Would it be possible to make my own motherboard?

[mrgoblin]

Im toying around with an old p2 processsor and mainboard and I was wondering, would it be possible to construct my own mainboard? It would be a neat project if I could but im not sure what it entails. Any info would be appreciated.

 

[Burner]

It would be possible to design one, and send it off to a pcb shop to get it fabricated. That would cost a little bit of money. If you wanted them to soldier everything on for you it would be even more.

Just doing it yourself with wirewrap may be possible but it would be quite difficult to get high speed busses to work like that.

I'd suggest you start with a simpler project like take a PIC chip and some schematics online and design something cool like a harddrive fan and light controller. You can also make a board for such a project and get it fabricated (small boards with 2 layers can be made for under $200 maybe under $100) A good free design tool for doing it is eagle cad. You can search for pcb manufactors with google and probably find a reasonable price for a small board. You can buy most of the parts at digikey. To do a project like this you have to read all the datasheets and understand what each part needs to have happening to it. Also look at other peoples designs and figure out what decisions they made to get there.

After doing a project like this you'll have a much better idea of the scope of doing a motherboard and if its possible.

 

[dkozloski]

Back in the old days when I went to NASA soldering school one of the first things we were told is that there is no component on a low volume circuit board that is worth anywhere near what the bare board costs. I wouldn't be surprised if the developement costs of a motherboard reach the mid six figures and then some. Modern motherboards are five layers plus and have traces as fine as integrated circuits used to have. I would say the project is more like a career.

 

[Evadman]

If you wanted to, it would be possible to design a 4 layer board for a p2, but it would not be a short project by any means. You would likely have TONS of stuff to learn. Doing it by yourself would likely take years. Not that I am trying to diswade you, but you should know ahead of time.

If you design a 4 layer board, expresspcb can prototype it you relitively cheap. I don't remember their 4 layer prices off the top of my head, but 3 moderately sized 2 layer boards were only a little over $50 total a few months ago.

If you do this, here is a quick hint: do not put 90 degree bends in the traces. do 2 45 degree ones.

 

[Peter]

Sure. You'll need to buy all the parts, get all the non-disclosure agreements from the chip makers to get the documents to learn what you're supposed to do, come up with a schematic, layout and route it, and then have a professional production line manufacture it.

Then, to find out whether the thing is even remotely working, you'll need to license and port a BIOS core to your design. Next you'll need advanced diagnostics tools, both hardware and software, to debug your prototype and assert its workingness and standards compliance in all aspects, from core busses down to every single peripheral I/O. Then enter round two of schematics/layout/prototype and see if that one got better. After a couple of iterations, when you think it's good enough, get serious about teaching your crude BIOS everything it needs to do to let a modern operating system even think of running, PnP, ACPI, blahblahblah. Done that? Good. Start writing/porting drivers for all the peripheral chips you used next.

It's not a spare time project. If you seriously want to do these things, start a career in that direction. In about five to ten years you'll have learnt enough to get assigned your first mainboard project.

Note: I do not want to sound discouraging at all - it's a rewarding path, and a really fine job with a perspective. Just know the dimension of the task.

I second what others have said: Start small. If you want to toy around in your spare time, try figuring out something much simpler like a PCI I/O card, made from parts and using busses and ports whose specifications are public.

 

[Mday]

Originally posted by: mrgoblin
Im toying around with an old p2 processsor and mainboard and I was wondering, would it be possible to construct my own mainboard? It would be a neat project if I could but im not sure what it entails. Any info would be appreciated.

if you're going to ask, you cant. it's a set of about 3-courses to be able to design high speed i/o boards, and those require a few years worth of physics, materials and of course comp sci, comp E and EE courses.

 

[dkozloski]

You might have more success if you try working with one of the simpler Zilog chips. There are tons of public domain information on those things and they are much more forgiving for the debugger. Who says you have to build a PC?

 

[wacki]

Originally posted by: dkozloski
You might have more success if you try working with one of the simpler Zilog chips. There are tons of public domain information on those things and they are much more forgiving for the debugger. Who says you have to build a PC?

Never touched the Zilog, but I agree, start simple.

The basicX24 microcontroller available at phanderson and basicx.com is the way to go for beginners.

You only need to learn visual basic.... easy.
And the chip is powerfull enough to do all of this Link

the next step up from there is AVRs

 

[KalTorak]

Just doing it yourself with wirewrap may be possible but it would be quite difficult to get high speed busses to work like that.

Flat-out impossible for anything modern, for the same reason you can almost never use 2 layer PCBs.
Signal traces need to be kept REALLY close (thousandths of an inch) to a power or ground plane, such that when current flows one direction in the trace, it can complete the loop by flowing in the other direction in the plane. In real motherboards, this is done by stacking signal layers next to power or ground layers.

I wouldn't be surprised if the developement costs of a motherboard reach the mid six figures and then some.

It's not quite THAT bad.

If you do this, here is a quick hint: do not put 90 degree bends in the traces. do 2 45 degree ones.

Ooh! Let's have a holy war over this "rule" The theory is that the trace is SLIGHTLY more capaciitive at the bend (since it's a little wider), and that the increased capacitance leads to an impedance discontinuity and thus a reflection. Frankly, I've never been convinced there's a noticeable impact or reflection here, even when you're looking with a TDR...

Sure. You'll need to buy all the parts, get all the non-disclosure agreements from the chip makers to get the documents to learn what you're supposed to do, come up with a schematic, layout and route it, and then have a professional production line manufacture it.

That's a little overly discouraging. The hard parts are getting the components and the BIOS. For Intel parts, at least, there're published reference schematics and routing rules that are available off http://developer.intel.com once a part is launched. Professional board fab houses can be found in the back of every EE magazine there is.

If you seriously want to do these things, start a career in that direction. In about five to ten years you'll have learnt enough to get assigned your first mainboard project.

Hardly 5 to 10 years. I'm 5 years out of school, and on my third mainboard (I keep moving to more complex stuff; I'm doing DP servers and workstations now.)

Note: I do not want to sound discouraging at all - it's a rewarding path, and a really fine job with a perspective

Actually, I'm not sure how much future there is in heads-down mainboard design; there're a lot of people in a lot of different places (Asia) with those skills. The guys who're getting and staying employed now understand not only board design, but also big-picture systemwide stuff.

I second what others have said: Start small. If you want to toy around in your spare time, try figuring out something much simpler like a PCI I/O card, made from parts and using busses and ports whose specifications are public.

The one reason I miss ISA - as a junior, I had a PC interfacing class where we designed and wirewrapped an ISA printer interface card. That's the class where things finally started to click for me.

it's a set of about 3-courses to be able to design high speed i/o boards, and those require a few years worth of physics, materials and of course comp sci, comp E and EE courses.

Yep - EM Theory's actually REALLY important to understand high-speed PCB design.

My most recent mainboard design

 

[Peter]

Well, you're five years out of school, and how long did engineering school take? My 5-10 years estimate was meant to include learning@school.

My most recent mainboard design ... darn, hasn't been press-released yet :-(

Here's one we made two years ago ...

(I'm the BIOS guy, mind ;-)

-----------------

not speaking for my company either ;-)

 

[KalTorak]

Aaaaah. For some reason I thought Peter meant 5 to 10 years out of school till one got one's first mainboard project, which I thought was a little pessimistic. 5 to 10 years from NOW, sure (And if you're gonna start learning now, shoot ahead of the duck and be sure to take at least one class to learn about microwave PCB design rules, since in 5 to 10 years EVERYTHING's going to have edge rates that fast.)

Peter's board clearly has more layers than mine, but I'm pretty fond of mine, anyway (and very glad it got publicly shown at IDF so there're pictures I can link to!) I'm fond of making fun of BIOS guys, too, for not having to learn EM Theory, but damn, having a good BIOS guy makes all the difference in the world.

 

[Peter]

Want to take a guess at how many layers? (The board shown is 160x230 millimeters. PIII-BGA, ServerWorks two-piece chipset, VGA, dual Gbit Ethernet, a P2P bridge, 2 GBytes of RAM, IPMI, SIO, two PMC slots, voltage regs, hot plug circuitry. And a BIOS of course.)

Making fun of each other, BIOS vs. EE people, is part of the job So, yes, we don't have to learn all the analog and high frequency stuff, right ... but we're the ones who (should) bring the compatibility and software concerns in, in the earliest possible design stage. I agree, a good BIOS guy makes for a better product - but unless the EE people know their stuff at least as well, it's not going to happen at all.

Isn't it fun when one's baby is going public? And scary at the same time?