NVIDIA Tesla A100 - GPU - Tensor Core - 40 GB compatibility

[DomZorro]

I’m looking for a way of getting into AI/ML (machine learning), LLM (Large Language Models) applications and testing and I'm planning to purchse one Nvidia Tesla A100. On my search to find the best possible, non-server chassis I have to run these testings in a HP Z840 (Serial Number: [personal info removed]).
Is this system compatible with the A100?
I’ll be running these AI applications under Linux.

[DGroves]

the nvidia a100 can draw up to 300 watts peak 

https://www.techpowerup.com/gpu-specs/a100-pcie-40-gb.c3623

while the z840 can run this card you need the optional 1250 watt power supply the stock 850 watt unit will not work due to how the power supply is made (it's a multirail unit not single rail design) google this if interested in details

next the a100 is a pci-e 4 card and the z840 is pci-3  so data xfers will be approx 30% slower

last this card being a server designed card will not be cooled properly in a z840 you would need to cobble/hack a cooling solution for it and seeing the cost of this card i would not do this if you do have this card look at several of the supermicro chassis and build your own or buy a used server

https://www.reddit.com/r/nvidia/comments/my15ao/a100_pcie_desktop_cooling_solution/

https://www.supermicro.com/en/products/chassis

https://www.reddit.com/r/nvidia/comments/my15ao/a100_pcie_desktop_cooling_solution/