USB Stick PC - Very tiny PC built for $25

[JODY]

This could have some uses for those wanting mini machines.

http://www.geek.com/articles/games/game-developer-david-braben-creates-a-usb-stick-pc-for-25-2011055/

[BobA]

Amazing what they can put into very small spaces now.   Will be following this one with interest.  Thanks

[alfonzotan]

You beat me to it, I was just reading about that on Slashdot:

http://tech.slashdot.org/story/11/05/06/122233/A-25-PC-On-a-USB-Stick

Definitely has huge potential for MAME and other emulation builders... vastly simplifies and makes cheap the PC component.  Plug one end into your controller (or USB hub) and the other into a monitor.  Load up an appropriate version of Linux MAME and your ROMs of choice on the SD storage card, and you're set (although the article doesn't address audio).  The PC horsepower described in the article should be plenty for classic-era games.

You could easily set up a 100% portable controller setup with one of these... no need for a full PC power supply, fans, etc.  Heck, the computer would be much smaller (and cheaper!) than a standard keyboard emulator...

[alfonzotan]

Sorry, butterfingers...

[MonMotha]

I'd hesitate to call this a "PC" on a stick.  It's supposedly a 700MHz ARM11.  The ARM11 series (if I had to guess, it's probably an ARM1176ZJS, as those were the most popular) is no slouch, but don't expect miracles.  Depending on what you're trying to do with it, expect CPU performance comparable to a 300-400MHz P6 era Celeron.  These ARM SoCs are usually designed around the assumption that software will make heavy use of specialized hardware acceleration facilities, which MAME especially just doesn't really do much of.

The 128MB of RAM is also a bit constraining for some applications, but you can probably deal with it.   It sounds like it's probably a PoP SoC like one of the TI OMAP chips, so don't expect to upgrade it.

You'll also have to deal with the fact that it's ARM and not x86.  This shouldn't be too terribly hard, but you will have to either compile MAME yourself or find a suitable ARM build.

The price also seems aggressive, though possible if you're willing to accept essentially no margin and have reasonable volume.  I'd be worried about it being vapor.

[SavannahLion]

alfonzotan the HDMI specification allows for audio to be transported on the cable.

How much value will this have in teaching the science side of computing. So many lower functions are wrapped, that many programmers never learn about how their pc actually works. not to name names, but there was a poster here that had a rather twisted view of computing, know what I mean?

[alfonzotan]

This particular item may well turn out to be vapor, but there are plenty of companies that specialize in "low-margin, high-volume" commodity hardware; netbooks sprung from that very philosophy, and have sold many millions.  Commodity tablets are next.

As a proof-of-concept, it's very intriguing, and given the prior history of PCs, it's close to inevitable that the dirt-cheap-computer-on-a-stick will be a reality sooner or later.

[newmanfamilyvlogs]

Not exactly $25, but it would be an adventurous project to get MAME running on a configuration like this:

Marvel 1.2Ghz CPU based "Plug Computer"
http://www.globalscaletechnologies.com/t-dreamplugdetails.aspx#features
+
16Mb USB VGA Adapator
http://www.newegg.com/Product/Product.aspx?Item=N82E16815158162

[RandyT]

The first thing that comes to mind when I see this is "Cell phone, minus the phone", not "PC on a USB Stick".  Still, if it can run emus and game applications better than the current crop of handheld devices, at a price point that low, it would be quite a nifty device, and open a lot of possibilities.  But if it's just a "dingoo" minus the screen and controls, then it's probably not worth getting terribly excited about.  It'll probably cost more by time it sees light of day as well, if it ever does.

[Vigo]

At the price of $25, I can think of a ton of projects to build. Now they just need to make cheap little lcd screens with mini hdmi input.

[Howard_Casto]

The first thing that comes to mind when I see this is "Cell phone, minus the phone", not "PC on a USB Stick".  Still, if it can run emus and game applications better than the current crop of handheld devices, at a price point that low, it would be quite a nifty device, and open a lot of possibilities.  But if it's just a "dingoo" minus the screen and controls, then it's probably not worth getting terribly excited about.  It'll probably cost more by time it sees light of day as well, if it ever does.

What he said.  The use of a hdmi port pretty much negates it's use as a low cost computer for third world nations as well. 

Now if you wanted to put... lets say a bootlegged version of pacman specifically designed for the device on it it might work for that... but something as complex as mame, or even a multi-game emu is going to be too beefy to ever run on that.  Processors aren't everything you know... there are still concerns like ram, video ram, storage space, ect...  Even if it's expandable via the usb port by the time you add all of that in it'll be too costly and big to be practical. 

[Necro]

This is cheaper than a damn Arduino...which is nuts.

[ark_ader]

The first thing that comes to mind when I see this is "Cell phone, minus the phone", not "PC on a USB Stick".  Still, if it can run emus and game applications better than the current crop of handheld devices, at a price point that low, it would be quite a nifty device, and open a lot of possibilities.  But if it's just a "dingoo" minus the screen and controls, then it's probably not worth getting terribly excited about.  It'll probably cost more by time it sees light of day as well, if it ever does.

What he said.  The use of a hdmi port pretty much negates it's use as a low cost computer for third world nations as well. 

Now if you wanted to put... lets say a bootlegged version of pacman specifically designed for the device on it it might work for that... but something as complex as mame, or even a multi-game emu is going to be too beefy to ever run on that.  Processors aren't everything you know... there are still concerns like ram, video ram, storage space, ect...  Even if it's expandable via the usb port by the time you add all of that in it'll be too costly and big to be practical. 

My Original XDA which was an ARM Cpu ran at 200 mhz and it played Mame just fine at full speed.  Not HD, but you can bring that HDMI port down to VGA land very easily.

[Gray_Area]

.....the trend, and arguably a preferrable one, is devices becoming so sophisticated, knowing how they function at a fundamental level would impede, or at least be irrelevant to, their use. This is already much the case.

[alfonzotan]

Oh c'mon.  We were running Ms PacMan at full speed on 386s back in the 90s before MAME turned into a bloatware nightmare that accurately emulates mah jong.

Ding-ding-ding.  The early versions of MacMAME ran just fine on a 120MHz PowerPC 601.  A throwaway DVD player has more juice than that these days.

[Necro]

Interestingly (that's a word, right?) it's actually going two different ways.  (Note - this is per a recent wired article I just read so....*shrug*) - The 'maker'/DIY crowd is actually increasing recently due to the reasonable availability of what used to be extremely high end equipment (i.e., 3d printers) to the 'masses'.  Also, the availability and support of things like the Arduino, adafruit, and make, are all increasing the interest in the 'hobbyist' realm and those developing new little 'gadgets' that may not be commercially viable (or, if they are, they need a prototype platform).  I actually just went out and bought one of those 'learn electronics' kits myself because I simply wasn't 'getting' it when trying to learn just by reading.  Definitely interesting, fun, and I'm already thinking of cool little things to do for my cab and just in general.  I always could code hack, hopefully I'll be able to hack together electronics kits as well.  (I use the hack term in the original, hack away at something to change it sense).

From the PC/IT side of things, yup.  Totally agree.  

[Minions]

Who will be the first to try to make an even smaller donkey kong cab!? But really, I love the concept, they already have small pc's you attach to the back of monitors, I'd love to have a linux box in my pocket. I don't see these retailing for less than $50-75 though. If product costs were $25, factor in production and labor, then a small amount of profit and you will get the end user prices. Such potential for DIY projects though.

[justinjstark]

For about $100 you can get the Marvell Plug Computer 3.0 with a 2ghz arm processor.

http://www.slashgear.com/marvell-plug-computer-3-0-updates-sheevaplug-with-wifi-bluetooth-hdd-0567674/

I'm wondering about using one of these in a bartop.  It should be plenty beefy to run the classics.

[Thenasty]

if you can put DOS OS then .36 MAME DOS would run good witht it.

[MonMotha]

You can't run DOS on an ARM machine.  DOS is inherently tied to the PC (x86 w/ BIOS and specific well-known IO/memory locations) architecture.  You can, however, run Linux on larger ARM processors quite readily, and there are usually ready-made ports for these development type boards, and MAME runs fairly decently on Linux.

[SammyWI]

Kinda shows the limitations we are hitting with shrinking computers.  It's not getting the processing power into a small package, it's the input and output.  This is a tiny processor, memory, etc but you still have to plug in separate I/O.  Cell phones and tablets are combining I/O with touchscreens but you still need a good amount of area.  Making the screen thinner only helps so much.

I'm thinking that we'll see holographic projection in the coming years to deal with this.  For input: Kinect style gesture recognition maybe combined with holographic virtual controls.  Yeah, it's movie stuff now, but the cell / tablet companies have a lot of $$ to develop the next big thing.

Or maybe I'm just feeling optimistic today. 

[RandyT]

.....the trend, and arguably a preferrable one, is devices becoming so sophisticated, knowing how they function at a fundamental level would impede, or at least be irrelevant to, their use. This is already much the case.

Preferable to programmers, perhaps, but it also leads to greatly underutilized hardware.  Look at the demo below and consider that it is running on a 1mhz processor, virtually no RAM and an 8-bit bus with very simple audio and video processors.  I don't know that we've yet seen something 1000x better on 1 ghz handheld devices, with gobs of RAM and advanced graphics processors.  All that isolation from the hardware has a big cost in performance.

[newmanfamilyvlogs]

Preferable to programmers, perhaps, but it also leads to greatly underutilized hardware.  Look at the demo below and consider that it is running on a 1mhz processor, virtually no RAM and an 8-bit bus with very simple audio and video processors.  I don't know that we've yet seen something 1000x better on 1 ghz handheld devices, with gobs of RAM and advanced graphics processors.  All that isolation from the hardware has a big cost in performance.

This. A thousand times, This.

I grew up in awe and wonder of the Amiga demoscene, and the really clever stuff I see any more in the modern demoscene always ends up being creative procedural generation. Placing intentional arbitrary limits on some part of the system to really make things impressive. The only really exciting stuff like the 16k, 32k, 64k demoscene, it feels like sometimes.

[Blanka]

Reminds me of another incredible C64 demo. Some guys ported the classic Unreal demo for a 286 PC to the C64. Amazing stuff!

[Vigo]

I love the C64 eye candy. It brings a tear of happiness to my eye. I think I'll just wash this down with some sweet, sweet SID music. 

[Necro]

Well, now...that's just an ignorant statement given you mean it to be such a blanket statement.  There's TONS of things you do on computers these days that are highly limited by the processor.

[RandyT]

Oh what a load of bunk.  What you can do on a computer these days is a limitation of your artistic vision, not the processor.

I never stated otherwise and I'm not exactly sure how you managed to get the message you are feeling compelled to decry from what I wrote.

My comment refers to the massive amount of horsepower needed to accommodate your "artists", when "low-level programmers" can do a lot more with a lot less.  The really impressive stuff comes from those who are both.  There is a ton of evidence to support this.  Look at every console launch and the titles which launch with it.  They all smack of the tools provided by the hardware manufacturer to provide basic functionality.  It's not until programmers learn to bypass these and take advantage of the hardware in their own ways, sometimes getting it to do things even the manufacturer didn't know it could, that the software really reflects what the hardware is capable of.  But by then, a new piece of hardware usually comes out, and the cycle starts over, leaving underutilized hardware in it's wake.

Those C64 demos are perfect examples of this.  They were done by ignoring Commodore's advice to "use the doors, not the windows".  And if you don't know what that is referring to, then we probably aren't going to have a very productive discussion.

[RandyT]

You sound old.

Thank you

[MonMotha]

He's right, at least to some degree, though.  All the modern programmer comforts like VMs, dynamic typing checked at runtime, other various runtime checks, interpreted languages, etc. do come with rather substantial overhead in terms of memory and CPU usage.  While they let programmers (even mediocre ones) quickly bang out "useful" programs, there's no doubt that it's at the expense of performance and size.  Now, on a modern PC, this often isn't a huge deal; they're so fast and have so much memory that nobody generally cares.  However, in some PC hosted scenarios, it matters: large data sets, highly interactive/real-time situations, slow IO channels like WANs, etc.  One has to balance the rather large costs associated with getting closer to the metal with the required performance.  It may be cheaper for all parties to just tell the user to buy a faster computer, much as it pains me to see a word processor running poorly on a 3GHz Pentium 4 with 2GB of RAM, but sometimes it's worth it to actually figure out why your program takes 30 minutes to load a 100MB dataset.

Now, when you get into the "embedded" world, things are a little different.  Other factors such as price and power consumption start to dominate, especially on high volume and/or portable (battery powered) products.  You can't afford a 3GHz, 100W CPU in either your power or financial budget.  8GB of RAM just physically won't fit on the circuit board in any real-world package along with all the other stuff the device needs.  The programmer has no choice but to move closer to the hardware for all but the most trivial of "apps".  The "$25 PC" described in this thread falls somewhere along these lines.

Things change even more when you're firmly in the embedded world.  In this case, I don't mean things like phones and tablets.  I mean things like your microwave oven or alarm clock.  These things often have micros with a whopping 256 bytes of RAM.  You don't have room for a VM, much less an interpreter.  You don't have room for dynamic typing.  Hell, you don't have room for much of an OS: maybe a simple scheduler and a few "tasks" with some locking/signaling between them.  The C language dominates, and it's by no means unheard of to bust out some inline assembly.

Heck, there exist some (very low cost) micros that HAVE NO RAM.  You get the CPU's registers and some ROM for the program.  Since they lack a stack, you basically can't program in C at all (well, there exist some hacks - you get one level of function call depth in most cases).

Don't knock on "lean and mean" programming just because you don't HAVE to do it on a PC.  There are plenty of environments where it's still required or at least a pretty good idea, and it can be useful even on a high-powered PC in some situations.

Then again, go ahead and knock it.  I make a decent living in part because of the fact that many "programmers" these days are utterly incapable of operating in the embedded environments I described.

[Vigo]

I think to sum up the key difference with this kind of "lean" programmer and the "modern programmer" is that on something as bare as such as C64 (or any other limited computer), a lean programmer has to stretch to every limit of the tiny box that he is given.

I don't know if modern programmers would always even know how to work within that small box, they would just tell you that you need to buy a much bigger box. 

[Lilwolf]

I think this is a cool idea.  Just as long as nobody tries to put windows on it.

But I agree... This is a phone without a phone.  Think how cool it would be for a little android device that you plug into your TV... an OS that was built around the speed of the processor but with a nice video card.  Would be great.

As a little embedded machine... this would be awesome!

Or its cheap enough that the could plug this into a every TV sold, and add a blutooth adapter... would turn every TV into a little computer for those times you need one. 

[ark_ader]

You can't run DOS on an ARM machine.  DOS is inherently tied to the PC (x86 w/ BIOS and specific well-known IO/memory locations) architecture.  You can, however, run Linux on larger ARM processors quite readily, and there are usually ready-made ports for these development type boards, and MAME runs fairly decently on Linux.

You got something that backs that up?

[ChadTower]

It's not until programmers learn to bypass these and take advantage of the hardware in their own ways, sometimes getting it to do things even the manufacturer didn't know it could, that the software really reflects what the hardware is capable of.  But by then, a new piece of hardware usually comes out, and the cycle starts over, leaving underutilized hardware in it's wake.

[MonMotha]

You can't run DOS on an ARM machine.  DOS is inherently tied to the PC (x86 w/ BIOS and specific well-known IO/memory locations) architecture.  You can, however, run Linux on larger ARM processors quite readily, and there are usually ready-made ports for these development type boards, and MAME runs fairly decently on Linux.

You got something that backs that up?

Um, that's just how DOS works.  The wikipedia article explains this to some degree (read the "history" section).  There may be "DOS like" OSes for ARM systems in that they have a similar shell and maybe a similar API, but it ain't actually DOS and can't run any DOS binary you'll find.  Even the "DOS prompt" in WinNT (NT4, 2k, XP, Vista, 7) isn't really DOS running as a typical OS.  It's either a virtual machine (read: emulator) running DOS, albeit with minimal virtualization (command.com in 32-bit NT, non-existent in 64-bit NT) or a native Windows application that's just a shell designed to look and fell like DOS (cmd.exe in all versions of NT).

As to why you can't run DOS directly on an ARM system:

First off, DOS programs as well as DOS itself and included utilities are all x86 binaries.  They will not run on ARM without an emulator.  The instruction set (and machine encoding) is completely incompatible.

Second, DOS itself makes HEAVY use of the PC BIOS, and the PC BIOS doesn't exist on anything other than a PC.  There in fact exist x86 systems that are not PC compatible and don't have anything like the BIOS (instead just having a lightweight bootloader that provides no facilities other than the ability to load a "real" OS).  DOS was in fact basically a thin wrapper around the BIOS with some libraries for e.g. filesystem access and memory management thrown in.  It didn't really do that much.  While there was some intent to make DOS run on x86 devices that didn't have a BIOS, this never actually happened as there was little to no demand for it.

Third, many DOS programs pretty much bypass DOS itself and bang directly on the PC hardware using the "well known IO and memory locations" inherent to the PC architecture.  Essentially, they include their own drivers that assume an underlying IBM PC compatible system.  None of these will be the same on an ARM system and will in fact vary depending on which ARM SoC is being used.

You could patch all this up by running a PC emulator, and in fact such emulators exist.  Bochs will run on ARM, for example. However, emulation of an entire PC, including the CPU, is not without substantial performance cost.  It also seems silly to emulate a PC just to turn around and emulate something else using MAME, in this case, especially given that one could just build a native version of MAME on e.g. Linux.  You'd need an OS in which to run the emulator, anyway.

One could, of course, port most of one of the open source versions of DOS to ARM if one REALLY wanted to.  It would be quite an effort, and it still wouldn't run any DOS programs in existence (due to the instruction set incompatibility).  Again, it would be more of a "DOS-like" OS than actually being DOS.  Since there exist more modern OSes that are already ported to many ARM systems and provide much more functionality, including ports of many popular applications such as MAME, with minimal overhead, there's little reason to do so.

[ChadTower]

You can't run DOS on an ARM machine.  DOS is inherently tied to the PC (x86 w/ BIOS and specific well-known IO/memory locations) architecture.  You can, however, run Linux on larger ARM processors quite readily, and there are usually ready-made ports for these development type boards, and MAME runs fairly decently on Linux.

You got something that backs that up?

ark_ader: 


Great post, MonMotha.

[Gray_Area]

 I don't know that we've yet seen something 1000x better on 1 ghz handheld devices, with gobs of RAM and advanced graphics processors.  All that isolation from the hardware has a big cost in performance.

This has me thinking of 'meat'. Which hasn't...yet....been superceded.

I don't know if modern programmers would always even know how to work within that small box, they would just tell you that you need to buy a much bigger box.  

The ability to pragmatically figure out solutions, in the face of incomplete or inadaquate knowledge, used to be called 'common sense'.

Regardless. There will come a time when the 'hardware' will 'figure out' the means to do whatever is requested, as (albeitly much faster than) biological systems have done.