I see considerable difficulties on sustaining your position on this.
Most x86 microarchitectures since williamette are still performant for daily tasks which do not involve 3D rendering. There are reports of people running light versions of linux and browsers on them, such as gnome web. gnome web runs even in tualatin, which has not the SSE2 instructions.
In regards of low power architectures, most people who know the mobile atom CPUs of the era mentioned by Nipos know Cherry Trail was and is still comparatively as fast or faster than a raspberry, uses as much or less power, and has a powerful gpu for the time and for smartphone and mobile chips. Providing a driver would not be dramatically harder than providing a driver for other intel GPU microarchitectures. There are also new fanless atom designs which are fit into industrial boxes with 32gb of ram, 8 or more cores, running below 15w, although they are more of a niche expensive device.
I use a pentium 4 524 with a i945G on haiku, which is an ancient graphics chip. In terms of GPU rendering, that is a thing. But, isnt haiku mostly focused on software rendering? How fast is it? Gnome Web runs far, far faster, in my pentium 4 machine, than in linux, no matter which distro (im not interested in LFS so i`m not able to test that)
I mean, i do understand why RISC enthusiasts like their chips. Easier development in some kinds of applied software or computer engineering scenarios, more registers, better to write low level stuff such as compilers, cheap, commodity hardware. But how much junkyard hardware is available that can run haiku (even on 512mb RAM !!!) on 32-bit x86 CPUs? And how good are in fact old x86 microarchitectures, especially the low power ones such as in this post, or any kind of them? how widespread are development tools and framework maturity on x86? aren`t assembly otimizations are still ubiquotous on x86 compared to other instruction sets? i see my P4 doing ultra-light office work about as fast as my ryzen laptop on haiku, what if netburst due to being so unorthodox has a deal with that? What if someone one day wants to port Haiku to old itanium/VLIW machines (god forbid the insurmontable and suffering tasks involved) or SPARC to try and see what would happen, should we also consider that exccentric and disencourage asking for help on development or support?
I think asking to move things is a little (with all due respect) very biased and detrimental to x86 users and developers in any sense, development, support, or information for new users.