Description

Deep Shadow is a high-end liquid-cooled workstation designed for deep learning and machine learning that we built for Pixel Scientia Labs.

This build is designed around an 8-core i9-9900K processor and twin liquid-cooled RTX 2080Ti 11GB video cards.

Hardware Selection Reasons

Processor (CPU): Intel i9-9900K

I chose this processor for the high single-core and multi-core performance.

The 9900K represented the most powerful single-core CPU that was available at the time, and fit within her budget well. Heather preferred the immediate stronger editing/modeling performance with Intel, and understood that needing more CPU threads down the road would require a new motherboard as well.

CPU Cooler: Corsair H150i Pro RGB

I chose this CPU cooler for the excellent cooling performance of water. While air is more reliable, water cools better and runs quieter. This cooler also looks nice in the case, making for a very sleek interior. Heather agreed with the visual design, and understood my explanation of the reliability of high-end water coolers.

Motherboard: Asus WS Z390 Pro

I chose this motherboard for socket compatibility, 4 PCI 3.0 slots, and extremely robust design. Asus is also one of my preferred motherboard manufacturers. Heather agreed with the visual design, and understood my explanation of the build quality.

Memory: Corsair Vengeance LPX 4x16GB DDR4-2666

I chose this memory kit for motherboard compatibility and robust design. Corsair is also one of my preferred memory manufacturers. Heather needed 64GB of memory for her workflow, and understood my explanation of the build quality.

Storage (SSD): Samsung 970 EVO 1TB NVME

I chose this SSD for robust design, high capacity, fast access times, and low cost. When Deep Shadow was built, SSDs were very cheap, allowing us to use a 1TB model for general work storage. Samsung is also one of my preferred SSD manufacturers. Heather understood my explanation of the build quality.

Storage (HDD): 2x Western Digital Red 6TB

I chose this HDD for robust design, high capacity, and low cost. Western Digital is also one of my preferred HDD manufacturers. Heather didn’t need more space for archival storage, and understood my explanation of the build quality.

Video Card (GPU): 2x EVGA RTX 2080Ti 11GB XC Hybrid

I chose this GPU for long-term performance, low noise, superior cooling power, and robust design. Heather needed lots of GPU power for her workloads, and understood my explanation of the build quality.

Power Supply (PSU): Corsair RMx 1000

I chose this PSU for wattage, robust design, and long-term stability. Heather wanted a workstation that would last, and understood my explanation of the build quality.

Case: Phanteks Enthoo Pro

I chose this case for robust design, airflow, and long-term stability. Heather wanted a workstation that would last, and understood my explanation of the build quality.

Log in to rate comments or to post a comment.

Comments

  • 5 months ago
  • 2 points

Do the RTX 2080 tis run in 16x/16x, or 16x/8x? Btw nice build!

  • 5 months ago
  • 3 points

These RTX 2080Tis run in 16/16, though the board can handle 4 2080Tis in 8/8/8/8. This is due to the special PLX chip on the board.

Thanks!

  • 5 months ago
  • 1 point

Ayyy thats dope

  • 5 months ago
  • 1 point

Looks like a great build. I took some inspiration here and made my own https://pcpartpicker.com/user/dokko1230/saved/vtZCLk. Can the motherboard here fit 4 of the RTX 2080 Hybrid Tis?

  • 5 months ago
  • 1 point

Thanks! The WS Z390 Pro board can fit 4 GPUs, all of which would get 8 lanes. PCI lanes don't matter much for deep learning, as long as it's not x1.

Make sure you've picked a good CPU though, for your workload. We used the 9900K because her workload preferred the higher speed of the i9.

  • 5 months ago
  • 2 points

Very solid build!

  • 5 months ago
  • 2 points

Thank you!

  • 1 month ago
  • 1 point

Just wondering. Why not put the hybrid 2080 ti in SLI? physical limitations with the watercooling?

  • 1 month ago
  • 1 point

Didn't need the NVLink bridge