One article to get the depth of learning entry-level computer hardware configuration _ Depth Learning

For alumni who have just contacted the deep learning, they may encounter computer configuration problems, such as graphics card model, memory capacity, processor model, etc. Good computer configuration, such as GPU accelerated graphics card, is able to effectively shorten the training time of the algorithm, which allows people to see the results of the algorithm as soon as possible, but also to facilitate the timely adjustment of the depth of learning parameters. However, as a beginner of the depth of learning users, Ding computer is too expensive, a cost-effective and stable brand computer is the most appropriate. Branding > assembly

Generally speaking, as an entry-level, I am willing to buy a brand computer, on the one hand, on the computer hardware (computer graphics, CPU, motherboard, fan and chassis) and other knowledge lack. On the other hand, the brand computer in the stability and a variety of hardware performance matching is considered. So I am willing to buy brand computer, also recommended to get started xueyou buy brand computer.
However, if you are a computer with a variety of hardware are more familiar with the people, and the collocation between them also have a certain understanding, the assembly of computers is also a good choice. GTX1070 > GTX980

Perhaps the most concern is the graphics card model problem, my main recommendation is the 2016 mid-June launched GTX1070,GTX1080 is its flagship version. GTX980 series is a better online wind rating card, but, whether in performance, or price, GTX1070 are superior to GTX980. Although the GTX1080 performance is stronger, but the host that carries this video card is too expensive, the price is about 15000-20000. So, as an entry-level, it's still too expensive.
The following table is the performance and price data for the 3 GPU. Model price core frequency memory bit width GTX1070 3500 1506MHz 8GB GDDR5 256bit GTX1080 5500 1607MHz 8GB gddr5x 256bit GTX980 4000 1126MHz 4GB GDDR 5 256bit

Note:
(1) GTX 1070 performance comparable to the top non-public 980 TI
(2) GTX1080 stronger than the GTX980SLI, even stronger than the Gtxtitanx, probably more than TitanX2 into.
The two conclusions from Baidu Encyclopedia: http://baike.baidu.com/link?url= Wfhqug0tzqramfhdw5l-lru3wb1r9cgihipl0eomqzq5ncmxl7cys8yyvtchkncv4-yegx1tb-z8hn1ltw9f8memttlwvwn48dlyzj36gsw
　　

On the other hand, since it is the entry level, the Titan and Tesla type graphics cards are not the focus of my recommendation for price reasons. The following article contains the Titan and Tesla, as well as the GTX980 of some of the introductions, if interested, can be consulted.

http://blog.csdn.net/feixue67600/article/details/51772384 GPU and memory

Since the choice of a powerful GPU, then the processor's performance must be matched to the line, so i7 is the first choice. Secondly, for the memory, 16G is the more appropriate choice. I selected the introduction of the mainframe

To sum up, Dell's XPS 8910-r29n8 is the author's chosen host, plus monitors and other accessories, the total price is 11777 yuan. Its specific parameters are shown in the following table: Brand model GPU CPU memory price Dell XPS 8910-r29n8 GTX1070 i7-6700 16G 10500

Finally, if everyone's budget is around 15000-20000, I would recommend that you use a computer equipped with GTX1080, and its performance can be improved to save your training time.