Welcome tuskiomi !
The power draw calculations are a little tricky, I seem to recall that the highest observed reading has been in the area of 17.5W. That’s with all the power management turned off, all CPU cores online, and all clocks turned to full speed. Like most modern embedded processors, there’s a sophisticated suite of power management tools that allows for different power configurations, such as max performance, “normal”, and energy saver modes. This is done in part by setting different clock rates for the CPU cores and GPU. When it comes down to it, the energy consumption profile has more to do with the application that is being executed than much else. The mix of the CPU and GPU use is a major factor, the number of cores, clock speeds, etc. For most people this is an observed value, rather than a calculated one. Obviously if you’re building your own board there’s a lot of sophisticated calculations you’ll need to do, but a back of the napkin sketch should be fine for figuring battery life. You can read more about power on the Jetson Wiki: http://elinux.org/Jetson/Jetson_TK1_Power . This includes information about how to connect the Jetson to a battery.
For the SSD, think about using using the SATA port. mPCIe is pretty useful for Wifi and USB expansion.
As I recall, there’s a diode which prevents the board from being powered from the Molex power connector.
As linuxdev pointed out JTAG debugging is really hardware oriented and is usually used for debugging hardware. General purpose assembly language programming is usually done “on the other side”.
The Jetson TK1 is a fast enough embedded processor where it can run a Linux desktop environment, which allows the user to use standard Linux development tools. In the case of debugging, the majority of users use the standard gdb debugger. This also means it is easy to use higher level languages such as C, C++, Python, etc. You can slip into assembler if need be, but you’re almost assured of getting lower performance than using an optimizing compiler for something like C++ unless you have a very thorough knowledge of the intricacies of the chip set (in this case ARMv7 Cortex 15) for a program of size.
This is especially true of a SoC such as the Tegra K1 (the Jetson TK1 processor) which also has a GPU with 192 CUDA capable cores in addition to the 4+1 32 bit ARM cores.
I hope you share with us some of the details of the autonomous drone command center when you get it up and running!