When are AMD and Intel going to learn from the mistakes of x86

ARM will save us

What's the most powerful ARM processor out there.
Are there any 16-core ARM processors?

File: Bugs_Bunny_King_Arthur.jpg (26KB, 482x381px) Image search: [Google]

26KB, 482x381px

>MIPS
>PowerPC
>ARM
>SPARC
>ARC

>>62310239
No one cares enough to go through the absolute pain of changing architectures when that means 98% of all software will run in compatibility mode anyways for the next 15 years.

File: Itanic.jpg (104KB, 672x371px) Image search: [Google]

104KB, 672x371px

>>62310239
The last time Intel tried, it was an EPIC disaster.

File: Processor_families_in_TOP500_supercomputers.svg.png (330KB, 800x569px) Image search: [Google]

330KB, 800x569px

>>62310271
nope

>>62310343
It's still being made though
https://ark.intel.com/products/122882

>>62310402
Yeah, but they have stopped all development.

Is it possible to have a little PCIE ARM co-processor for some easy hardware ARM dev?

end cisc

who's the artist?

>>62310239
Like MIPS? Actually, any arch is better than is x86

>>62310343
Why exactly did it fail so hard?
I read a little about it, in theory it sounds vastly superior.

>>62311013
How can I find a MIPS computer and install linux on it so I can truly be free

>>62311121
Some chineese university was making mips laptops, but you probably couldn't buy them anyway. Anything except arm and x86 right now probably costs huge money so only businesses can afford them

>>62311265
Is there anything like the raspberry pie that isnt x86 or arm?

Would love to touch some somewhat fresh-ish territory and have some homebrew fun on something.

My friend convinced me to get the Steam Link since its armv7 and has an official SDK but I want to mess around on something more elusive than x86 or arm

>>62311293
None that i know about,'cause i'd like to have one too

>>62311115
Chicken and egg. How good an architecture is matters very little if there's no ecosystem for it. And after AMD64 came out, there was simply no compelling reason for anyone to use Itanium, no matter how superior it might be in theory.

>>62310936
could swear it was this artist:
https://www.pixiv.net/member_illust.php?mode=medium&illust_id=46691557

>>62311115
one of the fundamental ideas of Itanium was to have the compiler do instruction reordering instead of doing that in hardware. On paper this sounds great, it saves a lot of very complex logic. In practice it turns out that its almost impossible to get instruction ordering optimally right at compile time, you basically have to do it at runtime.

There's also the fact that this was they heyday of the MS monopoly and closed-source software, and they put out a chip with really shitty x86 emulation (it was about as fast as a 100 MHz Pentium, for a chip released in 2000). As originally planned in the early 90s this wouldn't have been a problem, if it hadn't been for two things. One, Itanic was a big, complex chip, and it got delayed by years. Two, in the meantime, Intel's x86 department was in a running battle with AMD that was pushing performance of x86 much higher much quicker than anyone, including Intel, thought. The performance state of the art outran Itanium. This same phenomenon is also why x86 torpedoed a lot of other RISC chips. Alpha, MIPS, etc.

>>62311488

>>62311368
so x86 is popular because that got all the focus

meanwhile arm is becoming so common it might as well have a high performance desktop machine

if anything I think ARM is gonna be the new "wave" if anything rather than a totally brand new architecture

risc wins

>>62310330
But C/C++ are portable, in most cases it's just changing compilation target.
You don't even need to do anything for managed languages like Java, Python or C# except a few fixes.

>>62311013
"better" in the sense of "more elegant design", but that's not the only way to measure better.

Example, x86 has a bunch of standards (official and de facto) about how the system ought to be put together and boot. Things live at certain memory addresses and I/O ports and so on. This is why you can download a Linux ISO, put it on one of several different media (USB mass storage, optical disk, network-booting system) and expect that arbitrary x86 computers will be able to boot from it. OEM prebuilt desktops, laptops, self-built machines, pretty new or fairly old.

This is not at all the case in ARM land. Everyone does it differently. This is why phone ROMs need to be built for each device, instead of you having one generic LineageOS image that you can flash to any device with specs of at least such-and-such a level. Things being systems-on-chip also mean that you have a bunch of tiny Nvidias, each writing their own binary-blob drivers for their own custom bits of the GPU and the stuff that enables the CPU to talk to the rest of the system. Which is why your phone is likely running some ancient kernel, they don't bother rebuilding these things against new kernels.

Really, I'll take all of x86s creaks and clanks and rough edges and "Fucking hell, why would they do it THAT way?" moments and more besides in exchange for the world we have today where we can have PC components and whole operating systems as interchangeable parts.

>>62310239
because architecture isn't significant anymore unless you're a compiler writer, and most compiler writers aren't retards that only understand technology enough to make their video games go fast
besides, pretty much any significant architecture is loaded with "patchy" extensions as times change, that includes autism darlings like MIPS, POWER, SPARC and pretty much any other alternative platform that's been around for more than a few months
>>62311614
that's not really how it works, x86 had the focus, but it was also actively pushed towards that performance goal by a market desperate to get rid of the big, shitty overpriced RISC systems that were bogging them down
there's no such push for ARM, the only people who want high-end ARM desktops are autists and "hobbyists" who just want something that doesn't run an x86 chip and sure as fuck aren't going to pay the premium for it anyway

wheres that x86 bit bashing video where he found that one instruction that literally freezes up every modern x86 cpu and disassembler

Lets make Chip8 an enterprise instruction set

>hack the intel microcode
>reprogram an x86 processor to take a whole new set of instructions

is it possible?

>>62311706
>Which is why your phone is likely running some ancient kernel, they don't bother rebuilding these things against new kernels.

if it was about simple rebuild, things would be ok
but linux kernel update are so disastrous, you should be charged for new drivers

>>62311648
yeah tell that to the poor fucks who fell for Windows NT on Alpha in the '90s and barely got dick beyond the paltry shit Microsoft half-assedly ported to run on it

there's a retarded assumption at the heart of every shitty piece of technology

File: 1475096454540.jpg (46KB, 371x348px) Image search: [Google]

46KB, 371x348px

>>62311115
vliw was and will be forever a mistake

>>62312256
but why

>>62311937
>is it possible?
yes, but also useless

>>62310239
>When are AMD and Intel going to learn from the mistakes of x86
when the time is right, edgelord

Graphene or quantum CPUs when?

File: cropped-LI_profile.png (28KB, 512x512px) Image search: [Google]

28KB, 512x512px

Risc V is exactly that, I really hope the industry will adopt it, tho I realize it's quite unlikely.

>>62313608
15 years if we're lucky?

>>62313654
Are the technologies still being researched/developed or are they a dead meme yet? I don't follow the topic.

>>62313677
quantum is a hail mary
people don't know a way to make it work yet, or even understand how what they do have at the moment works
dunno about graphene, but nanotubes might be used as transistors in 5 years or less
this is all kinda off topic though

>>62313799
Not off topic. OP wants a new CPU architecture which may make more sense once we get the new technologies to play with.

>>62310246
been hearing that since 10 years now, arm hasn't saved anything. most arm devices are arm7 which has no acpi, that alone was and still is such a huge disadvantage to the platform that it will take several years. aarch64 market share is so tiny, i might as well buy some alpha processors on ebay and have a better chance running modern software on it than arm.

>>62310318
>MIPS
dying and only embedded on network devices
>PowerPC
dead/irrelevant
>ARM
only embedded devices like smartphones or development boards, few network devices. barely any market share for aarch64
>SPARC
dying/dead
>ARC
>iot
>32bit
>the future of modern computing
l-fucking-mao

>>62310239
>When are AMD and Intel going to make every program written in the past 30 years stop working
That's pretty much what you're asking, right?

>>62314513
It is really fucking weird how much /g/ pushes the 32 bit meme, seeing as back when everything was 32 bit people wanted 64 bit.

File: 1483817675158.png (155KB, 1200x898px) Image search: [Google]

155KB, 1200x898px

They tried to save us but we did not listen

>>62314880
>>62314513
Why the fuck would you want to go back to 32 bit??

>>62314847
new architecture with a x86 co-processor/compat layer

>>62310239
Cute anime girl, can I fug her?

>>62315080
>new architecture with a x86 co-processor/compat layer
because it doesn't fucking work. too many edge cases.

>>62315160
shes pixels, anon

>>62310246 Single threaded performance is shit.

>>62315209
You're pixels

>>62314984
are you literally retarded? can you not read the obvious sarcasm on ARC

what if we just use the intel microcode as an instruction set directly

Risc

>>62315298
I don't know much about ARC or IoT

Risc > Cisc

>>62310246
ARM is designed for embedded devices (at least it has been), and also is a RISC arch. It's not designed for high performances, but for simplicity and low consumption, even if there is a large gamme of CPUs.
I'd like to add that this is a proprietary architecture. Even if this doesn't affect a lot development on it. There is a good chance for the next generation of cpu architecture to be free. This could make cpu actually cheaper (or increase profit).

>>62317524
>>62317561
The thing is ARM is becoming more CISC with time and that CISC CPUs tend to get RISC. I think this will be interesting to see how it'll evolve in the future.

Read senpai
https://danluu.com/risc-definition/

>>62311293
Try to find the Creator board by Imagination, it has chinese MIPS. Of course, it is dumb to distrust western hardware and buy something from actual totalitarian regime that spies in the west like crazy.

File: C5kA5gZXQAAzl_d.jpg large.jpg (57KB, 750x735px) Image search: [Google]

57KB, 750x735px

>>62311706
>Really, I'll take all of x86s creaks and clanks and rough edges and "Fucking hell, why would they do it THAT way?" moments and more besides in exchange for the world we have today where we can have PC components and whole operating systems as interchangeable parts.
Amen, anon.
Nice to see somebody with enough insight to realize this, for a change.

>>62315080
>just add a compatibility layer!
Enjoy your $800-$1200 useless piece of shit with no native software and Core 2 celeron-tier performance. That's pretty much what happened to Itanium and it was fucking abysmal.

>>62313627
It's probably too academic, too autistically-RISC. It has really low number of instructions, so in reality everything will be slower and have bigger binaries choking CPU caches.
I doubt the CPU will be able to overcome that deficit through high clocks and bigger issue/execution width.

The arch will likely be okay for ucontrollers, but not for high performance targets, mobile phones, servers, or anything of note.

>>62313926
>SuperH
of course it's nonsense, but a patrician one.

>>62315298
Yeah, ARC was kinda >WTF
AVR32!