MOS Technology 6502
The MOS Technology 6502, or MCS 6502, is an 8-bit microprocessor that was designed by Chuck Peddle and Bill Mensch for MOS Technology in 1975. When it was introduced at around $25 it was the least expensive full-featured CPU on the market by a considerable margin, costing less than one-sixth the price of competing designs from larger companies such as Motorola and Intel. It was nevertheless faster than most of them, and, along with the Zilog Z80, sparked a series of computer projects that would eventually result in the home computer revolution of the 1980s.
The original NMOS 6502 turned out to have unusually few bugs during the development cycle, unlike some competitors, but there were still a small number of quirks that occasionally created trouble. An early bug was the malfunctioning 'ROR' instruction on pre-June 1976 CPUs. The instruction basically behaved like 'ASL' (Arithmetic Shift Left). Due to this problem the ROR instruction was only documented as "will be available on MCS650X microprocessors after June, 1976". Very early computers like the KIM-1 produced in 1975 were fitted with 6502 CPUs with the ROR bug.
A longer lasting issue (affecting all NMOS 6502 CPUs) was the 'JMP ($xxFF)' bug. This is a jump to an indirect address, where the LSB of the destination address is in location $xxFF ('xx' can be whatever), and the MSB of the destination address is in location $xy00, where xy = xx+1. The bug is that in the case where the LSB byte is in the last address of a page (as indicated by the FF) the MSB is not read from the first byte of the next page but instead it's read from the first byte of the current page. So, if we have JMP ($05FF) the address should be composed of an LSB from $05FF and an MSB from $0600. Instead the MSB is read from $0500. This bug would be rare enough that a user may never be hit by it, particularly when using an assembler which knew about the bug and could detect a problematic address.
The CMOS version
The JMP bug, and a couple of others, were removed when the processor was redesigned using CMOS instead of NMOS. The CMOS version was designed by Bill Mensch (one of the original designers of the 6800 and the NMOS 6502). The CMOS 6502 added additional instructions and addressing modes. It was licensed to many different companies for production, where some of them would use slightly different designs to fit their specific needs. The most well-known versions are the Rockwell 65C02 and Western Design Center (WDC) 65C02. Other variants were made by Synertek and GTE (the GTE version worked well as a drop-in replacement for the NMOS 6502 even in older Apple II computers, although later Apple would use Rockwell 65C02 chips in the Apple IIe and IIc models). WDC continued to design other variants of the 65C02 and are still at it as of 2016. There were other manufacturers as well, including but not limited to NCR, California Micro Devices, and Ricoh. Ricoh made a variant for the Nintendo game system, this version missed the BCD instructions - presumably never needed for Nintendo games.
Synertek and Rockwell were also second-source vendors of the NMOS 6502
Other variants of the 6502
The 6501 was the very first version, but did not last long (due to the Motorola lawsuit - because it was pin-compatible with he 6800). According to Chuck Peddle it wasn't meant to last either - who knows, but in any case the 6502 was already in the design and was on the market soon after. When Commodore bought MOS Technology they started to produce numerous variants of the 6502, and variants of other MOS parts too - almost at the whims of whatever the Commodore engineers felt they needed at the time.
Pin-compatible with the 6800, otherwise (except for the 6800 clocking style) similar to the 6502. DIP40.
Addresses 4KB. Leaves out RDY, includes /NMI and /IRQ. DIP28
Addresses 8KB. Leaves out /NMI and RDY, includes /IRQ. DIP28
Addresses 4KB. Leaves out /NMI, includes RDY and /IRQ. DIP28
Addresses 4KB. Leaves out /NMI and RDY, includes /IRQ and ϕ1 out. DIP28
Addresses 8KB. Leaves out /NMI and /IRQ, includes RDY. DIP28
Like 6510: With I/O port, but also with 256 bytes of internal static RAM, dual-mapped to zeropage as well as stack page ($0000-$00FF and $0100-$01FF).
"Microprocessor with memory management" according to the datasheet. By using three 6502 N.C. pins, one redundant VSS pin, and requring a two-phase clock input instead of a one-phase clock input and two clock outputs, room was made to include four signals P0-P3 which could be used to switch between 16 banks of 64KB memory. The address bus, data bus, and R/W signal are tri-state, unlike the 6502, and the state is controlled by the fifth available pin named AEC (Address Enable Control). This could be used to assist with DMA. Note that P0-P3 aren't tri-stateable.
The banking mechanism is not straight forward. The value of P0-P3 is set by two different internal registers, 'management execute register' and 'management indirect register'. The value of these registers are set by writing to location $0000 for the execute register, and location $0001 for the indirect register. Most of the time the value of the execute register determines P0-P3. However, when instructions LDY and STY are used the processor maps the value of the indirect register to P0-P3. This happens when the data transfer is to occur. After one cycle the processor switches P0-P3 back to the value of the execute register.
A 6502 with an 8-bit bi-directional I/O port with the data direction register at address $00 and the port itself at address $01. However all 8 bits were only brought out to actual pins on the 6510-1 and 6510-2 variants, the 6510 only brought out the first 6 bits to pins. The 6510 was used in the Commodore 64 where the main usage of the i/o port was to manage bank-switching (you could switch between ROM and RAM banks, for example). In the C64 the i/o port also controlled the electric motor of the tape recorder. The main reason for using this variant CPU appears to be that it saved the use of a separate 6522 VIA chip. The 6510 was also used in some other Commodore computers.
The 6510 series also used a tri-state address bus, unlike the original NMOS 6502.
This is a HMOS version of the MOS 6510. It appeared in 1985. Except for using HMOS instead of NMOS, it was exactly the same as the 6510, running at around 1MHz and compatible all the way down to the undocumented opcodes and original errata.
A 2MHz version of the 8500. Used in the Commodore 128.
Current status of the 6502
WDC continues production and licensing of the 65C02 to this day, in 44-pin PLCC packages as well as the good old 40-pin DIP variant, but now it runs at up to 14MHz (conservatively. 20MHz is easily achieavable and 25MHz has also been reported, when used with components that can keep up.) WDC also licenses 65C02 cores that run up to 200MHz. The 65C02 (including cores) still sells in hundreds of million units per year as of 2015, according to the WDC homepage. There's probably one or more 65C02 in your car, unless it's as old as my car. If you have an implanted heart defibrillator or a pacemaker then you probably (and certainly if it's a defibrillator) have a 65C02 inside your body, the only (AFAIK) processor so far approved for human body implants.