PC Graphics

[MoAbe]

Hi folks,

Having a problem with my XFX Radeon RX 6800 graphics card. The card crashes on a certain game (see below) and specific benchmark tests (and combination), and I’m not sure if the problem is with the GPU itself or something else in my brand new tower.

Problem Description:

• Crashes: The system abruptly crashes, resulting in a blank screen. Sometimes, the fans blast off (and won’t stop till I hold down the power button), or the system becomes unresponsive and eventually restarts on its own.
• Games: Found out the issue in Guardians of the Galaxy (gotg) on Steam, though I haven’t extensively tested other games.
• Stress Tests: Despite the crashes during gaming, the system remains stable during some intense stress tests (details below).

Troubleshooting Efforts:

• BIOS Update: Updated the system BIOS to the latest version.
• PCIe Mode Adjustment: Set the GPU to work in PCIe 3.0 mode in the BIOS for more stability. So while it was running the GPU on gen 4, I’ve downgraded it in BIOS for hopefully more stability for now.
• Driver Reinstallation: Completely uninstalled and reinstalled AMD Radeon Software using DDU in Win safe mode.
• Windows Update: Paused to eliminate potential driver conflicts and windows forcing its own drivers.
• Stock Settings: Everything is running at stock speeds with no overclocks.
• Temperature Monitoring: No signs of overheating during crashes. Monitoring using HwInfo up to the moment of crash with 100ms time intervals. Nothing really jumps out at me.
• Logs: Checked Windows Event Logs, but found no helpful information. The logs register the crash but don’t point in any helpful direction.

Stress Tests Conducted:

1. Prime95 (CPU): Ran for 12 hours without issues.
2. Cinebench (CPU): the CPU multi or single testing runs fine. Either 10 minutes or 30 minutes. On its own (without other tests running simultaneously).
3. FurMark & Heaven (GPU): Ran simultaneously with no crashes over a 4-hour period.
4. Aida64, Prime95, FurMark, Heaven, Memtest_Vulkan (simultaneously): System became extremely slow (could move the cursor slowly and with delay) but did not crash.
5. FurMark, Heaven, Memtest_Vulkan, Cinebench Multicore (CPU): Crashed after about 10 minutes.
6. OCCT Vram Test (80% default): Crashed within 5-10 seconds.
7. Cinebench GPU Test: Usually crashes within a few seconds, or close to 10 minutes after it starts.

Additional Observations:

• SAM (Smart Access Memory): Disabling SAM allowed the game (GOTG) to run longer without crashing. I tried for up to 50 minutes. Could still crash if I let it run longer.
• Driver Versions: Using driver version 32.0.11029.1008 (Adrenalin 24.7.1) DCH / Win11 64-bit.
• XFX Support: Suggested trying an older driver version or contacting the reseller for a warranty claim. To quote them: “Please try those old version driver to see if it can help. If not, please contact your reseller for warranty claim.”
• Once after a Cinebench (GPU) crash and another time after a GOTG crash, GPU disappears from Device Manager. One time it only reappears after shutting down the system for 30 minutes to an hour. Another time, it only came back after I upgraded and flashed the BIOS. Just observations, could be a coincidence. This makes me think it’s more hardware related.

Questions and stuff:

1. Is it the GPU? Could the card itself be faulty? Should I just return it and get a new one?
2. Is it the PSU? Despite passing stress tests, could there still be a power issue causing these crashes?
3. Is it the Motherboard? Could there be a problem with the PCIe slot or something else on the motherboard?
4. Should I try anything else to rule out the PSU or motherboard before returning the GPU?

System Specs:

• CPU: AMD Ryzen™ 7 7800X3D Desktop Processor
• GPU: XFX SPEEDSTER SWFT 319 RAD RX6800
• Motherboard: MSI B650 GAMING PLUS WIFI
• Cooling: Cooler Master MasterLiquid 240L Core ARGB
• Case: NZXT H5 Flow - CC-H51FB-01
• Storage: Samsung 980 PRO M.2 NVMe SSD
• PSU: Corsair RM750e (2023)
• RAM: CORSAIR Vengeance DDR5 RAM 32GB (2x16GB) 6000MHz CL30 AMD Expo
• OS: Windows 11 64-bit Pro

Any ideas or suggestions would be greatly appreciated! Thanks in advance!

[eebiii]

I think you need a larger PSU, 850w or more and make sure you run dedicated lines to the card. Should solve all of your issues. I would not recommend running Furmark. It destroys GPU's. 

Ed

[MoAbe]

UPDATE (1 Oct 2024)

Hi everyone,

I wanted to update you all on the issue I had with my RX 6800 GPU. After experiencing persistent problems, I decided to return it to the Amazon reseller within the first 30 days. Fortunately, it was still within the no-questions-asked return period. The reseller confirmed that the GPU was defective, and I received a refund of €390, minus €10 for shipping costs.

During the return process, I decided to upgrade my GPU. I purchased the RX 7800 XT from Gigabyte for around €500. Although it was a bit more expensive, I am extremely happy with its performance. The RX 7800 XT is a more capable card with better ray tracing capabilities. It has been stable through extensive stress testing, although the fan can be a bit loud at times. Overall, the performance has been amazing, and I am very satisfied with my purchase.

I also want to clarify that I tried running the original RX 6800 GPU at a lower voltage and lower TDP, but it still crashed. There were many potential reasons for the issues, but in the end, I was just unlucky with a defective unit.

Thank you all for your responses and support. If anyone else encounters a similar issue, I recommend returning the GPU within the warranty period to get your money back. It’s important to act quickly if you experience problems within the first 30 days.

[eebiii]

Cool! glad it worked out.

I still think you need to upgrade PSU. You are pushing that 750w to its limits. IMO