First PC Build for Machine Learning/Deep Learning Feedback

[Bryan Carty]

Hi, Happy New Year,
I'm doing my first PC Build and would appreciate any feedback/advice on what I currently have. It will be mainly be used for general programming/software development but will also be used for some machine learning and AI tasks, so for that reason I need a decent GPU and sufficient RAM. I'm told Nvidia is preferred when it comes to ML due to it having tensor cores. After spending some time searching I found a EVGA XC3 ULTRA GAMING GeForce RTX 3080 12GB LHR 12 GB Video Card that's within my budget. My PC build is shown below and I'd really appreciate any feedback you have to offer on it, but in particular your opinions on the EVGA XC3 ULTRA as I'm unsure as to how good it is for ML tasks.

CPU: Intel Core i5-13600K 3.5 GHz 14-Core Processor ($319.99 @ Amazon)
CPU Cooler: be quiet! Dark Rock Pro 4 50.5 CFM CPU Cooler ($89.90 @ Amazon)
Thermal Compound: Noctua NT-H1 3.5 g Thermal Paste ($8.94 @ Amazon)
Motherboard: MSI MAG Z790 TOMAHAWK WIFI ATX LGA1700 Motherboard ($319.99 @ Newegg)
Memory: Corsair Vengeance 32 GB (2 x 16 GB) DDR5-5600 CL36 Memory ($139.99 @ Amazon)
Storage: Sabrent Rocket 4.0 1 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive ($79.98 @ Amazon)
Video Card: EVGA XC3 ULTRA GAMING GeForce RTX 3080 12GB LHR 12 GB Video Card ($817.99 @ EVGA)
Case: NZXT H7 Flow ATX Mid Tower Case ($129.99 @ Best Buy)
Power Supply: NZXT C750 (2022) 750 W 80+ Gold Certified Fully Modular ATX Power Supply ($129.99 @ Best Buy)
Case Fan: NZXT F120Q 64 CFM 120 mm Fan ($14.99 @ Amazon)
Case Fan: NZXT F120Q 64 CFM 120 mm Fan ($14.99 @ Amazon)
Case Fan: NZXT F140Q 101 CFM 140 mm Fan ($25.99 @ Newegg)
Total: $2092.73

 

[Lutfij]

Welcome to the forums and a Happy New Year, newcomer!

You could go for a slightly higher frequency ram kit and a higher PSU with a lot more wattage like 1KW due to transient load spikes but outside of that the build looks fine by me.

 

[Zoid]

Welcome! You have put together a very nice build there. I don't have many notes right off the bat.

I might be able to give more advice on the GPU if I knew more about the type of machine learning you are hoping to do. Can you offer any more details about what types of models you will be running / what size data sets you'll be using / etc?

Is gaming performance important, or just scientific computing performance?

 

[Bryan Carty]

Welcome! You have put together a very nice build there. I don't have many notes right off the bat.

I might be able to give more advice on the GPU if I knew more about the type of machine learning you are hoping to do. Can you offer any more details about what types of models you will be running / what size data sets you'll be using / etc?

Is gaming performance important, or just scientific computing performance?

Sure, the main motivation behind doing this PC build is I'm a fourth year computer science student doing my final year project on an area of reinforcement learning known as Deep Q Learning. As part of this project I will be training a reinforcement learning agent to play the game of Atari Breakout using a number of Deep Q Network variants. As part of these projects I'll need to be able to store frame transitions of the game. A single transition consists of 4 84 x 84 frames stacked on top of each other. And I need to store approximately 1,000,000 of those transitions at any one time.

But after this year I hope to continue doing some machine learning stuff and possibly pursue a career in it but as for the specifics of the machine learning tasks in the future, I can't be sure.

Gaming performance wouldn't be a priority for me with this build.

 

[Bryan Carty]

Welcome to the forums and a Happy New Year, newcomer!

You could go for a slightly higher frequency ram kit and a higher PSU with a lot more wattage like 1KW due to transient load spikes but outside of that the build looks fine by me.

Thanks, for the reply.
Just in relation to the PSU, PCPartPicker says the expected wattage would be 661W. I was hoping 750 would be ok due to my budget being limited. Do you think 750 would suffice or would it cause problems?

 

[Zoid]

Sure, the main motivation behind doing this PC build is I'm a fourth year computer science student doing my final year project on an area of reinforcement learning known as Deep Q Learning. As part of this project I will be training a reinforcement learning agent to play the game of Atari Breakout using a number of Deep Q Network variants. As part of these projects I'll need to be able to store frame transitions of the game. A single transition consists of 4 84 x 84 frames stacked on top of each other. And I need to store approximately 1,000,000 of those transitions at any one time.

But after this year I hope to continue doing some machine learning stuff and possibly pursue a career in it but as for the specifics of the machine learning tasks in the future, I can't be sure.

Gaming performance wouldn't be a priority for me with this build.

Awesome, the 3080 should be a good pick then, given your budget.

 

[firedrakes]

check out deep learning sub on reddit. depending on what type of learning you do. the spec can heavily very.
also this link to might help.

Best PC for Machine Learning - An entry-level Guide

Machine Learning demands has extremely high performance demands if used right. Our Guide explains what hardware components are most important to speed up your workload.

www.cgdirector.com