PNY permanently denied my RTX 5070 RMA because it passed 3DMark
retained text
EDIT: Sorry, but this post went out faster than I thought it would. I want to change the title, but I couldn't. PNY stated that it passed Nvidia diagnostic and stress testing from Steel Nomad and Unigine Superposition. While I can't understand what "Nvidia diagnostic" included, since they mainly mentioned the two stress tests above, I've still put a lot of effort and money into testing the card in different setups as well, eliminating all possible variables, and the card still wouldn't stop artifacting in daily use. I’ve been dealing with a nightmare of an RMA process with PNY over the past two months for my RTX 5070 and honestly don't know what else to do at this point. The card keeps crashing with severe checkerboard artifacting, black screens, and hard BSOD reboots (0x133 DPC\_WATCHDOG\_VIOLATION). Every single time it dies, Windows Event Viewer logs the exact same nvlddmkm sequence: Event ID 13 (Graphics FECS Exception: UCODE Fatal Error), followed by Event ID 14 and dozens of Event ID 153 TDR resets. After a ton of troubleshooting, I narrowed it down to Windows 11 Hardware-Accelerated GPU Scheduling. Based on my research, when HAGS is turned on, GPU task scheduling is handed over to the on-die FECS microcontroller, and the card crashes constantly during normal desktop use or light-to-medium voltage transitions (like browsing with Chrome or waking from idle). When HAGS is off, it masks the issue because scheduling falls back to the CPU. Obviously, "just leave HAGS off" isn't a real fix on an RTX 50-series card since HAGS is literally required for DLSS Frame Generation to work. I paid out of pocket to ship the card back to PNY in August. Ten days later, they sent the exact same card right back to me claiming they couldn't reproduce the problem. I emailed them back with the raw Windows minidump files, event logs, and photos of the artifacting. I explained that the crash happens during dynamic desktop P-state transitions with HAGS enabled, not at 100% full load, and asked a simple question: did your test bench even have Windows 11 HAGS turned on? Their support rep (Don) literally replied: "We cannot confirm that our standard testing procedure specifically reproduces the exact HAGS-enabled, dual-display, DSC, desktop/DWM workload described in your message." So they basically admitted they have no idea if HAGS is even on during their testing, yet still refused to replace the card. To make sure I wasn't crazy and to kill any excuse that it was "my system," I swapped my CPU to a brand-new Ryzen 7 5800XT, swapped RAM kits, and even put the RTX 5070 into a second, completely separate PC with a different motherboard, CPU, PSU, and monitor. On the second PC, it immediately black-screened and threw the exact same Event ID 13 FECS Exception: UCODE Fatal Error at 100% factory stock clocks. I sent all of this cross-system proof to PNY. Instead of helping, a Technical Support Supervisor named Bruce P. stepped in and shut the whole thing down: "Your card was evaluated under RMA 526155 and passed NVIDIA diagnostics and extended testing. This included running Steel Nomad and Superposition simultaneously across two 4K displays for 24 hours without any issues. We acknowledge your reported results in another system. However, the reported errors do not, by themselves, establish a physical defect in the graphics card. Our evaluation found no hardware fault. PNY will not authorize another RMA, replacement, or advance shipment for this reported issue. Completing an online cross-ship form does not constitute approval. Our warranty evaluation for this issue is complete." Running Steel Nomad and Superposition for 24 hours just pins the GPU at 100% load in the P0 power state at max voltage. It completely bypasses the low-power P-state transitions and HAGS context switching where the card actually fails. Even after I proved the card throws hardware microcode errors and artifacts across two completely independent computers, PNY's supervisor just declared my warranty dead for this issue. Has anyone else dealt with PNY pulling this kind of move? Edit: There was one occurrence of a screen artifact while the system was booting up. So, clearly the card itself was faulty already, nothing to do with Windows nor the Nvidia driver since they aren't even loaded. But I forgot to take a picture.
document fetched 2026-11-11 00:36 · canonical https://reddit.com/r/nvidia/comments/1xsu222/pny_permanently_denied_my_rtx_5070_rma_because_it