Unsolved
This post is more than 5 years old
7 Posts
0
1420
June 9th, 2018 01:00
Dell PowerEdge C6220 || installation problem GPGPU Tesla K20
I have a system based on the C6220 083n0 board for which I bought a riser from Z8220 on flexible cables and Tesla K20m. Additional power is connected to the riser. When the server starts, the resource initialization error occurs How to fix this error? Or does it not affect anything and just do not pay attention?
My second problem is in Ubuntu 16.04.04 x64 I'm having a problem installing the driver on the GPGPU Tesla K20. The driver installs its X-server nVidia and wants to switch the output to it, but I need a built-in video card - without it IPMI does not work and I need all 3 expansion slots - 2 for Tesla K20 and one for SSD m2.SWAP OS.
I understand that my equipment is obsolete. Which version of the driver is best for him? Which version CUDA + cuDNN?
falanger
7 Posts
0
June 9th, 2018 01:00
Maybe I should stop using Ubuntu on the server and use another Linux distribution? Arch or even install a virtual machine and already in it? Ubuntu works very badly in RAID-array disks. The desktop version I was not able to make it work at all, even install on a separate disk and use an array to store data.
falanger
7 Posts
0
June 13th, 2018 00:00
The problem with Tesla is that for your work, Navidia-smi requires the operation of an iVidia X server, which does not work when the built-in graphics are running. You can certainly put there through the corner riser low NVIDIA GT 610/710/730 - but then IPMI will not work.
falanger
7 Posts
0
June 14th, 2018 15:00
I was able almost by accident to find out the cause of the problem. On my C6220 does not work PCIe Edge slot.
Updated firmware BIOS did not help. In two conventional slots PCIe 16x, different video cards and Tesla K20m work fine. The problem is at the iron level of the PowerEdge C6220 II motherboard 0083N0.