Intel and AMD Form an x86 Ecosystem Advisory Group (phoronix.com) 55
Phoronix's Michael Larabel reports: Intel and AMD have jointly announced the creation of an x86 ecosystem advisory group to bring together the two companies as well as other industry leaders -- both companies and individuals such as Linux creator Linus Torvalds. Intel and AMD are forming this x86 ecosystem advisory group to help foster collaboration and innovations around the x86 (x86_64) ISA. [...] Besides Intel amd AMD, other founding members include Broadcom, Dell, Google, HPE, HP Inc, Lenovo, Microsoft, Oracle, and Red Hat. Here are the "intended outcomes" for the group, as stated in the press release: The intended outcomes include:
- Enhancing customer choice and compatibility across hardware and software, while accelerating their ability to benefit from new, cutting-edge features.
- Simplifying architectural guidelines to enhance software consistency and standardize interfaces across x86 product offerings from Intel and AMD.
- Enabling greater and more efficient integration of new capabilities into operating systems, frameworks and applications.
- Enhancing customer choice and compatibility across hardware and software, while accelerating their ability to benefit from new, cutting-edge features.
- Simplifying architectural guidelines to enhance software consistency and standardize interfaces across x86 product offerings from Intel and AMD.
- Enabling greater and more efficient integration of new capabilities into operating systems, frameworks and applications.
Really? (Score:4, Funny)
One of the rationales is:
Enhancing customer choice and compatibility across hardware and software
If this is the case, then why are Microsoft and Oracle involved?
Re: (Score:2)
and why Linus? What value, beyond political, does he provide?
Re: (Score:3)
Who says this whole thing is not political?
Re:Really? (Score:5, Informative)
I imagine he has some insight into what features would be beneficial to things running well and which would be a waste of chip space.
He's only one perspective, but definitely one I'd want when planning out new features for a next generation chips.
Re: (Score:2)
They all seem to have bet on AI tech being the next big thing, but it won't be. I wonder how much die space is wasted on AI accelerators that very few people care about.
Re: (Score:2)
They all seem to have bet on AI tech being the next big thing,
In a weirdly half arsed way. Somehow even getting decent inference on their GPUs (integrated or discrete) is a huge pain, especially if you want better than a completely generic third party Vulkan backend which hasn't been even looked at by someone who owns the same GPU as you.
NVidia has bet on it big time. Intel and especially AMD appear to be maximizing the spend while minimizing the effect.
Re: (Score:3)
I read a tweet just now from a woman who signed up to speak at a conference. The organizer took her profile photo that was cut off at chest height and used AI to make it a bit taller. The AI hallucinated a visible bra and plunging neck line for her.
It's way to early to be deploying this crap for anything that matters.
Re: (Score:2)
brilliant!
There's plenty of good uses of it, if one rather unfashionably calls it ML or even worse, machine vision. AMD still suck, and so do Intel.
Re: (Score:1)
This is a wild ass guess: MS/Linus/etc are involved to get the hardware and software people talking more amongst each other.
The Win 11 24H2 update affecting Ryzen CPU performance, and it being released after the launch of the first Ryzen 9000 CPUs, is all the evidence required why there is a real value in discussion between parties in the x86 space (both hardware and software).
Re:Really? (Score:4, Informative)
Political value is still value. And if he wasn't involved and it was just Microsoft, Oracle, and Red Hat everyone would be screaming conspiracy theories.
At least we have a reasonable chance of him calling others on their bullshit, and we know he won't just shut up and take it if they're trying some shit. He has no problem with blowing whistles on bullshit (unless it's his own flavor of bullshit, but that's what the other guys are for - to call him on his too).
Re:Really? (Score:5, Insightful)
Re: (Score:3)
There is truth to this. In the days before we all used NT based Windows, the old Windows 3.x used illegal instruction errors to get back into kernel mode, as it was generally one of the fastest exceptions at the time.
So when Intel was describing their new chips, they asked Microsoft engineers what their desires were, and one asked for even fast
Re: (Score:2)
To avoid him ranting on the power virus [zdnet.com] of AVX-512. /s
Re:Really? (Score:5, Informative)
He wrote the entire firmware, "Code Morphing Software", for the Transmeta Crusoe. He knows more about the x86 instruction set and operation more than most people on earth
Re: (Score:2, Troll)
Because it is a lie.
Re: (Score:2)
Re: (Score:2)
Apple? You're kidding right? They won't ship their chips in any machine but their own. ARM on the other hand...
Re: (Score:2)
Apple runs Acorn RISC Machine CPUs.. ARM. was once ABC - Acorn Business Computers. Now it is Advanced RISC Machines. Apple has completely adopted ARM, as its licensing fees are easy to absorb, even easier than USB.
intended outcome #4 (Score:5, Insightful)
- weaken ARM
Re: (Score:2)
And arm based laptops are still kicking Intel's ass on battery life. Once you get to 12 hours there's no practical reason for more battery life for most users but a lot of users like only having to plug their laptop in occasionally and OLED screens guzzle battery life so having a more efficient chipset offsets that and lets you
Re: (Score:2)
Actually Lunar Lake is pretty good on battery life. X Elite isn't really any better.
Re: (Score:2)
The vast vast majority of users want laptops and they don't play games on their computers, at least nothing more complicated than a crossword puzzle.
[citation needed] (sure they are a *vast* majority?)
Maybe the users you speak of would be better served by tablets?
Re: (Score:2)
arm based laptops are still kicking Intel's ass on battery life
So? They aren't kicking AMD's. Intel is well known to be far behind on bringing down TDP for several generations now.
Re: (Score:2)
it's a conspiracy to compete!
What took them so long? (Score:2)
Re:What took them so long? (Score:4, Insightful)
Re:What took them so long? (Score:5, Interesting)
Many of the instructions were clones of the earlier 8080A [wikipedia.org], and that was one of the decisive reasons for the platforms success too. Specifically, DOS and BASIC predated the x86 as Z80/8080A programs. Other than some strange stuff with the CP and A registers having reversed byte orders, it was really easy to take 8080A code and make it into x86 code. (I made an emulator to do this.)
The roots of software compatibility run deep. The first successful microprocessor defined the major instruction set for the next 50 years, with many additions along the way.
Re:What took them so long? (Score:4, Informative)
Correction: The flags and A registers have reversed byte orders.
Re: (Score:2)
Re: What took them so long? (Score:2)
DOS and BASIC as names are both things of the 1960s. 8080/Z80 had them, but they're not the same programs as their 8086 counterparts. Okay, MS Basic for 8080 resembled that for 8086, but so did MS Basic for 6502.
Incidentally, there was a MS DOS for Z80 (sorta) created by the author of MS-DOS. It was a port of MS-DOS, was released much later and was called MSX-DOS.
Re: (Score:2)
If the first IBM PC used a Z8000 or a Motorola 68000 the history of microprocessors would have been decidedly different.
Intel didn't like very much that architecture, and they made the 8086 as a stopgap CPU, while developing the iAPX 432. They tried a lot of time to propose new CPUs ending with the Itanium, and while at it killing the HP and DEC processor line.
Re: (Score:2)
the basic instruction set that's not used any more? How old is 64 bit x86?
Also, ARM is 45 years old.
Re: (Score:2)
Re: What took them so long? (Score:2)
Not used anymore? You may still use it if you want. Modern x86 CPUs still support both real mode and 16-bit addressing, and using both means using the original 8086 instruction set.
Re:What took them so long? (Score:4, Insightful)
It took so long for Intel to weaken to the point where it needed to work with others (especially AMD) rather than just forcing everyone else to follow their lead. The AMD64 debacle did not cut deeply enough.
Re: (Score:2)
It took so long for Intel to weaken to the point where it needed to work with others (especially AMD) rather than just forcing everyone else to follow their lead.
Except they didn't have to do that, and you alluded to it:
The AMD64 debacle did not cut deeply enough.
You mean the debacle in which they had to follow AMD's lead?
The only time AMD tried to differ and failed was FP on the K6, and we had to have patches for it, and then they fixed it in the K6/2.
Re: (Score:2)
It was called Itanium. You people didn't buy it, so stop crying.
The Itanium was actually Intel's third failed attempt to create a clean sheet CPU. It was preceded by the iAPX 432 in the 80s and the i860 in the 90s. The people didn't buy those either, mainly because all three pretty much sucked.
Re:There was a clean sheet CPU. (Score:4, Interesting)
The Itanium was actually Intel's third failed attempt to create a clean sheet CPU. It was preceded by the iAPX 432 in the 80s and the i860 in the 90s. The people didn't buy those either, mainly because all three pretty much sucked.
i860 was a weird and hard to use chip that was never a viable x86 replacement. i960 was up to task, though. It was torpedoed by internal intel politics. It was pretty successful in the embedded space for a while despite intel's bungling.
Seriously, let it die already (Score:2)
And it has far outlived its usefulness. How much farther ahead could we have been in computer technology if we'd standardized on a better architecture?
Another requested bullet point (Score:4, Insightful)
Would have been nice to have at least some lip service towards secure chip design that doesn't require disabling performance features you paid for to not leak data.
Re:Another requested bullet point (Score:5, Informative)
Re: (Score:2)
Off topic, but regarding your profile quote. They seem to have finally fixed the issue in the last few months
It must be the last two weeks or so, because less than that long ago I pasted some text from a webpage that got mangled.
Re: (Score:3)
Good job, Slashdot. It only took about 3 decades to figure out how encodings work. Kudos.
Re: (Score:2)
As someone earlier pointed out, this is to weaken ARM. While the marketing terms imply this alliance will be 'customer' focused, it will be nothing of the sort.
ARM must be doing good (Score:2)
For Intel and AMD to join together, they must really be scared of ARM or even RISC V (even though RISC V sucks, it has potential).
Re: (Score:2)
There are at least half-a-dozen RISC-V cores out there at fabless semiconductor companies that (on paper, at least) should be able to compete pretty well with x86, if only they found their way into actual silicon...
"if only"... because that latter part is the harder part.
RISC-V Risk (Score:2)
RISC-V is on a trajectory to eat everybody's lunch.
If not for performance than for security, then cost.
Since the x86 abi sits on top of a RISC microcode these days I wonder what would happen if somebody tried it on the RISC-V architecture.
$99 laptops probably.
Re: (Score:2)
Oh! now there is an idea ! somebody should make a RISC extension that adds hardware acceleration for processing x86 emulation
Re: god fucking damn it (Score:2)
Why do you need it? You want to run Windows 98 on a post-Skylake Intel CPU?
Re: (Score:1)
2) Use dosemu on 32-bit Linux
3) Use 16-bit Windows binaries on Linux with WINE (it actually works on 64-bit Linux, contrary to Windows which relies on NTVDM for this use case).