OSDev.org

When VIA first released their Pico-ITX mainboard i was very impressed and for a time i actually considered buying one, but then i thought: "Hey, im not gonna you half of the features on this board for the purposes i'm going to use it for..."
I even wrote a e-mail asking if they planed to release a "Lite" version, explaining which components i wouldn't need and so on. Amazingly i recieved a reply and of course the answer was no, however it would be possible to make certain modifications if the order was big enough. I reached for my wallet, opened it and was not surprised to see that it was empty.

It is now 2:40 am and i can't sleep, so i lie there in my bed thinking about nothing and everything and suddenly find my self wondering how hard it would be to make your own customized MOBO. Of course the specifications would have to be very specific and minimalistic.

Personally i would need, for those weird ideas i get, an efficient CPU, but it wouldn't have to be very fast. I would of course allso need some ram and means of storage, preferably via usb or eth, and some sort of IO interface, eth, usb, com, whatever, just not too slow.
I wouldn't even need graphics! And that of course goes for sound and various other stuff as well.
Then all there is left is putting it all together. Of course it would have to be very small, power efficient and passively cooled.

It doesnt get much simpler than that. All the key components can be bought at Radioshack, well maybe not all of them, and if you have the means to put it all together, then you're on your way.

The question is how hard would it be? Would it be something that anyone who is determined enough could do or would you have to be an electronic engineer or something to pull it of?

I really have no idea what the answer is, but i suspect that some of you do.


Now it's 3:03 am
Regards

I'd expect it to be harder than you expect. You'd have to implement a south/north bridge, all the IO handling, the pulse/clock conversion between hardware. A motherboard is far more complex than just a board connecting pins together.

MessiahAndrw wrote:I'd expect it to be harder than you expect. You'd have to implement a south/north bridge, all the IO handling, the pulse/clock conversion between hardware. A motherboard is far more complex than just a board connecting pins together.

How can you expect to know what i expect?

But yes, of course it will not be a picnic, but remember that im talking very minimalistic and simplified here, yes it will still be complex, but i should think it could be done. (not by me that is)

I would imagine that you would first want to prototype the board you want to design, but to do so would mean trying to find a clean and easy way to pull all the pins out so that you may patch them from place to place as you test your design. Those chips contain a _lot_ of pins packed into a tight space.

The easy part would most likely, oddly enough, getting it running. I say this out of no experience in dealing with a general purpose processor on a low level, but likely all you need to do at first is find the pins that need power supplied and apply it at the correct voltage. Then find the clock pins and hook up a crystal and capacitor or what ever it may need to get it clicking. You will most likely also have to connect each and every other pin just so none are hanging around gathering charges that might confuse the processor. So connecting these to logic low, logic high, or more importantly having the processor specification sheet and doing what it notes if it even does note it.

So all in all I think it is possible, but your biggest hurdle is actually getting the chip into a position where you can connect up all the _tiny_ little pins. You might look for a socket of some type for the specific processor that brings the pins out onto breadboard like tracks or allow you to somehow plug it into a breadboard.

There chips would most likely run at a really low clock speed for testing and tinkering if trying to build a new motherboard as this would be controlled by the clocking circuit. I almost want to say for certain that you _should_ be able to step the processor in cycles with a push button and a special circuit to dampen the electrical noise of the button. Now that would be pretty neat, right?

As a matter of fact there might be a way to even test and simulate a motherboard on a processor using another processor such as a embedded one where you use one processor to generate the clock pulses (PWM, GPIO, .. whatever ..) and even feed it logic ones and zeros by writing code for the other processor. Of course the reason being is that an embedded processor is much easier since it has a lower pin count around 40 and might or should even have DIP connectors or some type.

I just read your post over again and realized that what you are asking for already exists, but maybe not exactly in the custom what you want aspect.

They are embedded processors and I spoke of them in the last post. There are many types, and many different companies produce then along with many architectures including ARM, 8086, PPC, MIPS, and plenty more I am absolutely sure of including one I just though of: dsPIC and PIC.

And as I said these different processors stem a broad range of boards that come with a variety of assorted devices. So most likely the customized idea you have has already been done and is just waiting for you to find it.

I have a ADuC7026 that is on a board with a 3-Phase PWM, GPIO, UART, SPI, ADC, DAC, EEPROM, SDRAM, PLA, and I2C which makes plenty of things to tinker with. It also makes a pretty nifty signal processor which is my main concern with it for my robotics projects. It is passively cooled and in total eats less than 25mA. I can power it from a solar panel, but in a dimly lit room it needs a little tinkering where I stop(form of sleep) the processor and wake it up in periods in order to decrease the power consumption to keep it running (along with a capacitor).

I remember looking around for people attempting to utilize older x86 processors in hobby projects..

Sadly, I only found 1!!

http://dubel.org/computer/

Unfortunately, The project above probably isn't exactly what you're looking for...

Good luck.....

@Kevin
Ill have to read that post a few times to understand exactly what to google for, but it seem that you are very close to what i had in mind.

@Brynet
You too are close in a sence, but you are right that it is not exactly what im looking for at the processor and maybe some of the other components are old and probably not vary efficient by today standards.

I will have to look very close at that site as it may hold at big part of the answer to my question.

Anyway, to take an example, imagine taking an old i386, and reproduse it with todays technologi and maybe optimizing it a bit using the huge amount of knowledge (cant find the right word) that im sure has been gathered since the release of the first i386.
I am fairly certain that something simular to whats discribed above has been done on severel eccation, for use in small devices and stuff, i am however not certain what to google for; actually i have no idea...

by the way, i have no intentions of trying to pull this of. not for some time anyway, i allready have far to many small project lying around in piles...

Hi,

Have a look at http://www.avrfreaks.net/ (especially the forums) for information on 8 and 32 bit AVR uControllers and a helpful community (just like the one here on osdev ). You will not achieve the same as you would on an IA32 processor, but they are pretty good fun to play with. My soldering skills are appalling, but even I can just about manage breadboarding a 40-pin DIP processor!

It seems that one of the most difficult things with them is driving a display device - you can forget 32 bit 1024x768 graphics


Cheers,
Adam

at first glance this looks too "alien" for me to want to get involved with, but i will take a closer look when i have the time, thanks for the link.

It's certainly a bit different if you are used to IA-32, but there are free GCC compilers, assemblers and IDE's available. The chip is flash-programmed via a USB dongle plugged in to your PC. I learned asm on these chips before moving to x86 - so it is certainly possible to move over quite easily.

I'll stop with the hard sell now! I have nothing to do with Atmel, honest, I just like the architecture

Cheers,
Adam

i will take a closer look at it sooner or later, but first i have something else to post


Amazingly, sometimes when you try, you actually find what you are looking for, or at least fairly close to it...

Now this looks interesting. Then scrap the video(input), touchscreen, audio, encryption, ata(leavin only the flash which should be suficient), GPIO?, and the other useless stuff. Provide support for a more... "interesting" cpu like the Via Eden ULV@500Mhz

This is allso interesting, but i didnt find it before i had writen all the other stuff

Oh, and while you are at it, consider if you would call This a desktop

Zacariaz wrote:This is allso interesting, but i didnt find it before i had writen all the other stuff

Hmm - I agree that looks interesting, I may have to investigate. I was originally interested in uControllers because I was having a go at building a media centre for my separates stereo system. I have found that some things are prohibitive because they are either hard to get hold of from a UK supplier or are double the cost over here. If I could get that board cheaply, it may re-awaken my interest!

Thanks for sharing the links you have painstakingly found!
Adam

Im kinda tired of searcing the net withoug results, it seems that even finding a board featuring the via eden ULV @500 Mhz, is problematic.

Anyway, there is two links that might be worth a look:
http://www.systems104.co.za
http://www.compulab.co.il

Good luck

Does > this < help?

Cheers,
Adam

Ive found yet another interesting site but still im not satisfied.

About the link from AJ, it was very interesting, especially one or two of the xscale boards, but id rather not begin messing around with other architectures...