First Deep Learning Build, Stuck on Intel vs AMD

[VitaminP]

Budget (including currency): Id like to stay under $2400 for new parts,

Country: US

Games, programs or workloads that it will be used for: Graphic Design, GAN Images, CS:GO

Other details Some of the parts are being reused (RAM, PSU, NH-D15, Case, SATA Drives)  I will be buying CPU, GPU, and Motherboard

 

PCPartPicker Part List: https://pcpartpicker.com/list/mx8Hgt

CPU: Intel Core i9-10900K 3.7 GHz 10-Core Processor  ($464.49 @ B&H)
CPU Cooler: Noctua NH-D15 82.5 CFM CPU Cooler  ($99.95 @ Amazon)
Motherboard: Gigabyte Z590 AORUS ELITE ATX LGA1200 Motherboard  ($219.99 @ B&H)
Memory: G.Skill Trident Z Royal 32 GB (2 x 16 GB) DDR4-3600 CL19 Memory  ($209.99 @ Newegg)
Storage: Samsung 860 Evo 1 TB 2.5" Solid State Drive  ($109.99 @ Best Buy)
Storage: Samsung 970 Evo 1 TB M.2-2280 NVME Solid State Drive
Storage: Seagate Barracuda Compute 2 TB 3.5" 7200RPM Internal Hard Drive  ($55.49 @ Amazon)
Storage: Seagate Barracuda Compute 2 TB 3.5" 7200RPM Internal Hard Drive  ($55.49 @ Amazon)
Video Card: NVIDIA GeForce RTX 3090 24 GB Founders Edition Video Card
Case: NZXT H710 ATX Mid Tower Case  ($139.99 @ Amazon)
Power Supply: SeaSonic FOCUS Plus Gold 1000 W 80+ Gold Certified Fully Modular ATX Power Supply  ($349.00 @ Amazon)
Total: $1704.38
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2021-05-13 02:51 EDT-0400

 

I am tired of farming out computing for GANs and want to be able to train them on my own machine.  I am stuck between an Intel and AMD CPU (10900k vs 3950x or 5950x).

The Intel chip fits comfortably within my budget but both AMD chips would be a little bit of a stretch for me, but I realize on paper they are better, but have also read that Intel is the go to for deep learning?

Does anyone with deep learning experience think this is a serviceable image generating build? I admittedly don't have a strong CS background  but have been using AI images in graphics with success for a bit now. 

[igormp]

CPU should be irrelevant for GANs per se, your 3090 is what matters. Also, go for 2x32gb, 32gb is not enough for most reasonable models.

 

14 hours ago, VitaminP said:

but have also read that Intel is the go to for deep learning?

Only if you're doing CPU training, which no one other than masochists should do, otherwise it doesn't matter. I used to train tons of models on my old FX system, and my azure machine has a crappy dual core xeon paired with 2x V100.

[VitaminP]

54 minutes ago, igormp said:

CPU should be irrelevant for GANs per se, your 3090 is what matters. Also, go for 2x32gb, 32gb is not enough for most reasonable models.

 

Only if you're doing CPU training, which no one other than masochists should do, otherwise it doesn't matter. I used to train tons of models on my old FX system, and my azure machine has a crappy dual core xeon paired with 2x V100.

Thank you for your insight, it helps a lot.  I think you have answered a couple of my questions this week.  I will tweak the build to get more ram into it.

[igormp]

If you go intel, try to look for the 10850k, should be cheaper and net you pretty much the same perf.

[VitaminP]

2 hours ago, igormp said:

If you go intel, try to look for the 10850k, should be cheaper and net you pretty much the same perf.

Yeah I was just reading that they are pretty much the same, I can find the 10850k for about $50 cheaper