Yesterday somebody said something about a "raspberry pie". So when I got 
home, I looked it up. It turns out Raspberry Pi is a kind of cheap, 
minimal computer system. It's powered by a 700MHz ARM11, which 
apparently is comparable in computational power to a 300MHz Pentium-II. 
It also has some kind of GPU onboard, which HDMI output. It's powered by 
USB, and the entire thing costs $35.

Naturally, I immediately asked myself whether you could build a cheap 
render farm out of these things. Having uttered the words aloud, at 
least one website about this subject must now exist. A trivial Google 
query verified this fact; there are indeed *several* such pages. They 
conclude that 100 Raspberries have roughly the same compute power is a 


In summary, you *can* do this, but the price per FLOPS is very poor. Or, 
it is if you use the *CPU*. My next question is what the hell happens if 
you try to use the *GPU* instead. I haven't seen an answer to this yet. 
(I also haven't looked very hard.)

Obviously the next important question is "can it run Haskell?" And the 
answer is obviously "yes" - although it seems it's difficult to compile 
anything particularly large on the actual machine itself, due to it 
having "only" 512MB of RAM.

You know, when I first learned Haskell, 512MB of RAM constituted a 
*very* high-end system. What the heck has the world come to when you 
need more RAM than that just to compile some source code?

(Then again, I haven't tried it personally. Perhaps they mean it can't 
really *huge* stuff like the entire Haskell compiler or something. They 
don't really quantify what constitutes "big".)

The goal of this project appears to be to churn out cheap computers that 
kids can mess around with to learn about computers and programming and 
stuff, similar to the way I did with the C64 when I was a kid. Trouble 
is, no kid today would find something as primitive as a C64 interesting, 
with its piffling 16-colour graphics and monophonic sound.

And there's the rub. It's got to be modern and powerful for anyone to be 
interesting. But if it's modern and powerful, it's going to be far too 
complex to tinker with at the level you would with the C64. If you have 
a C64, it's *feasible* to write a short machine code routine which 
performs a key-scan and pokes the video hardware in response to 
different key presses. You could, given a month of Sundays, 
realistically build your own micro-OS.

*Nobody* is ever going to do that for a Raspberry Pi. It takes something 
as complex as Linux to power it. Recall that Linux is the result of 
decades of effort by thousands of expert coders across the entire world. 
(And no, I'm not just talking about the kernel. The kernel, by itself, 
doesn't give you a runnable OS.) Nobody has the time or resources to 
write an entire OS in their spare time.

Similarly, with the C64 you can write a few POKE commands and watch the 
screen turn green or hear a sound play or something. The GPU on the 
Raspberry Pi are closed-source; you need to sign an NDA just to see what 
registers it has! Nobody is going to be experimenting with that anytime 
soon.

So, yes, it's a small hand-held computer with Python on it. But it's 
basically a normal PC, but smaller and cheaper. The fact that it's a 
bare PCB doesn't make it any more inviting to tinkering or whatnot. I 
suspect most kids are being to be really disinterested in this thing.

Now, the big kids who should know better? They might find this 
interesting...

Post a reply to this message