Problem with NVIDIA DGX-1 has 8 GPUs installed but only 4 GPUs are active

Accelerated Computing DGX User Forum
1

I installed NVIDIA DGX-1 with all 8 GPUs but when tested only 4 GPUs worked. I tried changing the GPU location but the result is still the same. There are only 4 active GPUS slots 0,1,2,3. I tried many ways but none worked. I don’t know why, hope someone can help

screenshot_1695115776
screenshot_16951157761514×1096 174 KB

2

Sounds like either the GPU tray or the GPUs are bad. If you move them around and the PCIe bus addresses stay the same for the “working” GPUs, then you’ve got a bad GPU tray. If the bus addresses change, then I’d suspect you have some bad GPUs.

I assume you don’t have a DGX-1 support contract, since the DGX-1 systems are generally end of sale and end of support?

ScottE

3

Yes, I don’t have a DGX-1 support contract. I think it got a bad GPU tray. So what should I do, replace a new GPU tray or do something else?
Thanks for replying me.