1
$\begingroup$

Is the laptop Asus ZenBook Pro 90NX0152-M02980 enough to do deep learning model?

Specs:

  • Processeur Intel Core i7-7500U (Dual-Core 2.7 GHz / 3.5 GHz Turbo - cache 4 MB)
  • 8 GB Memory
  • SSD M.2 SATA de 256GB
$\endgroup$
  • $\begingroup$ Welcome to SE:AI! I've added the "hardware-evaluation" tag. I suspect you're going to need a lot of memory for such a model. (Hopefully one of the DL experts will weigh in!) $\endgroup$ – DukeZhou Oct 15 '18 at 16:53
  • $\begingroup$ @DukeZhouz ,Does the question portray research or it's me pretending to confuse,just like how Kanye West does it. $\endgroup$ – quintumnia Oct 16 '18 at 17:52
  • $\begingroup$ @quintumnia I think the OP is trying to get a gauge on the suitability of their equipment, which information should be of practical use to many indie researchers and developers. $\endgroup$ – DukeZhou Oct 18 '18 at 3:13
1
$\begingroup$

Here is a little bit of aglimpse

Asus Zenbook pro,is quite good for technical nor software development tasks,despite having ordered one recently.

That said,you need to that each and every computer/machine for effective and efficient computing purpose,the first priority is to analyse critically the processor chip,which generational architecture/design does it have inline with GPU/Graphics ,Resolution,motherboard,hard-drive...e.t.c..

Therefore,basing on such technical specs,what matters most,is not the brand nor logo..! Go for any brand,keeping in mind your objective "DeepLearning"

$\endgroup$
0
$\begingroup$

Is the laptop Asus ZenBook Pro 90NX0152-M02980 enough to do deep learning model?

Not really.

  • It would be adequate if you want to learn the basics, just run TensorFlow or PyTorch and train some simple models, maybe run some tutorials on MNIST.

  • It would be adequate to run large models in prediction mode - e.g. you could run VGG-19 or InceptionNet-v5 and classify images.

  • If you can offload your larger training requirements to a separate server, and use the laptop to do data preparation and smaller scale tests, then a top end ultra-portable laptop would be fine.

I have a similar spec laptop, and have taught myself a lot of deep learning concepts on it. However, I cannot realistically train any modern image-processing, natural language model or interesting deep RL model on it, without committing to days of training time.

I still have managed to do some interesting things, such as play with Deep Dream (even making a Deep Dream video), but I often hit limitations where I would like to try something but the laptop is not capable:

  • Cannot enter Kaggle competitions on image, audio or language datasets.

  • Cannot perform style transfer in reasonable time.

  • Cannot solve larger, more interesting Reinforcement Learning problems.

  • Cannot train LSTMs on large text databases.

If you want to do those or similar things, you absolutely need a machine with a powerful GPU that can accelerate the deep learning library that you want to use. Typically that is an NVIDIA graphics card that supports CUDA. Even modest, relatively cheap (in modern standard) GPUs can accelerate deep learning model training very well, so a 1070 or 1060 card can make a big difference. The difference between the graphics cards is that better and more expensive ones are faster, can handle larger neural networks and larger problems overall.

If you are still keen on buying a laptop, you will probably want a gaming laptop, and pay attention to what kind of GPU it contains. You will probably also want a larger memory than 8GB, and expect at least one of the drives to be of the SSD type so you can load training data faster. That combination does not come cheap in a laptop - e.g. an Alienware laptop with i7 processor, 16GB RAM and NVIDIA 1070 costs £2,300 as of November 2018, and I would consider that to be an acceptable, but not top of the range spec. Also, of course it is not ultra-portable, but quite the opposite, you won't be happy to carry it for a long time.

The way to drop the price down is to:

  • Buy a desktop, not a laptop

  • Build your own from components. This is not without risk, and you should only attempt it if you have at least some of the knowledge and skills already (and could look up the rest). However, this can save a signifiant amount of money. You can search "build own deep learning machine" or similar to find many helpful articles on how to approach this, like this one.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.