Help w/ System build for GPU-based Machine Learning research (~$1000 + starting parts)

machinelearning · Feb 10, 2017

Hello all,

I could use some help please with building a computer system. I already have some parts and am able to spend about ~$1000 more.

The purpose of this machine is medical research, which maxes out CUDA cores on the GPU. I don't think I can afford an NVidia Pascal TitainX, but I am looking at the NVidia GTX 1080. Eventually, I'd like to SLI if possible, but I believe that is out of my budget for now.

So this is to be a build for a GPU-based Machine Learning research machine ~$1000 (+ some starting parts). Thanks!

Existing parts include:

PCPartPicker part list / Price breakdown by merchant

Case: Cooler Master HAF 932 ATX Full Tower Case (Purchased)
Power Supply: EVGA SuperNOVA G2 750W 80+ Gold Certified Fully-Modular ATX Power Supply (Purchased)
Generated by PCPartPicker 2017-02-10 12:32 EST-0500

Full ATX Case,
Power-suppy (750W, let me know if I need a new one with more Watts),
Hard drives (just generic SATA),
DVD drive (just generic SATA),
Keyboard & Mouse & Cables (HDMI, power, etc),

(update, I also already have)

Operating System (Ubuntu GNU/Linux),
Enough HDD space (>2TB)

Supahos · Feb 10, 2017

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Pentium G4560 3.5GHz Dual-Core Processor ($64.79 @ OutletPC)
Motherboard: Gigabyte GA-B250M-DS3H Micro ATX LGA1151 Motherboard ($73.98 @ Newegg)
Memory: Mushkin Blackline 16GB (2 x 8GB) DDR4-2400 Memory ($95.97 @ Jet)
Storage: PNY CS1311 240GB 2.5" Solid State Drive ($79.99 @ Best Buy)
Video Card: EVGA GeForce GTX 1080 8GB Superclocked Gaming ACX 3.0 Video Card ($559.89 @ B&H)
Case: Cooler Master HAF 932 ATX Full Tower Case (Purchased For $0.00)
Power Supply: EVGA SuperNOVA G2 750W 80+ Gold Certified Fully-Modular ATX Power Supply (Purchased For $0.00)
Operating System: Microsoft Windows 10 Home OEM 64-bit ($88.58 @ OutletPC)
Total: $963.20
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2017-02-10 12:42 EST-0500

Probably your best bet, good chunk of ram, fit a 1080, processor can feed things. If you need more processor it'll cost you your 1080, a Titan is $1600 by itself.

machinelearning · Feb 10, 2017

Supahos :

Thanks for the quick response + advice!

Yeah, Titan is off the table.

I don't need the SSD storage or the OS - do you think the difference should be put into RAM or CPU?

Supahos · Feb 10, 2017

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Core i5-7500 3.4GHz Quad-Core Processor ($197.55 @ OutletPC)
Motherboard: Gigabyte GA-B250M-DS3H Micro ATX LGA1151 Motherboard ($73.98 @ Newegg)
Memory: G.Skill Aegis 32GB (2 x 16GB) DDR4-2400 Memory ($166.97 @ Jet)
Video Card: EVGA GeForce GTX 1080 8GB Superclocked Gaming ACX 3.0 Video Card ($559.89 @ B&H)
Case: Cooler Master HAF 932 ATX Full Tower Case (Purchased For $0.00)
Power Supply: EVGA SuperNOVA G2 750W 80+ Gold Certified Fully-Modular ATX Power Supply (Purchased For $0.00)
Total: $998.39
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2017-02-10 13:01 EST-0500

Is the route I'd go. If you can research the program you're using I may can tailor it a bit more but your parameters were a bit vague. I don't know if less memory and a 7700 would be better or what I listed

machinelearning · Feb 10, 2017

Supahos :

Ok thanks. I'm not sure either. I plan on using Tensorflow and DIGITS https://developer.nvidia.com/digits .

Supahos · Feb 10, 2017

Appears 32gb or memory is recommended for high end rigs as is a Titan. So I think the i5/32gb/1080 build should be solid

Supahos · Feb 10, 2017

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Core i5-7400 3.0GHz Quad-Core Processor ($188.75 @ OutletPC)
Motherboard: Gigabyte GA-Z170XP-SLI ATX LGA1151 Motherboard ($114.99 @ SuperBiiz)
Memory: G.Skill Aegis 32GB (2 x 16GB) DDR4-2400 Memory ($166.97 @ Jet)
Video Card: EVGA GeForce GTX 1080 8GB Superclocked Gaming ACX 3.0 Video Card ($559.89 @ B&H)
Case: Cooler Master HAF 932 ATX Full Tower Case (Purchased For $0.00)
Power Supply: EVGA SuperNOVA G2 750W 80+ Gold Certified Fully-Modular ATX Power Supply (Purchased For $0.00)
Total: $1030.60
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2017-02-10 13:20 EST-0500

Bit out of budget but I forgot you might go sli.

Search

Help w/ System build for GPU-based Machine Learning research (~$1000 + starting parts)

machinelearning

Commendable

Supahos

Supahos

Expert

machinelearning

Commendable

Supahos

Expert

machinelearning

Commendable

Supahos

Expert

Supahos

Expert

TRENDING THREADS

Latest posts

Moderators online

Share this page