Re: the teething troubles... (not the launch)

[toxibunny]

Right, so there's a few people actually got their Raspberry Pis now, and a couple (or 3) major things have come to light.

1. Sound drivers. Fixable, eventually. Eventually, Raspi owners will be able to output Hi-Fi audio through HDMI, or passable audio through the headphone-style socket.

2. USB. It seems to be a common problem that people's USB stuff just won't work. Something to do with voltages and power and special resistor fuses. Like 'OMG, When I plug my keyboard in, my mouse doesn't work, and when I plug my mouse in, my keyboard doesn't work, and when I plug my wifi dongle in, neither the mouse, keyboard, *nor* the wifi dongle work' - Okay, I'm exaggerating, but you know what I'm talking about. _Bad_News_. This needs a hardware revision, probably. :/

3. Graphics acceleration on the OS. Without it, the pi seems crappy and slow. There's no reason for a new product to seem crappy and slow. There really isn't! Actually, even if the seeming crappy and slow *isn't* due to lack of graphics acceleration, it's still not okay for a new product to seem crappy and slow... This is fixable by having lower expectations for the OS. Someone suggested basing Raspi OS on a linux from the times when Raspi-level processing was the norm (i.e. 10-12 years ago). I think that person has had a very intriguing idea...

[johnbeetem]

Well, this is the developer release. It will take time for developers to tune their software to run well on RasPi. Some software is bloated and RasPi is simply exposing the truth. I also suspect that a lot of performance problems are due to individual SD cards. After a while it will be clearer what software runs better than other software with similar functions and natural selection will occur.

[rmm200]

Sounds like the Op is speculating without actually having a Pi.
Base your expectations on what the Pi is intended to do; not as a PC replacement.

[toxibunny]

Sounds like the Op is speculating without actually having a Pi.
Base your expectations on what the Pi is intended to do; not as a PC replacement.

It's true that I haven't got one yet, but I'm not _speculating_ - I'm calling attention to what I see are the most common problems reported by those who have.

I'm basing my expectations on what the Pi is intended to do: It's intended to output sound, it's intended to work when you plug a keyboard, mouse, and ethernet into it, and (I assume) it's intended to be pleasant to use...


Sorry for what seems like a very negative post, I'm just a bit worried by these, and am wondering whether or not to go through with my RS order when it comes through, which should be this week. I'm wondering if it would be better to wait 6 months for a model A, or some sort of future revision that fixes the USB power problem...

[nick.mccloud]

As mentioned above, the Pi is still very much a work in progress, however as the owner of one I beg to differ in a number of areas

1. Sound drivers. Fixable, eventually. Eventually, Raspi owners will be able to output Hi-Fi audio through HDMI, or passable audio through the headphone-style socket.

Yet to get sound as a no-brainer but people have various players running in X, I've got sound out of mine and the OpenELEC build I played with yesterday was pumping out some heavy tunes in the later half of the afternoon along with the sound from the Go Pro HD video of some glider training flights.

2. USB. It seems to be a common problem that people's USB stuff just won't work. Something to do with voltages and power and special resistor fuses. Like 'OMG, When I plug my keyboard in, my mouse doesn't work, and when I plug my mouse in, my keyboard doesn't work, and when I plug my wifi dongle in, neither the mouse, keyboard, *nor* the wifi dongle work' - Okay, I'm exaggerating, but you know what I'm talking about. _Bad_News_. This needs a hardware revision, probably. :/

You are absolutely correct, you are exaggerating. Many of the issues reported have been resolved. The PolyFuse issue is a red herring fed by over zealous data sheet reading. Those that have a cheap power supply and have gone hell for leather with plugging things in without any logical testing have had issues and once they've settled down they've managed to resolve things. I've a whole bag of hardware here, some of which causes issue, mostly the cheap stuff but I still get errors with a Microsoft Basic Mouse - go figure.

3. Graphics acceleration on the OS. Without it, the pi seems crappy and slow. There's no reason for a new product to seem crappy and slow. There really isn't! Actually, even if the seeming crappy and slow *isn't* due to lack of graphics acceleration, it's still not okay for a new product to seem crappy and slow... This is fixable by having lower expectations for the OS. Someone suggested basing Raspi OS on a linux from the times when Raspi-level processing was the norm (i.e. 10-12 years ago). I think that person has had a very intriguing idea...

For it's intended use it's fine - I get up Geany, write some Python, look at documentation with Chrome, hit run, it's all good. I shut them down and type up some stuff in AbiWord. Even fire up OpenOfficeCalc for some light spreadsheeting.
There's no denying the Pi could do with 512Mb of RAM to facilitate a bit more multitasking. Cunningly the designers can provide that as soon as the memory hits that price point without blinking.
If you want to shuffle windows around the screen at top speed or surf to overly bloated websites then you will be disappointed.
The idea of rolling back the clock on the distribution would be fine except people still want an up to date JavaScript enabled browser - not something that will run on an older distribution.
As to the idea that it is considered fine to be 'crappy & slow', this is you being rude - no one designed it to be or considers it to be OK to be 'crappy & slow' - this is your subjective opinion not based on first hand evidence. The Pi was made to an objective and it's met it's design goals admirably already - things can only get better as the OS's are refined.

[jamesh]

If I was in a sarcastic mood, I'd be thanking you for bringing all that to the Foundations attention...

But we are aware of all this stuff. We need to sort out the USB thing, but things are a bit busy at Broadcom at the moment. Sound drivers - this seems to be turning in to a community effort after Dom started everything off, so that's going pretty well. Performance - well, this was always going to be an issue, it's not the fastest Arm device in the world, and there are limits to how much graphics acceleration will help (but it will help) but again, that has to be a community effort. If you really need a really fast GUI, RISCOS may be a good bet, that's almost ready.

So, it's almost all in hand.

[Ed Raket]

Sounds like the TS is a bit frustrated and confused, i have the same issu... it's coused by the small amount of time someone is willing to wait for such a neat and special product. Tho... when it eventually arives it wil have the same effect as a tit on a crying baby, so when it arrives first thing i will do: is put it in my mouth wil this void the warranty?

Siriously; of course the hardware issu's are known, and i am sure the "Pi people" wil resolve them in time. And software issu's are all ready beeing resolved by a number of people (children of the Pi ) who are using the Pi for what it is, a fun board with loads of posibility's to tincker with and get things working. off course it will improve in time, but for now you either want to be a part of the devellopment, or you dont. it is as simple as that.

[plugwash]

The PolyFuse issue is a red herring fed by over zealous data sheet reading.

I see no evidence of that. I've seen measurements that the unloaded resistance of those polyswitches is 5-7 ohms and given how polyswitches are supposed to work i'd expect the loaded resistance to be higher. I will test this to be sure when I get arround to it but i'd be very surprised if the Pi can output 100ma at an acceptable voltage while being supplied with 4.75V (listed as the minium voltage for a charging port in the USB battery charging specification).

It's just part of a larger problem though that it is very common to play fast and loose with USB specs. When all but one device is following the specs carefully and leaving safety margins everything works. When multiple devices in a system start playing fast and loose with the specs things start to break.

But we are aware of all this stuff. We need to sort out the USB thing, but things are a bit busy at Broadcom at the moment.

Good to hear, gert seemed to blow things off when the polyswitch volt drop issue was brought up.

[zag]

I've got a raspberry pi and been using it extensively myself and with a select class of students.

The general feedback is its just too slow at the moment to be usable as a teaching tool.

I don't even think its the cpu that is the bottleneck, its the graphics on screen really and the feedback a user gets when clicking on something. For kids to keep their attention they need an instant response really.

Saying that after a few hours using the thing, they did start getting used to it. Maybe it is all about expectations. I am sure this will improve over time as well.

[rurwin]

When I was a youngster we were used to a fraction of a second delay every time we hit Enter. When one day the computer was very lightly loaded and it responded faster, it was rather disturbing. All this on a 110 baud teletype.

The power supply issue wont require a hardware change as such if it is a polyswitch problem. All that would be required would be either a Raspberry Pi branded power supply, and/or replacing the USB polyswitches with a wire link -- not ideal but probably acceptable.

I have a PSU I built with 1A polyswitches in the USB charging ports, and my phone takes about 700mA to charge. I'll give it a test tonight.

[AndrewS]

I don't even think its the cpu that is the bottleneck, its the graphics on screen really and the feedback a user gets when clicking on something. For kids to keep their attention they need an instant response really.

Currently the graphics are all being rendered (drawn) by the CPU (rather than being accelerated by the GPU), so "slow graphics" is because of the CPU bottleneck
As others have said, this is a software problem that should get fixed over time as parts of the graphics rendering get offloaded to the GPU.

[rurwin]

either a Raspberry Pi branded power supply, and/or replacing the USB polyswitches with a wire link -- not ideal but probably acceptable.

Replacing the polyswitch with one rated at 0.5A/1.0A would solve the problem. (If it is voltage drop related.) The PCB tracks would probably take that sort of current OK for short periods.

[tufty]

Those that have a cheap power supply and have gone hell for leather with plugging things in without any logical testing have had issues and once they've settled down they've managed to resolve things.

ORLY?

There's people who are having issues with name-brand power supplies that are labelled well above the spec nominally required for the Pi.

There seem to be 3 main failure modes (although there may be more, I'm not convinced the overheating SD card issue is unrelated):

1. USB device pulls too much power, overloads the polyfuse. Result - dead USB.
2. USB device pulls enough power to drop the PSU voltage. Result - Crashola.
3. Pi goes into high power mode, pulls enough juice (in combination with USB devices) to drop the PSU voltage. Result - seemingly random crashola.

As for "resolving things"; this seems, largely, to revolve around the following sequence of events:

1. Buy random USB device, carefully checking to make sure it's not labelled as pulling enough to blow the 140mA power budget
2. Plug in USB device
3. See if it works
4. Does it work? If not, take it back to the shop and start all over at 1

The other "resolving things" approach is to hack up a USB cable and provide separate power down that. Leaving aside the issue of whether such an approach is workable for schools or anyone who's afraid of cutting and soldering, that's a major pointer that the Pi is (at least close to being) fundamentally unable to provide enough USB power.

Given that the major issue here is given by an s/USB device/keyboard/g on the above list, and that the keyboard is, quite frankly, a fundamental device that should just work, there is a major issue somewhere. I personally suspect a large part of it is polyfuse resistance, but as low-resistance polyfuses simply don't exist in the range that's needed, something else needs to be done.

Simon

[AndrewS]

The other "resolving things" approach is to hack up a USB cable...

Or simply use a powered USB hub?

[tufty]

The other "resolving things" approach is to hack up a USB cable...

Or simply use a powered USB hub?

Not necessarily a panacea. Unless you want to also apply s/USB device/Powered USB hub/g to the previous list.

Simon

[AndrewS]

Not necessarily a panacea. Unless you want to also apply s/USB device/Powered USB hub/g to the previous list.

Oh, I naively assumed that self-powered USB hubs wouldn't draw any power from their upstream port

[nick.mccloud]

For kids to keep their attention they need an instant response really.
Saying that after a few hours using the thing, they did start getting used to it. Maybe it is all about expectations. I am sure this will improve over time as well.

Not being opinionated or anything but I'd say that 99% of the UK riots were all about kids expecting things instantly. And as soon as they learn that life is not actually like that the better!

[Bakul Shah]

Even with a good supply I had trouble with Ethernet (a short cable fixed it). And later the board wouldn't reboot for a while. Though now it is up for over a day. And there have been quite a few reports of problems.

Apparently different flash cards can take different amount of power and may be the heat problem is related to some specific models? A number of people have reported that putting a fan on helps. So there are a number of suspects and finding the root cause(s) and a better characterization of the scope of the problem will require a systematic study. For instance, do working boards fail to boot when the ambient temp is raised to 40° C or whatever is the design limit? Do they fail to boot when a different card is used? Do any boards in the first batch fail?

The community can help with software issues but IMHO the foundation has to sort this reliability problem.

[nick.mccloud]

Those that have a cheap power supply and have gone hell for leather with plugging things in without any logical testing have had issues and once they've settled down they've managed to resolve things.

There's people who are having issues with name-brand power supplies that are labelled well above the spec nominally required for the Pi.

Maybe I was oversimplifying but the impression I get is that they were whizzing about with a pile of kit almost randomly plugging stuff in without a logical testing method. Maybe they had a good power supply and dodgy USB gear. Either way, if you encounter a problem and take a couple of minutes to try out some combos and take notes, my experiences thus far show that it all comes good in the end.

[Joe Schmoe]

I think what's coming out of this discussion, and it is something I've felt for quite some time (reading about people's various problems with the Pi and their USB peripherals) is that, unfortunately, cheap hardware requires expensive peripherals. That is, the Pi itself was designed so tightly, so minimalist-y (heh heh), that it only works with luxuriously designed peripherals (starting with the power supply, then on to the two essential peripherals - the keyboard and the mouse).

Which leads to 2 comments:
1) It certainly sounds to me like the Pi designers are used to working with good power supplies, and non-cheap mice and keyboards - unfortunately, it sounds like they never really field-tested it with the kind of real world junk that most people have lying around.
2) In order to be a realistic product, it has to work with the "real world junk that most people have lying around" - and it can't be the sort of "Try this; if it doesn't work, send it back, get a refund, try something else, dah dah dah, dah dah dah" approach that has been suggested in this thread.

Yes, I know this is a "developers release" (and you will note that I have frequently posted to that effect in response to the various troll posts recently), but the fact remains that, absent a hardware refresh (suggested by the OP, but rejected by most of the thread responders), these hardware problems aren't going to go away. You can accomplish a lot with software, but these hardware problems are what they are. So, it may be a "developers release" forever.

I, for one, have always thought that a hardware refresh was inevitable.

Finally, regarding the OP's final sentence (suggesting that expectations be dialed back), let me note that 'twas I who suggested that it is looking like a marketing error to be pushing the idea that the Pi needs a GUI. I think the whole thing should be re-done as a (snappy, fast and good) command line computer - just like the 80s micros we are trying to re-create. And this is not just motivated by the prime law of marketing ("sell what ya got!"), but also (and more importantly) by the twin questions of "Who is your target market?" and "What are we trying to accomplish?"

And this, boils down to: Are we trying to compete with MS, whose educational goals are to produce Word, Excel, and Powerpoint mavens? No, we are trying to produce technies - and techies are not afraid of the command line. And this, my friends, boils down to what I've been saying for 15 or so years now (*): "You can't beat MS at their own game. You have to stake out new ground and win there."

(*) Ever since I gave up on the idea of OS/2 being a better Windows than Windows...

[mahjongg]

I think what's coming out of this discussion, and it is something I've felt for quite some time (reading about people's various problems with the Pi and their USB peripherals) is that, unfortunately, cheap hardware requires expensive peripherals. That is, the Pi itself was designed so tightly, so minimalist-y (heh heh), that it only works with luxuriously designed peripherals (starting with the power supply, then on to the two essential peripherals - the keyboard and the mouse).

To the contrary, the best keyboard is one that is as bare as possible, and therefore draws a minimum of current and is also cheap. As for the mouse, these normally draw acceptable levels of current already, except for maybe some wireless versions. And optical mouses probably draw more current than simple trackball versions. So for mouses too, simpler, thus cheaper, is better.

If the R-PI had used slightly larger rated polyfuses these problems would go away, without any price increase, so no its in no way true that "the Pi itself was designed so tightly, so minimalist-y (heh heh), that it only works with luxuriously designed peripherals".

The one exception would be the Chinese $2 USB chargers that are totally unregulated pieces of crap that immediately drop their output voltage whenever some load is applied, these are designed to power charger circuits of simple cell phones (powered from 3.6 Volt batteries) that don't mind that the input voltage drops to 4 Volt during charging. These are "5 Volt" power supplies in name only, even when labelled "5V 1A DC".

[Joe Schmoe]

Thank you for making my points beautifully - in each of your 3 paragraphs.

[nick.mccloud]

That is, the Pi itself was designed so tightly, so minimalist-y (heh heh), that it only works with luxuriously designed peripherals (starting with the power supply, then on to the two essential peripherals - the keyboard and the mouse).

Well it doesn't work too well with my branded MS Mouse!

I think the Pi was designed to work in spec - the problems arise because there are too many things lying about that aren't in spec and when you end up with a couple or three plugged in all hell breaks loose.

I've purchased a Nokia PSU (£3.30), optical mouse (£1.49) and a MS keyboard (£4.49) from eBay - all work just fine and I have change from a tenner!

But this is, after all, the developer phase, and we are here to figure out these wrinkles as well as play Quake and watch movies on the OpenELEC distro.

[Burngate]

Most chargers being sold to date are just that - made to power charging curcuitry within phones etc., which a) don't care if the volts are within spec. and b) take a reasonably constant current. The phone stays powered by the battery.
The Pi is relying on in-spec 5v, and takes a varying current depending what it's doing, just like a dongle hung on a hub.
So powering the Pi from a hub is going to be more reliable. Just so long as the hub will provide power even if it's got nowt up its backside. Some won't.

[adlambert]

That is, the Pi itself was designed so tightly, so minimalist-y (heh heh), that it only works with luxuriously designed peripherals (starting with the power supply, then on to the two essential peripherals - the keyboard and the mouse).

Well it doesn't work too well with my branded MS Mouse!

.

I've been using an MS branded Mouse, an optical wheel mouse. It's been absolutely fine.
My keyboard is a very cheap PCLine one, that's fine too. Never had any problems when using them together.
Now the Pi is only accessed using Putty from a Laptop over SSH so no keyboard or mouse needed. When in command line mode it is in its element and it's a great way to learn.