Just completed this build. This rig is not for gaming but for software development. Dual boot Windows/Linux. Got CPU from https://siliconlottery.com, it's bined to 4.4 GHz and works very well. Reservoir on the right and four fans on the bottom is just for show, otherwise the case looks pretty empty. Case is Lian Li V3000. I really like it both in term of functionality as well as quality.
For comments that reservoir for show is lame have no fear. This build will be converted to open loop very soon, with the same reservoir and fans/radiator at the bottom used for their proper functions. Fun continues.
Update as of June 10.
It turns out that having GPUs air cooled was a problem. Many of my machine learning tasks run non stop for days and sometimes 2-3 weeks. During this time both CPU and especially GPUs are maxed out. GPUs were reaching thermal threshold and were throttling. This added several days for ML tasks to complete. I'm in a process of putting water cooling loop, should be completed soon. There will be changes to some components as well. Stay tuned.
Update as of July 11
Watercooling loop completed. Will post pictures soon.
Also I want to make general comment about the cost of such a system. First keep in mind that I make a living working on various systems from AWS to local machines that I build. If algorithm I'm working on needs SSD with best random access I will get Optane. If my Machine Learning task takes a week to complete I will get Titan RTXs. Time is money, it's as simple as that. It's also useful to compare the cost of this build to something you can buy. I think the most useful comparison is with Nvidia DGX Station, which is a workstation build primarily for Machine Learning. That system has four Titan V cards. I also have systems with these cards and they are noticeably slower than Titan RTX. I estimate that Nvidia workstation is about 50% faster than my build. And you can get it for measly $70K. I also heard that there is a line out the door to get it.