Mining to Deep Learning build & Power Concerns

rschofield

Prominent
Jan 31, 2018
1
0
510
Hey everyone,

First off thanks for taking the time to read this. This is kind of an unique transitions that I'm trying to achieve with my build, and I just wanted a second opinion. I'm trying to initially build a mining machine with the end goal of becoming an optimized deep learning machine (CNN, RNN, Stacked-GAN, etc). Basically I want a nice machine, but I want to get a head start on the ROI given my current budget ($1250).

I'm currently in the possessions of:
2 x EVGA SC2 HYBRID GEFORCE 1080 Ti
1 x AORUS XTREME EDITION GEFORCE 1080 Ti
1 x PNG PNY GEFORCE GTX 1060 6GB XLR8
1 x GIGABYTE Z370 AORUS GAMING WIFI (rev. 1.0) LGA 1151 (300 Series) Intel Z370 HDMI SATA 6Gb/s USB 3.1 ATX Intel Motherboard
1 x Kingston A400 2.5" 120GB SATA III TLC Internal Solid State Drive (SSD) SA400S37/120G

First Build (Mining Build):
1 x Everything Listed Above
1 x G.SKILL Ripjaws V Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3000 (PC4 24000) Intel Z170 Platform Memory Kit Model F4-3000C15D-16GVKB
1 x Intel Core i7-8700K Coffee Lake 6-Core 3.7 GHz (4.7 GHz Turbo) LGA 1151 (300 Series) 95W BX80684I78700K Desktop Processor Intel UHD Graphics 630
1 x PSU (Discussed Below)

Second Build (Deep Learning Build):
4 x EVGA SC2 HYBRID GEFORCE 1080 Ti
1 x Intel Core i7-8700K Coffee Lake 6-Core 3.7 GHz (4.7 GHz Turbo) LGA 1151 (300 Series) 95W BX80684I78700K Desktop Processor Intel UHD Graphics 630
1 x ASUS X99-E WS/USB 3.1 LGA 2011-v3 Intel X99 SATA 6Gb/s USB 3.1 USB 3.0 CEB Intel Motherboard
4 x CORSAIR Dominator Platinum 32GB (2 x 16GB) 288-Pin DDR4 SDRAM DDR4 3000 (PC4 24000) Desktop Memory Model CMD32GX4M2B3000C15
1 x PSU (Discussed Below)
1 x Hard Drive TBD

Power Supply
I wanted to go with the EVGA 120-G2-1600-X1 SuperNOVA 1600 G2 1600W 80 PLUS Gold ATX12V & EPS12V Power Supply, but Im not sure if it's enough for my final build given each of the 4 GPUs is 250 Watts and the CPU is at least 100 Watts which puts me dangerously close to the 80% efficiency of 1600 Watt 80%. Should I get multiple GPUs and daisy chain them?

I'd appreciate and comments :)



 
Solution


You're fine with the 1600w. The PSU will still output 1600w of power, the efficiency rating is for how much power it will have to suck from the wall to produce that 1600w (nearly 1800w).

Especially for mining, you'll be lowering the power limit of those GPU's so they won't use as much power. And you wouldn't be mining with that CPU unless you get free power, so it wouldn't be pulling max power. Even if you did mine with it, you could undervolt that as well.

 
Solution