D4 Error with GPU?
D4 Error with GPU?
I own a Threadripper system that I've been using for some time. It has a 3990x with 128GB of RAM and four 3090s. Everything ran perfectly until recently. My AIO CPU cooler was failing, but the system would just shut down whenever I started rendering. Eventually, I discovered the issue was the CPU reaching temperatures above 95 degrees. The system likely shut down to protect itself. I ordered a new AIO CPU cooler and two 4090s for an upgrade. The new AIO took three days after the GPU arrived before showing up, so I installed the GPU first and everything worked fine. After installing the CPU again, the BIOS reset, and restarting the system resulted in D4 errors. The screen goes black, and it can't even reach the BIOS to enter.
I tried enabling Above 4G Decoding but it didn’t help. I re-seated the CPU just in case, but that didn’t resolve anything. I replaced all other components except the CPU, but nothing worked until I put in three or four GPUs, after which I got a D4 error. I’m unsure what’s wrong and suspect the motherboard might be faulty.
Additionally, it isn’t the PSU issue—I have two running units. The main system uses a 2050W PSU, while the two GPUs run on a 1600W unit.
PCI resource allocation issue. Out of Resources.
TRX40 motherboards offer 56 PCIe lanes. GPU #1 = 16, GPU #2 = 16, GPU #3 = 8, GPU #4 = 8. Combined usage is 48 lanes. You are exceeding this limit.
This error appears consistently across multiple systems and all have four GPUs. It seems to stem from a Gigabyte BIOS configuration problem. This issue is not limited to the TRX40 board; it also occurs on workstation models.
I found an example where resolving the D4 error involved disconnecting one power supply unit, using a 3090 GPU instead.
Thanks, but I've been using two PSUs continuously without any problems. Running a system with two 4090 and two 3090 units on a single PSU would require around 1600 watts just for the GPUs.
I made changes to my post. It seems the issue is more related to using four GPUs. I couldn't locate a clear explanation, though it's not unusual for problems with four GPUs and Gigabyte boards. The PSU problem was just a coincidence. I'd be happy to assist further if needed.
I think a BIOS reset might be the solution. With four GPUs running continuously for two years, it’s possible something needs adjusting in the BIOS settings beyond just the decoding level.
I notice no other details beyond that setting. It might be helpful to ask on this site, as the US forums appear to be inactive.
Thank you, I have already sent a support ticket to Gigabyte, waiting a response. Will try the forums also. This is so frustrating because I had it all working, not sure what changed that. I want to buy 2x more 4090's to replace the 3090's but if it's not gonna work now.....
Sure, I'll report the problem and resolve it. I understand there was a solution since it functioned properly before I replaced the CPU and reset the BIOS.