Computers

Here is some information about each of the computers used; more specific and private information is on the drive. Also, the information is on each computer; try to keep that organized and updated.

Pev

It is the toshiba gray laptop. It has its bad moments, but in general is good and tolerates hard use. It is a normal x86 computer with Ubuntu.

• Intel i5-3210M 2.50GHz, 2 cores 4 threads, 64 bits
• 6 RAM
• 775gb HDD
• Ubuntu 18

Epia

This is the raspberry pi-like computer that has connected the ssd disk; was given by the professor J.H. It is a "complete normal" computer, has a typical x86 processor (but it isn't intel) that makes it compatible with everything without doing any change. The processor is decent, it draws good amount of energy, and the cpu doesn't have avx/ssm instructions.

• Via embedded, Epia-p910 10Q, Pico ITX
• x86 64bits 4 cores intel-like (NO avx, ssm, etc.)
• 4 RAM
• 125gb external SSD
• Ubuntu 18

Power

It needs 12v that is connected via plug (jack barrel) to a special port. Its power consumption avg is 2A, min 1.5A, and very frequent max 2.8A.

CPU

It has a normal x86_64 processor (like an intel); you can use any program you would use in a normal computer. But, it doesn't have the vector arithmetic instructions like avx, ssm, etc. (very popular in ML frameworks). This means that some programs like TensorFlow, won't fully work by the common installation method, instead will need to be compiled from source without these instructions. This can be done by building the program (i.e. TF) in this epia computer (easier) and will automatically exclude these instructions; other way is to cross-compile it in other computer (faster) for the epia, specifying as flag to not include avx and what arquitecture is the target cpu: -march=nano-3000 (or supposedly -march=core).

Jetson

The Nvidia's Jetson is the raspberry pi-like computer with the black heat sink. It is the constrained model Jetson Nano that has an ARM CPU with a chip to accelerate ML. It supposedly offers a boost in ML inference, but easily is overwhelmed; also includes full ubuntu.

• Jetson Nano
• ARM Cortex-A57 (ARMv8-A) 64-bit 4 cores, GPU 128-core Maxwell
• 4 RAM
• 64gb external SD
• Ubuntu 18 64bits

Power

This needs to be verified.

The computer requires 5v and has two modes of power consumption. The modes can be choosen in the OS bar between efficiency (up to 2A) and performance (4A). It has three different ways to supply power: a micro-usb up to 2A, a barrel jack max 4A (to choose between these a short-circuit jumper needs to be added), and via special pin.

CPU

The computer has an ARM cpu and a GPU-like hardware accelerator mainly for ML. The platform has OK support because at the end has normal Ubuntu. The problem is that the desktop runs slowly, like the i/o seems the problem. Also, the GPU memory seems to be low and easily runs out in a not so-intense ML program.

Rapsberry Pi

Coming soon...