This build out is for a general purpose ML server. There are a few upgrades coming once PCPP price alerts get triggered.
The biggest limitations right now is the GPU. However for kaggle contest and entry level work it gets the job done. I'm waiting until the 30xx series comes up before upgrading to a ti.
Other limitations is the PSU and SSD space. Planned upgrades to an EVGA 1300w PSU to provide headroom for multi-gpu's. As well 2 1TB QVO SSD's in a RAID 0 to provide space for datasets that wont fit on the 1TB NVMe drive.
The OS loaded into the computer is Ubuntu Server 18.04.
CPU stays turbo-boosted @ 4.2Ghz even under full 20 thread load thanks to the AiO.