nothing unusual here, lenovo had a twisting display, around 2012 (?), one of the sub-notebooks, 12 inch, X201? Very good keyboard it was, in comparison with current (...)

Why there is no dual type C ports!

Also the controller is lacking of extra back buttons!

And dose it have system level on screen mapping?

Ibwas exited about it but lacking of above will makes me skip it

I did some testing and these are the speeds I get when not offloading any layers to my 4070 GPU:
-ngl 0 --lazy-mode on: 6.6 t/s
-ngl 0 --lazy-mode off: 6.9 t/s
And without -ngl 0:
lazy-mode on: 19.4 t/s
lazy-mode off: 20.5 t/s

Launching llama.cpp with -ngl 6 (as much as would fit into 12 GB VRAM), I get only about 8.0 t/s with --lazy-mode on/off, so one gets about 20 t/s when not using -ngl at all and let llama.cpp handle the offloading of those layers, that give the highest performance increase, automatically.

If you are deciding between a RTX 4070 (12 GB VRAM at 504 GB/s) and a RTX 5060 Ti (16 GB VRAM at 448 GB/s), then I'd recommend the 5060 Ti 16GB, as its additional 4 GB will go a long way and you won't be as memory constrained.
Of course, AMD GPUs also exist: RX 9060 XT 16 GB VRAM at 322 GB/s would be kinda the equivalent, but they don't use GDDR7, and as you see, they bandwidth is slower, so the 5060 Ti will have a 448 GB/s / 322 GB/s -> 39% faster token generation (output) speed than the 9060 XT. In prompt processing (input), which is based on compute performance, based on 3dmark.com/search scores, they are about equal (tho, since the 5060 Ti will use CUDA, it might still be 10-15% faster (in llama.cpp, CUDA is usually 20% faster vs a vulkan build, at least based on my NV GPU testing)).