Deep Learning PC - Feedback welcome

[chacho]

Processor (CPU)
AMD Ryzen 7 3700X Eight Core CPU (3.6GHz-4.4GHz/36MB CACHE/AM4)
Motherboard
ASUS® ROG STRIX X570-F GAMING (USB 3.2 Gen 2, PCIe 4.0) - RGB Ready!
Memory (RAM)
32GB Corsair VENGEANCE DDR4 3000MHz (2 x 16GB)
Graphics Card
8GB NVIDIA GEFORCE RTX 2060 SUPER - HDMI, DP - VR Ready!
1st Storage Drive
2TB SEAGATE BARRACUDA SATA-III 3.5" HDD, 6GB/s, 7200RPM, 256MB CACHE
1st M.2 SSD Drive
512GB INTEL® 660p M.2 NVMe PCIe SSD (upto 1500MB/sR | 1000MB/sW)
Power Supply
CORSAIR 650W VS SERIES™ VS-650 POWER SUPPLY
Power Cable
1 x 1 Metre UK Power Cable (Kettle Lead)
Processor Cooling
PCS FrostFlow 240 Series RGB High Performance Liquid Cooler
Thermal Paste
ARCTIC MX-4 EXTREME THERMAL CONDUCTIVITY COMPOUND
Sound Card
ONBOARD 6 CHANNEL (5.1) HIGH DEF AUDIO (AS STANDARD)
Wireless/Wired Networking
WIRELESS INTEL® Wi-Fi 6 AX200 2,400Mbps/5GHz, 300Mbps/2.4GHz PCI-E CARD + BT 5.0
USB/Thunderbolt Options
MIN. 2 x USB 3.0 & 2 x USB 2.0 PORTS @ BACK PANEL + MIN. 2 FRONT PORTS
Operating System
NO OPERATING SYSTEM REQUIRED

 

[Oussebon]

I'd suggest a better quality PSU, especially given the workload.

Why the X570-F over the X570-PLUS?

What case is this system being put in?

I'd probably go with the SX6000 SSD as it's a bit faster for the same kind of price

I'd probably also go with the H100x cooler which is even more effective so as to keep temps as low and therefore CPU frequencies as high as possible.

 

[chacho]

Thanks for your input.

I chose the x570-F due to an extra expansion slot for future.

I'm thinking of putting it in PCS P209 RGB MID TOWER CASE.

 

[Oussebon]

You mean for a triple GPU system? The bottom slot on the mobo is x4 only so you know

That case would be woefully inadequate for a 2/3 GPU system and tbh is not good for the system above either, as it will have very limited airflow.

 

[chacho]

Ahh I see, didn't know that, i guess then I'll have to settle with x570-plus in that case as triple gpu mobo will be expensive I guess .

Not sure about the cases any recommendations for this system?

 

[Oussebon]

Meshify C is great if you want good airflow without spending too much.

 

[markh]

I'd say that the RTX 2060 Super is really a good but bare-minimum GPU for serious deep learning. You'll be able to do a lot with it, but there might well be scenarios where you run out of memory. It really depends what kinds of models you're working with. Similarly, 32GB of RAM might be a squeeze (as ridiculous as that sounds) depending on the sizes of datasets you're working with and how much you might want to experiment with batch size. For example, I recently tried to run an audio generating RNN on an RTX 2070 with 32GB RAM and ended up having to reduce my dataset size, batch size and some other settings in order for the model to even train without running into memory errors. I'd also opt for a faster SSD like the Samsung Evo. The above really depends on what kinds and sizes of DL workloads you are involved with though (your setup would eat image classification tasks for breakfast) and only you will know that, so apologies if you already thought about everything I've flagged here!