The EVGA RTX 2080 Ti SC experienced a sudden failure (BSOD).
The EVGA RTX 2080 Ti SC experienced a sudden failure (BSOD).
Hello people, It seems I've stumbled into some unknown issue with my graphics card that is very difficult to diagnose. I've had an EVGA RTX 2080 Ti SC since March 2020, and it has worked flawlessly up until now. Two days ago i got a sudden crash playing Apex Legends, which has since then continued to crash my PC in multiple scenarios. What I've experienced 1. PC does not crash in windows, or in any non-game applications. 2. PC crashes during gameplay or while starting a game, i have tested many different games and the same goes for all of them. 3. Game crashes during gameplay, but PC does not crash. For example Apex Legends has given 2 similar errors: DXGI_DEVICE_ERROR_RESET and DXGI_DEVICE_ERROR_HUNG. 4. PC will crash during regular use, however, ONLY after i have experienced a BSOD crash from any game. It's like there are issues that reverberate after a BSOD crash that affects everything, even regular use. (only happens irregularly) What I've tested 1. different displayport cable 2. different displayport port 3. different pci-e slot 4. different pci-e cable 5. DDU driver uninstall and reinstall 6. Nvidia debug mode 7. Plug in gpu into entirely different PC (same crash happens) 8. Use my old GTX 980 (this works, no crashes at all) There are no thermal issues at all with the card as far as i can tell. The latest driver 537.42 was released 2023-09-21, which is more than 10 days before the crash. I can't wrap my head around why this would happen all of a sudden, I'd love to get some insight or feedback on how to further diagnose this. Unfortunately the card is 7 months past it's 3 year warranty period. I have spoken to EVGA and they can't provide me an RMA case for this.
it seems like the image conveys the strongest message... it resembles memory remnants. you could forward it to krisfix-germany, though he might not accept 2080ti anymore. fixing it could cost more than the benefits of a newer card. still, i bought my 2080ti for 16k nkr pr cards, and a 4090 would need at least 24k.
your cards come with a four-year factory guarantee... but it seems that might be expired already. did you purchase your 2080ti shortly before the 30 series release? in norway we actually have a five-year consumer warranty. some electronics are designed to fail early, so repairs can become costly. don’t let the price of the 4090s fool you—when issues arise, replacing the whole system is often necessary.
I ran additional tests using 3D Mark benchmarks. The card finished the Port Royal benchmark without any crashes, so I moved on to Firestrike Extreme. It stopped working after just two minutes. I’ve attached the relevant logs. There don’t appear to be unusual temperature spikes, power usage, or speed issues. It looks like the GPU is struggling under sustained heavy load. I’ve had the PC streaming YouTube for four hours without any crashes or visual problems. The files are GPUZ Crash Log.pdf and GPUZ Sensor Log.xlsx.
The card functions properly, but the second fan cord is damaged. It works perfectly when connected to an external fan for cooling that area.
During extensive testing at 3000rpm with the first fan and second chassis fan, temperatures stayed under control—peak GPU reached 65.8°C and the hottest spot hit 80.2°C. Using default factory options keeps things stable. If I increase power to 130%, the hotspot can rise up to 95°C, but once the water-block is installed, it should handle that issue.
I own two 2080Ti graphics with waterblock from the factory. I swapped out thermal pads and paste last month, but more than expected was needed. One card uses cheaper pads than the other—those are three times thicker. On regular and hotspot, my top temperature reaches 88°C at 110% power, which feels like a service is necessary. Make sure you get the correct thickness of the thermal pads. Check a test fit to see if the paste spreads evenly on the CPU; if not, it might be too thick. A pasta roller can help thin them down or adjust as needed.