering the upgrade to a ASRock B550M Pro SE motherboard.

 

[Toothless]

Yes, I've been eyeing upgrading to a Scythe Fuma 3 for that very reason. Also considering the upgrade to a ASRock B550M Pro SE motherboard.
View attachment 326658

That cooler will do good. I use two Fuma 2s myself. As for the board unless it doesn't have the features you need, no.

 

[Thebadgamer98]

Does a GPU heavy workload like furmark crash the system as well?

I ran the Burn-in test for 20 mins, temps peaked at 93C but it didn't crash.

 

[evernessince]

I ran the Burn-in test for 20 mins, temps peaked at 93C but it didn't crash.

View attachment 326667

93c is very high for only 20 minutes and it appears your GPU core clock is throttling as well. You are 80 - 100 Mhz lower than where you should be compared to other examples I've see of the exact same card running Furmark on stock settings. Looking at TPU's reviews of different 6800 XT models, the average for this card is between 63 and 73c with 73c being overclocked while having the card set at max power limit. Given that your CPU temps were also high I'm wondering if your case doesn't have enough airflow. If your GPU and CPU are reaching danger area temperatures individually that would create an environment where games that utilize both could lead could result in a crash due to them both dumping hot air into the case. Is it possible for you to run the same test with the side panel off? If your case has improper airflow temps should noticably drop.