[SOLVED] ZOTAC GAMING GEFORCE RTX 3090 AMP CORE HOLO 24GB GDDR6X constantly crashing while gaming.

[Cabeza2000]

Hi,

A few weeks ago I bought a prebuilt coming with a ZOTAC GAMING GEFORCE RTX 3090 AMP CORE HOLO 24GB GDDR6X and I am experiencing crashing (CTD, not BSOD or PC reset) usually around 15 to 30 mins of gaming, sometimes earlier. The longest session I could get without a crash is one hour.
The games I have been playing are Watchdogs Legion and CyberPunk 2077, with maxed settings. Both games are highly demanding on the GPU but also unoptimized. I play at 1440p.

These are the specs:

PC Case: XPG BATTLECRUISER ARGB BLACK
CPU: INTEL CORE i9 11900F
Motherboard: GIGABYTE Z590 GAMING X
RAM: 2 x KINGSTON HYPERX FURY DDR4 32gb
SSD: 2 x WESTERN DIGITAL SN550 NVMe 1TB
HDD: Western Digital 4TB WD Red Plus
Graphics Card: ZOTAC GAMING GEFORCE RTX 3090 AMP CORE HOLO 24GB GDDR6X
LIQUID COOLING: DEEPCOOL GAMMAXX L240 V2
PSU: GIGABYTE P1000GM 1000 WATTS
OS: Windows 10 Pro 20H2
Monitor: Samsung Odyssey G75

Background:

I am not really into overclocking or hardware. I just change my whole rig every 6 years and do not touch it unless something gets broken. So I did not think too much into getting this prebuilt with a RTX 3090, especially as I am coming from a 2015 custom build with a GTX 980.

So I just went for what I thought was one of the best cards available and just paid the hefty premium for it. I just did not expect this thing to be this unstable and right now I am regretting I did not get something less powerful but more stable.

This card have these key specs:

Interface: PCI Express 4.0
Boost Clock: 1755 MHz
Effective Memory Clock: 19.5 Gbps
Thermal Design Power: 370W
Recommended PSU Wattage: 800W


Initially I kept everything in stock settings. Only started to get into tweaking and tuning things when I noticed the problem.

Temperature: While gaming under max settings the card will go around 77c to 79c, sometimes hitting 80c or 81c. This is below the target temp of 83c. My fan setup is probably not the best (is what came with the prebuilt) and there is no waterblock on the GPU. I know cooling can be improved but so far I am not sure high temp is what is causing the crashing, especially not when I am into the specs and people buy this card for crypto mining and run it at 90c or even 100c.

Wattage: Usually around 360+ while playing the aforementioned games.

GPU Utilization: Never saw it above 97%.

Voltage: Usually at around 1 Volt or 1.1 Volt (always while gaming at max settings)).

Used FireStorm first but left it, that thing is only good for the RGB setup. Then MSI AfterBurner (set up a custom fan curve). Then tried the new Nvidia experimental feature automatic fine tuning (which actually increased the GPU core clock by 85 Mhz).

I am using the latest Nvidia Driver (471.96)

After crashing sometimes Event Viewer shows that the driver nvlddmkm stopped working and recovered, with Event ID



Things that I already did or tried:

1. Underclock both Core Clock (-300 Mhz) and Memory clock (-200 Mhz) which resulted in 2c less temperature and 15 or 20 Watts less GPU Power but still crashing.
2. Ran SFC and DISM (no problems found).
3. Changed the Power Management Mode to "Prefer Max Performance" at the Nvidia Settings (it did nothing but I kept it that way.
4. Windows is set to never sleep or shut off the display.
5. Ran 3Dmark tests "Time Spy" and "Port Royal" and all good there. Will probably get the stress test package to run it for longer and see how it goes.

Things that I still did not try:

1. Updating the BIOS.
2. Updating the card firmware.
3. Lowering max settings (this defeats the whole purpose of having this card).
4. Turning off XMP.
5. Undervolting (no idea how to do this).
6. Did not try other high demanding games that are less buggy (I am thinking Control, Death Stranding and Anno 1800). This is important to see if the problem is the card or the games that are unoptimized for the card.
7. Did not run memtest86.
8. Did not update the monitor driver (running as Generic right now).
9. Did not care to get the Resizable BAR which my system supports (this is a long shot but perhaps it may help reduce crashing?)

There is still much I can try, what I am looking for from you is some guidance on what I should try first, whenever you see any incompatibility in my rig or if you think the card could be faulty, which I guess may be the case (hopefully not).

As you can see, I just need some direction. I do not want to overreach, I just want to get some stability and would love to get some feedback on where I should start or what are the experiences of other users with the same problems.

Thanks for helping!

EDIT1: Ran 3Dmark tests "Time Spy" and "Port Royal" in stress mode (20 times loop each) and results are OK. The system ran at 81c temp and 370 wattage the whole time for one hour, no crashes. Next step will probably try some more stable game to see how it goes and discard the possibility of the card being faulty.

EDIT2: Just played two hours of Control, not problems at all and the game even ran at a higher temp (81c) and wattage (365 to 370). So guess this is enough prove that the card is not necessarily faulty and the issue must be something about Cyberpunk 2077 and Watchdogs Legion and the RTX 3090 series. Will post in those games forums to see if anyone else experienced issues.

EDIT3: By reducing the Power Limit with MSI Afterburner I stopped the crashing for Cyberpunk 2077 (still need to check Watchdogs Legion).

EDIT4: I was just able to play CP2077 for three hours without any issue after just disabling Nvidia Broadcast feature in the GeForce Experience overlay. All this with stock settings, no MSI AB or Firestorm enabled. Before this I was never able to play for more than one hour except by lowering the Power Limit with MSI AB. Will keep trying and see if the problem come back. Still need to check Watchdogs Legion.

 

[rgd1101]

test with a different psu? gigabyte psu isn't the best

 

[Lutfij]

GIGABYTE P1000GM 1000 WATTS
Yeah, that is something I wouldn't call a PSU. Nothing from Gigabyte. I would recommend that you use a Seasonic unit if at all possible since they are built better than pretty much all units out there.

ZOTAC GAMING GEFORCE RTX 3090 AMP CORE HOLO
The AMP's from Zotac were bad in terms of cooling since their cooling solutions were heavy/clunky and caused imperfect mating between the cooling plate and the surface of the GPU.

OS: Windows 10 Pro 20H2
You should ideally be on version 21H1.

As for your build, that 240mm AIO isn't enough for that i9.

 

[Phaaze88]

3rd. Psu is poop.

 

[Cabeza2000]

GIGABYTE P1000GM 1000 WATTS
Yeah, that is something I wouldn't call a PSU. Nothing from Gigabyte. I would recommend that you use a Seasonic unit if at all possible since they are built better than pretty much all units out there.

ZOTAC GAMING GEFORCE RTX 3090 AMP CORE HOLO
The AMP's from Zotac were bad in terms of cooling since their cooling solutions were heavy/clunky and caused imperfect mating between the cooling plate and the surface of the GPU.

OS: Windows 10 Pro 20H2
You should ideally be on version 21H1.

As for your build, that 240mm AIO isn't enough for that i9.

Thanks for your answers.

I do not have another PSU, before I change it I would like to make sure this is indeed the underlying cause. I do not want to end changing the PSU if it was not the problem. I mean, it could be, but I am not getting system shutdowns.

Updated to Windows 10 Pro 20H2, but still have the same issue.

 

[DSzymborski]

Thanks for your answers.

I do not have another PSU, before I change it I would like to make sure this is indeed the underlying cause. I do not want to end changing the PSU if it was not the problem. I mean, it could be, but I am not getting system shutdowns.

Updated to Windows 10 Pro 20H2, but still have the same issue.

Unfortunately, system repair doesn't work like this. You don't get to know what's definitely the problem before swapping out a part.

In this case, it's a no-brainer because you want to replace this PSU even if it is not the cause. It's a junky PSU series that Gigabyte threw together cheaply with a manufacturing partner with little experience making consumer PSUs. It's not just widely known as junky, it's become notorious for destroying 30 series GPUs.

View: https://www.youtube.com/watch?v=aACtT_rzToI&t=4s

 

[Cabeza2000]

Unfortunately, system repair doesn't work like this. You don't get to know what's definitely the problem before swapping out a part.

In this case, it's a no-brainer because you want to replace this PSU even if it is not the cause. It's a junky PSU series that Gigabyte threw together cheaply with a manufacturing partner with little experience making consumer PSUs. It's not just widely known as junky, it's become notorious for destroying 30 series GPUs.

View: https://www.youtube.com/watch?v=aACtT_rzToI&t=4s

Thanks for your advice and the video, I watched it and this is definitely concerning... Even if my PSU is a P1000GM and not the P750GM and P850GM... But I guess it has the same design flaws.

BTW, I just added an EDIT to my original post. I tried playing with another game that is more stable (Control) and had no issues with it.... I hope I can find a solution for my issues with Cyberpunk 2077 and Watchdogs Legion.