Freezing in games with new Radeon GPU, MCE-WHEA CPU Bus Error

[Bytorane]

I've sporadically familiarized with this throughout here and there, coming across videos as I mentioned earlier, but never had a complete picture and understanding when I needed it the most. I thought I wasn't doing any OC, I wasn't doing any tweaking, it probably doesn't affect me as much, so I didn't took it that serious.

That means, basically "oh it supports 128GB", but at "this" speeds, if you have "this" kind of DIMMs.

Here's the obvious question, why the hell do they even make and sell/advertise a Quad Kit with 3600MT speeds? Just for the Overclockers, as if enthusiast overclockers need that much RAM for content creation /development. ... Makes no practical sense.

However, now that we cleared that up, earlier I put it down to 2x2R DDR4-1866 and it still kept WHEA crashing.

Thanks guys, really appreciate the babysitting me through this, it's kinda of a rought times these past few months, I had so many breakdowns and RMAs with so many things, unrelated to PCs too.


I think testing with Nvidia GTX1070 8GB is good for now, I also have to return it soon, but I can re-borrow it perhaps next time if necessary. It should have crashed in 20-30 minutes, but it never did.
Also, I intentionally upped the graphic settings to max and kept playing with 15FPS. Then I simultaneously ran DCS and even turned on a benchmark to try to force something bad to happen, and it succeeded by crashing Fortnite with a standard application crash, due to out of memory, ofcourse due to Prime95/TestMem5 eating the rest up.

 

[Vincero]

Here's the obvious question, why the hell do they even make and sell/advertise a Quad Kit with 3600MT speeds? Just for the Overclockers, as if enthusiast overclockers need that much RAM for content creation /development. ...

Yes, pretty much... Same reason they sell kits with speeds over 3600MT/s for those who want to push as fast as their silicon can go.

Some dumbasses even pay for RGB LEDs on it... Because that's what you need on components that rely on least amount of power loading and clean electrical signals... Pulsing/flashing lights

 

[Bytorane]

Just found out, I still have the old RAM, 64GB of Crucial Ballistix. I think there were some errors found in the standard Memtest86+, thought I'm not sure right now. I thought I resold them, but guess wasn't successful yet (forgot about it and listings expired).

Now, first I'll just go back to AMD Radeon RX 6700XT and then re-test again, to double confirm. Then we'll see what PCIe GEN3 does.

 

[Zach_01]

However, now that we cleared that up, earlier I put it down to 2x2R DDR4-1866 and it still kept WHEA crashing.

I'm curious to see ZenTimings with this^^

As for all the other you are saying I cannot argue against. These people can write anything on their specs in order to get sales. Not Just AMD of course but anyone.
Not necessarily lying but stating them in a such a way to cause confusion. As if everyone is into every detail of this tech.
They wont point the weaknesses but only the "strong" points (marketing). Im surprised to the fact that they even state those stick/rank configurations now.
Intel still saying "up to DDR 5600" for their latest...

 

[A Computer Guy]

I'm curious to see ZenTimings with this^^

me too!

 

[eidairaman1]

Did you ever run the 6700xt in a totally different system?

 

[Bytorane]

I don't have another system anymore, my secondary PC has been slowly breaking down motherboard component by component, unexplained OS lag, weird errors, sata port malfunctions, data corruption, I decided to deconstruct it and most likely for archive or resell.

The other computers aren't mine, I need to schedule borrow time to do tests on them.

I've just reinstalled Radeon RX 6700XT to the primary PC now. However it's late evening, I might call it a day. Tempting to continue, but I don't want to get too exhausted into the night over this.

I'm curious to see ZenTimings with this^^

As for all the other you are saying I cannot argue against. These people can write anything on their specs in order to get sales. Not Just AMD of course but anyone.
Not necessarily lying but stating them in a such a way to cause confusion. As if everyone is into every detail of this tech.
They wont point the weaknesses but only the "strong" points (marketing). Im surprised to the fact that they even state those stick/rank configurations now.
Intel still saying "up to DDR 5600" for their latest...

Oh, I posted the wrong screenshot back on page 3. I meant DDR4-1866, that was back 2 days ago, before I took out 2 DIMMs.

It get's reported as 1860/930, not sure if this is just HW peculiarity or ZenTimings' ... I think it was similar in BIOS, I'll see later again.

Update1:

No wait, I did made a screenshot of ZenTimings with 64GB ...

And that above screenshot was on 3rd August? That far back, wait what, I don't know anymore, I lost track right now, this was a busy week I guess.

 

[eidairaman1]

I don't have another system anymore, my secondary PC has been slowly breaking down motherboard component by component, unexplained OS lag, weird errors, sata port malfunctions, data corruption, I decided to deconstruct it and most likely for archive or resell.

The other computers aren't mine, I need to schedule borrow time to do tests on them.

I've just reinstalled Radeon RX 6700XT to the primary PC now. However it's late evening, I might call it a day. Tempting to continue, but I don't want to get too exhausted into the night over this.



Oh, I posted the wrong screenshot back on page 3. I meant DDR4-1866, that was back 2 days ago, before I took out 2 DIMMs.

View attachment 358834
It get's reported as 1860/930, not sure if this is just HW peculiarity or ZenTimings' ... I think it was similar in BIOS, I'll see later again.

Update1:

No wait, I did made a screenshot of ZenTimings with 64GB ...

And that above screenshot was on 3rd August? That far back, wait what, I don't know anymore, I lost track right now, this was a busy week I guess.

View attachment 358835

That is what shops are for to troubleshoot.

 

[Bytorane]

I'm trying with RX6700XT and DDR4-2400 128GB like I used to have, and it doesn't want to crash for 2 hours, running 3 games simultaneously.

The only thing different with hardware is that I moved to a less powerful but more recent PSU (with good voltages) that I had in the secondary PC.

However we ruled out the PSU already. ... Then I drilled a bit deeper. I've disconnected the KVM Switch when I moved the system to the temporary Seasonic GX850W PSU, so that's ruled out.

I don't have the KVM Switch plugged in anymore since I'm down to just this 1 AM4 computer.

I remember some widely reported issues with USB connectivity with AMD several years back, requiring BIOS fixes, however I never experienced those.

I moved the AM4 system to a newer Case, which has USB3/USBC front ports, so I conveniently connected keyboard and mouse there.

So that's literally the only change I can think of right now since doing the RX480, GTX1070 and back to RX6700XT

So I'll move the pheripherals back to the back-io ports, ... all* crashes was during gameplay, when mouse and keyboard are actively used, ... almost all the time, however I remember it crashed once when I left Fortnite in spectator and I came back to see it frozen, so, might have been an exception, I doubt it.

Very annoying. Before I packed up the Corsair HX750W PSU I wanted to double-check with DMM, Volts shown 12.1 V for the +12VDC ... that prompted another real live test.

After hooking it up back to the same PC, readings in BIOS now show:

+12VDC == 12.07
+5VDC == 4.980
+3.3VCC == 3.280

Similarly in HWiNFO.

Is it possible the PSU "cooled off" after a week and is showing temporary good readings?

What if the BIOS update fixed the sensors? I forgot exactly when I updated the BIOS ... need to rollback my memory.

Worth downgrading BIOS to test the theory?

Anyway whatever, the old PSUs aren't ATX3.0 compatible so no big deal, I'll move on and forget about it.

 

[Vincero]

A not great contact on the 24-pin connector could cause a lower reading.
Motherboard makers have been known to correct BIOS voltage readings in the past with updates - I would sooner test the PSU on another board with known acceptable readings and see what it says.

 

[Bytorane]

Well it's done for, I unpacked and broken the seals of the new PSU yesterday, I still can return if I'd really wanted because of good return policies (the 14-day window), if it was really just bad readings I could have just waited a bit longer for proper ATX3.1 PSUs, I was negotiating about that in my mind ... but oh well.

I played a match of Fortnite for 20-30 mins, there were no dips on any of the key voltages, I had the digital multimeter hooked up at all times monitoring the 12VDC rail and it never went below 12.1 ... only now at the end, like 45 minutes since I hooked it up, it's 12.09V - I could keep it running idle for a few more minutes, and then I'm going to test the PSU on a different PC.

 

[A Computer Guy]

Well it's done for, I unpacked and broken the seals of the new PSU yesterday, I still can return if I'd really wanted because of good return policies (the 14-day window), if it was really just bad readings I could have just waited a bit longer for proper ATX3.1 PSUs, I was negotiating about that in my mind ... but oh well.

I played a match of Fortnite for 20-30 mins, there were no dips on any of the key voltages, I had the digital multimeter hooked up at all times monitoring the 12VDC rail and it never went below 12.1 ... only now at the end, like 45 minutes since I hooked it up, it's 12.09V - I could keep it running idle for a few more minutes, and then I'm going to test the PSU on a different PC.

So is the conclusion that RX6700XT and DDR4-2400 128GB works in your scenario?

 

[MrBanana]

Did you already RMA the gpu?
I had WHEA 18 errors after a month with a brand new rx 6700xt pulse from sapphire.
I changed the CPU and PSU, but it kept crashing, I eventually RMA the GPU and they sent me a brand new one, I don't have any issues since.

 

[Bytorane]

So is the conclusion that RX6700XT and DDR4-2400 128GB works in your scenario?

No. That's how I had it configured for since I started seeing these WHEA crashes. That's the configuration I had when I started this thread.

I just wasn't able to get the same configuration to crash again the other day, since I put RX6700XT back in after a test with Nvidia GTX1070.

Everytime I changed a GPU, I uninstalled the drivers officially, rebooted and used DDU in safe-mode to cleanup and then reinstall with "Factory Reset / Default".

Did you already RMA the gpu?
I had WHEA 18 errors after a month with a brand new rx 6700xt pulse from sapphire.
I changed the CPU and PSU, but it kept crashing, I eventually RMA the GPU and they sent me a brand new one, I don't have any issues since.

No, since it was never conclusive that it was the GPU, because RX7800XT was crashing too, and that card works completely fine in it's intended AM5 computer. It could have as well be something with drivers/mobo/AGESA, who the heck knows.

---

I've finished installing the brand new 1000W PSU now voltages look even better in BIOS and with a DMM, for what's it worth. The PSU factor should now be completely ruled out.
+12VDC is 12.17V for example, hopefully that's not too much, but that's idle high side, it flucuates with 12.07V on the lower side.
I know it's overkill currently, but I wanted it like this for the little bit more Euros it took, under load it'll also operate in the range where efficiency is the best.

Did you ever run the 6700xt in a totally different system?

Wanted to respond directly to that too ... similarly, the response I gave to MrBanana above. It didn't felt like a GPU issue ever since the swap to RX7800XT never made a difference.

I could do the tests if you guys insist, but I would have to borrow time with other PCs again which aren't mine and I've already had a bit of an issue in past occurences, because one of the machines is for active business office use and it happens to have a GTX1070 and lacks an iGPU (I wasn't the one building that PC, I wouldn't have made that mistake).

I will continue testing now that I can't seem to recreate the problem in Fortnite or any other game, one would say oh it's solved ... well not, I don't know what exactly solved it, I don't know what's different about the configuration now, it's kinda worse now because it's dangling and there's no guarantee it's completely solved.

Maybe I have to play a bit longer of a session, we'll see. I never got to the step of trying another fresh Windows 10 or Windows 11 installation yet. Currently all testing mentioned in this thread has been done on a second Win10 installation, as mentioned. Everything that happened last year was on the old Win10 installation (which had a lot more programs, tweaks, etc)

That is what shops are for to troubleshoot.

With the amount of back and forth, wouldn't that come at a significant price, I mean, the hours I spent into this potentially might be lesser, but not much, I'd say 20% to 30% ... especially because it requires playing and waiting for the game to crash.

 

[Bytorane]

I noticed I forgot to check out the Ryzen DRAM Calculator ... but there's no ZEN 3 option, am I missing something?

Perhaps it not strictly necessary that I have to pick Zen3, to get useful results for my configuration?

 

[Zach_01]

No you are not...
I think latest is v1.7.3



You can use it and see if it helps at all.
You have to import XMP profile though for more accurate settings suggestion

 

[A Computer Guy]

I still think your best bet is to get a ram kit from your QVL that specifically states 4 dimms and 128Gb capacity at whatever speed. I know what it's like going through 6 months of testing different stuff swapping out parts etc...only to have it end in wtf. If I recall properly your QVL doesn't seem to like Kingston. Even though QVL isn't perfect at least it shows a trend of testing ram that works.

 

[DaemonForce]

Here's the obvious question, why the hell do they even make and sell/advertise a Quad Kit with 3600MT speeds? Just for the Overclockers, as if enthusiast overclockers need that much RAM for content creation /development. ... Makes no practical sense.

It's probably for types like me that not only like to find the agreeable maximum for general+gaming performance but to push encoding, rendering and other content creation stuff balls to the wall.
I'm still on a 750W antique and these are my reported lines:

Load/Vdroop look pretty good and there's no instability symptoms but I sometimes wonder if it's time for a new power box.
Thing is, I don't see any valid successor on the market. I've put massive chonker cards in here before but couldn't initialize some and we still don't know why.
Obviously my rails are fine but if there are any power related issues, they're supposed to be obvious messages like sudden shutoffs, weird throttling and so on.
Some boards act very weird when you throw very specific hardware at them and that could always be the case.

it was never conclusive that it was the GPU, because RX7800XT was crashing too, and that card works completely fine in it's intended AM5 computer. It could have as well be something with drivers/mobo/AGESA, who the heck knows.

This part right here is so frustrating. We can say that PCs are like Legos or whatever but the moment crashes start happening with UEFI glitches, AGESA updates and everything else that starts up, all bets are off. Good luck. Can't wait to deal with the same nightmare when I hop over to 8000 series.

 

[Vya Domus]

Nothing suggested this is a PSU issue, I think you wasted your time and money there, those readings are not very accurate and it wouldn't matter anyway since the components are meant to work assuming a +-5% tolerance so 11.4V to 12.6V would be fine.

 

[Bytorane]

Nothing suggested this is a PSU issue, I think you wasted your time and money there, those readings are not very accurate and it wouldn't matter anyway since the components are meant to work assuming a +-5% tolerance so 11.4V to 12.6V would be fine.

Right ... well that HX750W was getting 10 years old. If that one's a good unit I wouldn't be surprised if it lasted probably 4-5 more years even ... but on average I'd say it lasted long enough, now I can resell it properly instead of just disposing it.
Edit: typos.

 

[A Computer Guy]

Right ... well that HX750W was getting 10 years old, now if it's that one's good I wouldn't be surprised if it lasted probably 4-5 more years even ...

It's probably best that you replaced a 10yr old PSU for use with newer hardware anyway. One less thing to worry about.