/tech/ - Technology★

>>924428 (OP)

>Basically, I want to build a computer using a POWER9 or similar non-botnet cpu then use an x86 add-on card for all software that would not run natively on the new architecture.

Why would you need real x86 hardware this powerful for legacy applications? Spend the money on a higher core count POWER9 and use QEMU.

Pretty sure that the CPU <-PCI-E-> card link would be the bottleneck. Running it via dynarec on your powerful CPU is a better solution. Now, some on-die memory just for this purpose (like a user controllable L3: L4) would make sense.

>x64

t. Winbabby

>>924574

To follow up on this, it's better to pool money to pay some devs to convert some software to use Altivec (maybe bountysource?). RPCS3 (maybe some other emulators), x264/x265/AV1 and gzdoom are probably the only software I use that might need it.

>>924472

You're still using x86 processors, which are botnet by design.

Why use x86 if you're planning to move away from x86 in general?

Just use a VM. Chances are if you can't compile your software for POWER it's probably shit.

>>924598

What about nonfree software? I'm not going to build my life around your licensing fetish.

>>924588

No they aren't, stop spreading bullshit. Even POWER can be backdoored.

Don't want backdoor shit don't use computers at all. It's that simple.

>>924600

x86 is botnet you fucking (((shill))). Intel Management engine. Even if it wasn't, x86 is such a shit architecture that it would still warrent a switch to POWER.

>>924624

>x86 is botnet you fucking (((shill))). Intel Management engine.

Huh. I didn't know my VIA CPUs had an Intel ME.

>>924625

Aren't VIA chips underpowered? Pre-PSP AMD processors would make a better choice.

>>924599

That's the life you actually chose! You have chosen to give control of your computing to the owner of the software. This is not a secret, you're expected to understand this for every single piece of non-free software that you choose.

>>924635

>You have chosen to give control of your computing to the owner of the software.

I put all nonfree software in docker, including vidya games. Your move.

>>924624

>x86

>ME

Pick one you faggot. What's next, you're going to tell me the Pentium MMX or K6 is backdoored?

>>924644

A virtual machine or some other kind of sandbox isn't going to give you any meaningful control over the software, the only control that gives you is to limit its reach to within the sandbox. Your ability to study its code and modify any part of it is practically impossible, that control belongs to the owner of the software and you have chosen this way for your own life. You can change this if you want but something tells me that you will refuse.

>>924659

I don't care about my ability to study or modify the nonfree software that I use. I don't trust it and I don't have to because I'm not using it for anything critical. I have other things to waste my time on.

>>924661

Then there's no problem for you. You don't care about POWER because you actually like x86.

>>924667

I'm interested in open architectures and there's nothing that I particularly like about x86, other than some of its implementations being fast. Why are you posting lies on the internet? Stop that.

>>924681

You have admitted that you don't use your computer for anything critical and that you refuse to change your x86 based non-free software. This implies that you like the x86 and that you're not interested in other computers except for minor academic reason. If you had a more serious attitude about other computers, you would not be so tightly wedded to your x86 computer as implied by your x86 only software.

>>924654

It is. Look up SMM.

>>924682

I have admitted that I use nonfree software for nothing critical. But even if I wasn't using computers for anything critical this wouldn't invalidate my point.

>>924644

What is your argument? You are still cucked by proprietary. You don't have the freedom to run it natively on POWER right now, how can you be so retarded.

>>924574

>Pretty sure that the CPU <-PCI-E-> card link would be the bottleneck.

Which is why some of these cards are using 10Gbps links instead. I don't really care about the bottleneck as long as I have support. Really all you need is the ability to output screen grabs every 60ms or so that can be embedded into the native GUI.

>>924588

>>924624

Here come the licensecucks right on time to stop any progress. No one cares that the software is proprietary/closed source. The idea is to give yourself a safe place with a condom wrapped around it to run that software inside of something safe/open source/free/auditable.

>Wincuck

I am considering I have machinery that requires early Windows/DOS/x86 software that doesn't run well in any emulator to run. This is a valid use case and one many are stuck in.

>>924685

Good point. Does 80486 not have SMM?

>>924822

>It would have its own PCI lane so connectivity isn't an issue.

this is where the problem is. the cpu has direct links to the ram, and everything else on the system. pci is fast, but no that fast. there is a reason why you cannot simply upgrade your computer by plugging in more cpu's on cards that are on the pci bus. graphics cards work because the i/o makes sense, they take an input, do a shitton of work, and output, a cpu on the otherhand needs to take in massive amounts of i/o for relatively trivial tasks.

tldr; this would work, but it would probably end up being slower than emulating, and even if it wasn't, you would take a massive performance hit. imagine putting a newer generation i7 on one of these pci cards, it would immediately be bottle necked by the pci bus.

>>924765

If you want to avoid SMM, you have to use a regular 386 (not the special laptop version that's got power management shit). I guess they made versions at up to 40 MHz, and you might even be able to play Doom on that (the 3-episode release, up to and including version 1.2, i.e. before they changed the engine for Doom II). Much fast, so games!

>>924958

The idea is to self contain the CPU on the add-on card though. If the CPU has RAM available to it on the card the bottleneck can be worked around. All you really need is an update of the output from time to time. This could simply be a screen capture every few ms that can be sent to the main GPU/CPU. I/O can be on the slow side as long as it runs fast enough once it's on the card. Plus, we do not need i7 level of performance. As long as it's good enough to run things like photoshop, old DOS software, and other proprietary software at a decent speed it'd be fine.

On your other point (emulation) the idea here is to use thing until emulation becomes good enough. With such an add-on card it doesn't matter how well the new CPU emulates x86 software. You have plenty of time to work on that and make it better with the peace of mind of having the real thing tucked away inside of the machine if it were ever needed. The ideal situation would be one where you'd eventually not need to the add-on card at all. But we all know that'll never be a reality because certain edge cases simply won't run in emulation due to reasons beyond our control. For those cases you use the card, for everything else you use the host machine.

>>925138

if the add-in card is running it's own ram and your envisioning this as something you'll vnc into you might as well just buy a seperate x86 system, stick 2 motherboards in the case, and pretend they are the same computer, vnc'ing into the x86 machine when necessary

>>924765

Some versions don't, but the high end ones do. You would have to get an early 486.

>>924630

The latest Isaiah architecture chips were pretty good. Athlon 64 level IPC (but with lowish clocks), up to quad-core. Quite functional for a basic desktop.

>>925183

SL-enhanced models have SMM, the others apparently don't have it: http://www.pcguide.com/ref/cpu/fam/g4I486DX-c.html

Ergo, I guess you're safe with a DX4.

>>924428 (OP)

Wouldn't it be possible to get one of those old Mac PC cards to run in a modern non-x86 mobo?

>>925391

Aren't those proprietary and/or super obscure? Don't think you can get anything non-x86 with ISA slots nowadays.

>>925672

I want them to be able to run in tandem or separately. I realize this card isn't exactly what I want and I'd never shill for it. It seems like a dumb idea but it does have the hardware in the form factor that I'm seeking.

It doesn't have to be a PCI port but it does need to interface with the mainboard/CPU somehow. I am not saying you need to limit it to just x86 either, you could have similar cards with a variety of things: MIPS, RISC-V, a M68K, Z80s, FM chips. x86 seems like the killer application though because you gain so much software support along with it.

Your argument boils down to software you can't run natively is shit and you shouldn't use it anyway. That's great but you'll never see mass adoption of a new architecture without support for the old. Emulation simply isn't going to cover everything and when it doesn't a cheap add-on card starts to look like a good option. Normalfags aren't going to pick up another computer or have a server laying around.

If such a thing exists it's going into my next build simply because I need real x86 hardware for a variety of things I do. Some will eventually be workable on POWER9 but until that software is optimized it doesn't make sense not to run it on the real deal. Plus I can think of a lot of tasks I could use it for while it was idle. It basically becomes free processing power at that point and you could utilize for a bunch of stuff outside of this one use case.

x86 emulation has been really good for a long time, you're better off keeping the money such a card would cost.

>>925711

>x86 emulation has been really good for a long time,

How's your emulation of x86 on arm/power/riscv going? Show numbers. In my experience x86 emulation on x86 is unbearably slow.

>>925807

>judging by his post op wants much power

>durr buy some 60$ chink shit and upgrade pc for the next decade ;-)

retard tier post