Is it possible to build your own CPU

[BrightCandle]

I had to design an entire 8 bit CPU as part of a 3 hour exam in VLSI architecture and algorithms (1 of 3 questions so technically I had an hour). I assure you its not hard to design a basic CPU given the knowledge, the book I linked to gives you a mighty good start on that and one of its outputs is you have designed yourself a CPU so I recommend you read it.

 

As to producing the CPU that is tougher but there are companies that will do outdated process runs of your hardware for a few thousands of dollars or as I recommended before using a FPGA gets you access to more modern processes and allows you to redo the design and provides interfacing so its possible to actually run some stuff without having to design an interfacing mechanism as well.

 

Its not actually that hard to do this, you could just do it and prove the design in software (a good idea before you spend thousands on a FPGA or more on actually getting it fabricated so you can test it) and then move from there.

[Spencer1559]

In my other post someone thought that I was building a CPU (I wasn't planning to) but now I'm wondering if it is possible. If it is can someone give me some links on how to do it? I don't care how old of architecture it would be or what it could run (maybe MS-DOS?) and if so, how would I connect it to say a home built GPU and motherboard (or bought ones) and how I could get it to post. Maybe a guide on building a PC from scratch. Thanks

Risky 

[Nena Trinity]

Ask a chip maker after you design one!

[Clanscorpia]

Well, the whole thing is about making it myself, not from a chip manufacturer. Can anyone find links? I want to do this for a city science fair and if you could find links to guides on how to do this would be awesome but guides on how to design your own 64 bit CPU and on who can build them for you that would be AWESOME!

[Stefan1024]

Well, the whole thing is about making it myself, not from a chip manufacturer. Can anyone find links? I want to do this for a city science fair and if you could find links to guides on how to do this would be awesome but guides on how to design your own 64 bit CPU and on who can build them for you that would be AWESOME!

 

64 bit is pretty steep to start with. And you know you will never be able to produce something that you can put in a motherboard and run windows on it?

All you can do is a simple CPU running of an FPGA with a reduced instruction set that executes a simple assembler programm you wrothe by yourself.

Everything else need a LOT of resouces, most likely way to much for a city science fair.

[Clanscorpia]

I just want it to be able to run a calculator. And I need the resources to build one in Canada if you have to be in the US to make it. I want to be able to build it by myself. I don't care if its 8 bit for runs below 1 mhz, I just want it to be able to run something!

[Clanscorpia]

And the 64 bit was for people recommending ordering customs CPU's. And I'm 12 so I don't understand some of the terms you guys are using as I am relying on you guys to decide if this a a worthwhile project to do. My Grandpa's an electrician so he said he might be able to help. He was around when most of the parts of a CPU were created. And please don't underestimate me because I'm 12. Computer's are my passion. My teacher even told me if people are going to ask to help me so often they should hire me. He wasn't' joking around when he said that.

[Stefan1024]

I just want it to be able to run a calculator. And I need the resources to build one in Canada if you have to be in the US to make it. I want to be able to build it by myself. I don't care if its 8 bit for runs below 1 mhz, I just want it to be able to run something!

 

Please quote on people so they get a notifiaction.

 

You have two options:

 

1. option: calculator only

For a calculator alone you can use an ALU (arithmetic logic unit) and hardwire the functionality. It's not really a CPU, but it will work.

An example for an ALUis the 74F181: http://media.digikey.com/pdf/Data%20Sheets/Fairchild%20PDFs/74F181.pdf

Skills you need:

- understanding of logic gates

- soldering

 

2. option: "real" CPU

For a real CPU you need a lot of controll logic for the ALU. It's way to much to do it with simple logic gates so you need a FPGA to make it efficiant.

The workflow would be:

- Decide what assembler commands you like to support (MOV, ADD, SUB, JMP, conditional JMP are the absolut minimum)

- Make a table how you translate them to binary (complier)

- write the functionallity in VHDL or verilog and programm it to an FPGA

- make a simple programm in assembler

- translate it to binary

- have fun

 

Skills you need:

- understanding of logic gates

- assembler

- VHDL or verilog

 

This is quite a big task!

[Clanscorpia]

Could I do this in four months if I had a team (say 4-5 people) working on this?

[Clanscorpia]

Please quote on people so they get a notifiaction.

 

You have two options:

 

1. option: calculator only

For a calculator alone you can use an ALU (arithmetic logic unit) and hardwire the functionality. It's not really a CPU, but it will work.

An example for an ALUis the 74F181: http://media.digikey.com/pdf/Data%20Sheets/Fairchild%20PDFs/74F181.pdf

Skills you need:

- understanding of logic gates

- soldering

 

2. option: "real" CPU

For a real CPU you need a lot of controll logic for the ALU. It's way to much to do it with simple logic gates so you need a FPGA to make it efficiant.

The workflow would be:

- Decide what assembler commands you like to support (MOV, ADD, SUB, JMP, conditional JMP are the absolut minimum)

- Make a table how you translate them to binary (complier)

- write the functionallity in VHDL or verilog and programm it to an FPGA

- make a simple programm in assembler

- translate it to binary

- have fun

 

Skills you need:

- understanding of logic gates

- assembler

- VHDL or verilog

 

This is quite a big task!

Where could I learn these things?

[Clanscorpia]

Oops, double post

[You_are_a_cunt]

Where could I learn these things?

You could try asking on the EEVBlog forum. It's full of electronics engineers and people passionate about electronics in general. You might be able to find someone with more knowledge on building processors

[Stefan1024]

Could I do this in four months if I had a team (say 4-5 people) working on this?

 

Are you aiming for option 1 or 2?

 

Personally I could do option 2 in several days to weeks, depending on who many instructions you like to support.

But I'm studying electronics since 7 years now and know a decent amount about VHDL. So yes, 4-5 people can indeed pull this off, but:

- are they working all day long on the project or only on evenings / weekend

- who skilled are the people? Any experiences in electronics? Assembler? VHDL?

 

 

Where could I learn these things?

 

What do you like to learn? There are some compleately different topics.

I can give you some documents of my school,, but a lot of them are in German. Also without the teacher talking about it, it's not the best material to work with.

I can help you, but my time is fairly limitted.

[Clanscorpia]

It would be 4-5 people working evenings and weekend's. I want to learn how to do option 2 and all I want is it to be able to run a calculator for now. Thi might be a long running project that will get better and better and most of the people will have basic electronic skills. Me and my best friend have the most skill.

[Stefan1024]

It would be 4-5 people working evenings and weekend's. I want to learn how to do option 2 and all I want is it to be able to run a calculator for now. Thi might be a long running project that will get better and better and most of the people will have basic electronic skills. Me and my best friend have the most skill.

 

Ok than you can split the work into two chunks.

 

A: Provideing binary code to execute.

This can be done by translateing assembler code to binary with a table. To generate the assembler programm you may can use a compiler to generate it from C code, as writeing complex programms in assembler from scratch is terrible.

 

B: Building (programming) the CPU

Learn VHDL, buy a board with an FPGA and build up the CPU. Fist think about how to get input signals and how to display the results (7-sement display, VGA,...).

 

First of all you need to agree which assembler instructions you support and how they are translated to binary. The more instructions the easier is the development of the software, but the harder the development of the CPU.

Also keep in mind the both parts are software, but the underlying mechanic is totally different. Especially FPGAs are completely different to utilize than everything you have ever worked with. Make sure you have a person with knowledge at hand.

 

As for the hardware, factor in at least 500$ for the project.

[Clanscorpia]

Wow, ok thats a lot. I think that might be waaaaaaaaay to much for a city science fair but maybe for high school this might be a better project. Thanks for all the help guys!!!!

[Stefan1024]

Wow, ok thats a lot. I think that might be waaaaaaaaay to much for a city science fair but maybe for high school this might be a better project. Thanks for all the help guys!!!!

 

If you are only interested in a calculator you can take a microcontroller board and build a calcualtor in C. Or something like an arduine / rasperi pi. That's way less work.

 

You could also build the calculator directly with VHDL and you don't have to make software for it.