Skip to content

Latest commit

 

History

History
26 lines (21 loc) · 1.54 KB

system.md

File metadata and controls

26 lines (21 loc) · 1.54 KB

System

For the ones curious about the system that I use.

I like to keep myself updated on the new hardware available in the consumer computing space. I started my journey on a laptop with an Nvidia 940M GPU. I quickly realized that the models created and the data that needed to be processed quickly consumed all the VRAM causing OOM (out-of-memory) issues. In 2020, I invested in a system capable of running my tests locally. My primary focus was not to incur any costs for processing and testing models using the cloud.

System configuration (November 2020)

  • CPU: AMD 3700X - (8-cores)
  • Motherboard: MSI MAG X570 Wi-Fi
  • RAM: G.Skill RipJaws V 32 GB @ 3200 MHz (2 x 16 GB kit running in dual channel configuration)
  • GPU: Nvidia GTX 1650 (4 GB of VRAM) - There was a GPU shortage at this time
  • Power Supply: Corsair RM 650
  • Primary Drive: Western Digital Blue SATA drive (500 GB)
  • Secondary Drive: Seagate Barracuda Hard drive (2 TB)

Upgrades

2023

GPU: (Out) Nvidia GTX 1650 ---> (IN) Nvidia RTX 3060 (12 GB of VRAM)

Wish-list

Primary Drive: M.2 Gen 4 SSD with a minimum 1 TB

Parts selection logic

  1. 8-core CPU for any multi-threading, multi-processing loads
  2. 32 GB is not required for a normal system but is required for "containers". This has also proved useful for testing LLMs locally using "LMStudio"
  3. RTX 3060 - effective price per GB of VRAM. AMD does have a good value but CUDA libraries are (currently) the industry standard and run only on Nvidia.
  4. X570 - Future-proofing by adding M.2 Gen 4 and I needed all the IO ports.