Amd Entered the CPU Market with Reverse-Engineered Intel 8080 Clone 50 Years Ago

Software or microcode emulation works pretty well.

So it would be faster and more efficient when sticking to the new subset and Nx slower then using the emulation path.

There are software compiled today without using MMX support. I was thinking the idea of something that is open or for licensing is an 86 ISA that is forward compatible. And for customers that requires strict backward compatibility they could still source it from AMD and Intel.

i.e Software compiled for 86 should work on x86. The value for backward compatibility is kept with both Intel and AMD. If the market wants something in between they now have an option.

I know this isn't a sexy idea because HN or most tech people like something shiny and new. But I have always like the idea of extracting value from the "old and tried" solutions.

> A subset of an ISA will be incompatible with the full ISA and therefore be a new ISA. No existing software will run on it. So this won't really help anyone.

This isn't an issue in any way. Vendors have been routinely taking out rarely used instructions from the hardware and simulating them in the software for decades as part of the ongoing ISA revision.

Unimplemented instruction opcodes cause a CPU trap to occur where the missing instruction (s) is then emulated in the kernel's emulation layer.

In fact, this is what was frequently done for «budget» 80[34]86 systems that lacked the FPU – it was emulated. It was slow as a dog but worked.

Wut?

>AMD and Intel Celebrate First Anniversary of x86 Ecosystem Advisory Group Driving the Future of x86 Computing

Oct 13, 2025

Standardizing x86 features

Key technical milestones, include:

    FRED (Flexible Return and Event Delivery): Finalized as a standard feature, FRED introduces a modernized interrupt model designed to reduce latency and improve system software reliability.

    AVX10: Established as the next-generation vector and general-purpose instruction set extension, AVX10 boosts throughput while ensuring portability across client, workstation, and server CPUs.

    ChkTag: x86 Memory Tagging: To combat longstanding memory safety vulnerabilities such as buffer overflows and use-after-free errors, the EAG introduced ChkTag, a unified memory tagging specification. ChkTag adds hardware instructions to detect violations, helping secure applications1, operating systems, hypervisors, and firmware. With compiler and tooling support, developers gain fine-grained control without compromising performance. Notably, ChkTag-enabled software remains compatible with processors lacking hardware support, simplifying deployment and complementing existing security features like shadow stack and confidential computing. The full ChkTag specification is expected later this year – and for further feature details, please visit the ChkTag Blog.

    ACE (Advanced Matrix Extensions for Matrix Multiplication): Accepted and implemented across the stack, ACE standardizes matrix multiplication capabilities, enabling seamless developer experiences across devices ranging from laptops to data center servers.

The "s" stands for "stupid".

But it's fortunate that they realised the main attraction to x86 is backwards-compatibility, so attempting to do away with that will lead to even less marketshare.

AFAIK, most of event x86_64 patents are largely expired, or will be within the next 6 years. That said, efforts for a more open platform are probably more likely to be centered around risc or another arm alternative than x86... While I could see a standardization of x86 compatible shortcuts for use with emulation platforms on arm/risc processors. Transmeta was an idea too far ahead of its time.

I don't think it works that way in practice.

Some distributions like Debian or Fedora will make newer features (such as AVX/VEX) mandatory only after the patents expire, if ever. So a new entrant could implement the original x86-64 ISA (maybe with some obvious extensions like 128-bit atomics) in that time frame and preempt the patent-based lockout due to ISA evolution. If there was a viable AMD/Intel alternative that only implements the baseline ISA, those distributions would never switch away from it.

It's just not easy to build high-performance CPUs, regardless of ISA.

If it's going to do that it's going pretty slowly.

In which space? Desktop and high performance servers? Why would it?

Mature gallery of software to be ported from TSO to weak memory model is a soft moat. So is avx/simd mature dominance vs neon/sve. x86/64 is a duopoly and a stable target vs fragmented landscape of ARM. ARM's whole spiel is performance per watt, scale out type of thing vs scale up. In that sense the market has kind of already moved. With ARM if you start pushing for sustained high throughput, high performance, 5Ghz+ envelope, all the advantages are gone in favor of x86 so far.

What might be interesting is if let's say AMD adds an ARM frontend decoder to Zen. In one of Jim Keller's interviews that was shared here, he said it wouldn't be that big of a deal to make such a CPU for it to be an ARM decoding one. That'd be interesting to see.

Why would the market jump from one proprietary ISA to another proprietary ISA?

Well, given some of the political/legal gamesmanship over the company itself the past few years, it could very well self destruct in favor of RISC-V or something else entirely in the next decade, who knows.

I was on this hype train when Apple switched about 6 years ago, but too much time has passed since then without much movement outside of Apple-land for me to believe it anymore. Apple is a special case; they have done this twice before (68k to PPC, PPC to x86), they have a smaller/tighter ecosystem, and they are vertically integrated. I think a lot is going to change in PC-land over the next decade or two, but I don't think it's going to look like Apple's transition at all.

20 years is half of x86's lifetime and less than half of the lifetime of home computing as we know it.

So this is kind of a useless question, because in such a timespan anything can happen. 20 years ago computers had somewhere around 512MB of RAM and a single core and had a CRT on desk.

>Does anyone seriously think ARM isn't going to obliterate x86 in the next 10-20 years?

Lunar Lake shows that x86 is capable of getting that energy efficiency

Panther Lake that will be released in around 30 days is expected to show significant improvement over Lunar Lake

So... why switch to ARM if you will get similar perf/energy eff?

It seems to me that interest in AArch64 for on-promise general-purpose compute workloads has largely waned. Are Dell/HPE/Lenovo currently selling AArch64 servers? Maybe there is a rack-mounted Nvidia DGX variant, but that's more focused on GPU compute for sure.

I'm talking about way more than just CPUs... And for your question, we'd pretty much miss out on modern-like mobile phones entirely. 90nm -> 18A/1.8nm is a LOT of reduction in size and energy... not to count the evolution in battery and display technology over the same period.

Now apply that to weapons systems in conflict against an enemy that DOES have modern production that you (no longer) have... it's a recipe for disaster/enslavement/death.

China, though largely hamstrung, is already well ahead of your hypothetical 2005 tech breakpoint.

Beyond all this, it's not even a matter of just slower, it's a matter of even practical... You couldn't viably create a lot of websites that actually exist on 2005 era technology. The performance memory overhead just weren't there yet. Not that a lot of things weren't possible... I remember Windows 2000 pretty fondly, and you could do a LOT if you had 4-8x what most people were buying in RAM.

One superpower being in 2005 for CPUs and another being in 2030, and at cold/hot war, would be decisive.

If society as a whole reverted to 2005, we would be fine.

Software is much, much more bloated today than it was in 2005. 64-bit CPUs were available, but not quite mainstream yet. A "high end" consumer system had a couple gigabytes of RAM and chipset limitations generally capped you out at 4 or 8 gigs. You were lucky to have two CPU cores.

If you took today's software and tried running it on a memory constrained, slow, 2005 era system, you'd be in for some pain.

[delayed]

I think there have been many improvements since 2005 that are not dependent at all on the process node.

EU or RU.

They said that they implemented x86 ISA memory handling instructions, that substantially sped up the emulation. I don't remember exactly which now, but they explained this all in a WWDC video about the emulation.

Apple didn't implement x86 ISA in hardware, they have a few instructions that change memory behaviour to make emulation faster.

Can you explain what you tried so say with that?

Customer needs don't really matter in cases where monopolist (ab)uses the law to kill competition. That's the MAIN reason why monopolies are problematic.

This article is not about that. This article is about the AMD Am9080, which was a unlicensed clone of the Intel 8080.

The licensing deals that legitimized AMD's unlicensed clones came later.

Weren't legal protections for semiconductor masks rather lax in the 70s, at least in the United States? You might need certain patent licenses for the manufacturing process, but the chip itself was largely unprotected.

That wasn't the first time they had similar products out-speeding Intel. I have the CPU from the first PC I owned tacked to the front of my current main PC with a Ryzen. That was clocked at 20MHz IIRC (I'm at parental home ATM so can't confirm) where the Intel units topped out at 12MHz (unless overclocked, or course).

Intel did indeed later license AMD to produce some clones, but it was not due to their good heart, but those were cross-licensing deals, with AMD producing clones of some Intel chips and Intel producing clones of some AMD chips, which could be used as peripherals for the Intel CPUs.

Then there was the big licensing deal for Intel 8088 and its successors, which was forced by IBM upon Intel, in order to have a second source for the critical components of the IBM PC.

European countries "acquired" quite a few Chinese trade secrets in the past. And from eachother to be fair.

IP is one of those things you invent once you made it to the top.

Given that market control is one of the few ultimate gating factors that makes you thrive or die as a company, it’s no surprise that anything that could be used as a mechanism to maintain market control would be.