Graphics Cards

[andrei93]

Hello.

  I own an RX 7800XT Nitro+ from Sapphire and it performs like a charm. The temperatures are low, the card is quiet and the performance is nearly double to what I used to get from my old Vega64. It's all well and good if left on its default settings, however, if I try to overclock, my PC will crash to a black screen and restart itself. I have stress tested my GPU in synthetic benchmarks such as Unigine Superposition and Heaven as well as games with built-in benchmarks such as Tomb Raider and Middle-earth shadow of Mordor with zero crashes after two hours. I even ran the Radeon settings stress test for an hour and everything was fine with zero artifacts, stutters or driver or system crashes. I have even tried undervolting by .050mv which led to more stable core clocks but still crashed at the same rate.

  It seems that placing the GPU in any other mode other than the default one will cause my PC to crash. The crashes are completely random and they occur more often outside of games than during gaming. Anything from watching a video on YouTube to just opening apps or browsing folders can cause this crash and I am unable to replicate it after the restart by doing the same thing that caused it to crash. 

  I have noticed that even just putting the GPU in manual mode, within Radeon settings, without touching any of the clocks, voltages or even fan curves will cause my PC to randomly crash. Leaving it on default settings has never led to such a crash.

  If this is a PSU issue not dealing well with a power spike, why does it lead to a crash if I simply enable manual tuning and not changing anything else and doesn't crash if I leave it on default settings?

  My system specs are as follows:

Mobo: Asus Crosshair IV Hero with bios 8601

CPU: R7 5800X3D water cooled

RAM: 4x G.Skill FlareX 3200 CL14 for a total of 32GB

PSU: Corsair HX850i 

GPU: As stated above with driver version 24.4.1 (previous versions had the same issue)

  I have even done a fresh install of Windows 11 as I thought Windows 10 was the problem but it still persists. Why?

[432hz]

A few things come to mind:

PSU:
Ensure your GPU is connected with separate PSU cables, not daisy-chained. Your PSU has single-rail and multi-rail modes, you'll want it to be in single-rail AFAIK.

As you mentioned, the issue at least at this point, appears to be with the Adrenalin software itself, and not necessarily a lack of power from the PSU.

Max Frequency Inconsistency:
Some have noticed when enabling Manual Tuning, the GPU Max Frequency slider is set significantly higher than the card's listed max boost frequency (see your card's specification page), which leads to various driver timeouts and crashes. The solution is finding the max boost frequency from your card's specs and setting the GPU Max Frequency to that number or below. Try the same and see if you can achieve stability in Manual mode before going for the overclock.

Once you're stable in manual mode, the overclock will consist of 2 parts: 1) increase the Power Limit slider and 2) increase the GPU Max Frequency. There are guides out there on this.

Drivers:

The 24.x.x drivers have had many stability issue reports here. 23.12.1 and 23.11.x seem to be the "most stable". You may want to try DDU / AMD Cleanup Utility to remove your current drivers and install one of these older driver versions using the "Factory Reset" option on the installer.

[andrei93]

  The GPU is using separate cables as those cables have been in place since I had a Vega64.

  Max Core frequency, listed when I enable Custom mode, is 2560MHz. That's 5MHz lower than the maximum boost frequency listed on the box and on the spec sheets. I tried lowering it to 2254MHz which is the maximum game clock listed on Sapphire's web page and it still crashed while opening Discor with a youtube video playing (could not replicate it a second time after restart).

  The only overclock I've tried applying was a memory overclock from 2438MHz to 2614MHz which was perfectly stable (all things considered) without any artefacts. I even bumped up the power limit by 15% and undervolted the core by .050mV. I've set a somewhat aggresive fan curve of 40% at 55C and 100% at 80C with linear increments in between. 

  I have also tested 23.12.1 today with the same result, only, it seemed like my GPU power consumption was around 5% higher with that driver.

  All driver updates are performed using DDU and having my PC disconnected from the internet during the new install.


  As a couple of side notes, web browsing crashes are almost non-existant if I have an overclock applied and I disable GPU acceleration in the browser settings. Also, could it be my motherboard being too old to properly deliver steady power through the PCIe slot? My motherboard aslo had booting problems after upgrading from a R7 1700 to the R7 5800X3D. The motherboard now, sometimes, gets stuck at startup on code 45 or code 15 with very rare instances of code B0.

  Also worth mentioning, the PSU is plugged in stratight into the wall and not in an extension leed.

[432hz]

There have been numerous reports here related to issues with graphics hardware acceleration in the browser. IIRC, the remedy was disabling graphics hardware acceleration in the browser OR installing Drivers-Only version of the GPU drivers OR installing 23.12.x / 23.11.x.

A few other things come to mind:

• For the mobo issues: are you on the latest non-beta version of BIOS for your mobo and have you performed a CMOS reset? If not, consider updating. If so, consider a re-flash.
  • What are the error codes listed related to?
• Have you installed latest chipset drivers from AMD?
• Is your PSU's 12V rail output sufficient for your components?
• Is your mobo's power delivery / VRMs sufficient? Might be worth an email to the manufacturer.
• Disable XMP / EXPO and retest.
• Verify your RAM is good with MemTest86.
• Try a Drivers-Only install of your choice 24.4.1 / 23.12.1 etc. and try the overclock with MSI Afterburner. If it works, it's an Adrenalin issue. If not, it may be hardware related; maybe PSU or mobo.
• Have you checked Windows Event Viewer for info when the crashes happen?

[andrei93]

  So I have performed a CMOS reset and reflashed the bios. I am on BIOS version 8601 as 8701 and 8702 have some security mitigations that I worry will lower performance. Version 8601 is not a beta, though

  I have installed the latest chipset driver when I did a fresh Windows install about a month ago.

  The PSU's 12V rail is rated at 850W with 70.8 Amps

  I doubt Asus will respond to such an email as the mobo has been discontinued long ago. It is the Asus Crosshair VI Hero X370

  I don't know what code 45 means on the mobo as it is not in the manual. B0, however, relates to either bad ram, bad seating of ram, a broken ram socket or poor CPU seating. I will have to try MemTest as soon as I purchase a USB stick.

  I tried overclocking with MSI Afterburner and so far, the system went a whole day without a crash. I didn't do this very methodically as I also reflashed the BIOS of the motherboard so it's hard to say which did it. If it crashes again, I'll be back with an update.

  Until then, thank you for your help.

[432hz]

I see BIOS 8601 is from 2022 and 8702 does mention some system stability improvements. They very well could've improved stability as it relates to your issue, but I understand not wanting to update.

Your PSU's 12V rail seems strong enough.

Bad RAM can cause all kinds of issues. Probably best to start there. Also, if your CPU cooler is attached too tight, it could also cause RAM stability issues. This can happen from simply over-tightening or if using a cooler that isn't specifically compatible with your socket / CPU (which can more easily lead to over-tightening due to height differences).

Glad to hear Afterburner or the BIOS reflash seems to have provided stability (at least for now). Do let us know how it goes.