Good laptops for Cuda development - my experiences

I thought I would share my experiences on good laptop platforms for Cuda for those of you who prefer working on laptops / want to do cuda on the move.

Microsoft Surface Book 2. This has a Geforce GPU with cuda capability 5.0, 384 cuda cores (Maxwell). Importantly it also has a Intel integrated graphics adapter. This means you can set the Intel as the primary graphics adapter in the Nvidia Control Panel (under Manage 3d Settings) and then use the Geforce GPU for cuda development (so debugging dynamic parallelism is possible).

Not the fastest system but super portable so you can code from cafes etc.

Alienware 17 R4 + Graphics Amplifier. This has a GTX 1080 (Pascal)

This is a reasonable setup for a less portable laptop. You can put a Titan X in the graphics amplifier and then use this for cuda development. A disadvantage of this laptop is it uses G-SYNC which means you cannot switch to the integrated Intel adapter (the laptop screen is wired directly to the GPU). This means you have to share the gpu between driving the display and cuda - not good and no debugging dynamic parallelism possible).

However, once you connect the Graphics Amplifier you can use an external graphics card for Cuda and are able to debug in the presence of dynamic parallelism. Titan X is the best card that the amplifier supports.

Hope this is helpful to someone out there - this took me a while to figure out.