PC Graphics

[Venderhain]

Behavior

• Crashes when rendering 3-D graphics in games and Heaven Benchmark.
• At 1080p, games may run for a time before crashing.
• Frame rates vary wildly and seem low for this card
• At 4K, games crash almost immediately.
• Desktop and YouTube video render without a problem.
• When games crash, they don't crash completely. Example: Fortnite will crash, but  I can still hear the audio. I can even chat sometimes.

System

• CPU: Ryzen 7 5800X (Precision Boost Overdrive enabled, SAM enabled, PCIE set to Gen4)
• Motherboard: Asus ROG Strix x570-I Gaming (latest BIOS)
• RAM: 32GB Patriot Viper Elite II (rated at 4000mhz, running DOCP @ 3600)
• NVME: 2TB Samsung Evo 970
• PSU: Coolermaster V850 V2 Gold (No, my PCIE isn't daisy chained. 3 cables, 3 slots on each end)
• OS: Windows 11
• GPU Drivers Tested: 21.10.2 and 22.2.3 (all changes done with DDU and factory reset option)

Steps Taken

• Tried multiple drivers (DDUed all and chose factory reset option)
• Downgraded to Windows 10
• Auto undervolt of GPU
• Plugged directly into the wall
• Memtest for 1.5 hours (the only errors were on test 13 because it isn't ECC RAM)
• Stress Test within Adrenalin software shows no errors
• Tried with and without SAM, Precision Boost Overdrive, and DOCP
• Raised Power Limit in Adrenaline
• Set higher minimum frequency
• Windows Power Plan set to High Performance

Additional Notes

• CPU, RAM, and GPU temperatures seem fine. Honestly, the system doesn't seem to be working that hard.
• I'm in contact with Powercolor and Amazon and will probably RMA it, but I just want to see if anyone has any bright ideas that might help.

[ThunderBeaver]

Running the Power Color Demon RX 6900 XT.

I experience screen artifacts and horizontal tearing on all benchmark testers.

Gaming is very stable and stress tests work flawlessly.

Key thing to remember is this is a flagship GPU so a lot of games are not optimized for it yet.

Also I was running an EVGA 850 watt PSU 80+ gold and it just couldn't keep up with the power demand.

I upgraded to a 1200 watt Thermaltake ToughPower 80+ gold and the crashing stopped.

This is a base 300 watt GPU with capability of pulling over 450 watts.

No joke this monster GPU is a glutinous power monger. Treat her right and she will blow your mind. 

[Venderhain]

So you think it's power draw? Even with no overclock?

The only other PSU I have at hand is 750 watts.

If I run Adrenaline's stress test thing, I've seen it pull 330 watts without a system crash. It's only when I try to run a game or a benchmark that it crashes.

[ThunderBeaver]

It is possible games will cause higher than average power draw even unexpected short lived spikes usually because of normal background programs suddenly ramping up briefly and benchmarks are made to stress your build and test for weaknesses.

Does this problem occur with all games you own are just certain genre or brand specific titles? 

If your not experiencing thermal problems and your PC is running great outside of these to areas that may be a GPU hardware or software issue or it could be the games themselves are the problem.

There are a handful of games that I own that worked well when I was running 2 MSI RX 590s. Now with my single RX 6900 XT that has 3 times the processing power of the 2 previous GPUs combined will run them but I constantly get screen artifacts and horizontal tearing and they are all early access games. Most of my games like Star Trek Online Knights of the old Republic Tropico 4 and 5 Homeworld Remastered 1 and 2 Supreme Commander 1 and 2 Sins of a Solar Empire Rebellion Dawn of Man Planet Base and Warhammer 40K 1 and 2 run beautifully.     

All the visual glitchy games have 2 things in common they are all early access and they all show when I go to the local files that windows 10 is reporting them as an incompatible program.

I've tried running in a window and various compatibility modes but the problem persists and the game support sites treat the issue like a joke are just flat  out ignore me. 

If you have another GPU try swapping out and if you have another monitor/tv try connecting to those and see if the problem continues. If it was a GPU hardware/software issue then the problem would also occur in Boot Up or in Idle even in web browsing or streaming video.

If it is only happening in those 2 sources you mentioned that points the finger at those games and benchmark testers as the culprit.   

[Venderhain]

Two games I'm trying to solve right now: Fortnite and Elden Ring.

At 1080p, ultra settings, when it runs at all, Fortnite stutters like crazy with frame drops down to 14 fps. That's insanely bad.

At 4K, the game crashes before I even get in the lobby.

This was all tested on Direct X11, X12, and Performance Mode.

Elden Ring crashes on startup at 4K.

When I first got the card, neither of these games were that bad. Within a week, they hardly run and don't run.

When I run CPU benchmarks like OCCT and testers like Memtest, the other parts of my system run fine.

So the three possibilities are:

1. Power draw,

2. Drivers,

3. Manufacturing defect.

I hate to return the card, but I'm pretty sure I have to. First and foremost, I want to play games. Trying to debug hardware is interesting, but it's not what I signed up for. I don't mind tinkering and tweaking here and there to get a bit more out of an aging rig, but this is a brand new setup and the Powercolor is the most expensive part. I can't eat the cost of a potentially defective GPU.

Thank you for your time.

[ThunderBeaver]

Glad to help.

I've seen multiple reports over the past 2 months of massive FPS issues with Fortnite on multiple GPU's both AMD and Invidia and Intel based systems.

As for Elden Ring that is a new game and new games even on full release take on average up to a year to become fully optimized for all PC hardware/software.

If you are running a VPN while playing these games it can severely reduce your bandwidth by up to 90% which can lead to disconnects lockups crashes or just horrible FPS.

Also Fortnite could be experiencing some severe server issues either overloaded or dealing with malicious software attacks with the current world situation. Online games are prime targets because people often have debit/credit cards linked with their accounts and these are prime locations for ID theft money laundering and mass theft of individual players bank accounts. If this is the case Fortnite support would not admit it to avoid legal liability. 

[Venderhain]

I don't use a VPN. I saw someone else post that he's running into issues with Direct X games. I'm going to try installing Red Dead Redemption 2, which uses Vulkan, and see if that works.

[ThunderBeaver]

Good idea to test different base software and compare performance.

Most DX11 games for me work fine. DX9 games run visually great but I do get small intermittent audio cutouts on them.

I've tried disabling the AMD HD Audio Driver and using the MOBO Realtek onboard audio and tried an aftermarket PCIEx1 Creative Labs Audio card and the audio glitch persists.

The thing is I never had these problems on these games until windows 10 opened DX12 support. Even had the same issue on 2 previous GPU setups (2 RX 290s and followed by 2 RX 590s). I don't know if its a DX issue or windows issue. DXDIAG shows no issues. The only thing I can think of is that we are in an OS transitionary time (like windows 98 to XP) and that makes for a lot of programming dead ends and CPU and GPUs are growing in performance at such a fast pace that either OS development is having trouble keeping pace which we see evidence of in windows 10 horrible driver support or the OS developers are focusing their efforts away from individual customer support and just lining the pockets of their investors.  

[Geforsikus_2021]

Try to download latencymon, and run it in a simple PC (exactly when the PC is not in use), press start and wait at least 15 seconds and then press stop, and look at the delays, if they are very large, it will be written there what exactly causes them (at least it will be approximately clear which way to move further) There may be a problem in AMD sata controller (in device manager) if you have an AMD platform. I have an intel platform and an rx6000 card, all games even sooo old (such as: Nox (2000) or warcraft 2) work without any problems, the sound does not stutter.

[Venderhain]

UPDATE:

I had very limited time yesterday, so I didn't get to test a lot.

STEPS TAKEN:

(When I say crash below, I mean only that the program crash. The desktop recovers. My system does not fully restart, which is why I'm not sure on the "It's your Power Supply" argument. I assume, perhaps wrongly, that if it was a power issue, the entire computer would crash/reboot.)

1. Installed Red Dead Redemption 2. Ran its Benchmark. 

2. A new BIOS was released for my motherboard (X570). I flashed to that BIOS.

3. I reran the benchmark, and it completed to a good result. (This was at 1080p. I normally run at 4K and didn't have the time to retest that. If it runs at 4K without issue,  it might very well point to a driver issue, since Red Dead runs on Vulkan and the other games use DirectX.)

4. I retested Heaven Benchmark.

• 1080 ultra, DirectX11 = Crash
• 1080 ultra, OpenGL = Crash
• 1080 low, DirectX11 = Completed
• 1080 low, OpenGL = Complete
• 1080 medium, DirectX11 = Crash

I'm not sure what to make of all of this. Could it be a bad VRAM module?

TODAY's PLAN:

1. Test Red Dead Benchmark at 4K.
2. Run Rise of the Tomb Raider Benchmark at 1080. If it completes, try 4K.
3. If the above fails, wipe drivers. Remove the GPU and install my previous GPU in the system. Install drivers and run all the above tests. (If this succeeds, that narrows it down to either the Red Devil, the Drivers, or the PSU.)
4. If Step 3 fails,
   1. Switch to different Power Connectors on the PSU and retest. (If succeeds, it means the PSU has a bad power connector.)
   2. If fails, install RAM from previous computer and retest.  (If succeeds, it means bad RAM may be the issue. Reinstall Red Devil and retest.)

NOTES:

• I would be very surprised if it ended up being the RAM or CPU, as I've run Memtest and OCCT Benchmark before with decent results. I'm not ruling it out, but I think it unlikely.
• I would also be surprised if it was the motherboard, since I've never had a boot error. All drives and devices are listed in HardwareInfo and BIOS flashes work without problem.
• There is a Timer on all of these tests as Amazon requires me to return the GPU by April 1st in order to receive a refund.

QUESTIONS:

1. Could a bad VRAM module cause the behaviors I've listed? Or would the entire card fail if any part of the VRAM was bad?
2. For those of you who have had Power Supplies that weren't able to meet the power demands of your system, what behavior did you see? Did the system reboot entirely? Did individual apps just crash?
3. What if the Power Supply is fine, but the PCIE power connector on the GPU itself is faulty? Is there a way to determine the difference without another power supply handy?
4. What kind of fee would a place like Micro Center charge to test my system with a different power supply? Would it just be cheaper to buy another PSU? (I don't want to dump too much more money into troubleshooting this. I could just get another GPU.)

[ThunderBeaver]

The few times I've had a VRAM RAMDAC issue the AMD GPU software windows issue detection software or even the MOBO BIOS will report such critical hardware issues immediately. 

In the case of VRAM RAMDAC failures the GPU for me has always ramped fans to max on bootup but I would get a no signal or no video device detected in a BIOS message with no Windows access at all. AMD Adrenaline stress tester should tell you how much video memory is working properly.

Or just windows task manager under the performance tab select GPU and if it detects the full amount of VRAM (should be 16GB that's what mine is showing) then VRAM is fine.

Thermaltake and Corsair make some real good quality 1200 watt PSU for around $300. Even if it doesn't fix the issue the PSU upgrade will better support future upgrade options.

Outside of that you've done some really good testing. Your just facing a choice swap out GPU while still have the refund option. Upgrade PSU and see if that fixes the issue. Or you may end up having to do both even with a different GPU. 

Like I said my GPU crashes on all benchmark testers but performs beautifully on 95 to 98% of the apps I run. There are just a hand full of problematic ones. Check the benchmark software and see if it is optimized for Invidia Intel systems or AMD systems. I've found a lot of times in the past benchmark software tends to be very biased to specific gaming hardware and is often highly overrated.

[Geforsikus_2021]

It is possible that the problem is with some kind of video memory chip, it is possible that the problem is in the processor, there is also a memory controller there, of course, ideally check the video card on another PC to be sure that the problem is in it or something else, have you run the occt test on video memory? My games crashed at the time, but the stress test for RAM was passed, but I suspected that the problem was in RAM, and I raised the voltage manually a little bit, and the problem went away. In general, if the problem is really in the video card, then there should be crashes in all games without exception, there used to be problems with some power supplies from seasonic (with a sharp load, the unit cut down the PC (went into protection), but I think this is not your case) try the video card on another friend's PC, maybe in a computer club or store if there is such an opportunity, if the problem manifests itself in them, then feel free to hand it back under warranty. And run the stress test for video memory in occt, if you didn't do it. I have a 750 watt power supply unit, in fact, 500 watts would be enough for my system, i9 9900k, rx6900xt. Everything that the manufacturer of the video card indicates, he takes with a huge margin (suddenly you have an amd Thread reaper processor that itself consumes under 300 watts +.) At my peak, the system does not consume even 500 watts.

[Venderhain · Accepted solution]

FINAL UPDATE

First and foremost, I want to thank everyone who helped me troubleshoot this issue. You guys have been awesome: super polite, super informative. I can't drop a six-pack at your doors, but I'm definitely hoisting a beverage in your honor!

It's fixed!

The Solution

I was about to throw my previous graphics card (a Radeon VII) in the system when I decided to snoop a bit in AMD Adrenaline. Even though I never applied an overclock to the Red Devil Ultimate, it does have a factory overclock. I wondered, "Maybe the default setting just isn't stable?"

So I turned on manual tuning, lowered the max frequency, and raised the minimum frequency. I also raised the available power.

Voila!

I completed Heaven Benchmark at extreme settings without a problem at both 1080p and 4K. I also tested Rise of the Tomb Raider and Red Dead Redemption 2's benchmarks at 1080 and 4K. My jaw was clenched every time I ran a benchmark, but no crashes occurred. I almost feel like I have PTSD from this whole experience (I mean in no way to disrespect anyone suffering from real PTSD. I wish you strength and calm).

And I was able to play a full night of Fortnite with my 78-year old father without incident. It's his favorite game, and though it's not my favorite, I know I don't have much time left with him, so I'm keen to keep him happy.

Again, thank you, thank you for all your help. I've learned so much from you all. I'm marking this issue as resolved.

[DeadRedDevil]

Mate hi, I’ve just purchased the XT6700 variant and found the first few hours it worked great but then started crashing on the Home Screen (not even during a game),

short story, which frequencies, GPU or clock, did you adjust and any chance you recall to what values?

cheers if you get this.

[mate1950]

Are you fix that? I have have The same GPU and DirectX Error on MW2 Game? Can you Help me

[ThunderBeaver]

Ran Latencymon and everything showed optimal no issues found no bad kernel events.

Let it run for 15 minutes.

I think its just the fact that I'm running a PCIE 4.0 card in a 3.0 slot which would understandably cause some bottlenecking.

However I didn't start experiencing this audio glitch in a few games only until version 20H1 of Windows 10 Pro 64.

I use GPE to block non essential and unwanted windows bloatware spyware and from updating my system hardware.

Outside of those few games no issues at all. 

Now that I think about it those older games with the audio glitch had issues back in Windows XP days and were never fixed. I can play the original Neverwinter Nights Icewind Dale 2 Diablo 2 and the original StarCraft with no issues. Homeworld  1 and 2 Remastered play well to unless I try to play classic rather than remastered. FPS is great just get very short rare audio cutouts from time to time.

[Geforsikus_2021]

What it shows in the windows log(Computer Management=>Utilities =>Task Scheduler => windows logs=> applications, or the system.) at the time of the error when, for example, the game crashed? Try disabling Precision Boost Overdrive. And

[ThunderBeaver]

You make a good point with that log statement.

In all my builds I use a few manual settings in the BIOS but otherwise I let BIOS auto set anything to do with voltages unless auto set function goes out of safety ranges.

As for GPU settings I select standard in the AMD Adrenaline setup and the use across the board manual custom tuning when it comes to clocking and fan speeds. I let Adrenaline auto handle the voltages. So far in all my builds doing this I've only experienced 2 out of range GPU voltage incidents. One was due to a bad PCIE power cable and the other was just a bad GPU (One of the old X1900 series) back when AMD was going from GDDR3 to GDDR4 VRAM.