Has Anyone Here Built Their Own Motherboard?

[halfpower]

I built my first computer not long ago. I soon realized that the motherboard is made from a bunch of off the shelf parts, the chipsets, the BIOS, the codecs, the various controllers. Can anyone say how hard it would be to build one these things yourself?

 

[Doctorweir]

Ummm...would you mind posting this in Highly Technical...?

 

[Peter]

You need an EE with experience in that field, a BIOS software writer, a PCB layout person, a contract manufacturer who'll do the prototype PCB, another who will populate and solder it. Next you need to convince every single one of the chip manufacturers you chose that you have a business case there that warrants them to even bother supply you with the chips, the documents, and the design support effort.

When you've sailed past all those obstacles and actually did come up with a prototype, you'll need all those expensive EE tools and machines that go 'bing' to debug it, coupled with even more experience. Respin the prototype at least once until you get a final hardware you don't see any more bugs in, then let the BIOS people finish their job. Meanwhile, put another bunch of software people on the support drivers for all the I/O you planted onto the board.

Now ship, and use.

All that can be had for $30 just 'round the corner.

 

[Doctorweir]

Honestly this would be pretty hard, especially regarding multi-layer soldering all the SMCs (surface mounted devices)...no chance without thew proper machinery...

I don't mind soldering on a mainboard for 1 Layer components like the capacitors, slot- / power-connectors, etc. Did that for several BX-boards where the capacitors went bad...

 

[imported_Kensai]

I believe one of my co-workers used to do that in university in the 8086 days.

 

[Silversierra]

:Q
'nuff said, lol.

 

[halfpower]

I figured a computer engineer would be able to pull it off. Though I've never known of anybody to compile their own BIOS.

 

[halfpower]

Honestly this would be pretty hard, especially regarding multi-layer soldering all the SMCs (surface mounted devices)...no chance without thew proper machinery...

Another alternative, correct me if I'm wrong, would be to just have a giant mess of wires soldered t a circuit board.

 

[Peter]

Wouldn't work. You can't do high frequency circuitry in a bunch of wires. And you'd still have to MAKE the circuit board.

 

[GuitarDaddy]

Virtually impossible! Building computers, is like building cars. It's easy to build a car from parts, but try and make your own tires , or fabricate your own A/C compressor, ain't gonna happen

 

[Steve]

Originally posted by: Peter
You need an EE with experience in that field, a BIOS software writer, a PCB layout person, a contract manufacturer who'll do the prototype PCB, another who will populate and solder it. Next you need to convince every single one of the chip manufacturers you chose that you have a business case there that warrants them to even bother supply you with the chips, the documents, and the design support effort.

When you've sailed past all those obstacles and actually did come up with a prototype, you'll need all those expensive EE tools and machines that go 'bing' to debug it, coupled with even more experience. Respin the prototype at least once until you get a final hardware you don't see any more bugs in, then let the BIOS people finish their job. Meanwhile, put another bunch of software people on the support drivers for all the I/O you planted onto the board.

Sounds like you've discussed this before

 

[classy]

It would be impossible. Unless they got a multi-million dollar fab in the basement.

 

[Insomniak]

Originally posted by: GuitarDaddy
Virtually impossible! Building computers, is like building cars. It's easy to build a car from parts, but try and make your own tires , or fabricate your own A/C compressor, ain't gonna happen

Best analogy so far.

Building a PC is like combining all the parts and assemblies to form a car. Building a motherboard is like smelting and molding the metal to actually create the engine block.

In other words, good luck.

 

[Slug]

Impossible. Among other things, all the tracings from the CPU to the RAM are exact lengths to eliminate timing issues secondary to the high frequencies involved. Placement is critical.

 

[Peter]

Well, it's not impossible, since we (those people who work in mainboard design) obviously do come up with mainboards that work

But it's not something you can DIY outside the professional network of knowledge, skills and tools that is the chipset/processor/mainboard business.

halfpower, the #1 skill in engineering is to dispose of silly ideas once recognized as such

 

[halfpower]

I'm not sure what type of circuit board you are referring to. They sell empty bread boards at radio shack. Even if they didn't, you could probably use some other material instead.

I don't see what the problem is with building high frequency circuitry with a bunch of wires -I'm no expert though. How would they differ from the wires that are built in to the board?

 

[halfpower]

Originally posted by: Insomniak

Originally posted by: GuitarDaddy
Virtually impossible! Building computers, is like building cars. It's easy to build a car from parts, but try and make your own tires , or fabricate your own A/C compressor, ain't gonna happen

Best analogy so far.

Building a PC is like combining all the parts and assemblies to form a car. Building a motherboard is like smelting and molding the metal to actually create the engine block.

In other words, good luck.

I disagree. Although it might be more expensive, I think it is not unrealistic for an engineer to build his own A/C compressor.

For something like an engine block, one draw up a diagram and then place an order with a machinist.

 

[halfpower]

Originally posted by: Slug
Impossible. Among other things, all the tracings from the CPU to the RAM are exact lengths to eliminate timing issues secondary to the high frequencies involved. Placement is critical.

This could present a problem. I know little about this stuff as I don't know what impact length has on timing, or what exactly is being timed. Do you happen to know how precise the lengths must be?

 

[Big Bunny]

Originally posted by: halfpowerI disagree. Although it might be more expensive, I think it is not unrealistic for an engineer to build his own A/C compressor.

For something like an engine block, one draw up a diagram and then place an order with a machinist.

Are you an experienced systems and circuit design EE? I think the point here is that it really IS unrealistic unless you are willing to throw very large amounts of cash and time at such a project. Essentially, you would need to be willing to put in the time and effort to learn the job of motherboard design. People spend their careers designing systems and circuits, and plenty of times even with extensive testing and QC they don't work as planned.

I am a mechanical engineer, and sure, I could design an engine block or compressor. However, because neither of those is my specialty, and I would be buying one-off parts, they would probably not work anywhere near as well as ones I could buy from regular auto parts manufacturers, and they would cost probably five time as much if not more.

I think even if you are an experienced EE, it's still not worth the time or very high cost to that it would take to build what would amount to a third-rate motherboard. And I suspect you are not, since you are asking this question in the first place.

 

[Peter]

Length matching: Read into signal propagation delays vs. switching frequencies. If you got a parallel bus to route, you don't want the least bit of phase shift inbetween bussed signals. What the tolerance for "the least bit" actually is? That'll be in the chip manufacturer's design specification documents.

DIY boards: Try soldering a 1000-point ball grid array chip to a Radio Shack "bread board".

May I point to rule #1 again?