Raspberry Pi 2: We're Giving Away 50 Units!!! - Review

Table of contents

RoadTest: Raspberry Pi 2: We're Giving Away 50 Units!!!

Author: k12h

Creation date:

Evaluation Type: Independent Products

Did you receive all parts the manufacturer stated would be included in the package?: True

What other parts do you consider comparable to this product?: Other parts: Power supply, power supply microUSB cable, Ethernet cable, desktop PC. Comparable products: Raspberry Pi model B

What were the biggest problems encountered?: Providing adequate power supply. In particular, proper microUSB cable is not trivial to choose.

Detailed Review:

Motivation

 

I like playing with embedded devices, but at the same time I do not have much time for it. I also like applications that contribute to the community. Two cases of such applications are contributing computing power to BOINC scientific computations and contributing network bandwidth to Tor. While So while I am not using a device, why not make it crunch numbers for a RPi-compatible BOINC project such as Einstein@home  and run a Tor relay?

 

I have already been running a Raspberry Pi model B for 2.5 years (well, two of them, one at home and one in a colocation) as a Tor relay, and tried it briefly for BOINC computations. It's limitations were that the Tor relay was often crashing because of lack of memory, and BOINC computations were competing for clock cycles with the relay on its single core. Pi 2 has 4 cores and twice the memory, so I am curious to see if how it performs in this applications. The obvious expectation is that more performance (at the same price) makes it better.

 

The Pi 2 I am testing runs without any peripherals connected, Raspbian, console-only through a 3.3V USB-to-serial cable (no X running), Ethernet connected.

 

I have spent more time looking into the power issues than into the initial applications.

 

Conclusions


  1. The biggest problem for me appeared to be the power supply.
    I must say that the power LED coupled to the power supervisor chip provides an easier way to detect power-related problems than on model B. Nevertheless, the choice of a suitable microUSB cable and adapter is not trivial. The problem has been scrutinized in the RPi community so reading appropriate materials quickly helps. Still, I would like to claim that from the usability point of view, the choice of microUSB port as a power supply connector is controversial. On one hand, microUSB is omnipresent. On the other hand, many cables in combination with many adapters will not provide adequate power, especially when significant computing load is present even without any peripherals connected.
  2. The default image of Raspbian at the moment of testing was coming tuned to conservative Debian wheezy repositories.
    These repositories include not up-to-date versions of applications in scope that are not efficient (Einstein@home) and out of memory (Tor). I have updated the system to Jessie repos, which gave the more up-to-date versions of the applications that do not have these problems.
  3. A moderate overclock (1 GHz ARM, 500 MHz core, 483 MHz SDRAM, overvoltage 2) works stable with full CPU load by Einstein@home.
    The default overclock option in raspi-config menu sets SDRAM clock to 500 MHz which was found to be unstable in Linux on Flash blog: A look at Raspberry Pi 2 performance and overclocking; I did not test the default configuration myself.
  4. As expected, 4 cores and 1 GB RAM facilitates running both Einsten@home computations and a Tor relay  compared to model B. Pi 2 is a stable platform for running these applications.
    For Einstein@home, the computing advantage is straightforward. For Tor relay, for the relatively small bandwidth (1 MBit upstream) of my home connection it does not matter. For a thicker connection, the benefit of more RAM and more computing power should provide a benefit, but I did not test it. Note that both applications are memory and clock cycle hungry, so they will be competing for resources if run together.

 

Recommendations


Nothing really new here, there are enough materials on the internet on all the points below.

 

  • If you already have a Pi, consider contributing its computing power to BOINC and your bandwidth to Tor.
    Though the efficiency of Raspberry Pi 2 is not higher in BOINC computations than that of a desktop with GPU, little is better than nothing, especially if this little is for free. For Tor, it enables a low-power relay which you can leave running while your more power-consuming devices are turned off.
  • Get a good USB cable and a slightly above 5V power supply. A 5.25V one from ModMyPi one is an overkill.
    Read Raspberry PI:- USB power cables, crashing and other problems. Out of the 8 cables that I tried myself, the winner was from a portable power storage by Gigabyte (though it did not have any parameters imprinted). See details below.
  • Do not waste money on a heatsink if you plan to run the device in a room temperature environment - spend it on a better USB cable.
    Read comments to https://helloacm.com/different-approaches-to-reduce-the-cpu-temperature-of-raspberry-pi/.
  • If using Raspbian, update to Jessie to get application versions that are more efficient.
    I consider it easier than building from sources. (Note however that I did not test anything apart form the applications in scope, in particular I did not run X.)

 

References


Here are links to additional materials that I find useful.

 

1. Power and stability issues

 

2. Einstein@home

 

3. Tor relay

 

Details on power supply


The device needs adequate current and voltage to work. When they are not enough, the device works unstable. In particular, the CPU is automatically throttled to 600 MHz and peripherals start failing. In my case this was the onboard network adapter (which is a USB peripheral), so the device lost network connection. The onboard red power LED is connected to a voltage supervisor; the LED blinks when the voltage drops below about 4.67 volts (R-Pi Troubleshooting - eLinux.org). The more current the device uses, the higher the voltage drop on the power cable, the lower the voltage at the device.


Though the article Raspberry PI:- USB power cables, crashing and other problems is in depth and conclusive, I did some extensive testing to illustrate the issue more. I was just curious how different power supply configurations compare.

 

I was measuring voltage between the PP2 and PP5 test points with a multimeter when the device was under full load (all 4 cores were running Einstein@home tasks). The measurements are an average over about 10 seconds (using a feature of the multimeter). A consumer power meter was used with the socket power adapters to provide a rough indication of the power consumption. The figure below shows my quick "test bench".

 

 

The tables below show measurement results. For two power supplies I tried all the cables I had. If the specs are printed on the cable, they are given in the table.

 

CableSamsung 5.3V 2.0AModMyPi 5.25V 2A
from Samsung Galaxy S4 mini (23AWGx2c + 30AWGx2c), 1m5.065.31
from GIGABYTE GZ-G90B0 portable power storage, 1m5.185.38
"USB 2.0 high speed mini cable", 1m5.035.20
"USB 2.0 SHUNXINYUAN" (28AWG/1p 28AWG/2c), 1m5.095.28
unidentified cable, 1m5.055.26
from ModMyPi black (28AWG/1P+28AWG/2C), 0.5m5.125.32
from modMyPi black (28AWG/1PR+26AWG/2C), 1.8m5.075.25
from ModMyPi flat ribbon, 2m5.085.28

 

Conclusions: if the cable has no specs, consider it as a poor one, unless it comes from a power storage (in which case there is more reason for it to be good). If the cable has specs, you know what to expect.

 

Then I continued only with the best cable (i.e. the one with lowest resistance), from GIGABYTE portable power storage, to compare the rest of the power supplies.

 

Power supplyVoltage at RPi, VPower at the socket, WPower at the socket with RPi disconnected, WNotes
Samsung 5.3V 2.0A5.183.4-3.50
ModMyPi 5.25V 2A5.383.8-3.90.3>5.25 V at device, not good
Samsung 5.0V 1.0A4.963.3-3.50
Garmin-Asus 5V 1.0A4.983.6-3.70.1
Apple 5.1V 2.1A4.983.2-3.40
USB port on a PowerCube extender5.11--
USB 3.0 PC (with AsRock AppCharger driver)4.83--
USB 2.0 PC4.78--RPi power LED blinking
USB 2.0 HG655D modem4.82--

 

These results confirm the conclusions from Raspberry PI:- USB power cables, crashing and other problems. A good USB cable is essential, a over-voltage power supply is not needed. It appears that a power supply from ModMyPi leads to more than 5.25 V at device even with some poor cables (while 5.25 V is the upper threshold in the USB standard).

 

Moreover, with a good USB cable, powering from a PC USB port that is capable of providing more than the standard 500 mA (as in my case of an AsRock motherboard with the AppCharger driver installed in Windows 7) makes it possible to run applications on RPi that fully load the CPU. In this case, my rough measurements lead to an estimate of about 700 mA.

Anonymous