cpu

122 Articles

One Bit, One Instruction Discrete CPU

November 30, 2016 by 26 Comments

There is a certain benefit to being an early adopter. If you were around when Unix or MSDOS had a handful of commands, it wasn’t hard to learn. Then you learn new things as they come along. If you started learning Linux or Windows today, there’s a huge number of details you have to tackle. You have the same problem trying to learn CPU design. Grappling with the design of a 16-bit CPU with a straightforward data path is hard enough. Throw in modern superscalar execution, pipelining, multiple levels of microcode, speculative execution, and all the other features modern processors have and you’ll quickly find yourself lost in the details.

[Michai Ramakers] wanted to build an educational CPU and he took a novel approach. The transistor CPU uses only one instruction and operates on one bit at a time. Naturally, this leads to a small data path, which is a good thing if you’re only using discrete transistors. His website is a ground-up tutorial in building and using the tiny computer.

Continue reading “One Bit, One Instruction Discrete CPU” →

1MHz, 2 Boards, 4 Bits And A Homebrew CPU

November 10, 2016 by 17 Comments

[Agp.cooper] saw a vintage 4Kx4 bit RAM chip and decided that it needed a CPU design to match. The TTL design fits on two boards and has a functional front panel.

This custom CPU project has a few interesting bits worth noting. First, it is small enough that you can wrap your head around it pretty easily. And [Agp.cooper] gives  a good account of the instructions set architecture choices he considered and why he settled on the final design.

Continue reading “1MHz, 2 Boards, 4 Bits And A Homebrew CPU” →

Weird CPU

August 6, 2016 by 24 Comments

How many instructions does [agp.cooper’s] computer have? Just one. How many strip boards does it use? Apparently, 41 five 41-track boards. While being one shy from the answer to life, it is still a lot of boards for a single instruction. The high board count is due to the use of 1970’s vintage ICs including TTL parts, 2114 RAM chips, and 74S571 PROMs.

There are several different architectures for single instruction computers and [agp’s] uses what is technically at TTA (transfer-triggered architecture). That is, the one instruction is a move and the destination or source of the move determines the operation. For example, the Wierd CPU (that’s the name of it) has a P and Q register. If you load those registers and then the ADD register will contain the sum of the two numbers.

Continue reading “Weird CPU” →

1000 CPUs On A Chip

June 20, 2016 by 42 Comments

Often, CPUs that work together operate on SIMD (Single Instruction Multiple Data) or MISD (Multiple Instruction Single Data), part of Flynn’s taxonomy. For example, your video card probably has the ability to apply a single operation (an instruction) to lots of pixels simultaneously (multiple data). Researchers at the University of California–Davis recently constructed a single chip with 1,000 independently programmable processors onboard. The device is energy efficient and can compute up to 1.78 trillion instructions per second.

The KiloCore chip (not to be confused with the 2006 Rapport chip of the same name) has 621 million transistors and uses special techniques to be energy efficient, an important design feature when dealing with so many CPUs. Each processor operates at 1.78 GHz or less and can shut itself down when not needed. The team reports that even when computing 115 billion instructions per second, the device only consumes about 700 milliwatts.

Unlike some multicore designs that use a shared memory area to communicate between processors, the KiloCore allows processors to directly communicate. If you are just a diehard Arduino user, maybe you could scale up this design. Or, if you want to make use of the unused power in your video card under Linux, you can always try to bring KGPU up to date.

A Dis-Integrated 6502

May 16, 2016 by 57 Comments

The 6502 is the classic CPU. This chip is found in the original Apple, Apple II, PET, Commodore 64, BBC Micro, Atari 2600, and 800, the original Nintendo Entertainment System, Tamagotchis, and Bender Bending Rodriguez. This was the chip that started the microcomputer revolution, and holds a special place in the heart of every nerd and technophile. The 6502 is also possibly the most studied processor, with die shots of polysilicon and metal found in VLSI textbooks and numerous simulators available online.

The only thing we haven’t seen, until now, is a version of the 6502 built out of discrete transistors. That’s what [Eric Schlaepfer] has been working on over the past year. It’s huge – 12 inches by 15 inches – has over four thousand individual components, and so far, this thing works. It’s not completely tested, but the preliminary results look good.

The MOnSter 6502 began as a thought experiment between [Eric] and [Windell Oskay], the guy behind Evil Mad Scientist and creator of the discrete 555 and dis-integrated 741 kits. After realizing that a few thousand transistors could fit on a single panel, [Eric] grabbed the netlist of the 6502 from Visual6502.org. With the help of several scripts, and placing 4,304 components into a board design, the 6502 was made dis-integrated. If you’re building a CPU made out of discrete components, it only makes sense to add a bunch of LEDs, so [Eric] threw a few of these on the data and address lines.

This is the NMOS version of the 6502, not the later, improved CMOS version. As such, this version of the 6502 doesn’t have all the instructions some programs would expect. The NMOS version is slower, more prone to noise, and is not a static CPU.

So far, the CPU is not completely tested and [eric] doesn’t expect it to run faster than a few hundred kilohertz, anyway. That means this gigantic CPU can’t be dropped into an Apple II or commodore; these computers need a CPU to run at a specific speed. It will, however, work in a custom development board.

Will the gigantic 6502 ever be for sale? That’s undetermined, but given the interest this project will receive it’s a foregone conclusion.

Correction: [Eric] designed the 555 and 741 kits

Hackaday Prize Entry: You Can Do Anything With A Bunch Of NANDs

May 10, 2016 by 6 Comments

Every few years, someone on the Internet builds a truly homebrew CPU. Not one built with a 6502, Z80, or a CPU from the 80s, either: one built completely out of 74-series logic chips or discrete transistor. We’re lucky enough to have [Alexander] document his build on Hackaday.io, and even luckier to have him enter it into this year’s Hackaday Prize. It’s an 8-bit computer built completely out of NAND gates.

Computers are just logic, and with enough NAND gates, you can do anything. That’s exactly what [Alex] is doing with this computer. It’s built entirely out of 74F00 chips – a ‘fast’ version of the ubiquitous quad 2-input NAND chip. The architecture of this computer borrows from the best CPUs of the 70s and 80s. The ALU is only four bits, like the Z80, but also uses the 6502 technique where the borrow is an inverted carry. It’s a small instruction set, a 2-stage pipeline, and should be able to compute one million instructions per second.

Designing a CPU is one thing, and thanks to Logisim, this is already done. Constructing a CPU is another matter entirely. For this, [Alex] is going for a module and backplane approach, where the ALU is constructed of a few identical modules tied together into a gigantic motherboard. [Alex] isn’t stopping at a CPU, either: he has a 16-byte ROM that’s programmed by plugging diodes into holes.

It’s an amazingly ambitious project, and for entering this project into the 2016 Hackaday Prize, [Alex] already netted himself $1000 and a trip to the final round of competition.

The HackadayPrize2016 is Sponsored by:

Crawl, Walk, Run: A Starter CPU

March 25, 2016 by 3 Comments

Last time I talked about getting started with CPU design by looking at older designs before trying to tackle a more modern architecture. In particular, I recommended Caxton Foster’s Blue, even though (or maybe because) it was in schematic form. Even though the schematics are easy to understand, Blue does use a few dated constructs and you probably ought to build your take on the design using your choice of VHDL or Verilog.

In my case, my choice was Verilog. You can find my implementation of Blue on Opencores.org. I made quite a few changes to Foster’s original design. For example, armed with semiconductor memory, I managed to get all instructions to operate in one major cycle (which is, of course, 8 minor cycles). I also modernized the clock generation and added some resources and instructions.

Continue reading “Crawl, Walk, Run: A Starter CPU” →