Question EVGA 3080ti FTW 3 is crashing in new build with a Ryzen 7800X3D, but not crashing in old intel i7 build ?

[rickyrich93]

The Problem: After 10-15 minutes under full 100% load the GPU crashes the game to desktop often displaying "Graphics Device has Crashed. This may be caused by out of date drivers or an overheating GPU. Please try a clean install of your GPU drivers and remove any overclock on your system. GPU Hung or device remove". I notice this crash happens after GPU is heatsoaked for a few minutes and reaches 80C -85C. HOWEVER as I will mention later, I don't think it's due to the GPU getting too hot. I will mention that if I immediately jump back into the game after a crash, the GPU will crash much more quickly (under 5 minutes).

The new build:

• AMD Ryzen 7 7800x3d in an Asrock X870e Nova Wifi motherboard (BIOS ver. 3.30)
• EVGA 3080ti FTW3
• Teamgroup T-Create 32gb 6000mt/s (2x 16gb kit)
• EVGA Supernova G6 850w 80+Gold
• New WD Black nvme 2tb
• Antec Flux case

Things I've attempted:

• DDU in safe mode and "clean" reinstallation of Nvidia Drivers (most recent drivers, studio drivers, and 566.36 [recommended in a Gamer's Nexus video]).
• I've tried Pigtail PCIE cables (1 8pin + 1 pigtailed 16pin) and three separate 8-pin PCIE cables to the PSU. An entirely new PSU (MSI MAG 750w 80+ Gold).
• Increasing Power limits in MSI afterburner.
• 2 fresh windows installs (not really a big problem since its a new build)
• Enabled Expo in BIOS to achieve 6000mt/s. Ram is in the correct a2 and b2 slots per manual
• The game I'm using as a test is God of War Ragnarok PC, it's also crashed in Horizon Forbidden West under similar conditions.
• The AMD asrock nova BIOS settings are a little intimidating and I'm not super familiar with all of AMD's lingo so maybe it's SMT or PBO related?

Why I don't think its a GPU temperature thing:

• I moved the GPU back into its old build to see if somehow I damaged the GPU when I took it out and moved it into the new build, and repeated the same 100% load test in GOW Ragnarok. In the old build, the GPU temperatures get even higher, 85-89C but it can literally run until for forever and never crash, whereas the GPU crashes in the new build occur much sooner and at lower temperatures! (88C in the old build and 82C in the new build).

Old build:

• Intel i7 9700k in a Z390 Gigabyte Aorus Pro Wifi motherboard
• EVGA 3080ti FTW3
• G Skill 32gb Ram 3200MT/s (4x 8gb kit)
• EVGA Supernova G6 850w 80+ gold.
• Intel 660 nvme
• NZXT H5 case

Context: The reason for a new build was because the GPU at full 100% was getting super loud and hot with it's fans spinning at roughly 90% (2500rpm) especially in the old NZXT case. What started as just me shopping for a new case ended up with me basically getting a new PC build put together because of some stellar Prime Day deals on amazon, newegg, and micro center. I'm sure there's some setting in the BIOS that I'm missing and I'm not very familiar with AMD processors and their settings. The new PC performs fine for basic computing but in gaming I'm truly perplexed as to what to do next and why the GPU won't crash in the older build but within 20 minutes of running in the new AMD build.

 

[Lutfij]

Welcome to the forums, newcomer!

EVGA Supernova G6 850w 80+ gold.
So you carried over the PSU from your older build. How old is the PSU?

EVGA 3080ti FTW3
The RTX 3000 series was notorious for having high transient load spikes. Users were advised to overprovision on their wattage for their PSU's when purchasing into a unit apart from buying into a reliably built unit. In your instance, as you have an RTX 3080Ti, you should be looking at a 1KW unit.

Source(borrow, not buy) a reliably built 1KW unit and see if the issue persists.

AMD Ryzen 7 7800x3d in an Asrock X870e Nova Wifi motherboard (BIOS ver. 3.30)
If you're not aware, ASRock was the motherboard you should've avoided when going for the Ryzen 9000 series. Might be a good idea to return the board and pick something up from Gigabyte or MSI.

 

[thestryker]

I'd suggest trying to drop PCIe to 3.0 first just to see what happens as that's a low complexity test and your old system only runs PCIe 3.0 whereas the new one will run the card's native PCIe 4.0.

Another simple thing to try would be setting your video card at 75% power. I actually run my 3080 12GB at that because EVGA's vBIOS defaults to 400W instead of SKU default 320W.

A couple of questions about the temps:

are you citing hotspot or core temperatures?

what is the memory temperature?

 

[rickyrich93]

Welcome to the forums, newcomer!

EVGA Supernova G6 850w 80+ gold.
So you carried over the PSU from your older build. How old is the PSU?

EVGA 3080ti FTW3
The RTX 3000 series was notorious for having high transient load spikes. Users were advised to overprovision on their wattage for their PSU's when purchasing into a unit apart from buying into a reliably built unit. In your instance, as you have an RTX 3080Ti, you should be looking at a 1KW unit.

Source(borrow, not buy) a reliably built 1KW unit and see if the issue persists.

AMD Ryzen 7 7800x3d in an Asrock X870e Nova Wifi motherboard (BIOS ver. 3.30)
If you're not aware, ASRock was the motherboard you should've avoided when going for the Ryzen 9000 series. Might be a good idea to return the board and pick something up from Gigabyte or MSI.

The PSU was bought in 2021 and still runs the 3080ti just fine in the old case with the intel i7 9700k. The temperatures are hotter in that old NZXT case but I just moved the GPU back in there to make sure that I didn't break in when installing it in the new AMD build and it still runs just fine without any crashes for hours. Which is why I really don't think a 1k unit is necessary especially since pcpartpicker suggests a 650 watt as a minimum, while Corsair suggests a 750w PSU as a minimum for a 3080ti, this is an 850w PSU which truly does feel like an overkill PSU for most situations.

As far as Asrock motherboards killing the 9000 series CPUs, yes, I did know that but Asrock allegedly did look into the issue and issue a new BIOS version that they claim fixes the issue, and also I have a 7000 series CPU. The deal for the motherboard and bang for buck was just too good to pass up IMO but now I'm not sure with these GPU issues!

 

[rickyrich93]

I'd suggest trying to drop PCIe to 3.0 first just to see what happens as that's a low complexity test and your old system only runs PCIe 3.0 whereas the new one will run the card's native PCIe 4.0.

Another simple thing to try would be setting your video card at 75% power. I actually run my 3080 12GB at that because EVGA's vBIOS defaults to 400W instead of SKU default 320W.

A couple of questions about the temps:

are you citing hotspot or core temperatures?

what is the memory temperature?

I'm genuinely confused where to go in the Asrock Bios settings to reduce the pcie slot from 4.0 to 3.0. I can't find that option anywhere

 

[thestryker]

I'm genuinely confused where to go in the Asrock Bios settings to reduce the pcie slot from 4.0 to 3.0. I can't find that option anywhere

According to the BIOS manual:
Advanced -> AMD Common Platform Module -> PCIe/GFX Lane Configuration -> then there should be something in there about Link Speed (there may be another option that's titled PCIe/GFX Lane Configuration inside the other one the manual isn't clear).

 

[Third-Eye]

The Problem: After 10-15 minutes under full 100% load the GPU crashes the game to desktop often displaying "Graphics Device has Crashed. This may be caused by out of date drivers or an overheating GPU. Please try a clean install of your GPU drivers and remove any overclock on your system. GPU Hung or device remove". I notice this crash happens after GPU is heatsoaked for a few minutes and reaches 80C -85C. HOWEVER as I will mention later, I don't think it's due to the GPU getting too hot. I will mention that if I immediately jump back into the game after a crash, the GPU will crash much more quickly (under 5 minutes).

Context: The reason for a new build was because the GPU at full 100% was getting super loud and hot with it's fans spinning at roughly 90% (2500rpm) especially in the old NZXT case. What started as just me shopping for a new case ended up with me basically getting a new PC build put together because of some stellar Prime Day deals on amazon, newegg, and micro center. I'm sure there's some setting in the BIOS that I'm missing and I'm not very familiar with AMD processors and their settings. The new PC performs fine for basic computing but in gaming I'm truly perplexed as to what to do next and why the GPU won't crash in the older build but within 20 minutes of running in the new AMD build.

I'm curious what your GPU and GPU Hot Spot temperature is at the same time when installed in the AMD system and running a game.

You can you get sensor log data or just a max temp reading from GPU-Z sensors tab for GPU, Hot Spot, and fan rpm/max speed. https://www.techpowerup.com/gpuz/

If you don't want to look through the sensor log and the PC hasn't shut down from testing a game, you can change the currently displayed sensor reading by clicking the title of each sensor or clicking the actual sensor reading of each one. It goes from Current, to Minimum, to Maximum, to Average, and then back to Current readings when clicked each time.

You can also check GPU-Z for the "limit reasons" labelled as PerfCap Reason in the sensors tab to find out why the GPU is limiting itself. Hovering the mouse cursor over each sensor reading will explain what the readings mean.

 

[rickyrich93]

I'm curious what your GPU and GPU Hot Spot temperature is at the same time when installed in the AMD system and running a game.

You can you get sensor log data or just a max temp reading from GPU-Z sensors tab for GPU, Hot Spot, and fan rpm/max speed. https://www.techpowerup.com/gpuz/

If you don't want to look through the sensor log and the PC hasn't shut down from testing a game, you can change the currently displayed sensor reading by clicking the title of each sensor or clicking the actual sensor reading of each one. It goes from Current, to Minimum, to Maximum, to Average, and then back to Current readings when clicked each time.

You can also check GPU-Z for the "limit reasons" labelled as PerfCap Reason in the sensors tab to find out why the GPU is limiting itself. Hovering the mouse cursor over each sensor reading will explain what the readings mean.

Accoriding to HWinfo64 the gpu temp during the last crash was at 80C and the VRAM temp was 86C max with an average of 84C. Limit was power according to GPU-z. My other thought was a PSU issue but the same PSU can run this GPU just fine when i had it in my i7 9700K build. I moved everything back into the old case last night to see if it was doing this, and under the same 100% gpu load test it could run for hours with no crashes at all. So think the PSU isn't a problem either. I did buy a new MSI 750W PSU that was supposed to go into my old intel build since I was taking the 850w PSU for the new one, but I tested the MSI psu in the new build and I was also able to replicate the issue.

 

[rickyrich93]

According to the BIOS manual:
Advanced -> AMD Common Platform Module -> PCIe/GFX Lane Configuration -> then there should be something in there about Link Speed (there may be another option that's titled PCIe/GFX Lane Configuration inside the other one the manual isn't clear).

I just lowered it to gen 4 and it still crashed 🙁 I'm about to try PCIE gen 3.

 

[rickyrich93]

are you citing hotspot or core temperatures?

what is the memory temperature?

I was using Core temps and the GPU memory is 84C with a maximum of 86C

 

[Third-Eye]

I was using Core temps and the GPU memory is 84C with a maximum of 86C

What is the Hot Spot temperature? That matters more with GPU temperature than just GPU alone. If Hot Spot is 15-20c higher or more than the GPU then that could indicate an overheating issue.

 

[rickyrich93]

What is the Hot Spot temperature? That matters more with GPU temperature than just GPU alone. If Hot Spot is 15-20C higher than the GPU then that could indicate an overheating issue.

According to hwinfo: GPU Hot spot was 86 with a max of 88C so definitely not drastically hotter than the 80C on the GPU chip

 

[Third-Eye]

Does the 9700K system have the same GPU driver version as the 7800x3d system? If not, I'd suggest trying 566.36, which is one of the last driver packages released in 2024, with the 3080 ti in the 7800x3d system.

https://www.nvidia.com/en-us/drivers/details/237719

 

[rickyrich93]

Does the 9700K system have the same GPU driver version as the 7800x3d system? If not, I'd suggest trying 566.36, which is one of the last driver packages released in 2024, with the 3080 ti in the 7800x3d system.

https://www.nvidia.com/en-us/drivers/details/237719

Yes that was one of the first things I said I attempted. I've tried studio drivers as well. The intel build has the latest everything: GPU drivers, Motherboard bios ver, XMP settings, and all the latest hardware drivers. For the new AMD build it started off with me updating all the drivers to the latest everything and then I began rolling back GPU drivers to earlier studio versions and 566.36 after I saw a Gamers Nexus video claiming that it was a stable driver.

 

[Third-Eye]

Yes that was one of the first things I said I attempted. I've tried studio drivers as well. The intel build has the latest everything: GPU drivers, Motherboard bios ver, XMP settings, and all the latest hardware drivers. For the new AMD build it started off with me updating all the drivers to the latest everything and then I began rolling back GPU drivers to earlier studio versions and 566.36 after I saw a Gamers Nexus video claiming that it was a stable driver.

Welp... that is on me for having missed that when reading everything. 😆

So you have tried increasing power limits, which is a weird thing to do if the driver is crashing. Have you tried decreasing the power limit to under 100% to say 90% or lower? See if lowering the power limit in 5% increments eventually results in the games and driver not crashing.

The 9700K system uses less power than the 7800x3d system, and although the PSU should be enough, it may be right on the edge of having or being an issue for stability.

Another thing to try is using the automatic default ram settings the motherboard sets before enabling EXPO, which would probably be 4800MT/s.

Power issues would normally result in a full system freeze/crash, much like unstable ram settings or faulty ram, so it's possible none of this will help.

 

[rickyrich93]

Welp... that is on me for having missed that when reading everything. 😆

So you have tried increasing power limits, which is a weird thing to do if the driver is crashing. Have you tried decreasing the power limit to under 100% to say 90% or lower? See if lowering the power limit in 5% increments eventually results in the games and driver not crashing.

The 9700K system uses less power than the 7800x3d system, and although the PSU should be enough, it may be right on the edge of having or being an issue for stability.

Another thing to try is using the automatic default ram settings the motherboard sets before enabling EXPO, which would probably be 4800MT/s.

Power issues would normally result in a full system freeze/crash, much like unstable ram settings or faulty ram, so it's possible none of this will help.

No worries, I will try default RAM speeds when I get home but if that does solve the problem, does that mean the ram kit is bad or the speeds are incompatible. They’re teamgroup t-create 32gb 6000 CL30 which are highly rated online

 

[rickyrich93]

According to the BIOS manual:
Advanced -> AMD Common Platform Module -> PCIe/GFX Lane Configuration -> then there should be something in there about Link Speed (there may be another option that's titled PCIe/GFX Lane Configuration inside the other one the manual isn't clear).

It’s crashing on both PCIE 3.0 and PCIE 4.0

 

[thestryker]

It’s crashing on both PCIE 3.0 and PCIE 4.0

Set your power limit to 75% or 80% and see if that changes anything. It's possible the new system is allowing your card to reach clocks which aren't stable. If this changes nothing then it's a safe bet the platform is the issue not the card.

The hotspot temperature is fine (only becomes concerning in the 15-20C variation range) as are the memory temperatures and those are usually the signs of repaste/pads being needed.