The end of x86S
In fall, we reported here that Intel’s Panther Cove CPU architecture coming in two years (not in Panther Lake CPUs though) reportedly adopts an upgrade to the x86 instruction set labeled APX and also the AVX10 SIMD instructions, which are a compromise between AVX-512 and AVX2 designed for hybrid processors. However, the x86S architecture, the third of these awaited ISA innovations, will not be coming. It has in fact been cancelled.
It’s unclear whether this is something that was long coming, or a result of the departure of CEO Patrick Gelsinger and policy changes introduced in its wake, but Intel has now announced that it is ending this particular effort to reform the x86 (or x86-64, since only the 64-bit version is relevant today) instruction set, which was referred to as x86S or x86-S. It was supposed to be a cleanup of the historical baggage and now unnecessary legacy features in the architecture that could help processors perform more efficiently (and better compete with ARM, for example, I guess we can add the unsaid). Or at least that was the hope. Such a reduction of legacy cruft and simplification and modernization of the x86 architecture has long been called for from various sides.
This was a first cautious step towards reforming the x86 instruction set and removing some of its obsolete parts, which would effectively break its tradition of backward compatibility in exchange for disappearance of some of the historical complexity complicating processor design, as well as software. This change, for example, assumed that the processors would only support new native 64-bit operating systems (as they would be booting straight away into 64-bit mode). However, support for 32-bit applications in user space would be retained, so bulk of the useful software compatibility would be preserved despite the partial breakage. We discussed the proposed changes in more detail in this article:
Read more: x86-S: Intel wants to drop legacy compatibility from processors
The x86S initiative was announced in May 2023 and Intel has been working on the concept and specification since then. Back in June, an updated 1.2 version working draft was released. However, Intel has now announced that it is discontinuing the initiative and is no longer working on it. The plan to modernize and clean up the instruction set in this way has been abandoned.
We remain deeply committed to the x86 architecture, as demonstrated by the creation of the x86 Ecosystem Advisory Group in collaboration with AMD and other industry leaders. This initiative reinforces our dedication to securing a strong future for x86, building on decades of software compatibility. While we have pivoted away from the x86S initiative, our focus remains on driving innovation and collaboration within the x86 ecosystem. (Intel pro Tom’s Hardware)
Was it AMD who opposed it? Or Microsoft?
The reason for terminating x86S development is not clear. The company’s announcement cites a plan to focus on instruction set development within the x86 Ecosystem Advisory Group joint body, which was formed by Intel in October together with its rival AMD. This consortium is intended to harmonise the compatibility of the two companies’ processors and future instruction set development. Therefore, an explanation that offers itself is that the x86S project did not have the support of AMD, or for some reason could not be reconciled with this collaboration between the two companies.
Read more: Intel working with AMD on the future of x86. Afraid of ARM?
The other possibility is that the two companies plan to develop something together that would replace the x86S in scope. Although in that case, we would expect that x86S would not just be unceremoniously scrapped, instead assimilated into the new development as one of the draft proposals to take concepts and ideas from. And in such a case, it is likely that the cancellation of x86S as a stand-alone program would only be announced after a replacement was publicly introduced.
Interestingly, Intel also mentioned that this decision was driven by feedback from the “ecosystem”, which may not be talking just about AMD. So it is possible that software partners like Microsoft and the Linux development community may have had reservations about x86S. Linus Torvalds is an important member of the x86 Ecosystem Advisory Group (his commentary can be found at Real World News).
And of course Intel needs Windows to support x86S, so a disagreement with Microsoft would be a showstopper. It’s possible that the very idea of breaking compatibility clashed with preferences of important ecosystem partners, with the PC industry, or an unignorable part of it, wanting full x86 compatibility to be maintained.
Brain drain or advantages not being there?
Lastly, it could be possible that Intel stopped development of x86S for simpler reasons – it could have shut down the team of engineers working on the project due to ongoing cost-cutting measures, or important authors left the company on their own (or were poached into other companies).
Also, however, cost-benefit analyses may have concluded that the changes x86S proposes to make to the processors will not yield significant benefits, and that it therefore makes no sense to break compatibility and require significant changes to operating systems if there is no performance or other feature improvement associated with it that would be worthwhile. This may have been what the objections of software companies and the ecosystem in general were about.
The company clarified in subsequent statements that unlike with x86S, it is continuing to work on the upcoming AVX10 instructions, so these are still on the agenda. We assume that the APX enhancement project that adds extra working registers has not been cancelled either – unlike x86S, it has the potential to directly improve performance while not breaking backwards compatibility.
Intel also confirmed that it is still working on the Flexible Return and Event Delivery (FRED) specification. This is another plan for reforming old parts of the x86 ISA that existed in parallel with x86S (and has been in the works for significantly longer). FRED is also focused on a different area than x86S: it proposes changes and improvements to the way interrupts and exceptions are handled.
Source: Tom’s Hardware
English translation and edit by Jozef Dudáš
⠀
- Contents
- The end of x86S
