Intel Continues Prepping The Linux Kernel For X86S

[jeisom]

Long time ago, I had an Amiga, and though how shitty x86 is.
Years have passed, and so many less shitty architectures already bite the dust. 86k, 88k, Alpha, MIPS, Sparc, Itanium, PowerPC, SuperH to name a few.

And yet somehow not only x86-64 works while retaining legacy compatibility (think about Apple that had painful 68k->PPC->x86->Arm path), but it's not that much worse than "latest and greatest".

It's really unfair to compare let's say Apple M3 with x86-64 parts. Apple is expensive, Like $2000 pro laptop with 8GB of RAM in 2024.
They can use 3nm process, and throw a lot of silicon while keeping voltage and frequency low, because they will make profit anyway.
Ryzen 5950x has 8B transistors. RX6900XT has 26B including huge cache. Apple M3 Max has 92B. This is much more than 4090 has.
It's not architecture that makes the difference, but money. PC parts are always overclocked far too much to save on silicon. For example RDNA2 chips should never exceed ~2,2Ghz. They are painfully inefficient above that. Same story with Ryzens (series 5000 should run at around 4,5-4.7Ghz max even when undervolted)

I'm not a fan of x86, I just don't care anymore, as long as it works as intended. I doubt there will be any miracles if we go ARM or RISCV. x86 cores are very efficient nowadays. But there will be problems for sure, if you want to use older software.

A few things I feel you are mistaken about. You can still get Power systems. PowerPC were just the Consumer line.

Comparing performance between arm and amd64 is fair. Also, Apple's highest price 8GB model is $1700(Before storage upgrades). At $1849 you get 18GB. I really don't get Apple's allergy to a decent amount of ram.

If you are going to compare transistor count it would be most fair to compare to an apu like the 7945HX, which has 13B. That said I believe that Apple has a more transistors allocated for the GPU in theirs as AMD has dGPUs as well to fill in more performance. Probably best to compare the 25B in the M3 vs to the AMD 8945HS which has 25B as well.

[sobrus]

A few things I feel you are mistaken about. You can still get Power systems. PowerPC were just the Consumer line.

I'm thinking about consumer products only, although I know some of the architectures I've mentioned never had any commercial consumer product.

Comparing performance between arm and amd64 is fair. Also, Apple's highest price 8GB model is $1700(Before storage upgrades). At $1849 you get 18GB. I really don't get Apple's allergy to a decent amount of ram.

If you are going to compare transistor count it would be most fair to compare to an apu like the 7945HX, which has 13B. That said I believe that Apple has a more transistors allocated for the GPU in theirs as AMD has dGPUs as well to fill in more performance. Probably best to compare the 25B in the M3 vs to the AMD 8945HS which has 25B as well.

Good point, especially since 8945HS also contains an NPU. So now we can see that top of the line mobile Ryzen 9 8000 series (which I assume will be powering higher end laptop models) has the same transistor budget as entry level apple chip.

Apple knows where to spend money (beautiful screen, speakers, aluminium case, battery life, performance, camera) and where to cut corners (hard to repair, no upgrades etc). 8GB memory is just what they need to sell even more. And it will be buyer's fault that he/she didn't buy enough.

[creative]

It's about time.

[Artim]

[...] and where to cut corners (hard to repair, no upgrades etc).[...]

That's where you're wrong. Upgrades (on the software side) happen for much longer time, especially when it comes to iOS. And iPhones are the opposite of hard to repair, espeically when you compare them to any other brand. You just can't have any repair shop do that because of software emposed restrictions. So what you list as cutting corners is where Apple actually spends the money. Their performance is highly questionable, as is their battery life. And they have nothing to do with hardware like the screen. That's pretty much always Samsung or LG spending the actual money.

[bofkentucky]

Isn't the 64-bit mode backwards compatible? Most x86 systems have been 64-bit for ~20 years.

While mainstream intel went 64bit at the tail end of the Pentium 4 generation and into Core, there were 32bit only celerons and other lower tier SKU's into the mid-2010s

[kiffmet]

Will X86S also bring changes to the register set? ARM and RISC-V having 32 general purpose registers is one of the biggest advantages they have in comparison to X86_64 (16 + SSE/AVX regs).

[marios]

Let me tell my point again. If you are willing to break backwards compatibility, you should go for something other than x86S. If you are not willing to break backwards compatibility, you should stay at x86_64. Arguments like "ARM SoCs aren't compatible either" are irrelevant.

That being said, since there are people interested on my hate on x86, x64_64 (and x86S as a subset of x64_64) has some potential major drawbacks. It being a CISC architecture, it gives the implementer two bad alternatives. Either implement everything in silicon (costly in terms of silicon), or implement microoperations (makes implementation more complex, without any performance gain). A RISC ISA can (not that it always does, but it can potentially) avoid both alternatives. X86, also has too few general purpose registers. This causes more loads and stores compared to an ISA with more registers, which hurts performance and efficiency. The TSO consistency model, used by x86, can reduce parallelism of execution when it is not needed. Also guaranteeing the TSO model requires more transistors and power. X86s variable sized instructions, make the implementation harder (it also translates to wasted transistors and power consumption, potentially even adding clock cycles to the execution). Instead using fixed sized instructions looks like a simpler approach, without obvious major drawbacks (unless you are restricted by memory of course). Albeit fixable in future standards, x86 only supports 4Kb pages. Using larger pages, could make the page table walk cheaper. AVX/AVX2/AVX512 etc is a shitshow of extensions. Needing a new extension every time a new vector length is implemented looks like the pinnacle of bad design. Last but not least, x86 assembly is much harder to understand compared to other ISAs. This might not be important for most people, but for those that need to deal with assembly, it might be a big problem.

I hope that I stated enough reasons, to consider x86 an inherently bad ISA.

[jeisom]

Good point, especially since 8945HS also contains an NPU. So now we can see that top of the line mobile Ryzen 9 8000 series (which I assume will be powering higher end laptop models) has the same transistor budget as entry level apple chip.

The 8945HS is kind of a weird product, mostly because of AMD's naming scheme. 8945HS is an 8C/16T part and the 7945HX is an 16C/32T part, both Zen4 and both 780M gpu. Naming would imply that the 8000 part is the higher end, but it is actually a rebranded 7940HS. The 7945HX has 13B transistors which would imply that the NPU takes up most of the budget on the 2 other chips. Gotta love marketing departments.

[Artim]

[...]I hope that I stated enough reasons, to consider x86 an inherently bad ISA.

Only if you accept the fact that if x86 is inherently bad, this means every other architecture must be worse. Fact is, x86 brings the performance and the wide usability. So whatever you dislike about the ISA, it obviously is still better than anything else.

[S.Pam]

QEMU. There's no need to run FreeDOS natively anywhere that's not ancient hardware.

Lot's of embedded systems use Dos. However, they will probably not run very new hardware.