Re:Are the New M1 Macbooks Any Good for Deep Learning?
by apohn on Tuesday February 16, 2021 @09:33AM
(#26154170)
>A 1660 Ti 4gb RAM laptop gpu, runs about as fast as Collab CPU does
According to the chart on that page you linked to the Lenovo Legion (with the 1660ti) is about 30% faster than the Colab GPU. The Lenovo 480s (CPU only, no GPU) is about the same speed as the Colab CPU. So if you can train your model on the on 4GB of VRAM, a laptop with a discrete GPU (if you exclude the really basic business laptop GPUs) may be useful.
Re:Are the New M1 Macbooks Any Good for Deep Learning?
by joshocar on Monday February 15, 2021 @09:47PM
(#26150370)
Are the Quadro cards any good for it?
Re:Are the New M1 Macbooks Any Good for Deep Learning?
by teruakohatu on Monday February 15, 2021 @10:36PM
(#26150631)
They are, other than being priced more expensively compared with consumer cards. When doing things like RNN's, RAM becomes an issue regardless of how powerful the card is in other ways.
I agree, but from what I have seen they also perform poorly against collab. A 1660 Ti 4gb RAM laptop gpu, runs about as fast as Collab CPU does. [1]
I have not been able to find how the RTX 2080 Super (Laptop) and upcoming RTX 3080 (Laptop) compares to Collab. The 3080 has over 3x as many cores as the 1660 Ti.
[1] https://towardsdatascience.com/google-colab-how-does-it-comp...