I love the RPi and what it has done to introduce computing/programming/linux to more people. However I can't get past how unreliable booting from an SD card has been for me.

I know reliability depends heavily on the SD card used and on the quality of board socket but I am disappointed that Rev 2 of the board did not come with the option of on-board flash.

Been lurking for a bit. Was hoping to say hello with a constructive comment but I have to jump in here.

I bought a Pi model B for my daughter. Nothing but terrible unreliability. Booting problems as you state, power problems with the USB ports, random hangs and crashes and problems getting half of the devices recommended to work.

I kind of expected that for the money but the thing that really pissed me off was when I raised these issues constructively on their forum, I was met with nothing but vicious denial and had my post deleted and account closed within an hour. Possibly the worst bit was the insinuation that I was clearly an idiot and this was all my fault. I'm technically literate (I'm quite handy with anything from an oscilloscope to GDB) but I can't imagine this doing well in the hands of teachers I.e. one of the target markets.

She now uses an old ThinkPad X61 and finds it a million times more useful and with all the extra bits required on the Pi to get it to work, considerably better value for money.

Not impressed. When your policy is to silence critics the product should be distrusted.

As for the initial design brief to produce a useful educational computer, I think it fails entirely based on the sheer amount of friction between opening the box and writing your first program. I was brought up on a diet of BBC Micros and Acorn machines and despite the brand association, the Pi is nowhere near the mark. Her school even bought 20 Pis that are wheeled out for open evenings and otherwise never used as well in favour of bog standard PCs running python.

I currently have 2 model B's and one A. I had an unsupported SD card in the beginning, which annoyed me (it wasn't mentioned anywhere that this could happen). But you can check here if a card will cause problems: http://elinux.org/RPi_VerifiedPeripherals. After finding a good one: no more problems.

I also blew up a cheap chinese power supply 1A/5V but since I have been using a 2.5A/5V power supply I have had no problems (I back feed it via the "front" usb port, the hub has power on the input line, although that is not recommended.)

After that I never had any problems, 1 has been running OpenElec (XBMC) for more than a year now. Another one switches my lighting, also never any problems (running 24/7 for over a year), it is fed via the USB in my router. One of them also spent half a year in a data-center (http://raspberrycolocation.com/), it never failed.

There is no excuse for being treated so poorly on the forums but I have the feeling like you just had bad luck so far. It happens, My Moto G had yellow pixels and a failing audio system. Still a good phone judging from the experience of others.

If you really just want it to work out of the box, buy the recommended power supply (2A/5V I think), a good SD card or a preinstalled one with NOOBS (http://www.raspberrypi.org/help/noobs-setup/).

My wife is on her 3rd Moto G due to charging problems so there are at least two anecdotes there.

As said elsewhere (see entire thread) it was most likely a duff one.

My problem is with the community and the support and the quality, none of which are suitable for the target market.

As one of the mods on the Pi forum, I can categorically say - if you had a post deleted and your username banned, it was down to your behaviour, NOT criticism.

We simply do not delete posts or accounts unless the poster is overly rude or clearly trolling. The forum is frequented by a lot of children, so it is essential that things are kept polite.

This is reflected in the number of people banned from the forum in the last 3 years which is less than 50 (excluding spammers), and the excellent reports back we get from people of the high quality of forum etiquette and postings.

So, I would have to take issue with your comments on the forum. If you were banned, you brought in on yourself.

Can I also suggest that if your daughters school has some Pi's that are not being used, then the school send a teacher to a Picademy session - they are two day courses, free, held at the Raspberry Pi offices, specifically for teachers to give them ideas on how to use Pi's in the classroom. Lack of use usually means lack of knowledge of what is possible, rather than lack of ability in the device itself.

Total unadulterated bullshit.

My first post (perhaps this was a grave mistake), I gave a concise rundown of all of the problems that I had during the process of provisioning the thing and explained my dissatisfaction. I was not rude or condescending. In fact I asked for suggestions and offered improvement advice. Perhaps advice (from someone who's being working in the embedded systems sector for 20 years isn't real advice?). If I'd bought a MacBook Pro and it had the same functional problems, I'd say the same things about it. If I bought a vacuum cleaner that didn't work, I'd say the same things about it.

One of the problems was that the thing wouldn't output on composite at all (we didn't have a device handy that took HDMI in) and that I complained about the fact you have to poke keys in the dark to configure video (no video output) when you don't know if the hub or keyboard is working properly. Then that started the whole process of hoop jumping to get Linux running on it. Perhaps you remember me now?

I had about 4 replies to my comment, two of which were backing up my claims and one which was actually helpful.

Then the next thing I know, I went to reply to the post by someone trying to be helpful to say thanks and bang, no account. That's poor moderation and trying to hide dissatisfaction and criticism which is not acceptable. That's entirey your problem as a moderator.

Well lack of use in their case is that they don't have the time to do the hoop jumping when it's pretty easy to fire up a python interpreter from an MSI package on a windows machine to get the same results. I thoroughly approve of this; the first experience of proper computer science education shouldn't be hoop jumping and trying to work out how to get the thing working. Much like the BBC Micro, it should be plug it in, turn it on and open the User Guide.

90% of people fall at the first hurdle and you made it a very high one.

In absolutely no way would I recommend it to them.

Now I'm going to take a stab: The Picademy thing sounds much like the old school "big vendor" sales approach. Come see our products and you'll buy and market them for us. Sorry but no banana. I can't recommend the Raspberry Pi in any way and would not enable any promotion of them.

Maybe you should look at exchanging your Pi for another one. Manufacturing defects are rare but can happen.

I've used dozens of Pi's and never had bad reliability problems. Always use a quality 1 amp or more power supply, good name brand SD card, never remove power without running shutdown command to halt card writes, and never plug or unplug USB devices while powered up (older model boards don't handle the voltage drop well). I have never seen a corrupted SD card when following those rules.

I probably should have but I am disinclined to bother with it or promote it after getting a face full.

If you can't hot plug a USB device, forget it.

Power supply was a Thurlby PL330 linear bench PSU (£350 worth) so I'm sure that bit was fine.

I'd rather pay £10 more for something that works. To be honest a lot of no brand Chinese Ali Express dev kit I've bought is more reliable so I'm not sure what they're doing wrong.

I've only had good experiences with RasPis, but I do find the community management troubling (I've heard others also mention perfectly reasonable posts that were deleted). It's a shame, really, they're shooting themselves in the foot.

I assume you did get the constructive feedback that you need a proper 5V power source? Because it will boot with lower, but then randomly crash all the time.

Yes. I did a V-I curve of the wall wart and it was fine. I inspected the ripple under load with my scope too. I also had it hooked to a TTI bench PSU at 5.000v and it still had problems (I have a large selection of proper test gear at my disposal).

They didn't like those facts.

Then you've got a dud. Lots of the first rev in particular had a few problems like this. That's ample evidence to exchange it for a new one.

The later revs, particularly the B+, seem much more reliable (although occasional duds definitely do still happen).

Well you're probably right but after that face full, it went in the bin as i don't have the inclination to see if my time will be wasted again.

Just to give a different perspective, I think booting from SD card is my favorite aspect of the RPi (although I do not disagree with your comment).

This is because I find it gives me easier flexibility in switching between kernels with embedded userlands. I run my OS from RAM with only tmpfs for secondary storage, and I make my SD cards read-only. I do not try to use an SD card like a HDD, but I am aware that many people probably do and I am probably in the minority.

In sum, I find booting from the SD card is just easier when developing than constantly re-flashing on-board chips.

I doubt that this was the intention behind having the RPi store the bootloader on an SD card, but I am thankfule for it.

For ease of development for people like me, if nothing else, I sometimes wish more ARM evaluation boards would provide a similar boot from SD card option (where I choose the bootloader), in addition to on-board flash.

Would r-pi work better with 1gb or more of ram, or would that push the price into a prohibitive range?

What about usb3 or sata/esata?

The SOC doesn't support more than 512MB of RAM (or USB3). Raspberry Pi is built to a price point above all else - you could keep adding/upgrading features but you'd just end up way out of their target price.

One of the stated goals of the RPi foundation is to keep the price as low as possible. Adding a few GB of NAND flash adds ~$5 to your BOM for the part, plus you have to redesign and solder that extra component to the PCB. All told, it would probably increase the price of the board by ~$10, and cause suppliers to have to maintain additional models of boards in their inventory.

If you want a board with storage built in, look at the BeagleBone Black or the CubieBoard (there are several others as well). The communities for these boards aren't quite as large as the RPi's community but should still be large enough that you can get help doing whatever you need.

I am of a similar opinion. Here's my story.

I wanted to use the r.pi for storing sensor data sent from an arduino over serial port. The arduino worked rock stable but there were so many problems with the r.pi. Let me describe some of them.

First had a problem of r.pi freezing every few days. This was the first revision. The community said I must have had a bad SD card or a bad power supply or a bad USB hub. I was determined to make it work. I went ahead and bought 3 more power supplies and 2 USB hubs and 5 SD cards all of different but recommended models for use with r.pi. The freezes did not go away. At this point the community just started shrugging or saying I must be doing something wrong. This got me very frustrated.

So one day I ordered a new r.pi (model A), plugged in place of the old one and freezes were gone, ha!

However, there were still problems I could not work out. For example running lcd4linux with dpf-ax display on USB would make the r.pi freeze/crash (with kdb session spawned on r.pi serial port) after a few hours.

Another problem was data corruption on power loss. For example I had a FAT formatted USB flash stick to which I was logging data. And one day after an electricity blackout I find 80% of the files on that stick corrupt. I don't know why that would happen, I would expect maybe the active files being affected but not the files that had not been read nor written to since months before. Here is an example of data corruption [1]. Can anyone explain how this can happen? Any ideas?

I addressed the lcd4linux problem by buying a beaglebone black which has no such problems. I am still scared that one day my data will corrupt again so now I am logging straight to github.

To summarize, I think r.pi was definitely not worth the time nor the money I spent on it. Although I really wanted to, I could not make it work in the end.

[1] http://pingec.si/corruption.JPG

Data corruption during power loss has also been one of my biggest headaches with it comes to single board computers. I don't think I can say this is an issue that afflicts RPis only. However if you attempt to log data into the same device that you boot from you are asking for trouble imho. Someone else in this thread mentioned running the OS from RAM. I didn't know this was an option but I need to do some reading on how to do this. In your case though, I think backing up to a remote device is about the only sensible way to attack the data corruption problem.

There's one big advantage of using an SD card only, and this turns into a really massive advantage when you consider the target market.

There is no state persisted on the board apart from on the SD card.

I have a set of SD cards, and can switch between them at will. My Pi completely switches personality, with no internal state to confuse things.

Kids can play and risk breaking things as much as they want. The restore process is simple: take the SD card to a (normal) computer, rewrite the image, and you're set to go again.

On-board flash completely ruins the guarantee that replacing the SD card is all that is necessary (unless there is a hardware fault).

Maybe I'm just lucky, but my model b has been nothing but rock solid.

In fact, it syncs files across my machines through git and runs crash plan 24/7.

Absolutely love the little thing

I have 3 RPi deployments and at least two of those at home, one as 'home server' and the other one for dev projects. Never had a single SD-card failure. But a friend of mine, had a couple of failures...

I too have had problems with reliability, though more so when the device has already booted. e.g. numerous times I've had the device completely hang during file access.

When problems such as this crop up it really puts a dampener on the experience.

brand of flash matters.

before the raspi came along and popularized arm computing, there was a very lively scene around arm hacking out of existing devices. sheevaplug started the trend and people found that things like the seagate dockstar (a nas based hard drive dock) could be easily modified to boot a linux distro from usb and run as a low power headless server.

we kept a list of results from our experiences of what issues we had when trying to use new/different thumbdrives in the dockstar.

various issues with some brands included being unable to cold boot (boot from a power off state) or warm boot (boot when rebooting the system from a power on state). the brands that gave us issues were fairly consistent in this regards. we had the best luck with sandisk based thumbdrives.

i imagine the situation and cause is probably similar, if not identical, in this case.

>sheevaplug started the trend

Not to be pedantic, but before the Sheevaplug, there was the Linksys NSLU2, which came out in 2004:

https://en.wikipedia.org/wiki/NSLU2

http://www.nslu2-linux.org/wiki/FAQ/FirmwareMatrix

Pedantry noted. It's probably better describing the SheevaPlug as the device that fostered the trend towards plug computing rather than the one that encouraged device hacking as my statement would imply. I mistakenly omitted the key detail that the Dockstar was a custom piece of hardware that was based on the SheevaPlug... Same SOC and features.

Those were the days! I was the first to hack in to get root on that little box but others took it much further than me!

http://batbox.org/nslu2-linux.html

I still have a couple around, but rarely turn them on.

True, USB IO deadlocks and power usage causing drops/freeze made me stop using it. It was a surprise how many things can go wrong when you're out of the comfort of an industrial brand built PC board.

That's where I see the issues crop up. "Uncontrolled shutdowns" seem to be a killer of SD cards for me even when using R/O only filesystems. Wouldn't this be enough of a factor to point the issue towards software and not hardware though?

Honestly I don't know. The rpi is a weird SoC though, and many things are handled in the GPU by a blob, even if it's software it's outside our control.

On-board flash severely complicates the setup and recovery process: how do you get the OS in there> Also, strangely, it's more compact than almost all Flash packages of comparable capacity.

Have you made sure your power supply is reliable?

That's a good question. I haven't paid much attention to the power supply. I have a regulated (lab grade) power supply I am going to check if there are any changes when powered by it

There's a semi official wiki someplace that maintains a list of"confirmed" working power supplies and sd cards. One thing of value on that same wiki is instructions of how to test voltage on the pi board itself. On my phone now but I'm pretty sure the stuff I'm thinking of is on the elinux.org wiki.

Where are you obtaining your SD cards from?

I have tried different brands for what I'd consider reputable vendors. I do photography as a hobby which has educated me on quality (lack there of) of some SD cards

Brands are certainly important, as are the vendors. Basically anything from eBay is a no go, regardless of brand, but I'm sure you knew that already.

Does anyone know of a Raspberry Pi type board (low cost, Linux capable, good hardware graphics) with HDMI input and output suitable for image processing?

Something like the intel NUC is possibly your only real choice, unless you want to work really hard porting libraries.

It reasonably cheap, 10watts at full tilt, and has 2 real core (plus 2HT)

I use one for my openCV things, that need real time(ish) performance.

Just looked it up - i5 cpu, DDR3, gigabit ethernet, USB 3.0, all in a small form factor (very nice). $300 price tag is major killer though.

What type of OpenCV do you use it for?

No HDMI input on this board. I suppose we could look into doing our own HDMI in to PCIe mini board to go with it. Not that hard, just a lot of work.

Maybe look at a Cubox-i4 Pro? (Or the equivalent Hummingboard.) Depends what you mean by image processing.

That's the question, isn't it. I can't define it very well right now. We have a board we designed in house that uses a large FPGA for a range of basic real time image processing on 1920 x 1080 video. It's about ten years old. It'd be nice to be able to replace it with a software based platform and Python.

Also, it seems this has the same issue as other similar boards: No HDMI input.

Due to DRM issues, HDMI input on any device is rare. You can buy HDMI input chips that convert HDMI to CSI-2 so they can be plugged in to a camera socket - I've got a prototype that works on a Raspberry Pi. But it won't work for encrypted HDMI in, so you are very limited in what you can take in. And enabling DCHP is a very expensive proposition.

Well, there are no DRM issues with HDMI. The board manufacturer has to be a signatory to the HDCP [1] contract in order to be able to decrypt content. There are other design requirements, such as you decrypted data bus traces can't be on an outside layer on the PCB. It's been years since I looked at that stuff, there's more but I can't remember it. In terms of licensing, yes, if I remember correctly there are two tiers, either a $20K or $5K licensing fee and a couple of levels of per-device fees.

In general terms, from what I remember, the content protection agreements in place pretty much favor designs where the HDMI receiver is build into the main board as opposed to daughter cards. Again, in an effort to prevent illegal copying. Which is funny because to a knowledgeable engineer capturing the decrypted data is an almost trivial exercise. The only thing that can make it more difficult are SoC's where the HDMI I/O is built into the chip and you just can't get at raw data anywhere.

Now sure how I feel about the system Hollywood put into place. I do understand the need to protect their product from mass duplication. Perhaps they thought this was the largest threat to their industry. I strongly disagree with this position. I have not purchased a DVD or BlueRay disk in years. I haven't because there are plenty of content sources, including Netflix. I have just-about zero interest in "owning" their content. For the most part there's vastly more useful and interesting content freely available on the Internet covering a myriad of subjects. They may have shot themselves in the foot by wasting so much time and effort with HDCP.

[1] http://en.wikipedia.org/wiki/High-bandwidth_Digital_Content_...

Is small size the only thing the A/A+ has over the B/B+ board? Is size really such a big issue for prototyping boards like the Raspberry Pi?

Model A consumes less power than B, because it doesn't have the ethernet chip or port, does not have an onboard USB hub, has a single USB port and has less RAM.

A flatter profile is an advantage, when trying to tuck it into those narrow altoid boxes along with other circuits and battery packs.

Is is also cheaper. And if don't need the internet conectivity and so many usb ports why use the B?

Smaller size, smaller price and smaller power consuption. Also smaller memory, no ethernet and less USB ports.