Could a Chinese company viably produce an x86 processor with a fuckton of cores and some optimizations? To my knowledge...

>32 bit architecture in the current year
no, user. it would be terrible and extremely limited.

This is where you're wrong, kiddo.
Planted right on a hardware level, we won't be doing some shitty binary translation.

You're confusing what Faildozers weak points were.

listen sweety, stop pretending and go back to your games.

>8 core 16 thread 28 nm IA32 processor with SSE
>4 GHz
>TDP ~300 W
>~25 GFLOPS
I think this is a bad idea but I am intrigued.

Attached: sakuya wants to know more.png (337x481, 204K)

>brute power
More cores doesn’t translate into actually using them. Applications that can’t multi thread to shit and back will be awful on this kind of chip, provided they even run at all without extensions developers have come to expect as a bare minimum.

I just want someone to make a decent 486 asic motherboard with 16bit isa slots and modern memory management.

>isa
Why

for the same reason he wants a 486, probably retro gaming where you'd want a real sound blaster or whatever else

You would be horrified if you realised how much modern manufacturing runs on this vintage equipment.

Nah, I remember all those CNC machines running on Pentium in my school. It just werks.

I’m more perplexed as to why people are so horrified about it, honestly. How broken are we as a society that the idea of a piece of technology still doing what it was made to do after decades of service “horrifies” us? Is a quad-core processor with gigabytes of memory somehow better at the same simple math?

The problem is when (You) are tasked with bringing a special-purpose machine into the 21st century and you've waited so long that upgrades are a nightmare. Computers in any sort of production shouldn't be there more than 10 years because reliability and upgradability become massive hurdles.

On some systems upgrading is impossible.
The isa bus allows direct access to the hardware which the pci bus does not allow.
Some dos applications won't work on processors over a certain speed.

The manufacturers don't make replacement parts, so you become a specialist in board level repairs and making all sorts of weird contraptions and work arounds for obselete devices.

If you want to specialize in obsolete tech the DoD needs some PDP specialists to keep the ballistic missile program running for a couple more decades.

>The manufacturers don't make replacement parts
Trust me, if the government wants a part, they can get it. They have unlimited resources.

The general rule in manufacturing is that you need to sell at least 10,000 of anything to be profitable. Now if you can find anyone willing to purchase a minimum 10,000 486 motherboards in 2018, then we'll talk. A small handful of retro game neckbeards is not enough to be profitable.

If you can't use any of the modern features it would be better to go with RISC-V

too bad chinks hate doing actual r&d

Attached: 1509188624739.jpg (221x250, 5K)

>The isa bus allows direct access to the hardware which the pci bus does not allow.

There's a reason that's not done anymore and it's because direct hardware access of that sort is undesirable on modern operating systems. Maybe it was ok on an Apple II, on a Windows 10 box not so much.

>too bad chinks hate doing actual r&d
They can't do it. It's just not in them.

Perhaps you heard that Athana still produces 8" floppies for US government legacy systems.

>Some dos applications won't work on processors over a certain speed.
DOSBox?

RISC fag here, I'll offer my two pennies. I think that RISC machines are superior in every single way except for two big areas, which are software availability and emulation abilities. The software issue can be fixed right up with some FOSS and some neckbeards to compile and maintain binaries. The other issue is trickier, since can emulate x86 just fine, but a lack of hardware acceleration makes it slow as all fuck. The ARM and POWER hardware you can buy today simply lacks the instructions in silicon to do it. This is one place where CISC is better.

So this means you can either take the performance hits of software only emulation, buy some real CISC hardware, or buy expansion cards for your machines that have an Intel or AMD processor on them to be utilized by VMs.

My personal solution is to just use RISC hardware for everything that doesn't specifically require x86, and then remotely connect to a small AMD desktop on my network that I use for compiling and testing x86 software and running Windows VMs. Everything else can be done on a PowerBook G4 and PowerMac G5, both of which still handle web browsing and office stuff well with Debian.

Sorry m8, it's just not feasible to try to make x86 hardware. Two big companies have a monopoly on it and that won't change any time soon.

How good was Excavator if given an IPC boost via shrink?

Intel sells unlisted cpus to the government ( much higher clocks )

I thought one of the reasons they retired the Space Shuttle was because it had obsolete parts that were not available anymore.

You go off on a rant about how you use old macs that struggle with javascript to explain why people can't make x86 hardware.

BTW the latest snapdragon ARM SoC support hardware emulation of x86.

Attached: fedorafag.jpg (400x533, 41K)

Even if that were true, they could have easily replaced them with modernized substitutes. I know that the original Shuttle computers used core memory and were later replaced by all-digital units.

>tripfag
>fedora image
Hypocrisy at its finest.

>BTW the latest snapdragon ARM SoC support hardware emulation of x86.
Like I said, it's done in software, you retarded tripfag. It might include some extra instructions to speed things up a bit, but they're probably basic, generic instructions from decades ago. You do not get meaningful hardware acceleration on ARM for x86, if any at all. The bulk of the emulation is still done through a proprietary compatibility layer.

And you're filtered.

Didn't some Czech hobbyist start a business making his own nixie tubes because no new ones had been made in 40 years?

sse4 wtf?
is that sse4_1 and sse4_2 ?
that's all my processor has... (other than sse2)

The 70s vintage units the orbiters originally had were replaced with digital units in the early 90s which were still quite limited, but extremely durable. Astronauts often still brought along laptops during missions, but they weren't really designed to military/aerospace specs and weren't that dependable. For one thing, a consumer-grade laptop is susceptible to cosmic radiation in space outside the protective blanket of the Earth's atmosphere.

>You do not get meaningful hardware acceleration on ARM for x86, if any at all.

Snapdragon has had hardware accelerated x86 in both the 835 and 845, guess if you could afford something other than a 15 year old mac that stays perpetually connected to some AMD shit you would follow hardware news that didn't involve Intel BTFO in the headlines.

>and making all sorts of weird contraptions and work arounds for obselete devices.

Attached: gotek_amiga_600_2.jpg (1024x661, 67K)

From what I understand, the military considers 8" floppies to be more secure than modern storage devices.

>steal 8" floppy with nuclear launch codes on it
>now where can I stick this in my MacBook...

x86 is an antiquated 1970s CISC piece of junk anyway.

There are cases where they have to train new people to manage ancient bank systems designed around COBOL because the original programmer(s) retired and Millenial-aged tech guys don't know how to use it.

Additionally, old tech is well understood. Old floppy disks can withstand a pretty high magnetic flux without being wiped. Tube-based logic is immune to an EMP, If you don't need speed, but that operation will continue under extreme circumstances, then old tech can keep on chugging along.

I remember visiting an old Nike site, and part of the fire control system was all mechanical clockwork logic. It just had a simple computation to do (when to trigger the warhead on the missile) and they wanted to be absolutely sure that it wouldn't be disrupted by an EMP (since Nike Hercules were nuclear armed anti-aircraft missiles, they had to be sure to harden all of the control systems)

1: why would you want outdated and poor qc tech

2: why would you want our strategic opponent to have an upper hand in an extremely critical tech field?

To give another aerospace example, Russia has been very reluctant to retire the Soyuz/R-7 combination because although decades old, it's well-understood and all possible hardware bugs were worked out a long time ago. They did a few years ago replace the 1970s vintage Soyuz-U rocket with a modernized variant (Soyuz-FG) that has all-digital computers, in this case because it could handle aerodynamic instability generated by flared payload shrouds that wasn't possible with the old analog flight control system.

The Soyuz FG has had a number of flight failures since its introduction in 2010, all of them involving the upper stages, which does go to show how upgrading hardware introduces a whole series of new problems that have to be solved.

They should train them on how to rewrite the programs

But they were using them since the 70s-80s when floppies were far more common.

8" disks were never really used in consumer-grade computers and there were far fewer computers in general around back then.

Military is probably using something like a pdp11 or similar minicomputer

Uh...yeah. Even if this were 1985, the average person didn't have the ability to read an 8" PDP-11 floppy.

*blocks ur path*

Attached: altair-floppy1-1280x720.jpg (1280x720, 106K)

Now those things were costly as fuck (in 2018 dollars, approaching $10,000) and mostly used by businesses. In fact the 5.25" floppy was originally developed by Shugart because 8" disks were too expensive and unwieldy for the nascent microcomputer market.

It's been done, Lemote Loongson processors are MIPS CPUs with an x86 translation layer

I wonder why they didn't just replace them with 3.5" 1.44MB disks? Cheaper and much more readily available.

>and they only exist because they have some ancient patents from a US acquisition long ago, which Intel needed to cross-license.
FTFY

>Cheaper and much more readily available.
That was his exact point as to why they don't do that.

and

Attached: download.jpg (225x225, 6K)

If they're old PDP machines, they can't use 3.5" drives anyway because they're not electrically compatible with an 8" controller and besides, knowing the reliability of the average 1.44MB disk, I wouldn't trust my valuable data to them either.

>I think that RISC machines are superior in every single way except for two big areas
Yet you've provided zero evidence for this.

The problem with emulation is there is a delay between the software and hardware access that can cause timing and communication issues.

There is USB to isa solutions but they tend to be hot and miss.
Some of the hardware works others don't.

I should have said, but 8" disks are significantly more reliable because of the larger physical disk surface meaning that data isn't jammed in as tightly.

Isn't this basically those intel coprocessors? Would be interesting to see a chinese knockoff.
en.wikipedia.org/wiki/Xeon_Phi

Attached: Intel-Xeon-Phi-Family-1024x603.png (1024x603, 330K)

He also forgot that new processors aren't CISC or RISC anymore. They have decoders to translate x86 in micro ops native to every arch of the processors.
This is happening from Pentium III era if I'm correct.

Xeon Phi uses Atom cores

That's still an x86 architecture though, unless you specifically want 386.

That's x86_64, a whole different patent situation

Be that as it may, desktop motherboards still have all the floppy/parallel/serial hardware on them, although it's not connected to anything. Removing that stuff is more expensive than just letting it be.

RISC systems have never had a problem with FOSS, the problem is just that nobody in the x86 space wants it. They want the commercial applications they like using, developed by businesses that care little for CPU politics running on hardware built by OEMs that don't find the idea of retooling and re-negotiating contracts just so someone like you or I can have a different sticker on their case.
Emulation isn't even really that difficult either, NT Alpha systems with FX!32 were reasonably faster running emulated 386 code than actual x86 systems, for example. The likes of SoftWindows and Virtual PC were good enough to have an audience in the Mac and Unix spaces for over a decade.
But the whole exercise is just pointless, because all you're doing is putting in this effort with the ultimate goal of... what again? Putting the monopoly hat on another vendor/architecture with its own several years of "inelegant" baggage? Saving a few tens of dollars? Appealing to nostalgia for a time when high-end RISCs were actually worth having?
Why go through the effort when you can just contract a Chinese company to shit out a couple more runs of 8'' disks and not have to touch what isn't broken?

>Why go through the effort when you can just contract a Chinese company to shit out a couple more runs of 8'' disks

>US military
>relying on a Chinese company to supply items needed for the nuclear missile arsenal
I'm pretty sure that's not even allowed by law.

>Could a Chinese company viably produce an x86 processor with a fuckton of cores and some optimizations?
From scratch? No. They can't even made photoresist.

>NT Alpha systems with FX!32 were reasonably faster running emulated 386 code than actual x86 systems, for example
That's because of frequent design and architectural changes on x86 CPUs. Code optimization tricks on one CPU type might not work on another.

It's really irrelevant to the point where the supplies are coming from, the point is that obtaining 8'' disks would be pretty trivial and far less invasive than trying to rig a Series/1 to support a slightly less obsolete storage medium.

consumer qubits when

>the point is that obtaining 8'' disks would be pretty trivial and far less invasive than trying to rig a Series/1 to support a slightly less obsolete storage medium

IDK, a lot of retro computer collectors are increasingly switching to Goteks and things like that.

Keep dreaming.

The government =/= neckbeards. If some random neckbeard wanted new 8" floppies for his retro computer collection, they'd tell him are you serious?

The government has more money to spend than neckbeards.

I'm confused. What does this have to do with FX!32 being decent at 386/486 emulation?
>Goteks
Miss me with that ugly shit. And my $20 a pack budget doesn't have shit on the DoD.

FX!32 runs x86 code more efficiently than the real thing because you don't have to deal with the numerous hardware differences in the various CPU iterations that affect performance.

I'm still not really following how the differences between these iterations impact the performance of software emulating a specific iteration. FX!32 was also not more efficient, it was /faster/ on the brute strength of the Alpha it ran on. The point I'm making is that emulation isn't really an unsolved mystery inhibiting widespread adoption of non-x86 architectures.

When does X86_64 patents expire? Someone should make open X86 once it expires, replace SSE, AVX and everything with an open SIMD instruction set.

>replace SSE, AVX and everything with an open SIMD instruction set.
enjoy none of the software you'd use x86 to run in the first place ever using it

Except both x86 and x86_64 are a fucking mess. We'd be better off focusing on improving ARM to match x86_64 in performance, and you can already do that without traversing a legal minefield.

Process shrink doesn't give any IPC boost on its own.

With PAE, you get a 36 bit address bus, which can handle up to 64GiB of memory. It's not too bad considering most CPUs are 48 bit nowadays. It could become a problem soon, but for now it should be good enough.