Strange Problem With New System

[Jarhead0331]

I'm really not convinced this is a DCS problem, but DCS always seems to trigger it.  Maybe some of you guys might have some ideas.

 

I recently received a new build from CLX with the following specs:

 

Processor: AMD Ryzen™ 9 5900X 3.7GHz 12-Core

Motherboard: MSI B550-A PRO - ATX

Memory: G.SKILL Z Trident Total 64 GB Dual Channel Memory

Graphics Card:AMD Radeon™ RX 6900 XT, 16GB of dedicated GDDR6

Operating System Storage: 1TB Patriot P300 PCIe 4.0 M.2 NVMe

CPU Liquid Cooling: CLX Quench 240 Closed Liquid Cooler

Secondary Storage: 4TB 3.5in Seagate Barracuda HDD

Chassis Selection: Set Eclipse P400S Mid Tower Black/Red

Chassis Fans: CLX Red Ring Fan

Power Supply: EVGA 850 B3 - 80+ Bronze 850W

 

Shortly after installing DCS and running it in VR, I started getting green screen crashes followed by reboots, artifacting or signal loss to all monitors. Eventually, I sent the system back to CLX and they replaced the GPU and reinstalled windows. When I received the system back, and ran DCS, the same thing happened. Green screen crashing, artifacting and sometimes signal loss. CLX had me run DDU and do a fresh install of the Radeon software. This seemed to solve the problem and everything was running great....until this week. I put in a second SSD and to do that, I had to take the GPU out. As soon as I put everything back in and went to run DCS, same damn problem.

 

CLX insists it’s a GPU problem and they are sending me another one. When I express disbelief that two GPUs can go bad, they claim that it is an industry wide problem due to covid shortages and work restrictions. QA has been very poor and parts shortages are requiring manufacturers to source from more questionable sources, and they are seeing record numbers of bad cards as a result. They claim this is a problem with AMD and Nvidia.  They had me run my card with furmark under full load and they do not believe this is a power issue. What else could it be? MOBO? Ram? I’m getting to my wits end. CLX customer support has been great, but this is getting old. I've spent more time on the phone with tech support than I have using this $3200 system.

Edited April 24, 2021 by Jarhead0331

[silverdevil]

taking out the GPU and putting it back in really should not matter EXCEPT is it possible that something is loose or not completely seated? the problem here is DCS seems to spur this issue and taking bits and pieces out to test really is not viable. because if say you put in only one RAM DMIMM at a time the game is going to suck. how does it do in 2D? maybe the VR headset has the issue? just thinking of possibilities.

[Jarhead0331]

1 minute ago, silverdevil said:

taking out the GPU and putting it back in really should not matter EXCEPT is it possible that something is loose or not completely seated? the problem here is DCS seems to spur this issue and taking bits and pieces out to test really is not viable. because if say you put in only one RAM DMIMM at a time the game is going to suck. how does it do in 2D? maybe the VR headset has the issue? just thinking of possibilities.

I don't think taking the GPU caused the problem, per se. The problem existed before I did that, but the DDU and re-install of the radeon software seemed to correct it, or at least stabilize it. Once the system is "corrupted" so to speak I can get green screen crashes, reboots, artifacting and no monitor signal at random times...even from the desktop with nothing running. Really the only time it will be stable and trouble free is if I boot it in safe mode.

[silverdevil]

success in safe mode could still be a root hardware issue. however it more likely points to a software component loading at start up. it probably sounds silly but how about go through the motions to reinstall video drivers again? i had an issue where i had a 1070 card installed and DCS was only thing that crashed. i thought for sure it was the card so i got a 1080. when i installed i was watching the file / install progress and saw a really old DLL replaced. the 1080 always worked from then on. i sold the 1070 to a colleague and he still uses it without issue. also in case you do not know, nvidia has two main components,. one is the control panel and the other is basically an overlay software called experience. i had problems with experience before and removing it helped some.

[Jarhead0331]

Just now, silverdevil said:

success in safe mode could still be a root hardware issue. however it more likely points to a software component loading at start up. it probably sounds silly but how about go through the motions to reinstall video drivers again? i had an issue where i had a 1070 card installed and DCS was only thing that crashed. i thought for sure it was the card so i got a 1080. when i installed i was watching the file / install progress and saw a really old DLL replaced. the 1080 always worked from then on. i sold the 1070 to a colleague and he still uses it without issue. also in case you do not know, nvidia has two main components,. one is the control panel and the other is basically an overlay software called experience. i had problems with experience before and removing it helped some.

 

I've run the DDU many times. Also, as noted above, this is an AMD card. Not Nvidia.

[silverdevil]

i understand it is not nvidia. i am only suggesting that AMD has similar components that could cause issues. and perhaps those components are not loading at startup when in safemode.

[Jarhead0331]

3 hours ago, silverdevil said:

i understand it is not nvidia. i am only suggesting that AMD has similar components that could cause issues. and perhaps those components are not loading at startup when in safemode.

 

For the record, I did try installing only the AMD display driver without all the bloatware. Still, no dice. Green screen of death. Very very frustrating.

Edited April 24, 2021 by Jarhead0331

[XPACT]

1 hour ago, Jarhead0331 said:

For the record, I did try installing only the AMD display driver without all the bloatware. Still, no dice. Green screen of death. Very very frustrating.

 

I would try to reseat the GPU again and also change DP or HDMI cable that you are currently using.

 

Also do you have any type of anti sag card support/holder installed? I am asking because RX 6900 XT is a huge GPU and pins could possibly not make good contact inside PCIe slot due to sheer weight of the GPU.

 

If you don't have card support/holder after GPU reseating  try placing your case horizontally on the ground and see if the issue remains.

EDIT: with gravity pushing GPU down into the slot of course

Edited April 24, 2021 by XPACT

[Leg2ion]

Any motherboard BIOS updates available? Also check for any manufacturers utility software - as they can also install 'stuff' that may not be the best. I have a MSI Nvidia 3070 - and as a for instance MSI utility Dragon Center installs a 'network optimiser' - killed my internet speed both up and down.

 

Good tip from @XPACTabove.

 

Is your cooling working correctly on CPU/GPU, check nothing has been switched to silent mode or has disabled the fans allowing things to overheat? Make sure the cooling is on Auto, or if possible manually crank it up and then run DCS.

 

One other thing - kick out the VR and go with a monitor and see if it is stable.

[Jarhead0331]

I'm not breaking out any cigars by a long shot, but I did a manual install of the AMD drivers through the device manager and so far the system seems to be running stable. I've watched some youtube, loaded steam and downloaded a file. Otherwise, just let the thing idle on desktop and so far so good. I still only have one monitor plugged in. Next I'm going to plug in my VR headset and see if anything changes. Then, I'll run DCS and see what happens.

Why would installing the driver directly from the device manager, rather than through the radeon install program, which gives me the option to just install the driver make a difference?

Edited April 25, 2021 by Jarhead0331

[silverdevil]

@Jarhead0331 - the only defense i can provide is this is this is the way Windows would like the user to perform the driver update. Sometimes with manufacturer setup programs, there may be issues with the process. for example a DLL is locked and cannot be overwritten because the manufacturer code has no built-in logic. if one does it via the device manager, it takes that possibility into account. in my 30+ years of IT i have learned that things happen just because... i am guessing you are running the install program by right-click and run as admin?

[Jarhead0331]

The plot thickens...I had another crash and/or loss of signal not long after plugging in my VR headset. I don't know exactly what happened because I left the room briefly to take out the damn garbage and, of course, that is when the event occurred.  Regardless, I'm thinking one of two things at this point. It either is another bad GPU, or perhaps the VR headset adds to the power draw and there is an issue with the PSU?

 

As I said above, CLX is confident it is not the PSU, but I don't really know what to think at this point.

[Jarhead0331]

OK...I am now willing to bet that this is a VR related issue. There do not seem to be any problems until I plug in my headset and run Windows Mixed Reality. I do not know if it is a GPU or a PSU issue, but it has to be one of the two and it is triggered by the VR Headset and/or WMR.

[Leg2ion]

There are a few third party diagnostic tools that will monitor power output, etc. Might not be 100% accurate but might let you see any spikes when you plug in VR? How is the VR headset powered?

[Jarhead0331]

15 minutes ago, Leg2ion said:

There are a few third party diagnostic tools that will monitor power output, etc. Might not be 100% accurate but might let you see any spikes when you plug in VR? How is the VR headset powered?

 

It is a reverb g2, so it has AC power and plugs into the gpu with a display port and a usb-c. I’ve read about possible conflicts with reverb g2 and AMD MOBOs, so this could be causing the issue too, but the symptoms of that issue seem to be different from my own, so I still think it is gpu or psu related. 

Edited April 26, 2021 by Jarhead0331

[Leg2ion]

Real long shot - but are all the power connectors plugged into the Motherboard? So both the 24pin and 8 pin connectors are plugged in. And is your GPU powered properly? In that it has 2 x 8pin connectors. Is it powered by one power cable with a 'piggy back' 2nd connector, or 2 separate cables? If the 1st do you have a spare PSU cable available you can plug into the GPU. Is the card in right slot - PCI 1?

 

I see the PSU is the recommended 'minimum' for a 6900XT? Out of curiosity what make is the GPU.

 

Saw a post on the AMD forum whereby someone was running one of these with a underpowered PSU and had a lot of issues, but managed to address most of them by downgrading the PCI-e slot from 4.0 to 3.0 in the BIOS. 

 

One other thought, and please forgive me for this, but you are plugging into the GPU display port and not the motherboard display port? Again apologies.

 

And finally - have you tried checking the Windows advanced power settings - make sure they are at high, ultimate if an option.

Edited April 26, 2021 by Leg2ion

[Leg2ion]

Hi @Jarhead0331 - any progress?

[Pocket Kings]

PSU, pretty certain from what you've described and sorted out already. You don't happen to have a spare PSU to test? Also, try to flip the PC case 90° since these graphic cards weigh a ton. Then look at the 6 or 8 pin connectors and find out if they wobble.

 

How is the VR headset powered? Via USB only or dedicated power cord?

[Leg2ion]

On 4/26/2021 at 11:26 AM, Jarhead0331 said:

It is a reverb g2, so it has AC power and plugs into the gpu with a display port and a usb-c.

 

[Pocket Kings]

That means without the VR headset, the PC runs a prolonged stress test fine?

If the AC/DC Converter of the Reverb G2 has grounding issues, ground flows over the PC case, which is really bad. Just an idea.

[Jarhead0331]

Thanks for checking in Leg2ion. The retailer sent out a new GPU and since installing the new card, the system has been working. No green screens, no artifacts, no reboots or loss of signal. I’m not calling it a victory yet, but so far, so good. When replacing the old card with the new, I noticed the slot lock in the back of the pcie slot was missing. I don’t know if this could have been playing a role in the old card not being seated properly or not, but it is not impacting performance of the new card, which happpens to be noticeably smaller and lighter than the old. In any event, thought I’d mention it. 

Edited May 3, 2021 by Jarhead0331

[Leg2ion]

Good to hear @Jarhead0331. Must admit with the size of the new cards I found it a pain to seat mine properly - just built a new system and was only when I was exerting some pressure putting the power cables in that it actually clicked and locked in. 

 

Interesting difference regarding size/weight? Same card different manufacturer or different card?

 

 

[Jarhead0331]

Same card, different manufacturer. They are both 6900xt cards, but the old one was XFX and the new is ATI. I was surprised by the difference too. 

[Pocket Kings]

Yeah those three pounders. PCIe slots were never designed to hold so much weight. Cards often get micro cracks on the PCB when being shipped. Had many DOA cards due to rough handling. I've seen broken off capacitors, wobbling 6 Pin power connectors, broken fans, the whole nine yards. I've once heard from a retailer that up to 15% of cards are DOA depending on the PC case model. Some are too sturdy, some not sturdy at all. A shame for the industry imo.

[Jarhead0331]

2 hours ago, Pocket Kings said:

Yeah those three pounders. PCIe slots were never designed to hold so much weight. Cards often get micro cracks on the PCB when being shipped. Had many DOA cards due to rough handling. I've seen broken off capacitors, wobbling 6 Pin power connectors, broken fans, the whole nine yards. I've once heard from a retailer that up to 15% of cards are DOA depending on the PC case model. Some are too sturdy, some not sturdy at all. A shame for the industry imo.

 

Very interesting. Thanks for the insight. CLX tech support was telling me that it is surprisingly common for GPUs to be DOA. I was skeptical, but they seem to have been correct here...