Machine for protein structure analysis (emphasis on storage speed)

We received a request from a customer to consider a machine that uses the tool "AlphFold2" to calculate the three-dimensional structure of proteins with a budget of about 100 million yen.
In order to use other Windows software such as MOE (Molecular Operating Environment) and Autodock, you want to dual boot Ubuntu and Windows, and Ubuntu requires CUDA Toolkit 11 to be installed.

AlphaFold's computational processing is greatly affected by GPU performance, and because it frequently accesses a database of about 2.7 TB even during computation, the performance of the storage that stores the database has a large impact on usability.
All the storage in this configuration is SSD, but for the above reasons, the storage for database development uses a 4TB SSD that supports NVMe to further increase the storage speed.

【Main Specifications】

CPU Ryzen9 5950X (3.40GHz 16 cores)
memory 64GB
Storage 1 500GB SSD S-ATA
Storage 2 500GB SSD S-ATA
Storage 3 4TB M.2 SSD
video NVIDIA Geforce RTX3090
network on board (10/100 / 1000Base-T x1)
Housing + power supply Middle tower case + 850W
OS Windows 10 Professional 64bit
Others Ubuntu 20.04 dual boot setup (CUDA Toolkit 11 installed)