And yes, those options probably make more practical sense than building your own computer.But if youre like me, youre dying to build your own fast deep learning machine.
At the very least, this setup will easily outperform a 2,800 Macbook Pro on every metric other than power consumption and, because its easily upgraded, stay ahead of it for a few years to come. Best Laptops Hine Learning 2019 Free Trial TodayGet a free trial today and find answers on the fly, or master something new and useful. Since I didnt want to use multiple GPUs, the cheapest and smallest standard size is called mini-ITX, which will be fine for this sort of project. My minimum requirements were a PCIe slot to plug the GPU into and two DDR4 slots to plug RAM into, and the board I went with was an ASUS Mini ITX DDR4 LGA 1151 B150I PRO GAMINGWIFIAURA Motherboard for 125 on Amazon. It comes with a WiFi antenna, which is actually super useful in my basement. The size should match the motherboard, so it needs to have mini-ITX in the name. If you build a computer in your basement and you dont embrace your inner Burning Manteenager aesthetic, you are going to have a really hard time finding components. But I couldnt bring myself to build a whole computer with a CPU three generations old. I didnt spend the extra 20 for the I5-6600K, which is the same exact chip but overclockable because the whole notion of sacrificing reliability for a 10 speed increase seemed insane to me. I do have to admit that as I started getting into the idea of building my own machine, I started to regret the decision. So who knows Building a computer can change your outlook on life. Probably fewer regrets if you go with the overclockable chip. On the other hand, maybe its best to protect yourself from yourself and take the option off the table. A solid state drive would be faster, but much more expensive, and, typically, my deep learning programs have not been disk IO bound, since they generally load batches of data into RAM and then crunch on them for a long time. If you wanted to use your computer for heavy file transfer workloads or just wanted to make sure that its clearly faster than your friends Macbook for all applications, I would get a solid state drive like the Samsung 850 EVO 250GB 2.5-Inch SATA III Internal SSD, which is 250Gb for 98. For pretty much all machine learning applications, you want an NVIDIA card because only NVIDIA makes the essential CUDA framework and the CuDNN library that all of the machine learning frameworks, including TensorFlow, rely on. If TensorFlow cant fit the model and the current batch of training data into the GPUs RAM it will fail over to the CPUmaking the GPU pointless. The difference between the architectures really matters for speed; for example, the Pascal Titan X is twice the speed of a Maxwell Titan X according to this benchmark. Most people doing machine learning without infinite budget use the NVIDIA GTX 900 series (Maxwell) or the NVIDIA GTX 1000 series (Pascal). But a 980 card is still probably significantly faster than a 1060 due to higher clock speed and more RAM.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |