Jan 21
Configure Nvidia M40/K80 on your Home Computer
Solve D4 - PCI resource allocation error. Out of Resources
New secondhand video card available because not for miner! We can find with good price some economical GPU like Nvidia M40 or Nvidia K80, that are very good for Deep Learning but also Machine learning.
That GPU are not good for Mining but can be used for gaming by using a CPU integrated GPU and some configuration (like explained there).
I find a Nvidia M40, 24GB version with 300 EUR that is a very good price so I buy it and I add on my 1080 workstation.
Problem
Connect the Graphic Card with the CPU Cable like said in the M40 manual, but the herror heppen: D4 — PCI resource allocation error. Out of Resources.
I describe the problem in that thread.
This problem can happen also with a K80 or if you and two or more GPU and go over the 24 GB.
Solution
Changes in the Bios:
- Above 4G decoding : Enable
- CSM support : Enable
- Fast Boot : Disable
In particular the Above 4G decoding for my Asus Rog Maximum X Hero (with bios of 2017) that setting can be find into the boot option of the bios (not pci as in the recent bios).
Other guides:Aorus Guide , Gigabyte Guide , new Rog Guide
Other minor problems
The first problems is that the screw do not fit the normal case, so you can put your GPU into an horizontal case or use something to hold up it.
Another problem could be the dissipation because it is fanless, but you can find online to many solution: 3D printing or others.
Conclusion
This GPUs are not competible with the 30xx but they are very powerfull and economical to buy so for Deep Learning or Gaming can be an economical but powerfull choice!
Train your deep models and share it!
