Game developer David Braben creates a USB stick PC for $25
geek.com
geek.com
The OLPC Hardware could have been a huge commercial success if it had been gone into production soon after the orignal presentation. But no, it needed Acer and their EEE-PC to satisfy that demand.
Why develop a "universal purpose something" specifically for children? You're not a child anymore so its way easier to build something YOU want than imagine what exactly a child wants or needs. In fact, I'm sure this guy wants the USB-PC more than anyone else on the planet. Also, if you mass produce you bring down prices anyway.
But most important, everyone knows what children want in general: They want to be just like their adult idols! Don't tell kids: "This is for you. I wish I had it when I was a kid. Now, have fun, while daddy works on his Mac."
> 1) How long do you think it will be before the boards become available?
I'd say three or four months. As you can see from the screenshots, we
have usable Linux, but we're waiting to get final versions of the the
chip from our supplier.
> 2) Are there any plans for a version with onboard ethernet?
I don't think we're likely to do onboard Ethernet; we will have an
onboard 3-port USB hub so people can add an external adapter.
> 3) Are there any plans for a version with onboard wifi?
Yes. The final version (though maybe not the first distributables)
will have onboard WiFi (probably 802.11n) in the price point.
> 4) What are the power requirements, both under load and at rest?
At rest I'd say 50mW (we could trim this if it was really important,
but it gets a bit fiddly below this point), under serious load
(original XBox class graphics or 1080p30 H.264), 700mW.
I wonder if this means you actually can power it from HDMI's 5V pin (which is limited to 50mA ie 250mW) for some uses?
Why? Because an old PC with equivalent specs is essentially free (anything from ~2000 onwards.) Thousands of them, desktop and laptop, will be being recycled or landfilled every day.
Linux on x86 hardware is standard enough that you can effectively call it a standard platform for developers, already, same as this.
The remaining key advantage (I guess) is size. And maybe the fact that it's easier to get kids excited about playing with a tiny brand new board than with a 10 year old computer.
Also, the size difference could be significant, as well as allow novel ways of using it. For instance, if the goal is to get this machine into areas outside cities were computers are more common, shipping costs could be significant if recycled desktops/laptops are used instead.
I can imagine that you can conduct a class where kids take these computers home, along with a school-supplied keyboard, for use with the TV set they have at home, to work on computer (programming) assignments. This would allow the school to avoid the cost of acquiring monitors.
Last week my coworker bought an ex-government ultra-small-form-factor Dell PentiumD 3Ghz for $30 in an auction, they were selling dozens that way in single quantities for between $20-$40. 3Ghz PC with 1Gb of RAM!
A friend showed me the disposal tunnel at the local unversity, they have palettes stacked 1m high with old Pentium4-era computers and old CRT monitors waiting to be picked up for recycling. The university pays for them to be taken away.
I'm in Australia, but I expect if you look around most developed countries you'll find this is the situation - old computers all over the place scrapped, resold or shipped off-shore. The developing world is where a lot of those computers end up, so in those countries a $25 USB-stick-computer with no excess shipping cost may still be a better option than paying for shipping a crate of old machines. However, I guarantee you there is no shortage of old hardware out there.
If educational "demands on the market" don't exist for P4-era used PCs now, are they necessarily going to exist for the USB stick PC?
In the developing world though, I imagine there are some huge advantages to the power consumption of this. These may make it feasible to run a computer lab off of solar panels.
There are already computers that draw low enough currents to do this now but they are quite expensive. ($3-500 as I recall)
On the power front, the power requirements are minimal, allowing them to take advantage of the solar infrastructure in places that rely on car battery power (central america is covered in solar). Also to run on 12V SLAs which are a common feature in places with unreliable power.
You might be surprised to learn that your city may have a recycling contract with the landfill. Those thousands of landfilled PCs are effectively unavailable to you or anyone else, no matter how feel-good the project.
I think this is a Good Thing and I hope it becomes available to everyone.
With this, they use the exact same machine at home and at school, and it is easy for them to carry back and forth. The only downside is that they would also be easy to lose, so there would need to be a big lanyard attached so they can wear them around their necks.
EDIT: I added some explanation of the problem.
http://news.ycombinator.com/item?id=2517136
I honestly don't see how this is gong to encourage people, especially kids, to further wonder about how computers work, and what actually happens. It's smaller, and if anything even less accessible than a desktop machine.
I learned about computers and computing by building my own from a Z80, 8KB RAM, 8KB ROM, random logic, etching my own circuit board and soldering the components and sockets in place.
Should kids do that these days? Would it help?
Other options in this space (far more expensive! but similar ARM on a board style)
http://gumstix.com/store/catalog/index.php (although sadly they don't seem to carry the basix/connex, which are the same kind of power class as this)
The board he's built would have lower BoM cost than either of those. And if he's manufacturing them in similar quantities then maybe he can get the cost down.
If he's looking at production runs of similar numbers to wireless routers then I think $25 is completely believable, as cheap end wireless routers approach that cost new now.
[1] http://s.dealextreme.com/search/wm8505
For now, the Trimslice is pretty exciting (and shipping, and of course in a very different world)
Thanks for the pointer to the Mini2440 -- i haven't seen that one yet.
The FriendlyARM stuff looks damn cool--thanks to the grandparent for pointing this stuff out!
I don't know, I mean presumably he's thinking OLPC-esque. I could see this being at least as popular as the WM8505/VT8500 devices just on price (price having being those devices _only_ selling point), and then the education angle picking up some more bulk deals with education departments.
is he really going to contract out to a chinese factory?
Does anyone not contract out to a Chinese factory of some kind, these days?
For now, the Trimslice is pretty exciting
That is exciting, I hadn't seen those before! Thanks for the link.
Of course, as you say that is a "different world". The Trimslice is four-generations of ARM along the line, in x86 spec equivalents it's a Pentium III compared to a Core 2 Duo.
Cheaper, lower power consumption, and almost the same processing power. It works for me, sure I can't use proprietary software, or something like dropbox that requires a proprietary daemon, but I don't really miss it much. At first you notice but as you use it more, you come to realize that you really didn't need it.
Sorry if this was a huge wall of text, I am responding via my phone.
I meant the analogy the other way. That the USB stick computer is the Pentium 3 equivalent and the Trimslice is "four generations along the line", something like the Core 2 Duo.
Believe me, I'm extremely excited about the performance/costs power nexus of these new ARM cores. :). And well jealous of your Cortex-A8 build farm.
(I did get at least one thing wrong in my reply, which is I somehow thought the USB stick computer was ARM9 not ARM11. So it's only 3 generations of ARM before the Cortex A-9, and that's disregarding all the other lesser-used ARM cores that came between.)
PS Please don't make sweeping statements like There are probably other things incorrect in your comment, that's just plain rude.
And pretty much any game that runs on an iPhone could run on an ARM desktop. Off the top of my head, popular "mainstream" games that I can think of would be Street Fighter 4, Mirror's Edge, I'm sure there are more, I just happen to have those 2 installed on my iPhone. ARM machines can also run Android, as long as you are willing to put in the kernel work, and most companies do these days.
The EfikaMX has a "desktop" version as well as a netbook version. They are both almost identical, although the netbook version doesn't have any video out. It also runs off of a 3 cell battery, and gets somewhere in the range of 6-8 hours of solid usage, not just sitting there almost suspended. They(i.MX515) pack the AMD(ATI)z430 3D unit, which runs at 133MHz (the i.MX535 version will run at 200MHz); This is also known as a Qualcomm Adreno, although the Adreno uses a z180 instead of z160 for the 2D unit, not sure if it uses something other than the z430 for 3D. The difference between the graphics card in the Xbox360, and these is that the 360's runs at (i believe) 600MHz.
I'm not sure what graphics card will come with the ARM11, but if it does 1080p and OpenGL ES 2.0, then I'd guess that it's a Mali. I could definitely be wrong, but really, at 25 dollars, even if you only played with it every once in a while, it's probably worth it. And if they do a buy one give one like OLPC did, I'd definitely order a minimum of 4. I hope they are able to, and I'll be watching closely.
There is much more to a GPU than simply clock speed. Since modern GPUs do everything with programmable shader units, the number of units acting in parallel makes a big difference, as well as the efficiency of the units (cycles per instruction). Many mobile GPUs' shader units can be counted on one hand[1], while a modern AMD or nVidia GPU has hundreds (though the different architectures make a direct comparison of numbers almost meaningless).
[1] I had a hard time finding exact specs last time I looked, but I seem to remember reading about one popular SoC whose GPU had in the neighborhood of 4 pixel shader units.
Idling MacBook Pro (screen off) uses 8-9W according to Apple (http://images.apple.com/environment/reports/docs/MacBook-Pro...).
If you already have laptop running most of the time, addition of such device probably doesn't save much (if anything).
The brighter kids will already start trying to make something 'work' out of that. You continue further about how a CPU works, completely in the language of boolean components. And then they get to design and build one themselves.
That will really teach and show them there is no magic and how it works. Even with the insanely small and complex systems you have now, you can rest assure that in the core they are basically just like the one you made when you were in school. This kind of practical way of working with hardware (even simulated), gives a lot more pleasure and teaches kids computer hardware design, cpu design, microcodes, assembly and tons more.
I was taught from 'Micro computers' by A.J. Dirksen (ISBN 9789021015934); it (+ the teacher) gave me enough insight to quickly learn Z80 assembler, 68k assembler and add / replace hardware on my '80s computers.
EDIT: I forgot; it would be really stupid to not release this to the masses. Everyone wants one, but only the kids can it; I agree with the rest here; that makes no business sense. Charities are businesses too.
The preliminary specs are listed on the website:
That's what it looks like if you look at the pic on http://www.raspberrypi.org/
1. http://www.greenfoot.org/ 2. http://bluej.org/ 3. http://processing.org/ 4. http://www.squeak.org/
I like the idea, but I don't think the place for this device is in schools.
In fact, I hope computer science never gets into high school curriculum, it will only make students hate programming. Just look at math. Programming is even harder; and chances are if you're a programmer you're not really likely to become a high school teacher. Conversely most high school teachers will not know how to program.
There's room for improvement, but the classes are widely successful. I have a number of friends who learned to program from high school alone, without a serious effort to learn additional material by themselves.
I'm a bit of touch at 41, but are you sure that it isn't already available as an elective for students? Perhaps you meant making it mandatory, which I agree would probably not be very productive.
I took programming in High School, but back then it was all using BASIC. We learned the concepts of variables, memory, loops, and even simple algorithms. And if we finished our homework ahead of the rest of the class, we got to play games. Quite the motivating factor for me at the time.
> Just look at math.
> Programming is even harder;
> ... chances are if you're a programmer you're
> not really likely to become a high school teacher.
> Conversely most high school teachers will not
> know how to program.
You just can't "fake" it. You either get it, or you don't.
Most high school teachers won't "get" it.
> Ditto mathematician. Most high school teachers I know of who are teaching math have no idea what math is really a
I know, but they can fake it, because the curriculum is designed as a set of rules that must be learned.
> So, no different from math, then.
The difference is that programming cannot be taught as a set of rules that must be learned and then applied without much understanding.
The computer will apparently be able to have network access, presumably wireless, and enough storage to run standard desktop software.
Oh, well, as long as it runs Emacs... ;-)
Even if the PC only costs $25, will schools have to buy extra equipment to allow use of the above accessories?
I don't see how most keyboards, mouses, and monitors will be able to connect to it.
This device has one USB output. A mouse and keyboard will need one USB output each, so to use this $25 computer you need to buy an additional device that allows the connectivity of multiple accessories. Not all keyboards and mouses connect via USB, so some users will have to buy new keyboards and mouses just to use this "cheap" computer.
And not all monitors will be able to connect via HTMI. Users with another kind of monitor will need to buy some additional technology that allows these other monitors to plug into the $25 computer.
So how do I connect my equipment into this computer? It's not obvious just by looking at the picture.
Additionally, any new monitor for the past few years ships with HDMI.
I've got to say though, if you succeed in making a computer so cheap that people complain "but I have to buy a mouse made in the current or previous decade???", then you've done a damned good job.
Also, all monitors besides the cheapest / oldest models will have DVI input, which you can easily get from an HDMI port. The only displays you can't easily drive from this device are those which only take VGA input.
You get a bunch of these devices and arrange them into a cluster/cloud of nano-nodes and teach kids (teens) how to setup clusters of webservers, memcache, micro-dbs etc.
You effectively use the devices to teach large technical infrastructure with tiny physical representations.
Here is how I would do it:
I would whiteboard out a network diagram showing the different layers of tech infrastructure in contemporary cloud designs.
Draw the nodes and connections.
Define the role of nodes in each layer:
Firewall, load balancer, web server, memcache, app server, db server
Then do a micro-config for each of these and run them on the little buggers.
Get velcro tape and tape them onto the diagram in the appropriate location.
Physically wire them all together.
Have it run a website that the students build.
Run analytics...
This would be the ultimate in showing them how The Tubes actually provide them with the data they view every day.
By drawing this, then physically attaching the units to the diagram and running it there will be ZERO confusion and I guarantee that every student will grok the internet more fully.
The thing does not need to be high performance - but it does need to be fault tolerant.
Have the students pull devices out while its running and they are watching monitoring. Have some subjectively F5-ing the little site they made while you drop a few nodes.
Hell, I am ready to do this... anyone want to help me get some of these devices/similar devices?
2. Personalization. These are cheap enough that each student could have their own, for a fraction of either iPad or OLPC.
Of course, the fact that it requires a connection to a not-cheap HDMI display mitigates both these somewhat.
As a few people have pointed out, their site says Composite or HDMI.
developing apps for android phones?
I guess now that Android phones are coming down under $100 as well then this is becoming a possible alternative. You don't get the "nuts and bolts" aspect of it, but you do get to teach both programming & fun.