GPU received a return merchandise authorization, same FurMark issue with updated GPU model.
GPU received a return merchandise authorization, same FurMark issue with updated GPU model.
NVIDIA 3090Ti kept failing, could get a replacement directly from NVIDIA, but still crashes during FurMark. Everyone, unsure what to do now. Story: During an Unreal class, my screen went black completely. I managed to chat with students for about a minute before the machine crashed again that night. I continued teaching later using notes and memory (students were understanding and asked lots of questions). Once Unreal was out of the picture, no more issues. I searched online, tried different setups: 3DMark, PCMark, OCCT, FurMark all showed crashes. After moving the GPU to another PCIe slot, the crashes stopped. Then I removed it and tested with integrated graphics—no crashes! I contacted NVIDIA, submitted reports and proof, and they approved swapping the card (bought new from them in 3 years). Today I got a refurbished one. Installed it back in its original slot, ran FurMark, still got the same crash. Now I’m trying to figure out what’s causing it. Thanks. Specs: MSI PRO Z690-A WIFI DDR5 LGA 1700, Intel Z690, SATA 6Gb/s ATX, motherboard DDR5 4800MHz Corsair RMX (2021), RM1000x, 1000W i9-12900K.
EvGA advises against using this software, calling it a GPU killer. It can harm hardware by applying excessive stress, which may not pass standard tests.
Thanks for sharing your experience! It seems you haven't encountered any negative aspects of FurMark. The NVIDIA team suggested using it to identify the problem, and it worked perfectly. You mentioned you were able to replicate and fix the issue with vanilla Unreal Engine 5.4.4, which was a relief. It appears the fix might have come from something unexpected—like the 16 Pin Power Adapter swap included in the RMA package. Running FurMark with the original cable caused crashes, but after replacing the adapter, the problem disappeared. Could this adapter have been the root cause? What alternative tests could you have tried to support NVIDIA in their decision to exchange your GPU? Your insights are really helpful!
It's questionable to hold FurMark responsible for discovering issues with your graphics card. You might argue against playing New World at its launch since it also negatively impacted GPUs. However, the reality is that it simply identified pre-existing problems in graphics cards.
Hi there, I faced a similar problem with a 3080ti. It ran smoothly under normal Windows use and most games, but after about 30 seconds of Furmark it would freeze completely, showing a black screen before crashing. When I tried it on another machine it worked perfectly. After installing a different PSU in the existing setup, the issue returned. Upon inspecting, the problem was with the power cable—its PCB melted inside the PSU. Replacing the PSU resolved the issue. As you noted in a previous post, adjusting the cable seemed to help. I’d have secured it directly on the PSU itself so it wouldn’t look like the fix was just from the cable.
This situation clearly causes the system to fail under stress, forcing the card to exceed its capacity. Since this isn't meant for everyday use, it's a performance test. Just like New World addressed the problem, Furmark must resolve the issue here. Because you mentioned it wasn't the cards themselves but the software that failed.