How far can a 6502 be pushed?

It’s an 8 bit computer, and even the chip in the IIGS had a fairly limited address bus that capped its ram. But…

What could it be pushed into with clever programming, perifferals, and other tricks?

1 Like

a 2Mhz 6502 can render monochrome “Bad Apple” at 640x512 plus 44Khz sound in real time…

Given 2MB of linear address space … which it doesn’t have so imagine 16 bytes of dual-ported RAM that gets STA instructions poked into it as fast as the 6502 can execute them plus a JMP at the end, so it can do beam racing.

Not mine but I have seen it “live” …



A new 6502 computer is coming, this fall on NBC… err the internet. Not sure how long this link is good
for. This is what I consider a pushed 6502, since for the most part nothing new came out after 1989.

The 6502 is after the bigger stuff.

1 Like

I think the great majority of 6502 systems from back in the day ran at about 1MHz. Acorn’s BBC micro and Master ran at 2MHz, and allowed for a second 6502 processor running at 3MHz or 4MHz. There were, even back in the day, 6502 systems running at 20MHz (including TTL reimplementations.)

These days, 6502 parts are spec’d at 14MHz and can be seen running at 20MHz, maybe even more…
W65C816 overclock experiment

With an FPGA implementation, we can add a 64MHz 6502 as a second processor to a 2MHz Beeb or Master, and that’s pretty special. This sort of setup can run a smooth and impressive game of Elite, for example.

(Although in the video, the fast 6502 is emulated on a Raspberry Pi - that’s an approach which can now get up to 500MHz, but I don’t quite count it for the purposes of this discussion.)

(The most highly upgraded Acorn machines these days are festooned with multiple Raspberry Pi, acting as second processors, storage, video subsystem, network interfaces, and probably more.)

As noted, the first obstacle in trying to do more with 6502 is in the directly addressable memory - 64k at most, until you get into flexible memory mapping, which means some degree of breaking up code and data into chunks.

The aforementioned FPGA 6502 can support a 1MByte memory banking scheme which can run a surprisingly large and fast Conway Life program:

With the 65816, as seen in the iiGS and also in Acorn’s Communicator, a program can have access to a (nearly) flat 24 bit address space - 16MByte. But it starts to become clear that this is a lot of memory for the amount of bandwidth, so any temptation to make a large memory mapped screen will show up a limited ability to draw or clear the screen in reasonable time.

There are lots of discussions over on the 6502 forum about various ideas to build bigger and faster machines - most of them are of course pipedreams or speculation, but still of interest.