Why Doesn't Intel Expose the Internal RISC Core to Programmers?
Hello everyone, starting with the Pentium Pro processor, Intel introduced a microarchitecture that translated x86 instructions into micro-operations (uops) executed by an internal RISC core right, so that there is significant performance improvements while maintaining compatibility with existing software.
I know the reasoning behind using a hidden RISC core is clear but why not expose the RISC instruction set directly to programmers? This could potentially lead to further performance gains and a smoother transition away from the x86 instruction set in the future. I understand that maintaining backward compatibility is crucial, yes. Since Intel is already committed to supporting legacy x86 modes alongside 64-bit operation, wouldn't exposing the RISC core allow for a more natural path towards a future RISC-based architecture?
It seems that Intel continues to extend the x86 instruction set with new extensions like AVX and SSE. Is there a specific reason for not leveraging the internal RISC core more directly?
3 Replies
It is interesting, it could benefit the performance but complicate the processor design and fragmenting the software ecosystem. This could prove more expensive than the advantages would justify...
Yh , I am curious to see if ARM architectures make inroads into the desktop space, it could push Intel towards a more significant change for real
I think you’ll see Intel processors integrate more and more different cores over time. They’re already doing it with their version of bigLITTLE architecture in the newer Core processors, integrated GPUs, and now a bunch of accelerators for AI, vision, etc. They were also active with RISC-V for a while (but that may have had more to do with Foundry than anything).
At the same time I think they are preparing for a world of application developers and APIs. Those devs may be abstracted enough from the silicon that they don’t care.
Curious to hear others’ thoughts.