Need suggestions for new build to be used for ML/AI

[DoubleHelix747]

Hi there -

My last build circa 2008 is no longer adequate for what I plan to do in Fall. I am back in school and will be starting with hard core machine learning and artificial intelligence course work that would require lot of computing power (my peers tell me that even on their 2-3 year old fast computers, each project run takes a day or two).

I have not built anything for 10 years and need some advice with choice of the chip, motherboard and gpu. Since RAM is expensive, how much should be enough?

I think my computational needs will only increase over time so upgradability is crucial.

Budget is ~$500 - $1000 and I have all the software I need (including OS).

TIA for any tips and suggestions.

PS working with cloud service such as AWS is not an option right now.

 

[Ketchup]

I don't see you getting anything noticably faster than 2-year-old machines, unless they are referring to retail store machines, for less than $1000. Here is a well-equipped i7 machine I put together with "decent" graphics. Pretty close to target:

PCPartPicker part list / Price breakdown by merchant

CPU: Intel - Core i7-8700K 3.7GHz 6-Core Processor ($347.00 @ SuperBiiz)
CPU Cooler: Deepcool - GAMMAXX 400 74.3 CFM CPU Cooler ($21.89 @ OutletPC)
Motherboard: MSI - Z370-A PRO ATX LGA1151 Motherboard ($100.56 @ Newegg)
Memory: Team - Elite Plus 16GB (2 x 8GB) DDR4-2400 Memory ($140.98 @ Newegg)
Storage: Samsung - 850 EVO-Series 500GB 2.5" Solid State Drive ($145.59 @ Newegg Marketplace)
Video Card: MSI - GeForce GTX 1050 Ti 4GB GAMING X 4G Video Card ($214.99 @ B&H)
Case: Corsair - 270R ATX Mid Tower Case ($59.99 @ Amazon)
Power Supply: SeaSonic - 520W 80+ Bronze Certified Fully-Modular ATX Power Supply ($49.99 @ SuperBiiz)
Total: $1080.99
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2018-06-16 14:00 EDT-0400