Raspberry Pi M2 Unboxing and Setup

OK, I’ve got my new Raspberry Pi M2 all set up and moved in. My working space has had the old EVO removed to the “hacker office space” and the Chromebox has been unplugged and moved next to it. Both, now, in the “someday pile”. Yes, I’m that happy with this little guy.

I’m making this posting from the Pi M2 and the photos in it were edited with GIMP on the Pi M2. I’ve had a browser, GIMP, 3 or 4 terminal windows one with top running and a few other bits all running at the same time and performance was fine.

WordPress editing is very “chatty” and the load on both communications and processing can be significant (the perils of interactive graphical interfaces over an internet link…) Right now, for example, while typing I’ve got about 20% of one CPU used by IceApe (browser). It can drop down to near nothing if you stop typing (and no ‘save’ is happening) or rise to nearly 50% at other times (typing and a ‘save’ in progress.) Do note that with 4 CPUs, it is often the case that a 50% use of one by IceApe is about a 25% of total CPU cycles used for everything.

So there is a little bit of ‘type ahead’ rather like on all other platforms mostly due to network lag to WordPress.

On some machines, like the original R.Pi, it can be painful. On this box it’s about like on most other machines; or just a bit slower.

Well inside acceptable.

With that, on to the unboxing:

Unboxing The Pi

A very small package arrived from Amazon. In it were several other things the spouse had ordered, and a small plastic bag stuffed with bits. I opened the bag, and spread out the bits. Here’s what’s in the bag:

First Opening of the package

First Opening of the package

Here, upper left, you can see the nice little cardboard box that has the R.Pi inside. Bubble wrap over antistatic bag in a box. Very well thought out packaging. Just below that is the R.Pi M2 sitting on the antistatic bag (obviously after removal from the box…) and to the right is the other “inner bag” of small parts sitting on top of the shipping bag (that was inside the Amazon box with airbag padding..) Through the inner bag you can see the WiFi dongle, the mini SD car and adapter to full size for use in other machines, and the power brick in the background on the HDMI cable.

Inside the small parts bag:

Small Parts Bag Contents

Small Parts Bag Contents

Here you can see the HDMI cable (for connection to a modern TV, or in my case, to an HDMI capable monitor) in the upper left. Just below it is the power adapter in it’s bag. The power adapter is oversized for the power draw of the Pi. Nice. Added USB devices are easily powered (such as my keyboard, mouse, Wifi dongle and a SD adapter all at once as I uploaded pictures from the camera SD card).

To the right is the “Quick Start Guide” as background under the small parts. Those are the mini-SD card with adapter (in a very nice carry case), the two tiny sized heatsinks, the WiFi dongle wrapped nicely, and the clear case. On top of the clear case is a tiny bag with two microscopic screws. More on them below.

At first the case was a bit of a puzzle. It was not clear how to open it. Being clear, there were a lot of reflections and speckle from it. I figured out there were 4 snap hooks near each corner on the two long sides, but what to push, pull, or twist? Do the wrong thing with big Farmers Hands (as my Dad called them, or Smith’s Hands that was more accurate as we were a family of working smiths for many generations prior to the one generation of farming…) and you have several small bits of broken plastic…

As gently as I could, I tried things. What worked easily was to push on the long side of the top case just under the very large hole where the expansion pins are located. That ‘unhooks’ the clips from the sockets on the bottom and then the top can be rotated up on that side and the other side clips easily leave their sockets. Putting it back together is the reverse, more or less, and easier as the ‘clips’ are wedges that can just be pushed into place.

It would have been very nice if the “Quick Start Guide” had anything at all about mechanical assembly, but it doesn’t. It is a generic (and low skill level) intro to the setup and software of a R.Pi Linux system and board. Mechanicals left to the imagination. Not good. It tells you what the sockets are, and where to stick the micro-SD card, then goes into config land.

Pressing on…

Case open with R.Pi M2

Case open with R.Pi M2

Here the case is opened. At the very far left is the thin strip I pushed (GENTLY!) to open the case. The “clips” are the ‘smudges” at each end of that strip. Yeah, hard to see. And this is with it already open… The R.Pi card sits in the bottom section. It ‘clips’ in sort-of. It’s a snug fit without any actual clips, but took a tiny bit more force than I expected to ‘seat’ it properly.

In the foreground of the R.Pi are, left to right, the Ethernet RJ-45 connector, and 4 USB connectors. Along the right side are the expansion pins for anyone wanting to do that kind of direct hardware work. On the left side are the small round connector used for headphones or other audio out (and where, but some method I’ve not explored, RCA video out can be plumbed if needed). The crosswise white bit is for plugging in LCD display panels, then there is the HDMI connector and in the fuzzy distance the power connector. Along the far edge is the (unused by me) camera connector.

Next, one gets to guess what to do with the rest of the parts…

I guessed that the two micro-screws might fit into 2 of the R.Pi mounting holes and bind it to the matching holes in the standoffs below (as is often done). There were no threads in the holes, and the screws didn’t look ‘self tapping’ and that ought to have given me some clue. But it didn’t. So I set up to install the screws via finding my Jeweler’s Screwdrivers.

Screws, we don't need no steenking screws!

Screws, we don’t need no steenking screws!

I laid the small screw bag on top left (despite the small static risk from the plastic) and set my screwdriver set next to it. Then carefully selected the right size driver. #3 was loose. #4 was ‘about right’, but #5 at ‘a bit big’ was also nicely snug.

Attempting to install the screws (in the small round gold colored screw holes near the corners) I discovered that they didn’t ‘grab’ anything. Just spin in the holes. Too small to do anything. I can only assume they are supplied to mount the board to something else, later, maybe…

Oh, and you get to figure out how to install the heatsinks. One gets to guess that it’s a peal off the cover and just stick on, and what orientation. I put the fins aligned such that air coming in the big hole could easily flow in, through, rise and exit.

Peeling the heat sink

Peeling the heat sink

Getting the ‘cover’ off the heat sink is not easy for large fingers, but doable. Then oh so carefully align and press onto the chips of the same size on the top of the board.

After that, time to close up.

So the screws were set aside (back in the R.Pi box with the bubble wrap and antistatic bag) and the top was simply clipped into place. This took a bit more force than expected since I’d not fully seated the board… but clipped together nicely. Especially after I specifically checked each clip and ‘homed’ the ones that were not fully seated… (i.e. squeeze each corner specifically).

With that, assembly was done. Time to hook it up and turn it on.

Here is a view from the bottom showing the power brick connected:

Bottom View of R.Pi M2 with power adapter connected

Bottom View of R.Pi M2 with power adapter connected

I’m a little concerned that the bottom has vents on the very bottom, but no clear exit for the warm air at board level. Maybe I’m not seeing the leakage areas, or maybe that chip doesn’t make much heat. Note at each end of the vent holes places to wall mount on screws. Also note the four tiny legs in the corners so the vents are off the table surface. In a few hours of use, I’ve not yet had any detectable heat issue, so it seems to work.

Also note on the far right the micro-SD card. I’d stuck my 64 MB chip in for the picture as it is a nice red/gray mix. In practice, I’ve used the supplied 8 MB card for bring-up as I didn’t want to take time loading the “NOOBS” code onto my card (yet).

(NOOBS is the Raspian bring-up install code that makes it an easy click-click …)

On the right side you can see the WiFi dongle in place.

Here’s the finished set-up with all the connections in place. I used wired Ethernet for the initial install (as I think the WiFi dongle likely doesn’t work with NOOBS, but I didn’t test it…)

Finished Raspberry Pi Model 2 on Chromebox

Finished Raspberry Pi Model 2 on Chromebox

HDMI and power connectors out the top. Keyboard and mouse in the middle USB slots on the left, WiFi dongle in the bottom one.

It is sitting on top of the Chromebox for a sense of scale. About the size of a pack of cigarets. Or like a double thick pack of playing cards. In this image you can also see the orientation of the heat sinks and where I installed them. I hope that’s the right choice…

In Conclusion

And with that, the unboxing and connecting was done.

Next posting will cover software set-up and installation.

As of now, it’s my “daily driver”. We’ll see if I can live on this box “most of the time”, or if for some odd reason I revert to the others. The ‘type-ahead’ in WordPress is on the edge of an ‘issue’; but that might be networking (through a 10 mbit 20 year old hub, to the WiFi router, to the boundary router, to the internet, to…) or it might be that IceApe isn’t the best at this use. I’ll try Chromium next and see if anything changes. I’ve had ‘type-ahead’ problems with WordPress on many boxes that were fairly fast, so I suspect network issues as most likely. CPU running only 27% even as I type this with live spell check enabled. Though saving the draft runs IceApe at 100% of one CPU, and hitting ‘reload’ on another page things are slow. Clearly IceApe doesn’t know how to use more than one CPU.

All in all, I’m quite happy. Maybe not an ideal workstation for high end power users, but quite adequate for most uses. The quantity of available software for free, and the lack of Google-over-my-shoulder already make me happier than the typical Chromebox experience. (The Chromebox is likely better for folks who don’t want to deal with things like hardware, set-up, software installs, and such; and who don’t care about all the “Tell Google Everything You Do” mis-feature).

I’ve already installed GIMP, a couple of browsers, Libreoffice and more. Now it’s just a matter of trying it and seeing how good it is.

Subscribe to feed

About E.M.Smith

A technical managerial sort interested in things from Stonehenge to computer science. My present "hot buttons' are the mythology of Climate Change and ancient metrology; but things change...
This entry was posted in Tech Bits and tagged , , , . Bookmark the permalink.

38 Responses to Raspberry Pi M2 Unboxing and Setup

  1. E.M.Smith says:

    Well, from the “why have several browsers” department.

    IceApe reads this posting fine. Chromium ‘slews’ it a lot and it’s a mess. Yet loading the top level “chiefio.wordpress.com” works fine, only individual pages don’t. Tallbloke.wordpress.com worked fine, but wattsupwiththat.com (also on WordPress) didn’t even load the top page right in Chromium.

    So it looks like Chromium on the R.Pi M2 is “young” and “needs work”. We’ll see if it has issues on other sites, or if it is peculiar to WordPress (perhaps some ‘themes’ send out quasi-broken code…).

    So no Chromium WordPress edit testing for now…

  2. beng135 says:

    Just a tip, for your other posting I was reading:


    You can speed up Ext3 and 4 Linux filesystems using either noatime or relatime in the mount options in /etc/fstab — less disk I/O:


    Might be worth testing…..

  3. Larry Ledwick says:

    Great break down, I was looking at those on the web and was having difficulty finding good enough visuals to get a feel for size and construction details.

    The clear plastic is ok for “show and tell” of what it is, but I would prefer a metal enclosure for RFI and electromagnetic shielding purposes.

    Not that anyone would want to build a tempest certified system, but for Ham radio operators stray RF plays hell with plastic case electronics. Looks like it might be sized so that it could be fit inside one of the die cast experimenter boxes with a bit of effort.

  4. Steve C says:

    Very nice. You seem to be heading (and rather effectively) in the opposite direction sizewise to me; I swear my current 4-core, multi-GB desktop box is the biggest one yet. (Although, to be fair, its PSU is nearly as quiet as your Pi’s!)

    Re browsers, I’ve been using PaleMoon for a while now and am very pleased with it, which makes me officially a Lunatic. It’s a Firefox fork, optimised for modern processors to avoid the ‘legacy code’ that holds FF back, and feels very much like FF used to, along with using (at least most of) the same plugins, including the important ones. Although it doesn’t inherently include Linux, Sourceforge hosts pm4linux, which works just fine. Try a little Lunacy!

    @Larry Ledwick – Oh, boy, are you right about RF interference. Sadly, most of it these days IME is neighbours’ built-down-to-a-price, barely compliant (if at all) electronic tat – the only time I’ve heard the background noise in years now is by taking a receiver out on a country walk. The EM spectrum is a horribly polluted resource now, in the city at anyrate.

  5. p.g.sharrow says:

    @Larry and SteveC; you guys are very right! plastic boxes are nice eye candy and not worth a darn for RF shielding. I am looking forward to your suggestions as we move along with this experiment. pg

  6. Steve C says:

    Unfortunately, p.g., my modest suggestions tend to be seen as “a little draconian” by non-RF-aware types. Here’s what a fellow hammer chewer found necessary about a decade ago, and it ain’t got any better since: note also that he was only trying to quieten the thing on one band, not necessarily across the whole spectrum.
    Making the PC computer quiet at 144MHz.

    Right. Screening like this for all digital electronics worldwide please, and an immediate death sentence for anyone designing, manufacturing, distributing or using the atrocity known as “Broadband over Power Lines”. All CFLs and LED bulbs to be recycled, and, and … and I’ll get back to you afterwards and let you know how many dB the noise has dropped. If I can find a quiet PC! ;-D

  7. John Silver says:

    Meh, just wrap it in aluminum foil.

  8. LG says:

    @ E.M.:
    Did you have a chance to test audio streaming and playback ?
    That was one my stumbling block on the previous rev. of the board…..

  9. Larry Ledwick says:

    @Steve C
    I can relate, when I first moved into my apartment I put a mag mount antenna on a small metal table for 144mhz near the window. Unfortunately one of my desktops was on the same table. Every time I keyed the mic to talk on 2M for the severe weather net, my computer would reboot.

    Take a memo, physical separation is a big help, and so is a bag of ferrite filter choke beads to put on all the leads.

    Aluminum foil wrapping can be very effective if you can fully wrap the entire component but with cable connections to USB ports etc. you have lots of entry points where those sockets have to penetrate the foil due to poor physical contact/grounding at the penetration.

    I used to mess around a bit with RDF (radio direction finding) to support a local ham radio group that went out to try to find sources of RF interference and crashed airplane beacons etc. It is really difficult to build an RF tight enclosure that will allow you to cool electronic equipment.

    As mentioned above, there are so many common things which are very dirty at the RF spectrum in our homes today, that they add up to an over whelming fog of RF noise. Things like using CAT7 shielded Ethernet cables, ferrite chokes and choke coils in cabling, along with lots of consideration for how to physically separate certain items is the general solution. Electrical gaps like using optical fiber right at the input terminations and wave guide beyond cut off penetrations, and tightly sealed Faraday shield enclosures are required to get high isolation. The good news is that those protective measures also provide good protection against line surges, lightning strikes and EMP.

    The usual result is you end up buying lots of stuff to try it out and find out that it is not suitable for one reason or another. Different manufactures are vastly different in their success at building quiet components. Some computers are broad band noise makers, as or light dimmers, lots of components which have switching power supplies etc.

  10. Paul Hanlon says:

    Very nice review. I’m green with envy. It looks gorgeous.

    I’ll look forward to seeing the software install post just to see what things I’m missing. It really is amazing just what these tiny boards can do.

    The major pain point nowadays seems to be the browser. I currently have 13 tabs open on Iceweasel. The average page weight is about 500KB, so round that up to 7MB for all. I look at top and Iceweasel is currently using 1550MB of Virtual memory. So for every one byte of “page”, it needs 200bytes of background information storage? Plus it’s using 9% cpu on a four core chip? So I suppose it’s no surprise that the small boards struggle a bit. Maybe Wayland will change that.

    Where they really seem to shine is when they’re headless. I notice no difference in speed between the webpage that is pushed out by my cubie server to the webpage that is pushed out by my working computer. There’s also no difference between the two when I stream a piece of music, despite the fact that from the cubie it also has to traverse the network.

  11. E.M.Smith says:

    OK, weekends I tend to have more obligations than week days, so I’ve been AFK in terms of internet for a while. Finally, midnight-ish, I have some time.

    During the day I also got to move a few dozen GB of data around… cleaning up a few years worth of accumulated duplicate copies of stuff I never needed to save, really…

    At any rate, a quick update on the “usability” of the R.PiM2 before I go on to trying to get a coherent posting done for the night. A few points:

    1) I still love the little sucker, but he’s not my “Daily Driver” now. I’m back on the Chromebox for general browsing stuff. (More on why below). Lot so good stuff in the R.PiM2, but “has issues”.

    2) It works great for a lot of things. A couple “not so much”, and those two are presently deal killers for me as “daily driver”.

    3) I sunk a good full day into trying to get sound to work. No Joy. The Alsa sound mixer has some kind of lethal error. I can get a test tone out the speakers directly if ALSA is bypassed; unfortunately everything else goes through ALSA. Attempts of all sorts to reset the master sound level failed (including the half dozen ways to ‘fix it’ found on the internet for pre-M2 boards). The sound level is stuck at -10,xxx or some such (meaning “mute”) and can not be changed, near as I can tell. Likely due to the sound kernel “module” continuing to be bcm2835 when the chip is a bcm2836 … or maybe not. At any rate, no sound means not usable for youtube or a dozen other normal ‘daily driver’ things.

    4) Plugging in my USB Toshiba 1 TB disk caused the keyboard to go flakey. No idea why. Well, I have an idea, but no idea if the idea is a good one… Maybe it sucks USB volts just a tad low? Or maybe the disk and KB argue as there’s only really one USB port that gets mapped to 4 sockets. In any case, I can’t use it to be an NFS file share of my existing disks to the ROO (Rest Of Office ;-) without working something out…

    So, with that, it become a much much faster version of what the Pi B has been: Nice toy, good small services server, barely functional web browser services and small size NFS server (i.e. from the SD card or second SD card in USB adapter).

    I’ll work out more uses for it later, over time, or when the Raspian port for it gets cleaned up enough that sound works without a week of screwing around…

    On to the CubieBoard as my next attempt at a very low end “daily driver” as the R.PiM2 goes to “small server” duties…




    Thanks for the tip. I’ll give it a try.

    @Larry Ledwick:

    Any serious radio work is done with the computers shut down. I’m a SWL myself and ANY computer in the area causes so much noise on AM it’s just a PITA beyond belief. It’s also part of why I layed in an inventory of incandescent bulbs… nothing like a switching power supply and arc in a CFL to make RF noise…


    IF you really want the shielding, and don’t want a DIY aluminum foil job, there are metal cases for the R.Pi. for about $10 more than the plastic.

    ought to pull up one for about $27 and the suggestions bar I got included a dozen more with variation in vents, fans, and such and from $16 to $20.

    @Steve C:

    I tend to be a minimalist at heart. That, and much of my professional career has been spent in things like efficiency reviews and improvements, and finding ways to do more with less. Then add in that it just offends me that most boxes today have more total computes than a Cray of the ’80s and can’t run a graphics window full tilt boogie without a GB of memory? Uh, somethings wrong with this picture…

    (With that said, our Cray XMP-48 cost $40,000,000 and had an add-on $1/2 million or so frame buffer to let it do full motion animation in full color in real time… with NO sound at all… and about the same added cost to put a Gig-E speed network on it… and I have effectively the same for $60 now… maybe I ought to stop complaining ;-)

    Just seems to me that with some decent care, it ought to run rings around the Amiga of the Battlestar Galactica era… yet it doesn’t…

    Don’t get me wrong, I still love the little guy. Just not quite ready to be a Chromebox killer. (That the Chromebox “just works”, reliably and well, for general purpose web stuff including decent video and nice stereo sound to my bluetooth headset is “enough”. That it makes it a PITA to try to edit a file on your own computer without sending a copy to the internet (and to Google) is a sin and makes having that Linux box on the side essential.) So the two, together, make a nice team that does all I really need. And eventually I’ll settle on a small cheap SBC that does it all.

    I also have an absolute hatred of fan noise. There are KW stereos without fans. A fan is an admission of engineering failure, IMHO. Just bolt on a nice chunk of aluminum or a heat pipe and be done with it… but no… have to stick in a a cheap, failure prone, POS fan to disturb thinking clearly at 2am… So one of my goals is a silent computer. The Chromebox is very close… the Pi is perfectly silent ;-)

    @John Silver:

    Yes, but… that plugs up the vent holes. I’d rather put it in a nice bronze window screen Faraday cage… (anything with a wavelength shorter than 1/8 cm I don’t care about leaking so much ;-)


    Couldn’t get the sound to really work at all. Only a direct .wav file test direct to audio made it off the box. Video was a tiny bit glitchy over WiFi.

    I’m going to try some of the other OS builds and see if they are any better (Arch and others). The hardware ought to be able to do it “no problem”, so I suspect “young software” that isn’t very efficient and maybe is not matched to the actual chipset all that well.

    But that will likely be a few weeks out.

    @Larry Ledwick:

    Easy to get RF shielding and cooling air, if you don’t mind a cut-off frequency. Look at your microwave oven. You can see in, but the RF can’t get out. Tiny little holes in a metal mesh inside the glass lets light out, but not RF (both EM spectrum…). Take away the glass, you have an RF shield for up to Mwave oven freqs but with air flow. So… make your enclosure out of tight metal mesh (i.e. ‘window screen’) and you get shielding with ventilation.

    You can also form it into nice little ‘tunnels’ the input and exit wires use, with a dogleg or two in the middle so most RF gets blocked.

    @Paul Hanlon:


    Yeah, the little sucker is cute. Also fun to work with. General web browsing is fine (as long as no video or sound is involved), and all “terminal mode” unix / linux stuff is just dandy. IFF it would work with my TB USB disk, I’d make it my main file server, but it doesn’t, so I can’t (unless I can ‘fix that’ via my USB hub… presently powering my older R.PiB as DNS server with hosts.block file so I can kill any site that annoys me ;-)

    I was surprised to see that it had a 1 GB “swapfile” defined. Never did roll onto swap, even with a half dozen pages open in IceApe (a variation on IceWeasel that is a variation on FireFox that…). I’d been hoping to get the USB disk on it (which has a 4 GB real disk swap partition on it…) but that’s on hold now. That Cubieboard with SATA is calling my name ;-)

    So I had no issue with the browser (other than Chromium being a pill… and Arora not working) as Midorie and IceApe worked fine. Epiphany was also quite nice. Video would sometimes have a bit of ‘glitch’, but I’m not sure if that was the browser, board, or WiFi and what all else I was doing.

    I did have one odd case when IceApe decided to go open loop on CPU demand. Pegged on CPU at near 100% at all times. (A not that uncommon mode of bug…). Yet, with 4 cores, everything else just ran along fine. Didn’t even realize it was a ‘runaway’ even while using it, but got to wondering why the CPU meter showed is using almost 100% all the time. Exit and restart fixed it, and don’t know what triggered the bug, but it didn’t come back in further testing. (And I’m not sure if my ‘video has a glitch’ was during that episode or not…

    Oh, I’m also going to try porting GIStemp to it. It might be amusing to have a package of GIStemp, a full set of GHCN data, the needed USHCN data and other bits, all on one SD card for the R.Pi. and an all up package cost of under $100. Would also be a bit of a ‘nose tweak’ to show the “massive CPU power needed for NASA models” as a R.Pi SBC ;-)

    The “how to set it up” posting isn’t done yet as I was screwing around with the sound problem, but it’s “up next”… unless I do one on soggy corn first ;-)

  12. Another Ian says:

    Larry Ledwick says:

    19 July 2015 at 3:20 pm radio service rooms


    Re shielding and cooling

    One of the Royal Flying Doctor Service bases that I saw had its radio service room walled with brass gauze with many solder joints.and well grounded.

  13. Paul Hanlon says:

    Hi Chiefio, that’s a real bummer. I think you are probably right that the problems you’re having with the HDD and Keyboard are power related. You’ll need at least a 2Amp wall wart to power both, or a separate powered hub.

    There’s two sound interfaces on the Pi, the HDMI one and the Composite one. If you go from a HDMI to DVI connector, I think you might have problems, as I think that some sort of conflict arises with that setup. When they went over to the four USB ports and 40 pin GPIO, they made some changes to the composite and audio ports. Instead of a normal headphone jack with three sections, and a composite out, they went with a four section jack where the fourth section provides the composite (PAL-NTSC) output.

    As far as I know, you can plug in an ordinary headphone jack and get sound out that way without the composite out, so if you had a set of those powered speakers that they used to ship with new PCs 10 years ago, they should work from the headphone jack, albeit at the expense of another power connection. You’ve probably read it already, but this link might be some help.

    As for the cubie, the two new things that it brings to the table is the SATA connection and the NAND flash memory, but with only two cores, I think you might be disappointed if you want it as your daily driver. The only time I saw the GUI interface was through rdesktop. I found it great to work with, but my version of great to work with and your version could be two totally different things.

  14. Larry Ledwick says:

    Yes very familiar with shielded enclosures. I used to have one similar to this web page at work for the EMP protection at the EOC. It had 80 db – 100 db isolation if no one did anything stupid with the penetrations (which happened at some local EOCs where local hams who did not understand high level shielded enclosures drilled holes in the walls to pass coax. (really really pissed off the emergency coordinator)


    A microwave is “fairly good” but not an excellent RF shield. Good expedient test is to put a small portable radio in the microwave tuned to a strong local station and see if the signal goes away when you close the door.

    Doing that test on my microwave on some stations it completely kills the signal but the strongest stations you can still hear the audio with the door closed. On the stations it does work on you only need to crack the door a fraction of an inch about a mm or so and the signal comes back. Moderate isolation is fairly easy, high attenuation is very difficult. That is why the charge so much for enclosures which achieve 80-90 db at common frequencies.

  15. John Silver says:

    Prick holes in the vents with a needle to emulate the screen you mentioned.
    You can afford many iterations to test for a satisfactory result.
    Cheap n’ easy.

  16. E.M.Smith says:

    @Paul Hanlon:

    PWR wallwart for the Pi is a 2 A job. I think the disk is just too much, so needs external supply. (Which I have… but can’t find my uplink cable for the hub at the moment… and won’t be looking for a day or two as it isn’t a priority). I expect in the long run it will be working.

    Yeah, I got sound out the headphone jack. HDMI is HDMI all the way, but to a monitor. Doing the mod recommended to /boot/config.h ? to let sound work with monitors that lie to HDMI (not sure mine does) resulted in a “black screen at boot” and a login from a remote computer via SSH to undo the config change… so not going there again any time soon ;-)

    Near as I can tell, the problem is 100% located as the ALSA mixer insisting sound ought to have the volume set on mute. None of the “fixes” makes it budge (even the very nice script I found and installed that does all the icky bits for you…). It LOOKS like the value changes, then you query it and the return is “mute” (minus 10,000 and something).

    So my guess is that the ALSA code has “issues” with the BCM…36 chip as opposed to …35 and they don’t have it worked out yet. Or that my install ( I just added a bunch of stuff without incremental testing…) scrogged it. All on the “test someday” table now. Needed to knock out some postings first ;-)

    So next up is “make new NOOBS” chips and try again. Both with vanilla Raspbian and test sound before adding junk like XMBC, and with alternate OSs like Plan9 that I’ve wanted to play with and Arch and…

    But I really think it is just a mixer bug… as the “bypasses mixer” test tone comes out the headphone jack fine.

    At my next big payday (and when I have time) I’m thinking that Octo-Core Cubieboard has my name on it ;-) Yeah, I could get far more computer for the same money at Weird Stuff used… but it’s just so darned cute… and I like saying “Octo Core!”… 8-}

    @Larry Ledwick:

    Oh, so you want a good shield… why didn’t you say so in the first place!

    ;sarc> ;-)

    @John Silver:

    Hey, I’m happy running it open on the RF. If I’d wanted it tin colored, I’d have ordered the grey case… I’m from that expedient school that just shuts off the computer when I fire up the SWL rigs.

    Besides, here in Silly-Con-Valley there’s so much RF smog already that it’s just hopeless anyway. My back yard has a power line across the far end, and telco distribution up the side. Airport radar about 10 miles away, and EVERYBODY has computers, microwaves, CFL bulbs, etc. etc. running everywhere. Not to mention the RF squibs from the “smart meter” and gas meter and the folks running LAN over power lines and…

    In short, when I really want to do SWL anymore, I pack up and leave this soup bowl of RF. If it’s just “can I get radio Havana again?” as might be interesting given recent events, then I just wait until 2 AM, turn off the CFLs in the house (and the dimmer incandescents) and with ONE regular incandescent on, fire up the shortwave… Oh, and sometimes need to turn off the satellite receiver boxes and / or TVs if someone left them running. Oh, and make sure the exterior CFL lighting is off. Oh… hell with it, I’m just going to drive up toward Mt Hamilton ;-)

    But yeah, for ‘quick and easy’ RF reduction, a roll of foil is your friend… I recently moulded it around a “canning funnel” (used to fill jars when canning food) and put a WiFi antenna in the middle to get signal from a little further away. Poor man’s parabolic reflector, sort of ;-)

    Then there’s the hat collection ;-)

  17. Pingback: Raspberry Pi Software Setup | Musings from the Chiefio

  18. E.M.Smith says:

    Well this is interesting…

    I’m making his comment from the RISCos operating system on the R.Pi.

    A couple of added notes:

    When doing the NOOBS install, it has a tiny bar at the bottom of the screen where you can change your location default ( UK ) and keyboard. I’d missed that the first time which was why I was given a UK keyboard. This time I saw it. It didn’t change the default kb in RISCos. We’ll see about Raspbian when I boot it.

    Also I explored the “Overclock” menu item in NOOBs. It had a specivic call out of “Pi2”. I suspect that the default if you do not set that in Overclock is to give you Pi processor speed. That would explain the ‘less than impressive’ perception of speed gain as there wasn’t any. All I got was more cores at Pi speed. I’ve set it to Pi2 and things are zippier to about the amount I’d expect for a 1 GHz processor as compared with a 700 MHz. Not obvious…

    I installed ALL the OSes. Thus this being from RISCos. RISCos is a non-Linux non-Unix all unique operating system designed for the ARM chip by the ARM chip designers… It wants a three button mouse and I’ve only got a 2 button attached at the moment, so we’ll see if I can do things like, oh, leave ;-)

    The NOOBS installer was downloaded fresh from the Raspberry Pi site. It was different from the one that came on the chip in one very important way. It told me some operating systems would not work…

    In particular, the Pidora port of Fedora is flagged as an “incompatible OS” as is the Arch port. I installed them on this chip anyway to see how many bugs they have. Both give a black screen on sttempts to boot and I powerfailed out of them. I’m going to swap this chip into my Pi1 and see if they boot well on it.

    This is a 64 GB fast micro-SD chip (card). It took about 12 GB to install all the OSes in one go (and several hours… I just went to bed and let it run…) At boot time, you get a nice selector panel and can tell it what to boot. I tried a quick Raspbian (just enough to set some bits, like the overclock) and then on the reboot launched ( or attempted to launch) Pidora, Arch, and now RISCos. RISCos looks to be working fine. I’m hoping the other OS levels work on the Pi1.

    FWIW, between the original install and this run of NOOBS the choices of OS presented changed. NOOBS has a “network only” version and an “offline” version. The offline version has Debian embedded in the NOOBS and can install a minimal Raspbian / Debian without a network connection. ALL the other OSes available are picked up from the internet. In the original install, I only had 1 choice each for the two media station OSes listed. Now it lists 2 of them. One for RPi1 and another for RPi2. So they figured out something needed to change between the two board types. Also, it now flags that Pidora and Arch don’t work on the Pi2 and warns you off.

    OpenLEC and some other name are the two media center OSes. I’ve not tried them yet to see if they make sound work, but presume they do since that’s kind of critical to their function… that implies it is only a matter of time (and likely short at that) for Raspbian ALSA to be fixed.

    With that, I’m going to reboot into Raspbian. Why? Well, this browser works, but it’s an alien (if pleasant and very fast) operating system and I’ve not had morning coffee or tea yet ;-) I need something familiar for a few hours before running naked in the fountain of new knowledge…

  19. p.g.sharrow says:

    Lol,” running naked in the fountain of new knowledge…”

    quite a word picture! ;-)

    Glad to hear about your progress on the R pi 2. My lady tells me that the Dog Bone stuff has arrive at the shop. Soon I will Have to quit putting off upgrading my brain. Your travails and solutions may save me a bunch of wasted learning mistakes. pg

  20. Steve C says:

    @EM – “only a matter of time … for Raspbian ALSA to be fixed”. Hmm. Having recently encountered the Gordian knot that is Linux sound infrastructure for the first time, I respectfully suggest that it’s not just Raspbian ALSA that needs fixin’ … :-|

  21. E.M.Smith says:


    Yeah… I get a bit poetic when I’m short on sleep, have morning dream-state carryover, or have a bit of hypoxia ;-)

    Just think of it as a metaphor for risky behaviour that can be exciting but often ends badly ;-)

    As in I had to powerfail the board to exit RISCos as I could not raise an exit panel with my mouse despite the browser letting me find a web page that said how you are supposed to exit…

    FWIW, it looks like IceApe has been made a bit more multicore aware. It’s now showing 222% of CPU used as I type this (and with a Lady Gaga video running in another IceApe window) so it can now “walk and chew gum”. The IP stack likely needs a bit more work (in IceApe) as when this window gets heavy use / load the video gets a few glitches / pauses. (Other things in other apps just keep on trucking though, so not a Raspbian issue IMHO). Still no sound, but when it gets fixed, the video experience here is “good enough”. Not silky smooth yet, but getting there.

    As you might guess, with 222% of 1 core at 1 GHz, the demand is much higher than 1 core at 700 MHz can deliver without painful pauses… so the R.Pi1 was just not going to cut it, ever, as Youtube box. This one will end up “good enough” eventually. The CubieBoard with the A15 cores in it will be just dandy. It is also possible that software other than IceApe with a plugin would be more efficient and make decent video too. I.e. the OpenLEC and similar dedicated XBMC box installs. (To be tried later…)

    So a bit more progress.

    That the various NOOBS option were changing “now” says that the software is still quite young for the Pi2 and folks are still finding and fixing bugs and variances from the Pi1 (but making progress). With the size of the user base, things will move quickly IMHO.

    At this point I’m happy with the performance of each core (now that I set the speed right) and find that I just can’t load up all 4 cores in my normal activities with at least one core worth of compute capacity idle even when doing a lot of stuff. ( I’ll “fix that” later ;-) with more compute intensive uses as I find my torture, er, “exercise” suite ;-)

    For example, right now with “top” running in a window, 2 browsers open (Epiphany running as blog manager and IceApe for “me stuff”) and with IceApe running a Lady Gaga video I’m at 63% idle on the CPUs (sometimes dropping to 40ish percent).

    IceApe does have some small pause / skip artifacts in the video when I’m typing (that is interleaving web based spell checking into the IP stream) but how many folks type and spell check in the middle of videos? So it’s not an 8 core Intel processor. I’m fine with that for $60.

    Now I just have to find out how to fix the audio (or wait for someone else to do it)

  22. E.M.Smith says:

    @Steve C:

    Yeah… talk about things that are way broken from first design point on up… part of why I’m thinking “run videos on the ChromeBox and let someone else fix ALSA”…

  23. E.M.Smith says:

    Just an update on the USB drive with USB keyboard issue:

    Putting the USB drive onto a powered hub has fixed it. I now have a 1/2 TB of USB disk mounted onto the R.PiM2 and have NFS shared it to the Centos box. I’ve also moved the whole shebang behind it’s own router so that I can use PXE boot (that must be it’s own DHCP server and I don’t want to be ‘experimenting’ were the spouse does her shoe shopping… or I’ll get all sorts of hyper-expidited-panic-stricken-bug-reports when the latest shoes don’t show up in 10 seconds and maybe she needs to set a static IP as I upgrade…).

    I got it to NFS export, even though it is an NTFS file system, but it looks like I’ve got a bit more to go before it is fully usable. The export only mounts Read Only. I suspect I need to install ntfs-3g to get read-write. (Either that, or the NTFS file system is an issue and I’ll need to format the sucker EXT3 or 4).

    So that’s what I’m working on today. Polishing up the 1/2 TB disk mount / NFS export business, installing dmasque DNS server, and setting up PXE boot; all in a secure pod a couple of routers in from the internet… (posting from there at the moment. A bit of detectable lag from being behind a really old left-over router…)

    At this point it looks like the RPiM2 will make a very decent file server, PXE server, etc. etc. server and be adequate for direct use as a browser box / home directory place. Just need to get the sound fixed for it to be all-around useful and youtube watching station ;-)

  24. E.M.Smith says:

    Now have NFS with read / write working.

    apt-get install ntfs-3g

    then a
    service rpcbind restart
    servicd nfs-kernel-server restart

    and all was good.

    I now have a very nice 1/2 TB up to 4 TB (depending on what I plug in) NFS / Samba file server and I can stop moving crap from machine to machine should I feel like it. Yay….

    I’ve also installed a bunch of other stuff, but that will be reported under the ‘SW Setup” posting. I’d only had this issue here as it was unclear if the disk issue was a hardware limit or not, and now that it’s on a hub, it’s clear that power was a hardware issue, but from there on out it’s all software.

    So for the stories of installing Apache, dnsmasq, PXE servers and more, I’ll be putting that in comments over here:
    or in additional postings if things are so big as to warrant it.

  25. p.g.sharrow says:

    @EMSmith; I am delighted to hear of your progress toward this nodal concept of computer systems.

    Inexpensive, nodes, SD cards for firmware, USB “drives” for storage. All easy to replace. Free, open source software with a world of people working to improve it. All beyond the control of government officialdom.

    Now what is needed is for the nodes to communicate via broadband on a ground plain. WiFI without the over the air RF. There must be someone out there that can make that work. pg

  26. Pingback: Raspberry Pi Build Script | Musings from the Chiefio

  27. E.M.Smith says:


    I’ve got a “build script” now:


    so all the “exploration” and futzing about is embodied in a final answer. (Well, final until I add something else to it ;-)

    I’ve only used other machines in the last 4 days for putting a FAT32 format onto the 64 GB card ( oddly XP wouldn’t do it so I booted the SystemRescueCD Linux that has gpartd on it… I’d likely have tried to figure out how to do it with the R.Pi – i.e. install gpartd – except it was shut down as I had the chips out of it!) and not much else. ( I’d left the NOOBS on the XP partition and needed to copy them off to the 1 GB SD card where they now live).

    At this point, a few SD cards, a USB disk & Hub, and a $40 generic card, and I’ve got a “build it any time any where” generic system with very little worry about “buggery”. No BIOS crap to worry about. No “PRISM” crap to worry about. LOTS of eyes looking at the code (and lots of hackers trying to break it and failing – mostly). Only real risk I see is that it is possible with this method of install (via remote archives over the internet of precompiled packages) that an “agency” could spoof the IP addresses and deflect you to their ‘special’ versions of code. So an eventual improvement would be a download of the code source repository and set up of a local build repository server system. On my “someday” list.

    For now, were I really really worried, I’d go to a public access location and do the build there. Just takes a Pi, my build SD card ‘kit’, a laptop as proxy monitor, and a network connection (that can be shared through the laptop). (Plus some time working out any bugs in the laptop-as-monitor set up… laptop with HDMI input would help…) Or just use a VPN to some foreign country where my exit node would be hard to sort out from the 10,000 other users of that service…

    The RaPiM2 is “good enough” at this point for just about everything. Could be a tiny bit faster for video and the WordPress posting editor (that is sluggish anyway and very network chatty) can have pauses in it. So at some point that “Octo Core!” CubieBoard is calling my name ;-0

    But I’m getting plenty of mileage out of this guy first. It’s quite reasonably comfortable for most of what I do (and doesn’t share a damn bit of my stuff with Google, either ;-)

    For that, I’m willing to wait a month or three for someone to fix the ALSA sound mixer …

    Don’t know what you mean by “broadband on a ground plain” or over the air without RF.

    IFF I wan’t private communications, a wired ethernet switch is very cheap and effective. IFF I want private wireless, the typical home access point router is fine (the encryption is good)… and IFF I want to assure it isn’t pre-buggered, I can use OpenWRT software on a variety of commercial hardware, or just put it on a R.Pi also. ( The R.Pi makes a decent wireless router / access point ).

    What is the goal you are trying to achieve? Communications inside 20 feet? (Bluetooth can do that) or inside 200 feet? (Regular WiFi dongles and open source software for secure and encrypted and not PRISMed..) or further? (Packet Radio or shotgun antennas on WiFi). Wireless and through the ground? Um, why? (ULF can do that, but the bit rate on ultra low frequency radio is horrid – fundamental physics of bandwidth vs fequency…)

  28. E.M.Smith says:

    I was having sporadic crashes of browsers ( All three of Epiphany, Chromium, and IceApe / SeaMonkey / FireFox(the clone) / whatver they call it today) when “up” for a long time. Sometimes they would just close for no apparent reason even while I was just looking at a page, or sometimes when I was doing something in another program. The other programs seemed fine, as did the core OS. Just the browsers would halt. I did have the whole windows environment lock up once and powerfailed out of it ( I did not test remote login so the OS might have been fine under the windows manager) but primarily it has just been browsers doing an instant exit / crash. (Firefox / IceApe does a nice job of presenting a recovery panel and asking to open all prior tabs. So that’s why I call it a crash instead of an exit. THEY think it was a crash.)

    I don’t know that this fixed it, but I changed the “overclock” from the M2 setting to “medium” ( 900 MHz instead of 1000, lower core rate, etc.) and left IceApe open all night.

    It’s still here in the morning so I think that might have been the issue. I’m going to run this way for the foreseeable future. If all stays good, this note will stay here. If browsers continue to just close for no reason (the OS seems to stay fine…) I’ll come back and note it, or just remove this note.

  29. p.g.sharrow says:

    @EMSmith; That problem of browser freezes might be a “Feature” of the provider’s servers. We have been noteing freezes from time to time. XP, Windoz 7, using Mozilla and MS, also my ICQ connection with my kids. Leave them on for an extended period and they crash, must be shut down and then restarted. Often the computer must be rebooted as they will not restart even though the other OS functions seem to be fine.

    The Ground Plain-Broadband comment has to do with the communication between units within a group connected to the same base. A kind of WiFI or crosstalk without the wires or RF broadcast to non-connected devices. Ethernet, serial and parallel require dedicated wires for communication between units as well as hardware/software management of the communication links. Over a single conductor or Groundplain the computers could “sing” to each other. Kind of like very lowpower WiFi, where all of the antenna are connected to a common conductor.
    Just a brain exercise, not something for you to worry about. Bread Crumbs for others;-)

    I have read all of these RasPi posts and comments, often several times! to get as much out as I can. You have really gotten into this project and I am delighted with the results so far. It appears that every time you perceive a potential roadblock, it falls before you. After you grasp the requirements and their solutions you can K.I.S.S. it so that the rest of us can grasp it. It would appear that you are the right man at the right time for this part of building the foundation of the new age. “The Net that covers the world” Is the tool that is as important as Gutenberg’s press. The Greedy Evil Bastards will no longer be able to suppress knowledge and hide their machinations. An inexpensive, low energy consuming “Beer Can” computer with a Clean OS and Browser are parts of this. We will pray that GOD will Bless your efforts. pg

  30. LG says:

    @ p.g.sharrow:
    In windoz_7, do apps fail to relaunch even after a Ctrl_Shift_esc > Process tab> choose process>Right click>end process tree ?
    Just curious…

  31. E.M.Smith says:


    Well, the back off to 900 MHz has ended the browser crashes. (It is more of a crash than a freeze. The browser just packs up and goes away in an unclean fashion.) I’m pretty sure it’s a question of the “overclock” setting (and maybe I’m not supposed to have set it to anything anyway ;-)

    Ethernet Hubs are basically just one big shared wire. Ethernet already knows how to deal with this via back-off-retry on collisions. What you are talking about, then, is using the local dirt (local ground-plane) as ethernet hub. Hmmmm….. Most of the ground based communications are VLF, ELF, or even ULF and have very low bit rates. Just was reading about that last night… ( you can make a VLF receiver out of an audio amp and some bandpass filters to “tune” it, plus some metal for the antenna… from round ball about 1 ft in diam to a 1 m whip to a 6000 m coil (wound about 1 m diameter) for a magnetic loop…

    I’ll dig up the details later… but…

    tells how to turn your PC with a sound card into a selective (tunable) VLF receiver. Now add an audio output into a shifted frequency into a rod into the dirt ground plus antenna coil and, you will have a bi-directional VLF comms link over which you can layer data (as the military does to submarines…. One of the bigger problems with VLF and related ( ULF ELF) receivers is the need to tune out the very high military power signals to hear the other interesting stuff. A non-selective receiver gets swamped with the mil traffic. Lots of articles on that in the VLF ham radio space.

    has decent info in it including a long list of international VLF transmitters and this about the PC radio:

    VLF signals are often monitored by radio amateurs using simple homemade VLF radio receivers based on personal computers (PCs). An aerial in the form of a coil of insulated wire is connected to the input of the soundcard of the PC (via a jack plug) and placed a few metres away from it. Fast Fourier transform (FFT) software in combination with a sound card allows reception of all frequencies below the Nyquist frequency simultaneously in the form of spectrogrammes. Because CRT monitors are strong sources of noise in the VLF range, it is recommended to record the spectrograms with any PC CRT monitors turned off. These spectrograms show many signals, which may include VLF transmitters and the horizontal electron beam deflection of TV sets. The strength of the signal received can vary with a Sudden Ionospheric Disturbance. These cause the ionization level to drop in the atmosphere. The result of this is that the VLF signal will reflect down to Earth with greater strength.

    So the R.Pi being easily headless would likely make a decent one. Would need to port the PC software over.

    Depending on the range of communications desired, you would shift the frequency used up or down. (Down goes further, but at slower rate, up doesn’t go as far, but carries more information). Below about 1 kHz the baud rate is horrible, so most “production” (i.e. military) tends to hang out above this in the 10s of kHz. The list of sites in the wiki ranges from about 11 kHz for Russian ‘navigation’ to 24.8 kHz for US submarine communications. (They need to say “Don’t Fire!” in less than a day ;-) No, honest! A recent record set in ULF was down in the 8 kHz ish area and sent 25 characters from the USA to Europe is about a day… ) ULF is likely too slow to be of interest to you, and ELF isn’t much better.
    http://www.lwca.org/ has the article:

    Communication Landmarks: Across the Atlantic Below 9 kHz
    Third test on 1 January results in 25 character message.
    Over three consecutive nights, noted British DXer Paul Nicholson in Todmorden, UK, received a series of very slow BPSK transmissions from Dexter McIntyre W4DEX in Stanfield, NC, culminating in a 25 character message in the wee hours (UTC) of New Years Day 2015. Distance from W4DEX to Todmorden is 6194 km, or 3840 miles, and ERP is estimated around 150 μW.
    The first signal on December 30th was Dex’s grid square, “EM95,” the first intelligence ever conveyed across the Atlantic on such low frequencies by private individuals. The next day, a 12 character message (“PAUL HNY DEX”) was sent and received over a six hour span commencing at 0000 UTC. Then, in Paul’s own words: “The third test and best result so far was a 25 character message ‘8822HZ 2015 JAN 1 TA TEST’ sent from 2015-01-01 00:00 using 8 second symbols. This was received with Eb/N0 = -0.1dB. In the 0.125 Hz bandwidth of a code symbol, the S/N was -13.2dB.”
    Frequency used was 8822 Hz. Earlier in December, tests yielded copy of Dex’s steady carrier near that same frequency by DL4YHF in Bielefeld, Germany (4300 miles), by IK1QFK in Cumiana, Italy (4770 miles), using custom Linux-based spectrum analysis software written by Nicholson. He also wrote the BPSK applications. Reception has also been reported in the US using Argo displays. A selection of those reports is available at Dex’s Web site.
    Back on June 2 (UTC), 2014, Nicholson received transmissions from McIntyre on 8970 Hz, consisting of a continuous GPS-stabilized carrier that was shifted in frequency after some hours of integration time. Paul’s detailed reception report of that earlier feat, along with details and pictures of the transmitting setup, can be found at the W4DEX Web site.
    Previous experiments in the sub-9kHz region have mainly been conducted in Europe, such as the kite-based transmissions of Stefan Schaefer DK7FC. His page contains a good introduction to what has been done so far, and links to additional resources. Although frequencies below 8.3 kHz, or 9 kHz in the US*, are not formally allocated in most countries, some administrations do require their citizens to obtain special permission to operate there. (*Following the last WRC, an increasing number of countries now have an explicit allocation down to 8.3 kHz for meteorological aids such as lightning detection networks.)

    http://www.vlf.it/ has interesting stuff too…

    Personally, I’d likely expect that I didn’t want the signal to cross the Atlantic and go up frequency into another band…

    In the United States, there is a special license-free allocation in the longwave range called LowFER. This experimental allocation between 160 kHz and 190 kHz is sometimes called the “Lost Band”. Unlicensed operation by the public of any mode that falls inside the 30 kHz bandwidth is permitted, except where interference would occur to licensed location service stations located along the coasts. Regulations for use include a power output of no more than 1 Watt, a combined antenna/ground-lead length of no more than 15 meters, and a field strength of no more than 4.9 microvolts/meter. Also, emissions outside of the 160 kHz–190 kHz band must be attenuated by at least 20 dB below the level of the unmodulated carrier. Many experimenters in this band are amateur radio operators.

    Here’s a guy made a LowFER transmitter for / using his Arduino (a cheap low end computer unlike the expensive high end Raspberry Pi ;-)


    It doesn’t look too hard…

    You could try some even higher frequency open bands, but the propagation through soil drops off pretty fast and you are essentially starting to do “funny band WiFi”… And do remember the bit rate is inverse with the frequency (which is why WiFi is moving up GHz and getting faster). For really local “box talking to box” I’d be more likely to go to an IR transmission system. No leakage outside of line of sight, doesn’t propagate to the other side of the State, etc. etc. Then again, a REALLY trick system would be an IR main high speed comms with a LowFER backup link for things like saying “My GPS is xx.xx.xx point there”…

    Per Problems and Like Water:

    If there is one thing I have as an advantage it is that I don’t get flustered much. That Buddhist thing of being the empty vessel… I “become like water” and flow around roadblocks. Often works out well. Sometimes fails spectacularly, but then you just make a new puddle from the raindrops and flow on…

    “Never give up, Never Surrender!” coupled with “Be at peace and flow with the energy around you”. A Samurai Sword with an Aikido movement…

    So I hit a roadblock, and look what is behind it. On each side of it. Under or over it. Where does the water flow? What do I need to do to be the water? Learn Ruby? Or ask a friend? Or just size that the problems are large, so no longer of interest, and take a different path to the ocean…. but reach the ocean I do…

    It is a highly valuable way for a Tech P.M. to work, and has helped me fix a lot of broken systems in a lot of companies over the years.

    Sidebar Story: One night I’d been unscrambling some seriously broken network design issues for a major client. I, and the “Network Manager”, had worked all weekend. He had inherited the mess, and then layered on some more issues trying to cope. (Among other things they had multiple IP numbers running on the same wires and had to route from one machine to another on the same wire… which tends to cause collisions…) Well, it was a long weekend, and at about 11 PM Sunday we were ready for the last step. That was just to reboot the boundary router so it would pick up all the local changes. No real changes to it. We rebooted. It laid there. Dead. WTF?!? Not even blinky lights. Network Manager looks at me with that “What Have YOU Done?” look. I walk him through “We made NO software changes to it, and we agreed to power cycle, as you know it doesn’t break anything and can happen at any time with a power failure.” and eventually he accepts that this was just marginal hardware that died on the power cycle. Staff show up at work in 8 hours. NOTHING is open to sell you stuff. What do you do?

    I looked at him and said “Well, lets look in your boneyard and see what we have to build a router.” He was dumbstruck. The idea that you could build a router was a bit alien to him. As luck had it, I didn’t need to make one with Linux and a PC… they had some older CISCO parts in the closet including a chassis. After a couple of hours of moving hardware bits around, that chassis would boot. Then we just needed a config. The most recent one was not on backups, but “new enough” was. From it, I made what they really needed. At about 7-something AM Monday we had a config done and likely good hardware. Time to boot and find out… At about 7:59:59 AM it was up and running fine. A couple of ‘early birds’ had to be told to get coffee but when the bulk of folks turned on their desktops at 8 AM things were working very nicely. He was astounded, but relieved, having thought the task impossible, he thought I was of impressive skill. ( It isn’t that hard, really…)

    I left it as an exercise for him to get the old router hardware fixed under contract and swap back later…

    Now being able to “make a router” is a nice skill set to have, but I’d not done it under those conditions before, nor with that particular mix of stuff. The “magic sauce” was really just the willingness to “go there” and flow downhill around the broken hardware…

    On a later contract I made a 5 port router out of a Windows box (XP or NT or some letters…) with QOS on it for another client who specifically wanted that set up. I’d never dreamed of making a router out of a Windows box before, but “be like water…” the stuff can do it. Not too bad a router, either ( it was a high end box). They did video router testing with it…

    The constant thing is the willingness to flow, and not give up so soak into the dirt…

  32. p.g.sharrow says:

    :-) Ground plain is in the box, not in the dirt Lol! Often used as an electrical “ground” or noise dump or in concert with shielding to prevent RF radiation leaking out of the box. Computer devices connected to the common ground plain should be able to communicate with one-another without the need for additional connections. Get rid of the router. Modern computers work at such high speeds that the router is a bottle neck. Let the computer act as it’s own router/server. Sing to and listen to the others connected to the common ground.

    The ability to examine a problem from All sides is very rare. Most only approach from the front and if stymied there, give up. Often after examining the problem upside down and backwards, the solution becomes obvious! The way to the solution is in thoroughly understanding the problem. Then you can formulate a new way to solve that problem. pg

  33. E.M.Smith says:

    Ah… I am more into radio and antenna stuff where the “ground plane” is the area of dirt under the antenna (where the ‘virtual antenna’ is reflected for vertical antennas) and that thing in the box with the tubes ( hey, I learned this long ago ;-) is called the “chassis ground” (connected by wire to the ground rod). What you are talking about is “ground loop currents” IIRC. Typically a great deal of effort goes into suppressing those as they generally cause all sorts of grief.

    Yes, you could send signals through the “chassis ground plane” ( splitting the difference ;-) as a ground loop current. Would need to be careful with what all else it leaked into.

    I would protest, though, that either network speeds are limiting or that using a large chunk of noisy iron is more effective than a clean bit of copper…

    Modern network speeds are quite high, and you can strap them together for more speed. It isn’t just Ethernet either… There are a bunch of things like fibre channel and more. And for just getting chips to share with each other, the fastest thing to do is put them all in a mesh of communications links “on the die” like the Epiphany chip.


    ABSTRACT Research on Networks on Chips (NoCs) has spanned over a decade and its results are now visible in some products. Thus the seminal idea of using networking technology to address the chip-level interconnect problem has been shown to be correct. Moreover, as technology scales down in geometry and chips scale up in complexity, NoCs become the essential element to achieve the desired levels of performance and quality of service while curbing power consumption levels. Design and timing closure can only be achieved by a sophisticated set of tools that address NoC synthesis, optimization and validation.

    The basic point being that network tech is so good and fast that it beats doing dedicated interconnections (and all the wires THAT takes) on a chip with multi-cores and using a network on the chip is better…

    It does depend a lot on the class of problem, and processors, though. SIMD is Single Instruction Multiple Data and is the old Cray approach. You set up an array of 64 data items (called a “stride” and handed it to the vector unit. One instruction (say “multiply”) and you get 64 A x 64 B to give you an array of 64 C results. MIMD is Multiple Instruction Multiple Data and is more like the modern cluster of 4000 CPUs inside one skin or a Beowulf cluster built at home. Each one has its own data and its own instructions being executed. (Yes, you can think of a classical single core single thread PC as Single Instruction Single Data).

    For SIMD machines, the data pathing is very important as you must set up the arrays and hand them off as fast as the vector units can take them. (Or the multiple chips in an array or the GPU in PCs today). But that takes time, and often a network can deliver plenty of work (think of how you can drive a low end GPU to the wall with a networked video feed…)

    For MIMD machines, they are typically connected with at least a Gbit of network speed to each processor, and each one might be chewing on any given set of math for quite a while. (And if a 200 Mb ISP connection can drive your GPU to full use with video, a 1 Gbit can do more… and some folks are faster than that….)

    So not to rain on your parade, but I think the idea has more use for lower speed lower cost communications or communications between things that are not interested in wires or being bolted to a chunk of metal together…

    FWIW, the supercomputer guys are steeped in this stuff up to their eyeballs. It is ALL about optimal pathing for a given problem set and given hardware available. Buckets of money get pushed at whichever part is the “bottleneck” at any one time. Then it isn’t anymore and the money moves to the other parts… then back again…


    A Department of Defense supercomputer network is being upgraded to speeds as high as 40 gigabits a second.

    CenturyLink and LGS Innovations will enhance the five supercomputing sites that comprise the Defense Research and Engineering Network (DREN), according to a CenturyLink news release. Current DREN speeds range from 50 megabits to 40 gigabits per second.

    The Navy DoD Supercomputing Resource Center at NASA’s John C. Stennis Space Center in Mississippi, which is used for rocket engine research, will be the first site to be upgraded.

    Yeah, 40 Gbits… and you could strap those in parallel for even more speed if the computers needed it…


    But for now, Linpack rules the roost, and by its measurements, the top five computers on the new Top500 list, announced Monday, are unchanged from June’s list:

    The Tianhe-2 machine at the National Super Computer Center in Guangzhou, China, with a performance of 33.9 quadrillion flops, or 33.9 petaflops.
    Titan, a Cray machine at Oak Ridge National Laboratory at 17.6 petaflops.
    Sequoia, an IBM machine at Lawrence Livermore National Laboratory, at 17.2 petaflops.
    The K Computer, a Fujitsu machine at the RIKEN Advanced Institute for Computational Science in Japan, at 10.5 petaflops.
    Mira, an IBM machine at Argonne National Laboratory, at 8.5 petaflops.

    Next came the only new entrant on the top 10: Piz Daint, a Cray machine at the Swiss National Supercomputing Center.

    Basically, until you are in that class of computing, you have no worries about communications being the limit. These things are typically MIMD machines with a load of cores and comms between them. You can check the top one at any one time and see what tech they use to keep up on what changed from last year (as it always does).

    One specific problem is that “accelerators” — special-purpose add-in processors including Nvidia and Advanced Micro Devices graphics processing units (GPUs) and Intel Xeon Phi chips — boost Linpack scores a lot but aren’t always practical. As Dongarra and Heroux said:

    For example, the Titan system at Oak Ridge National Laboratory has 18,688 nodes, each with a 16-core, 32GB AMD Opteron processor and a 6GB Nvidia K20 GPU. Titan was the top-ranked system in November 2012 using HPL [Linpack]. However, in obtaining the HPL result on Titan, the Opteron processors played only a supporting role in the result. All floating-point computation and all data were resident on the GPUs. In contrast, real applications, when initially ported to Titan, will typically run solely on the CPUs and selectively offload computations to the GPU for acceleration.

    Accelerators do help in some circumstances. And meanwhile, researchers are adapting their supercomputing software to take advantage of these accelerators, and software tools are helping them.

    That kind of “where does my problem really get work done” is what really dominates, not network connections.

    See: http://www.top500.org/

    for what’s happening at any one time. Then just follow the leaders back to a description and you will usually find what is limiting at the moment, too. (Like now it is GPU vs CPU work partitioning)


    For the fifth consecutive time, Tianhe-2, a supercomputer developed by China’s National University of Defense Technology, has retained its position as the world’s No. 1 system, according to the 45th edition of the twice-yearly TOP500 list of the world’s most powerful supercomputers. Tianhe-2, which means Milky Way-2, led the list with a performance of 33.86 petaflop/s (quadrillions of calculations per second) on the Linpack benchmark.

    At No. 2 was Titan, a Cray XK7 system installed at the Department of Energy’s (DOE) Oak Ridge National Laboratory. Titan, the top system in the United States and one of the most energy-efficient systems on the list, achieved 17.59 petaflop/s on the Linpack benchmark.

    The only new entry in the Top 10 supercomputers on the latest list is at No. 7—Shaheen II is a Cray XC40 system installed at King Abdullah University of Science and Technology (KAUST) in Saudi Arabia. Shaheen II achieved 5.536 petaflop/s on the Linpack benchmark, making it the highest-ranked Middle East system in the 22-year history of the list and the first to crack the Top 10.

    The other nine systems in the top 10 were all installed in 2011 or 2012, and this low level of turnover among the top supercomputers reflects a slowing trend that began in 2008.

    A detailed analysis of the latest TOP500 list will be presented Monday, July 13, at the 2015 International Supercomputing Conference in Frankfurt, Germany. At that time, the full list will also be published at TOP500.org.

    Although the United States remains the top country in terms of overall systems with 233 (up from 231 in November 2014), this number is down from 265 on the November 2013 list. The U.S. is nearing its historical low number on the list.

    Other highlights from the 45th list

    Total combined performance of all 500 systems has grown to 361 Pflop/s, compared to 309 Pflop/s last November and 274 Pflop/s one year ago. This increase in installed performance also exhibits a noticeable slowdown in growth compared to the previous long-term trend.

    There are 68 systems with performance greater than 1 petaflop/s on the list, up from 50 last November.

    The No. 1 system, Tianhe-2, and the No. 7 system, Stampede, use Intel Xeon Phi processors to speed up their computational rate. The No. 2 system, Titan, and the No. 6 system, Piz Daint, use NVIDIA GPUs to accelerate computation.

    A total of 90 systems on the list are using accelerator/co-processor technology, up from 75 on November 2014. Fifty-two (52) of these use NVIDIA chips, four use ATI Radeon, and there are now 35 systems with Intel MIC technology (Xeon Phi). Four systems use a combination of Nvidia and Intel Xeon Phi accelerators/co-processors.

    Ninety-seven percent of the systems use processors with six or more cores and 87.8 percent use eight or more cores.

    HP has the lead in the total number of systems with 179 (35.8 percent) compared to IBM with 110 systems (22 percent). Last November, HP had 179 systems and IBM had 153 systems. In the system category, Cray remains third with 71 systems (14.2 percent).

    Note that number of cores and how the GPU / CPU work together (and how fast each) are the concerns? Notice the lack of any mention of network / communications gear? The networks are not the limit… so get no notice.

    I know, not exactly relevant to a guy and a garage, but it does clearly show the state of the art and what works (and what is a problem).

    Were interconnects and networks the limiting factor, we’d not have thousands of cores communicating with each other…

    Oh, final note:

    SOME problems do still need One Large CPU. And that needs One Large Data Path. But that class of problem will not be distributed between machines, so the network doesn’t matter… the disk, memory, and cache speeds matter…

    So that’s why I was thinking the goal was not speed, but ease of connection (no hardware dedicated) and / or long distance…

    If you just want to connect computers and data fast, this is the common method:

    Light, being higher GHz, carries more data faster. (Yes, FC can run on copper if needed).
    Equipment widely available and not too expensive (but not cheap…)

    Personally, I’d still use the LED / light path for communications. Lots of data bandwidth available. Use it with a fibre if you want privacy, or in the air (even inside the chassis) if ‘sharing’ inside the box is OK… It will be faster (and maybe even cheaper) than through the chassis.

    Per problem solving:

    I’ve often noticed other folks give up a lot. Didn’t think much about ‘why’…

    I usually call my method “middle out and ends in”. I look at both ends, and the middle steps, and build out from each toward the others… Sometime some segments get set aside for their own “middle out and ends in” if sticky… I guess that is different…

  34. E.M.Smith says:

    Well, having set the “overclock” back down to the 900 MHz that is the advertized rate for the M2, I’ve been able to leave IceApe open overnight for a couple of days without failures. At this point, I think it is pretty clear that the instability issues were entirely related to the overclock setting of max available. Lesson learned…

    Since I can’t tell any preceptable difference between 1000 MHz and 900 MHz, I’m not bothered by it. And since the volts and “core” settings doen’t seem to matter to me, either, I’m happy with the “medium” of 900 MHz clock.

  35. Pingback: Cloning USHCN Daily Data | Musings from the Chiefio

  36. Pingback: Well That Was Fun, sort of… | Musings from the Chiefio

  37. Pingback: RaTails – Draft High Level Steps | Musings from the Chiefio

Comments are closed.