6 ms·
A 100 MHz Pin-Compatible 6502 Replacement
- necovek 5y agoA discussion on the original post (referenced in this article) was on https://news.ycombinator.com/item?id=28852857 https://news.ycombinator.com/item?id=28852857
- hulitu 5y agoToo bad that the memory and peripherals are working at a 1/100th of the processor speed.
- vardump 5y agoTFA says memory is mirrored on the FPGA. So while peripherals don't operate much faster, memory does.
- tenebrisalietum 5y agoSo that's a big problem for the some of the big 6502-based computers and video game systems of the 80s, if you ever wanted a 100Mhz Commodore 64 or NES for example. Those systems had more than 64K of addressable items, and various techniques were used to map regions in and out of the 64K 6502 address space. Commodore 64, for example, has 64Kbytes of RAM, 16K of OS and BASIC ROM, 4K of memory mapped I/O, and 4K of character generator ROM. A register at memory location 1 controls what's visible at a given moment. By default you have 40Kbyte of RAM visible along with the ROM and I/O devices. Certain games like Impossible Mission will turn off the ROMs and I/O devices at various times to get at all the RAM "underneath." NES is even more of a mess. The original Super Mario Bros. and a fair amount of "NROM" games fit in to less than 32KByte of ROM, which is by default all that's available to ROM on that platform. Then "mapper" ASICs appeared and there are well over a hundred, and that's a challenge to anyone who wants to implement an NES emulator - emulating all the mappers that had to be reverse engineered. Some are common like the MMC1 and 3 and some are unique to specific games. Now I can see it working in the Apple II because the original platform I think had 16K of RAM, with the possibility to upgrade to 48K of RAM, and 8K of ROM near the top. BASIC was always loaded in from tape or disk. Video was writing to framebuffer and straightforward. You did have 128K Apple systems which came later but plenty of software that would work on an unexpanded Apple. Still awesome though.
- LocalH 5y agoThe problem was solved in the end of the C64's commercial life - the SuperCPU adds a 20MHz 65816, and using memory mirroring and wait states when accessing the 1MHz bus, it indeed effectively gives you nearly the full 20MHz in non-I/O situations, and even in I/O situations, the CPU can, with appropriate code, execute several cycles of instructions between 1MHz I/O accesses. The SuperCPU is the only way that I know of on the C64 to be able to literally write a new value to a memory location or chip register on every single 1MHz cycle.
- londons_explore 5y agoHow many old 6502 systems are still in use, and people want them to run faster? I'm thinking embedded systems like elevator controllers. But few of those systems need to get more performance out of a 6502 than it originally had. I would guess that many computer systems of this era will use delay-loops for timing some things, which means that running the CPU faster is going to break some code. That means there isn't much benefit to running the CPU faster unless you're happy to make source code modifications to the application you want to run... And at that point, it really might be easier to just run the whole thing on modern hardware...
- tssva 5y agoIf you look at the original announcement article for this project, linked in a comment above, it was developed mainly for use in chess computers from the 1980s.
- mdp2021 5y agoMany of us can still do 6502 assembly naturally. Fast 6502? Simplified computing. Understandable systems. Oh, and trustable. Less concerns about layers of unchecked code... Actually, it's a pleasant dream, the idea of an evolution of, say, a C64, keeping the projectual simplicity as much as possible, as a laptop. "Know your system byte by byte".
- stuaxo 5y agoWould like to see this in a BBC Micro
- analog31 5y agoA challenge with any such project, is that most code running on the old micro's took the processor clock speed as an implied system constant. So, timing of things like video refresh, and sound, would become unworkable if you tried to change the clock speed. The Apple ][ had a very particular clock speed, to implement a video controller that shared the memory bus with the CPU. I know that I'm guilty of managing the timing of some peripherals by using loops full of NOP instructions. As processors have gained more hardware features such as sophisticated timers and memory controllers, it has become easier to write code that runs independently of the clock speed. In the case of old systems, it's probably possible but might require access to the original source code. Some early PC-compatible motherboards with higher clock speeds had a front panel switch, so you could run the machine at the old speed for software that needed it.
- CGamesPlay 5y agoI am reminded of this similar but unrelated project: "Reverse-Emulating NES" https://hackaday.com/2018/05/31/reverse-emulating-nes-nintendception/ https://hackaday.com/2018/05/31/reverse-emulating-nes-ninten...
- BoxOfRain 5y agoI remember when this first came out, it's an astonishingly cool hack! What a triumph of "because it's cool and I can".
- bitwize 5y agoIt's really more like "forward emulating a NES cart", and it's been done in things like Wolfenstein 3D for Game Boy Color, and even since the dawn of the console: The Mega Man developers discussed using an ICE attached to an HP 9000 workstation to develop the game with. But that demo/presentation is cool, and definitely an interesting medium in which to present with lots of hack value.
- pwned1 5y agoCan I run MS Basic on it to revive the Tandy Color Computer III programs I wrote in 1992?