Dark Hero + 3080 + 5950x = graphics crashes? I don't know what to check for!

[edplaysdrums]

Dark Hero, 5950x, 128gb 3200 Corsair, ROG Strix 850W, TUF OC 3080

I've been having this problem since first building this PC:

Whilst gaming the game will randomly freeze & crash to desktop. Very occasionally I'll get the same thing outside a game whilst working - desktop freezes for a couple of seconds, the output to my second screen will go dead (can reactivate it by replugging HDMI), and then everything carries on as normal.

For all these crashes I get the same report in Event Viewer: "Display driver nvlddmkm stopped responding and has successfully recovered."

The same thing happens with a 3080 or a 3090 so I'm pretty sure it's not a graphics card issue.
I've run a lot of memory tests so I'm also sure the ram is fine.
I've also done a lot of CPU/ram torture tests and never had any errors.

The one thing I've noticed is that a crash is more likely in a game that stresses the GPU: Cyberpunk is far more likely to crash than a 2d platformer, for example.

Could this be a power issue? A motherboard issue? I don't really know what to test for anymore, so I was hoping someone here would have some ideas...

Thanks!

[edplaysdrums]

HK-47 wrote:
Have you tried to run the PCIE slot the GPU is in at GEN 3 instead of auto? I have to run like this because the riser cable I have installed or i get the same kind of issues as you. That why I asked about the riser cable. Im wondering if maybe your CPU has issues with GEN 4 even though you aren't using a riser cable. Anyways something to try.

I did try that back when I first built the computer using the 3090 but I've changed cards and reinstalled windows since then so I'll definitely give it another try.

[edplaysdrums]

In the process of doing every possible test I could think of I ran memtest again overnight and got several errors so have RMA'd the ram and replaced it with some older 4x16gb@2400 sticks in the meantime.

Is it possible that memory errors could result in graphics driver crashes at higher power loads?

[Silent_Scone]

edplaysdrums wrote:
In the process of doing every possible test I could think of I ran memtest again overnight and got several errors so have RMA'd the ram and replaced it with some older 4x16gb@2400 sticks in the meantime.

Is it possible that memory errors could result in graphics driver crashes at higher power loads?

Hello,

Yes memory instability can quite easily cause driver crashes in this way.

There likely wasn't any reason to RMA the memory. If using DOCP/XMP then the applied overclock wasn't stable and the instability needs to be dialled out.

Once the new kit arrives, test the system at optimised defaults without an applied overclock and see if the issue persists.

[edplaysdrums]

Silent Scone@ROG wrote:
Hello,

Yes memory instability can quite easily cause driver crashes in this way.

There likely wasn't any reason to RMA the memory. If using DOCP/XMP then the applied overclock wasn't stable and the instability needs to be dialled out.

Once the new kit arrives, test the system at optimised defaults without an applied overclock and see if the issue persists.

Hi, thanks for the advice... forgive my lack of understanding but I thought the whole point of DOCP was that the ram had ideal overclock presets meaning people who don't know what they're doing (me) don't have to dial in any settings? I wouldn't really know where to start with tweaking from the DOCP presets, but I'll happily try and learn, if there's a good resource for this.

[Silent_Scone]

edplaysdrums wrote:
Hi, thanks for the advice... forgive my lack of understanding but I thought the whole point of DOCP was that the ram had ideal overclock presets meaning people who don't know what they're doing (me) don't have to dial in any settings? I wouldn't really know where to start with tweaking from the DOCP presets, but I'll happily try and learn, if there's a good resource for this.

XMP or DOCP are predefined frequency, timings and voltage denoted by the memory vendor to offer users the best possible chance of stability. Different memory vendors use varying levels of guardband, which refers to the amount of tolerance in place to account for varying strength and differences in CPUs. So whereas they do offer some reassurances, manual tuning can sometimes be needed depending on conditions at an electrical level.

On the platform in question, VSOC is the rail that impacts memory stability the most. You may find the replacement kit does in fact work, however this is more down to variance between kits.

[transdogmifier]

I'm inclined to believe that most of these issues would go away if people would just buy a better PSU...and not the 100 to 150 dollar special.

[edplaysdrums]

transdogmifier wrote:
I'm inclined to believe that most of these issues would go away if people would just buy a better PSU...and not the 100 to 150 dollar special.

Hi, thanks for your contribution. I've already replaced the Strix PSU with a Corsair HX850, which exceeds the power requirements of an RTX3080, and seems to be very well reviewed. Should I have bought a bigger PSU?

[neikosr0x]

transdogmifier wrote:
I'm inclined to believe that most of these issues would go away if people would just buy a better PSU...and not the 100 to 150 dollar special.

I have a Corsair RM1000X Gold and also having the same issue... Tried with 2 different PSU and the same happens.

thomas_yiu@ROG wrote:
Our AMD chipset has all the drivers need for the board. We also qualify the AMD chipset driver that we have.

Emmm, Are you suggesting that the chipset drivers from Asus website are a better choice than the ones at the AMD's website?

regards

[The_Tick]

I recently had some issues with games crashing. I was going to flash my Asus Tuf 3080 back to V2 from V4 but I decided to DDU my Nvidia drivers instead that resolved all the crashing. I spent 4 hours last night testing not a single hiccup while gaming. Try that if your jumping from a 3080 to a 3090 could be a driver issue for sure.

[Doco]

Ive recently been having similar issues as yourself but not completely the same- rhe main difference is I have not gotten - Display driver nvlddmkm stopped responding and has successfully recovered. Nor any apps crashing, however over the past 2 weeks I have been getting alot of games just crashing to desktop, games I've been playing for months suddenly crash to desktop no warning

To cut a long story short what I changed has given me 2 days straight issue free, I had enabled PBO from within the amd overclocking tab, with the limits set to motherboard- ai made no changes to the curve optimizer, I had previously had PBO enabled from I a different way - the tab that let's you enable the dynamic OC for the dark hero, I had forgotten I had changed where I had enabled PBO after a disassemble and reassemble for some water cooling maintenance

Just thought I'd post this on the off chance the issue you are experiencing can be solved in the same way