ML/Deep Learning Rig Upgrade Recommendations

[thefantasticspud]

(1) Budget and Location

Budget: $2500 AUD

Location: Australia

 

(2) Aim

(2.1) Current System:

• i7-4790k, GTX 970, 1TB 2.5' hard drive (from a laptop), 250GB SSD (sata)
• Dual monitor setup (1080p, 28')
• ATX motherboard, in oversized corsair case...
• Stock heatsink (that runs too damned hot)

(2.2) What parts I need (within the budget):

• Ideally a 1TB SSD (or larger)
• RAM
• CPU
• GPU
• CPU cooler
• Power Supply
• (Iff it can be within the budget, maybe 2 newer, better resolution monitors - assume that the existing 2 Philips monitors can be sold for about $300 if need-be)

(2.3) Requirements:

• Primarily needed for machine/deep learning, and want to have a fairly low training times on larger models. Current system is quite slow with large models and large training sets. Need better GPU for acceleration, but not sure which is the best to balance with a new CPU at that price limit. Assume that it's just primarily research and rapid prototyping
• I would like to decrease the machine size to something smaller (e.g. Micro-ATX/Mini-ITX), the empty space in my atx case annoys me
  • I'm eyeing off the H200i NZXT case as a prime candidate. I would love SFF, but thermals take higher priority
• Dual boot machine - linux and windows (spend most of my time in linux)
• Solidworks modelling
• No gaming
• Want performance and reliability rather than RGB RAM

(3) Monitors

• Running 2 monitors, (see point 2.2)

(4) Peripherals

• Already have keyboard, mice, and the OS's I could ever need

(5) Why am I upgrading?

• Primary Reason: I want to accelerate the speed at which I can train my models, so that I can progress faster, rather than waiting for models to train and run for hours on end, even with scheduling them to run overnight.
• Secondary reason: Want a new setup, that runs cooler, quieter

(6) Conclusion

Any and all recommendations are greatly appreciated! Thanks

Edited December 21, 2018 by thefantasticspud
Unnecessary bracket removal