6502.org

I mean, so much has been written for PC/Windows/x86. Pretty much all my games, CAD software, video editing software... some industrial control equipment running ISA/PCI stuff, it goes on and on and on.

Nobody talks about building a 6502 based machine for modern gaming, meanwhile you have 100s of different tech Youtubers going over Intel vs AMD, all these ATX/ITX mobos, NVIDIA vs AMD GPU, etc etc.

Yes, x86 is a major platform for existing software. But it's very complex, and so verification is difficult. We've seen with 6502 and with Z80 that once a verification suite appears, the great majority of implementations are found to have bugs. The verification of an ambitious design (pipelined, with caches, branch prediction etc) is even more difficult.

So, you'd need a solid verification plan.

And a solid plan for how to sell something which competes with Intel: you'd need to demonstrate you understood the legal landscape around the production of x86 compatible parts.

I note that Transmeta and Montalvo each took tens of millions of funding. These are very big numbers!

But I also note that I still don't understand the scope of what you have in mind: business, or pleasure?

They yammer about these things because this is the hardware they can buy today, and part of the direction the industry is heading.

Microsoft Windows can now run on ARM and runs most x86 applications. ARM is essentially universal in the mobile space.

GPUs dominate the discussion because it's mostly the GPU that governs display quality, speed, and responsiveness. Most games today are not CPU (as in host CPU) bound, but, rather, GPU bound.

The other problem is that the GPUs are being dual purposed and used by the crypto currency miners -- to the point of affecting pricing and availability that the gamers are sharpening pitchforks and wrapping torches.

I certainly used to care. I was on the front lines between the 68K and x86 wars about memory and addressing. The MHz wars between PowerPC and x86. All new and exciting, there are certainly differences between architectures and such.

But in the end, especially after my server side Unix experience, for the work I do, I've long past cared about the underlying machine. I was more interested in storage utilities and features (like raid, striping, mirroring, etc.) than I was about architecture. My clients were more interested in bang/buck including licensing, support, appropriateness to their workload, etc. Nobody I knew bought an RS-6000 because of the pipelining and such of the microprocessor. In the end, the CPU was but a single component of the overall system.

Today I buy Macintoshes. Partially because of the "all-in-one" nature of them. They're "good enough". I'm so far past the "build your own" scene. I simply don't have the interest, time, or cognitive load to keep up with who's overclocked liquid cooled CPU/Memory/GPU combo gives the extra percentage points on some arbitrary benchmark.

Having the machine power up at all and work smoothly and reliably is far more important. My current work machine (an iMac) is pushing 7 years old. My home machine didn't last me 10 years, but it came close, and it mostly failed because Apple deprecated it due to it's 32b graphics bus. The machine was fine, it was simply left behind (and, no, it doesn't bother me). So my current one is probably 3 years old, and going strong. Nothing like playing a high FPS video game with the other 11 cores are busy ripping a DVD...you can almost hear the fan from the machine running 1200% CPU.

And, yea, I run DosBOX to run old PC games on my Mac. I'm a sucker for Masters of Orion 2. But even then, many companies have put work in to their older games to do a reboot. Same game, running on modern OSes with modern graphics and networking. So that you can play a game that was once for a 640x480 screen on a modern 4K monitor. A good game is a good game.

When MOS (Motorola ex employees) had introduced the 6501, it was pin compatible to the Motorola 6800.
MOS then was sued by Motorola, so MOS slightly changed the 6501 pinout, introducing the 6502.

The 8008 was binary compatible to the Datapoint 2200 (TTL CPU),
but for some reason Intel assembler notation for the 8008 was slightly different from the Datapoint 2200.
When Federico Faggin did the Z80, he went for a slightly different assembler notation than Intel
for copyright reasons.

But those were more innocent times, and I'm not sure that using such tricks would prevent you
from getting sued by Intel nowaday.

Would be hard trying to beat giants like Intel at their own game.
Also, it would be hard nowaday to find a "niche" like Transmeta did.


Hmm...
If your CPU wouldn't be exactly pin compatible to the Pentium 4, and it somehow would "leak out by accident"
that cutting a pin here and desoldering a tiny SMD component there at the chip package would make it possible
to plug the CPU into a PC motherboard, this probably won't be your fault.

If the CPU would be sold as an "accellerator for 6502 gaming stuff", and sort of "a little microcode patch"
would show up later "from a dubious source" that would make the CPU able to understand 80x86 machine code,
and this can't be tracked back to your company... this probably won't be your fault, too.

But you better ask a lawyer about this.

Nitpicking mode.. the 6501 and 6502 were ready at the same time (they're both in the first datasheet).

Yes, probably this datasheet from August 1975.

But who would doubt advertisements and Wikipedia.

Yes, but the 6502 was sold on Wescon too.. (Woz bought one there.) Chuck Peddle said somewhere “(The 6501) was never supposed to be a real product anyway… it was just for demo’s.”
Anyway, here's something a bit more interesting: "Chuck believes that the worlds first microprocessor is not the much ballyhooed Intel 4004 or 8008, “(I am) not trying to be negative about the guys that did it… they are nothing more than calculator chips”. He believes the worlds first real CPU is Tom Bennett’s 8bit Motorola 6800 “…it’s terrible that guy never got any credit.”

Sometimes, from some of the things Chuck says, and indeed some of the things Bill Mensch says, I feel there's a bit of retrospective editing going on. It's so much easier to have been right if you are retelling the story with hindsight. (And it isn't necessarily conscious editing.)

I've the same feeling - at least about the 6501. The truest part is probably where he said that the 6501 was made 6800 pin-compatible in order to try out the territory. It would make sense to try probing a market which already existed.

I'm going to be a little less polite here, because it looks like something I've seen time and time again. You don't have a team, funding, or experience in VHDL or chip manufacturing processes. You've looked up what vendors are in the space, and saw all the big industrial equipment you can name but can't afford and don't know how it all works together (and lol at fab costs vs individual guerrilla startups, even with VC funding and older processes). Your link has all negative responses to your questioning, and then talking amongst themselves, without response from yourself.

What actually do you have, besides a vague notion that VLIW will somehow magically run x86 code faster?

It's one thing to be curious to learn & hypothesize about some massive project, or discuss the "why"s of alternative industry progression of technology. It's another to claim that you're going to start a big commercial project, while basically requiring everybody else to bring everything to the table, both technically and business-wise.

White Flame made quite a few good points there.
Not to sound discouraging, but I felt a need to mention just one of the technical problems to come:

68020 had 200k transistors, and the test software required had half a million lines of code.
80386 initially had 275k transistors.

I'd say, that the amount of test code required might increase exponentially with the complexity of the design,
Pentium 4 in 65nm technology had 188M transistors (about 684 times the transistor count of the first 80386).


Back in 2017, I gave a forum member an old book about 80x86 assembly language programming,
and the book contained a capter about the known bugs for everything 8086..80386.
IIRC the list went on like this in short form for about 30 pages.

IMHO just writing the test code required for something at the size of a Pentium 4 might become a bigger project
than writing your own C++ compiler from scratch (including the libraries).

For sure. However it's not like I have to do literally everything myself here - I definitely have genuine, quality connections in the industry and believe I can coordinate our efforts well. I am familiar with the use of logic diagrams to solve bugs like that (ie. under states x, y, z with this test vector what are possible correct and buggy solutions (?), how can we probe the prototype chip and compare its oscilloscope readouts to what our simulations say they should be?)

Of course, will I and the people I work with do it well enough?

Thanks for the link - it's a trailer for a documentary: Could be quite interesting: from Soul of a New Machine onwards, there are many good stories about efforts to make a new processor.

Just supposing a person were going to be successful in initiating a funded startup. What qualifications might they have?
- business experience
- leadership experience
- domain experience

I've worked for a couple of startups. From what I've seen, to convince the money people you'd need 10-15 years experience as an engineer, with some seniority, and excellent communication skills. You'd need to recruit a credible team, and convince both the money people and the recruits that you can do that. You'd need the organisational skills to grow an effort from 6 to 60 to 600 people, and to convince your hires that you can do that.

Everyone around you, investors, recruits, customers, acquirers, are all evaluating you, with skepticism, to see if you can do what it takes. Self-confidence helps, but is not a substitute for experience and competence. Experience and competence are not enough either: you need to be convincing.

After all that, you need a good idea, and a lot of luck. You'll probably end up doing something different from the idea, too, so you need flexibility.

So: 10 years experience, a year developing an idea, a year seeking funding.

One possible qualification for having the domain knowledge: can you get hired into a startup? And then, can you be a key hire, one of the first after the founders?

eton975, don't want to sound rude, impolite or even discouraging, but:

If one would say:
"I'm out to bring a man safely to the moon and back before the end of the decade.
I need collaboration and funding.
I have seen the related tools, and I know some guys who were involved in the Apollo program."

He could be happy to get enough collaboration and funding for building skyrockets to be used for fireworks.

;---

People building CPUs as a hobby here in the forum are not convinced,
convincing people building CPUs for a living would be more difficult,
not to mention attracting investors.


Don't give up your dream, but try to reflect a bit about what you are out to do, and why.

Maybe it would make sense to aim for a smaller project first,
something like a system on a chip,
just to test the legal\financial\logistical\technical terrain.

IMHO it would be helpful, if you would have something for "show and tell".
Maybe an 8085 (or 6502 because it's a 6502 related forum) on a collaboration site like github.
Having an 8085 would be a base for building 8086 what would be a base for building 80386,
but I'm not sure about the Pentium.

Good luck.