Death of x86

Originally posted on June 13th 2024

by Gethyn “Xylemon” ThomasQuail

You know it’s funny, I was thinking about writing this for a while now, and then Microsoft made a special recent announcement and well, I knew it was time.

Yes, I think x86 is finally starting to kick the bucket as it should. It has been for sometime, but at an incredibly slow pace. However, with some recent events in the last decade, I think the industry is finally going to start ramping up efforts to move beyond x86, and honestly, good riddance.

What’s the big deal?

A friend and I were talking the other day about how modern operating systems have all sorts of evil hacks in place to make sure old junk runs like it should.
Specifically they mentioned Windows and surrounding drivers have lots of glue for old games which result in games like Lego Island still ‘running’ on modern machines. A great game for sure - but it’s just wild to think about that a modern OS has specific code inside its foundation to support 25+ year old games. The x86 architecture is kind of like that as well. It has special instruction sets for tons of wacky and now uncommon things.
For instance, MMX was a touted feature for Pentium CPUs, and its implementation was seen as a welcome addition, because John Carmack of id Software expressed interest in it for Quake with colored lighting.
However, id and many others never used it, since the rise of GPUs happened shortly after. At least it was used by Half-Life and Unreal (Tournament) when playing without a 3D accelerator card, but imagine how many instruction sets bloat the CPU just because of some twenty something year old decision that benefits only old software.

We are being held back by including so much legacy cruft not just in our operating systems - but our hardware. That’s part of why they’re so big, expensive, and waste so much power. They can be built more efficiently, it’s just that since then programs have tried to take advantage of x86’s cacophony of instructions and so you’ll run into knives rather quickly.

Surely there must be some compelling reasons x86 has stuck around right?

Why x86 chugged along

While it’s easy to be cynical and say x86 has only been so dominant because the big players who produce the hardware want to milk it as long as they can, it’s only slightly true in the grand scheme of things. Sure Intel/AMD has a ton of awful patents, install bases and financial interest in x86, but ultimately with CPUs we have to consider:

1. The cost of researching development of CPUs, building new factories or making adjustments to them, employment, etc
2. Enterprise is always slow to adopt new untested hardware and software. Most programs would have needed to be rewritten or even worse, emulated poorly with little to no improvements in speed or power consumption
3. Most consumers would have been in the same boat as above just in a different way, and for a long time, big beefy ultimate PC builds were the cool thing when it came to gamers/enthusiasts. And that’s certainly still true today, but perhaps not as it once was.
4. The speed and incentive over x86 just wasn’t there

This was the reality for a long time in the 90s to even early 10s.
But then, something magical happened. RISC finally found its footing, with low power consumption it was a natural fit for portable phones. While it wasn’t the first smartphone - as both my parents had a Blackberry or Windows CE phone - when the iPhone launched it undeniably changed the phone market forever. Soon followed with Android by Google, we now had two UNIX based or UNIX-like operating systems running on RISC based CPUs, and almost everyone in the entire world uses these systems. Similar to Disco/House music in the 1990s, the modern smartphone was UNIX’s Revenge.

With tested and desired platforms, and the push for native apps instead of WebApps, we had developers learning to write for different hardware from their own for the first time in ages.
Compilers got smarter, developers got smarter, and people saw the need to support other platforms beyond just “Mac”.
Even Wine (the Windows software layer that most of the Steam library is powered by) tested out Android support years ago. Flash forward a little over a decade and every component in Wine has been abstracted to work regardless of CPU architecture.
Moves like this are big, people who knew saw the writing on the wall.

But it’s still x86 World

Walk into any household in the world that has a desktop or laptop and odds are it is an x86 based machine. Probably every “PC gamer” out there has x86 hardware. This is of course because every software in that market is released for x86, and then there’s compatibility with all the other old software. This has been Microsoft Windows its biggest selling point, that older apps will always “work”. It’s probably why in part their original RISC effort, Windows ARM. failed with the Surface tablets. Most people use Windows because old stuff loads and it’s all they ever knew, but even with that the cracks are starting to show.

More programs every year fail to run on Windows, with some even now using DirectX to Vulkan layers on there, or simply choosing to run Wine/Proton (UNIX’s revenge indeed).
The common folk have been used to multiple operating systems for a while now, and how stuff is (and is not) compatible to some degree. A decent amount only use smartphones (gasp). This is no longer in the hands of only enthusiasts (PC Gamers).

RISC-V recently happened, creating a powerful open base for other RISC CPU manufacturers to utilize.
The recent Mac Airbooks have shown to out-perform their x86-based laptop competitors all with great battery life and higher performance, imagine what it could do for the Steam Deck? Only time will tell.

In a somewhat surprising move, Microsoft announced they have renewed interested in not just Windows ARM but that they want to tackle RISC based laptop and this time, they’ll have a transitional layer for x86 apps. Perhaps it will be a big failure, it won’t run well and no one will use it or gamers will hate it. I can’t see the future, but if Microsoft is serious about this, then we might finally be heading into a RISC transitional era.

Looking Forward

So, how will we run our old games and software in the future? One solution will be transitional layers, whether it be emulating Operating System calls from Windows or x86 CPU instruction sets. With patents finally expiring (and/or Microsoft willing to pay) we’ll see x86 running more-or-less at native speeds like we see on Mac OS 12 RISC machines. Worst case: we can always emulate selectively and still have an integrated experience that doesn’t reek of discomfort from the traditional “emulate a whole OS” approach. It worked for games consoles for decades. We’ll be fine.

Also a more recent phenomeneon has been complete reverse engineering efforts. People rewrite code from scratch or clean up decompiled binaries for providing a true portable experience with enhancements. Editors note: We do not condone breaking copyright law in your jurisdiction, we are merely reporting that it happens.

Once that is done, they’re basically in the same boat as games like DOOM and Quake which were blessed with official, open-source releases. These will probably be ported to everything, forever.
The gist is, so many people are finding incredible inventive ways to run their old favorite software, or have created entire replacements in the open source space that you don’t need that old EXE binary any longer. Unless it’s for studying, of course.

Conclusion

Will RISC based CPU’s be the future architecture powering all devices a hundred years from now? It’s hard to say, but at least we know in thanks to smart emulators, transition layers, and open-source efforts that we will see our favorite games live long past us. And if you are using a RISC-based desktop and reading this in the far future, then you know when the stone age of computers started to finally end.

Ultimately what will be the biggest challenge now is not getting old software to run, but “the last mile”.
Getting hardware in the hands of consumers, with a smooth transition where they notice little to no change, or the change is so great they don’t care about old software.

This seems near impossible in the PC market, but someone would’ve thought the same in the early smartphone days where Blackberries and Windows CE ruled the scene. Now, they’re nowhere to be found in anyone’s pockets.

Don’t worry. While the old hardware might go away, software lasts forever. If you want to truly preserve the past, you need to embrace new technology as well. For example, there are now low cost FPGA based hardware devices available which we can use to re-implement videogame consoles, or obsolete/unreleased/fantasy hardware.

Imagine that one day, you too could stick a USB-drive sized component into your machine and flash a core for any old hardware of your choosing. Cycle accurate, hopefully open-source, hardware replication for soundcards, graphics cards - the list goes on. This will be possible, since people have already successfully re-implemented videogame consoles of the 1990s with great results.

While we could still be working on Pentium processors, with 3dfx Voodoo graphics, the fact is that those are also no longer made. If you want to preserve these things for generations to come, help by re-implementing them in software with applications such as PCem or 86box or by re-implementing them using modern FPGA hardware.

And no, you don’t necessarily have to implement a whole system. The point is that you can create and apply a custom chip design to accelerate certain tasks without spending money on extra hardware. Due to the advancements in FPGA technology coming down in price (such as the Tang Nano 20K) we might very well see modular components in our computers taking advantage of this new-found freedom. That way being able to run even a subset of specialized hardware instructions is not a problem. That’s the theory, anyway.

The sooner we switch, the less work we have in front of us.