Advanced 72-bit processor design
Advanced 72-bit processor design
AMD developed the amd64 architecture, a 64-bit x86 instruction set. That’s impressive. We’ll rely on that technology for many years ahead. I’ve been exploring first how i386 and then amd64 processors function, analyzing their data handling. (To be honest, little has changed much over time except size and transistor density.) After some discussions with colleagues, I’ve put together my own research and designed a 72-bit x86 instruction set processor. Note: It’s not yet a finished CPU ready for motherboards, but simulations and virtual tests show promising results. Of course, 72 bits is quite ambitious—especially for memory usage. Don’t worry, I’m also working on solutions for that too. Currently, my focus is refining the design for better efficiency and speed. The architecture I’m testing is a single-core unit, yet it outperforms its amd64 predecessor in tests. That said, these are simulations; building it at this scale would require specialized equipment. If you’re interested, let me explain more about my development process. I can’t share visuals here, but I have plans for future sales if things progress. Such a processor could shine in server and supercomputing applications rather than everyday computing. It might also help with tasks like folding simulations. I’m currently developing a dual-core version, but architectural tweaks will be key. Also, running these simulations is extremely demanding—my system heats up significantly during loading. I’ll likely need better hardware if I want to enhance this further.
Could be because of current limitations, and I also believe it isn't necessary at this stage, which might explain why it hasn't been developed...
The advantage of a 64-bit processor lies in its ability to handle vastly larger amounts of memory compared to 32-bit systems. While current servers don’t reach the extreme limits you mention, increasing bit depth could provide future scalability and performance gains. A 128-bit architecture would certainly be more practical for demanding applications, though it’s still a topic of interest for advanced computing needs.
I believe architecture simulation is feasible, though figuring out how to run 72-bit code on a 64-bit processor presents challenges. It seems there isn't a straightforward method, especially without relying on emulation if it's accurate.
I'm here to help with your idea. Presenting at a big event or writing a paper could really impress you. Let me know how you'd like to proceed!
The choice for 72 rather than 128 comes from my current limitations in knowledge and focus. While 128-bit seems logical, I’m still developing my understanding. The shift happened because it would let me adopt a new RAM type that closely matches CPU cache performance without the delays of longer data paths. The challenge lies in adding more 'bridge' connections between modules; each data unit (1/0) grows exponentially due to the 3D setup of storage and logic, making it roughly 3-4 times more efficient than HBM while maintaining very low latency (around CL-8 at 4800MHz – purely theoretical based on simulations). This journey was humorous. I began in a program relying heavily on GPU calculations using floating point, which became unwieldy as complexity grew. I switched to dual CPU boards, built a custom Linux OS from scratch, and spent three months just getting it to boot. Eventually, I rewrote the code to run entirely on the GPU, but now I’m limited to that approach since both CPUs can’t handle more than we’ve already achieved. Currently, I’m running three Titan Xs, which leaves me stuck with a total loss of functionality for most tasks. The downside is that I won’t be able to use the hardware beyond this point.
That thing would be AWESOME!!!! I can only imagine the use case scenario where you need a ton of memory (like in a datacenter), and for consumers would mean a bit of performance boost! I guess you can contact VIA about it, so if they make a 72bit CPU, that would make them (and you) some money!
I just want the feedback and thoughts on the concept. No pressure, but I’m focused on finishing it properly.