Upgrade my old PC with GTX 1080 for NLP deep learning. Is it good enough?

Mar 26, 2018
8
0
10
Hello,

I'm doing some researchs in Natural Language Processing using Deep Learning (Tensor Flow).
I want to upgrade my VGA Card from GTX 650 Ti to GTX 1080 to use the power of GPU in deep learning.
I have low budget so I want to upgrade VGA card and PSU only, not build an entire new PC.
My current PC:
- Main board: P8H77-V
- CPU: Core i5 3470
- RAM: 8GB DDR3
Is it good enough if I replace VGA card only?
I heard that my mainboard's chipset does not have enough PCIe lanes for GTX 1080, so it will not be good. Is it right?
I'm looking forward to your advices. Thank you very much.
 
Solution
hlongsoft you will get a bottleneck as there is not enough bandwidth for the GPU to operate fully. You will need to upgrade your motherboard to something with more PCiE lanes. As for the PSU, a 650W would do just fine so long as you only using one card for the foreseeable future.
Mar 26, 2018
8
0
10


Thank you very much.
I have checked the H77 chipset manual and saw this: Max # of PCI Express Lanes: 8 (https://ark.intel.com/products/64018/Intel-H77-Express-Chipset)
Is this affects the performance of GTX 1080 (bottleneck)?
And one more question. What PSU should I pick for:
- i5 3470
- SSD 256GB
- SSD 128GB
- HDD 500GB
- GTX 1080
 

henryjamesguest

Prominent
Apr 21, 2018
113
0
710
hlongsoft you will get a bottleneck as there is not enough bandwidth for the GPU to operate fully. You will need to upgrade your motherboard to something with more PCiE lanes. As for the PSU, a 650W would do just fine so long as you only using one card for the foreseeable future.
 
Solution

TRENDING THREADS