>2017 >he still dual boots If you don't use GPU passthrough

>having 2 monitors

Settle down Neo.

If my CPU had an iGPU I might try it.

But for now I'd need a 2nd dGPU and getting two dGPUs to cooperate i've heard can be a pain in the ass depending how the PCIe lanes are configured.

>>58647080
Is there a way to run the host with graphic output while a VM with GPU passthrough is running?

can win10 still spy on me even in a virtual machine playing my manchild games online?

>>58647109
The host is running off a 2nd less powerful GPU, usually an integrated GPU. So whatever you're doing needs to be able to run on the iGPU, or whatever 2nd dGPU you decide to use.

I don't need Windows often enough to care.

I thought you needed to know some neckbeard wizardry to be able to pull this off. Has it gotten easier to do this?

>>58647144
Much easier, assuming your hardware supports the virtualization passthrough of a PCIe graphics device.

>tfw vfio masterace

>>58647144

you need hardware support otherwise it has been really easy to setup for years now, it has only got easier since all of the gaymers jumped all over it

>using windows at all

>>58647080
but nigger how am i supposed to do this with a single gpu when my mobo/cpu does not have a iGPU either

>>58647080
My whole system freezes when I install nvidia driver to guest machine. Can't figure out what to do.

>>58647080
i always run into the problem of needing the graphics power on both systems
i want to do 3D stuff and gaming. If I do both in the VM, what am I even using Gnu/Linux for?

>not just having a big desk with a windows gaymen pc and a loonix thinkpad for shitposting on the side

>>58647569
what do you use linux for? Also, if you want to use your gpu on your host, you could just grab that script that changes your gpu from vfio to nvidia (or amd) drivers and viceversa.

>>58647080
How is the CPU usage?

>>58647664
I heard it's quite difficult to find graphics cards that behave well when you swap like that without a reboot. Has this changed?

>>58647080
My setup is very similar to this except with only one monitor and a KVM. My question is how do you run xrandr to turn off a monitor and then let the VM take over? If you're using 2 GPUs, doesn't one of them then need to be using a different output?

>>58647898
If his 2nd monitor is plugged into both GPUs when he boots up the windows VM he could have it set to auto detect and switch when a new input is active or run a script to change to the proper input or similar trigger.

>>58647898
From the xrandr usage printout:

xrandr --output <output> --off

Then like >>58647929 said, his monitor probably auto-detects the new input.

>>58647929
That's what I was thinking, but would the monitor really switch over that fast? I haven't gotten a new monitor in almost 6 years so maybe I've just never experienced it, though. What about something like a VGA hub? Would that be capable of allowing one GPU output to display to 2 independent monitors?

>>58647080
>2017
been doing this since 2012

File: 20160828_133647.jpg (2MB, 3264x1836px) Image search: [Google]

2MB, 3264x1836px

>>58647101
Do i win?

https://m.youtube.com/watch?v=GhNBQFGFLfs

>>58647109
Yes

>>58647852
I don't know, I have a 960 and it works fine.

>>58647080

I did it, but the keyboard and mouse latency kills it.

Yes, I did pass my USB devices directly into Winblows and I still felt the latency.

And how's the input hooked up? Don't you need 2 sets of mouse and keyboard? A KVM is also required or it can be done via software?

>>58648015

holy shit

File: 1473980398454.jpg (3MB, 3500x2300px) Image search: [Google]

3MB, 3500x2300px

>>58648120
No. I have two monitors. I hook up both to my iGPU (Intel). I have a third cable from my AMD card to my main gaming monitor. It's literally an extra cable to one of the monitors. I load Win7 VM and I can play virtually 1:1 performance as I would if Windows was on hardware. None of that dual booting shit too which is always a waste of time. I run Manjaro and i3 atm.

>>58648120
>Don't you need 2 sets of mouse and keyboard?

that's one option

>A KVM is also required

another popular option, but not required

>or it can be done via software?

the most popular software method is to use synergy which should give almost no latency (given that networking between the guest and host can easily achieve 10gbps, but synergy isn't an ideal solution)

given that the host can dynamically remap usb ports that are passed through, the least supported option but arguably the best is to write a script to do this and have some communication between the guest and host so you can toggle this from inside the vm and on the host machine

>>58648091

you shouldn't be getting *any* latency when passing through usb devices, how are you passing them through exactly?

>>58648120
There are different approaches,

you can setup you VM without the gpu at first, install synergy, make sure it works and make it run at boot, and then passthough your gpu

You can also passthrough your mouse and keyboard

you could also passyhrough jist the mouse for the initial setup, although it's not very comfortable.

You can also, as you said, ise a KVM switch

Or you can use one of those fancy bluetooth mouses that hook up to more than one computer and just passthriugh the keyboard


And that's what I can think off now, there might be some other way to do so, though.

So, If I only have 1 video card and an iGPU, does the card automatically switch back to Linux when the Windows VM stops?

I was doing passthrough up until a week or so ago, now I just have 2 separate machines.
>>58647109
Yes, I used my integrated graphics for the host and passed an RX 470 to the VM. I even plugged them into the same monitor so I just switched inputs on the monitor to go back to the host. I also forwarded apps over X to the VM so I could run programs from the host in the VM as desired

>>58648204
this guide assumes my CPU has an iGPU, xeons and HEDT chips don't have one though.

>>58648224
No, you have to use your igpu for linux, the gpu needs other drivers to be loaded. It doesn't do that automatically.

You can use a script to attach the gpu drivers so you can use it on your host again, bit you'll have to kill the x-session first, or go to another tty.

>>58648242
It doesn't assume anything and goes over using two third party GPU's.

>>58648224
You need two monitors (or two cables into the same monitor) and your monitor can usually be set up to auto-switch inputs when the VM stops sending signal

>>58648015
is this a parody?

>>58648268
>>58648266
I misread question, sorry

>>58648015
Nothing. 2 monitors are cool. 3 are for gaymers. +3 are for spergs.

>>58647080
I still use ivy bridge.

>>58647107
That's the thing they don't need to to. It's not like xFire or SLI, they do their own thing

>>58648362

sandy bridge supported vt-d, pretty sure a few generations before that supported it as well

it was the k chips that lacked vt-d up until about devils canyon on the higher end cards

>>58648206

Tried passing the whole USB controler and the devices themselves. I did the usual way, like it's explained in all the tutorials.

It's totally noticiable in Osu.

>>58647080
>GPU passthrough
Why not just tun loonix inside a vm on windows host?

>>58648015
Why is your hair so long? Don't you have a father?

>>58648385
There must be something wrong with your setup, or it could also be some placebo effect.

>>58648381
Fuck off dumbass, if your computer doesn't separate the PCI lanes properly it wont work. This isn't an issue with newer hardware usually, but something 1-2 generations old will potentially fuck you.

Unless it's a xeon or HEDT setup.

>>58648385

something must be screwey in your setup or you've found an obscure hardware bug because there really shouldn't be any added latency when passing through usb devices devices, there's an obvious performance overhead but it should all mostly be running as close to bare metal as far as vms get

>>58648465
>>58648446

The game itself shows the input lag.

On a Windows install I get around 10ms, but on passthrough it goes to around 80ms.

The latency may not be noticiable by you, but it is by me.

I have macOS and Parallels running fullscreen on my second monitor. Literally the exact same thing.

>>58648526
1- what game is it?
2- again, don't you think there could be wrong with your setup? There shouldn't be any input lag.
3- have you tried other software besides that game to test the input lag?

File: 1459853707055.png (43KB, 850x570px) Image search: [Google]

43KB, 850x570px

>>58648567 Parallels is miles ahead compared to other software 3d acceleration but it is still much slower then gpu passthrough. And it still do not support dx12.

>>58648697
yes but OSX is for work not for retard gaymurs

>>58648726
linux too, that's the why people use the gpu-passthrough thing

>>58647144
roadmap for Fedora 26 includes making Boxes (gnome VM manager) having hardware passthrough a "point and click" adventure

>>58647080
I tried doing this but couldn't get it to work. So I just did the other way around. Its not bad because I use Linux daily anyways for my laptop and only need the desktop VM for certain assignments that are required to run on the specific VM they use in the computer labs as well.

>>58647144
>editing text files is wizardry
neo/g/ everyone

>>58647080
Don't dual boot (separate PCs) but I've been planning to use passthrough VMs for my windows work so I can have as many dev/test environments on hand as I please but with proper performance instead of the cucked performance of shit like vmware or virtualbox

>>58648526
>The game itself shows the input lag.

how do you get the game to display 'input lag'? I downloaded it to verify but I can't find anywhere to test input lag, the closest I can find is displaying frame times (which is obviously not input lag)

>The latency may not be noticiable by you, but it is by me.

I'm not saying that you're not getting input lag, I'm saying there shouldn't be *any* input lag, it should behave exactly like a native setup with an extremely minor penalty on total performance (talking about 95-99% compared to 100% native here)

>>58647080
>Windows on the main monitor
Spotted the manchild.

>>58647080
The VM is always going to run worse and waste resources.

>>58647080
Here's my adventure with gpu passthrough
>mess with lunix for some time
>hey i have all the right hardware needed for gpu passthrough
>x99 so no iGPU
>get a second card for lunix
>plan to pass my gaymen gpu to windows
>ok lets do this
>install second card
>turn on PC
>no display
>spend 5 hours trying to boot into linux
>bios works but once it loads the OS it just goes to black screen
>test the gpu in other system, everything's fine
>reinstall windows

>>58649102
>how do you get the game to display 'input lag'?
Options, scroll down to section "Mouse".

>>58649156
>too dumb to install linux

neo/g/

>>58649175
You have to press "raw input" also

File: 20170125_101330.jpg (925KB, 2592x1944px) Image search: [Google]

925KB, 2592x1944px

>>58648444
So i can try to look cute in angle shots :/
>>58647080
Is the advantage that u can run linux applications in the background, because seems like a dual boot would make much more since, with ssd boot times.

>>58649178
>reading comprehension
autism

File: 1476735176023.jpg (54KB, 680x598px) Image search: [Google]

54KB, 680x598px

>>58649156
>tfw to smart to determine which GPU to use

>>58649203
You've got manjaw, hon. No amount of hair will cover that.

>>58647080
>tfw bought mobo before hearing about PCI/PCI-e passthru
>my mobo doesn't support it

just pass the Colorx senpai.

>>58649178
It has to do with PCIe mapping, some motherboards just make it a real pain in the ass to get it to work properly. Which is why using an iGPU is the recommended way, but with x99 that isn't possible.

>>58649223
Both GPUs didn't display anything
I removed the second and returned to original configuration but shit just broke somehow

File: [email protected] (171KB, 1093x675px) Image search: [Google]

171KB, 1093x675px

>>58649256
I'm sorry you paid for a computer you don't know how to use, then.

Maybe you should stick to Dells?

>>58649303
>it's user's fault shitnux cannot detect a newly installed GPU

nice meme

File: muh privilege.jpg (260KB, 688x1434px) Image search: [Google]

260KB, 688x1434px

>>58649338
You'll get there some day, champ. ;^)

>>58647080
Show me how to do that with Optimus and it's a deal. Forget that, just show me how to get Optimus working period.

>>58649338
>It's the OS's fault the user is too stupid to add new hardware

nice meme

>>58649415
I already did, using the superior OS as we speak :^)

>>58649430
Oh please tell me the right way, I'm really curious what I did wrong

>>58647080
How do I do this.

>>58649430
>2002+15
>Plug and Play is not a thing on linuks
Keep your terminal cancer.

>>58649203
L O N D O N
O
N
D
O
N

>>58647080
That's great, but last time I tried I couldn't even get the VM to launch.

>>58649465
tell me what you did then.

What's the input latency like? Does it vary mobo to mobo or should any z170/x99 chipset mobo be fine?
How does Linux handle the audio on the VM? For example, if I had the graphics card outputting to a receiver that's hooked up to my TV, would Pulse fuck with the audio signal somehow?

>>58648526
RIP Street Fighter on passthrough
but the promised linux version will come right, guys?

>>58649475
>the current year
>plug and play exists
>still fucking up when installing

FTFY

>>58649548
For most setups including mine you'll use Synergy. I have absolutely no noticeable latency but I imagine depending on how your LAN is set up you might get variable latency.

>>58649550
why? is that the game he was testing? why would you use a mouse to play a figthing game?

>>58649418
http://bumblebee-project.org/

>>58649547
I tried adding every boot option in GRUB like nomodeset and bunch of other variations. I tried GRUB rescue or what's it called.

This fixes the problem for 99% of people but not for me.

>>58649573
>claims plug and play exists
>still calls it installing
pick one faggot!

>>58649607
no you dingus, tell me exactly what you dod, you canct be this fucking vague and expect someone to help you, what's the hardware you're running? waht distro? what' gpu are you trying to use for linux? do you get sound or something? feel free to add info that you think it's relevant and I didn't asked for, thqt is, if you really want the help and you're not just a boring teenager trying to be an epic troll

>>58647114
pls respond

>>58649548
If you run through a SPICE head or Synergy, you'll have noticeable input lag. SPICE does software rendered cursors too so its response time is awful.

The only way to make it playable is to passthrough a USB device directly in my experience.

>>58649654
>putting new hardware in the socket isn't installong said hardware

are you okay? did you hit your head?

>>58647080
I have no idea what's going on in this webm

>>58649683
If you call "plugging in" shit installing, then I regularly install my electric razor before I shave, then uninstall it. Why the fuck do you think it's called plug and play, not install and play?

Can you install it without uefi bios?
I wasn't able to dump it from my GPU as the guide wanted, and eventually gave up.

>>58649607
I'm also pretty sure it has to do something with Secure Boot and UEFI in general.
Because I had problems installing every single distro. Installation would show black screen as soon as I chose "install" in the menu.

I forgot how I even got it to install in the end, it was pure cancer.

>>58649660
I was running debian on x99 witj 5820K and with GTX 980. I added GT630 for linux.
>put gt630 in primary gpu slot
>turn on pc
>i have display right up until it starts loading the OS

>>58648015
I bet you don't even need a helmet to drive a motorcycle kek

>>58649600
Its shit with black screen boots and you know it.

>>58649709
If you put new component into anything, it's called installing.

>>58649722
>debian
found your problem, can you go to another tty? can you use the other gpu for loonix to see what driver is loaded to which card?

>>58649722
do you boot directly into x? if so your x config might be the problem

>>58647080
I can't because I only have one GPU and fell for the AMD CPU meme.

Somebody send me their old nVidia GPU pls

>>58649745
>I like to feel important, so I call populating a socket installing.
Have it your way, Mr. important tech wizard.

>>58648015
So many monitors but not one good one

>>58647138

And yet you felt you needed to add your filth to this thread.
I am assuming the hunt for friends is still continuing?

>>58647107
This is why SLI bridge exists

>>58649887
what's feeling important having to do with anything? Is that really the best you can do to try to invalidate somebody else's point? that's sad.

File: anakata.jpg (763KB, 1024x768px) Image search: [Google]

763KB, 1024x768px

>>58648015

>>58650069
typical arch user

>>58650016
What's sad is you trying to derail the subject, which was that your shitty OS doesn't have plug and play, which is the operating system's ability to automatically take a hardware into use, you just "installed", without any further user interaction, much less any terminal wizardry.

File: trannyboot.png (5KB, 283x283px) Image search: [Google]

5KB, 283x283px

GPU passthrough Windows VMs really are a giant leap towards true freedom machines.
You have the host OS that respects your freedom, but you can also run anything potentially malicious (from games to 'industry standard' production software suites) in an isolated and controlled environment.

One more important step would be to make all this work with deblobbed Coreboot / Libreboot.
Has anyone here succeeded doing that?

I've been rocking dual monitors for like 5 years now. I can never go back.

At work I started with 2 monitors, and then talked my way into getting a 3rd, and now I have 4. Its very nice, and useful.

I do mostly remote sessions to end users, and I have a constant remote session going to our AD server, and exchange, etc.

I have one monitor dedicated to serve access
One main monitor for everyday use of ticketing system, and the rest
One monitor for remote client support
and the 4th for email.

>>58650098
But anon, I thought it was clear by my meme arrows that it does indeed have plug and play, and that the other anon still managed to fucked up at adding new hardware for some reason.

>>58648353
>3 are for gamers
There's so many retarded kids on this board. Some people actually work with computers for a living and need as much display real estate as possible, crazy concept I know.

>>58650139

Why is their not a single person like the tor project team, Assange, Applebaum ever talk about this ?

i wouldn't have a single reason to have winblow machine if i knew about it.

>>58650227
>I-im not a gaymer, I totally need the monotors for work mom!

>>58649682
I never had any issues with input lag while running synergy

Maybe ditch the wireless setup or check your router?

I have a Dell Optiplex 7010 with a i5 3470 & a GTX 970.

Would it work properly? Have no idea.

>>58650151
You have no idea what this thread is about, do you?

File: 1440946501761.gif (916KB, 494x303px) Image search: [Google]

916KB, 494x303px

>>58650277
look up your CPU. Does it support vt-d ? then what are you waiting for f.am

>>58650259
You ever seen NASDAQ traders? Something tells me you're under 18 with that retarded response

>>58648799
There's nothing but fucking children on this board now

>>58648463
>if your computer doesn't separate the PCI lanes properly it wont work
Are you fucking retarded?

>>58650132
Wait, what didn't even work for you?

>>58650277
/g/ really needs a list of all the hardware that supposedly support passthrough and are also confirmed to really work without issues.

>>58650350
see >>58650301

>>58650317
You don't need the at your home kiddo, that was the point, I may have been unclear about it. Sorry if you made your mom waste money on your 3rd monitor.

>>58650350
PCIe passthrough is an IOMMU thing (chipset/cpu stuff), which means that all PCIe devices Just Work[tm].

File: Screenshot_20170125-130603.png (267KB, 1440x2560px) Image search: [Google]

267KB, 1440x2560px

>>58650301
Yup, looks like it does support it.

>>58650350
This.

>>58650160
shit, meant to
>>58650132

File: 1440949817742.gif (264KB, 500x275px) Image search: [Google]

264KB, 500x275px

>>58650392
I knew already

>>58650350
why? can't you just use your favourite search engine?

>assassin's creed
stopped watching there

Help me pls passthroughfags. Is there a good way to get sound from both the vm and the host os at the same time? If i emulate a sound card it takes away a fuckton of cpu performance so right now i've set up a keybind to switch the real soundcard back and forth between the vm and the host os.

Also, i've turned off mouse acceleration in linux and share my mouse with synergy, but if i try to modify the mouse sensitivity (which is very low by default) synergy just stops working. Right now i also just toggle the real mouse between the host and the vm.

>>58650259
>>58650374
You realize most IT professionals have to be able to work remotely right, dipshit? Mods underage ban.

>>58650411
How does one do it? I am willing to switch to Linux.

I don't even remember the last time I fired up my wi**dows VM

>>58650444
Enable IOMMU
Install Linux patch for Qemu or KVM or whatever the cool kids use today
Configure VM to use PCIe passthrough
Install Windows as VM guest
???
Profit

>>58650444
https://bufferoverflow.io/gpu-passthrough/

just werks (arch)

>>58648015
what's your electricity bill my man?

>>58650439
Your insisting too much on the underage thing boy, it's definetily not suspicious at all.

IT professionals can work fine with just one or two monitors, 3 is for gamers. You'll understand that the day you get a job.

>>58650469
you don't always need the patch though

>>58650439
dvtm

>>58650469
tried that, doesn't work

those instructions are way too simplified

>>58647080
I was retarded and bought a motherboard that doesn't support GPU passthrough. Waiting for Ryzen before upgrading to something that supports it.

>>58650489
Never had a real job I see. Two monitors is a minimum while working in IT.

>>58650503
True

>>58650511
They aren't. Seriously, just google an HowTo. It's easy as fuck.

But in order to help you out, do this:

dmesg | grep DMAR

And post output. If empty, IOMMU is not enabled either in kernel or in BIOS.

>>58650511
https://vfio.blogspot.hu/2015/05/vfio-gpu-how-to-series-part-1-hardware.html

This is the site i have used, might be a bit outdated now though.

>>58650520
Again, once you get a job you will see how the stuff works, I can understand how people your age can be stubborn about things they've never even experienced before, just wait and you'll see.

>>58650530
>>58650541
thanks, folks

i'll recompile my kernel and try it again

>>58650545
You're just being ironic, right?

>>58650489
>IT professionals can work fine with just one
Define "fine"

I'm not the guy you're sperging out on, but I definitively work better with two monitors than with one. One monitor dedicated for code, and the other for email client + slack client + browser with API documentation.

Ever since I started using two monitors, I stopped using workspaces in Linux. They were a poor substitute for more screens.

>>58650555
>i'll recompile my kernel and try it again
No need, just change bootflags and run grub-mkconfig

>>58650557
About what?

>>58650160
Meme arrows don't make a wrong statement true.

>>58650560
You're right, two monitors are better than one, you stil can get the job done in one monitor, though is not as comfy, my point was, 3 monitors is for gamers.

>>58648015
Well you certainly don't lose your virginity, so in a way you win.

>>58650593
I would see the benefit of three monitors, especially if I tilted the middle one vertically and/or one of the monitors were small. However, three horizontal 27" would be an overkill, I agree.

>>58650582
I inserted my gpu on it's slot and booted, it worked without the need of installing aditional software, that's what plug and play means right? If so, then yes, it does indeed have plug and play.

>>58650555
why? are you using gentoo and forgot to add something?

>>58650469
>>58650478
I wanna use a riced Xubuntu, hopefully it works all good.
Thanks for the link.

>>58650620
Second GPU? If you are talking one single GPU then fuck off, anon had problem installing a SECOND GPU IN ADDITION to his first.

>>58647080
Tell me more, I didn't had 2 OS on same machine since long time. Who must be host windows or unix?

>>58650380
Yeah and if your GPU is splitting PCI lanes with the chipset and the CPU instead of just one or the other, passthrough usually wont work.

>>58648799
>writing in machine code is just typing 0s and 1s

neo/g/ everyone

>>58650649
yes, second gpu

>>58647080
I intend to do that once Zen lands

How do you switch mouse/keyboard input between host and vm though? synergy?

>>58650641
Well, the procedure should be the same on any distro, it all comes down to whether or not the given distro's package manager can obtain the same packages.

You're welcome, good luck anon

>>58650661
>splitting lanes with the chipset and the CPU
What kind of retarded motherboard layout would combine lanes from chipset and CPU into the same slot? This would cause all sorts of problems so I don't think this is an actual issue.

Anyway, the IOMMU is located on the northbridge which is essentially located on the CPU itself.

>>58647080
does this setup require dual monitors in addition to dual GPU?

I recently went from a multi-heard configuration to a single 43" UHD and don't have space to add one of my old displays back.

File: aaa.png (23KB, 834x894px) Image search: [Google]

23KB, 834x894px

>>58650575
the flag doesn't show up on dmesg

>>58650639
:^)

>>58650730
>the flag doesn't show up on dmesg
Show output of

cat /proc/cmdline

>>58648419
Windows on bare metal is harmful

>>58647080

Is speed and compatibility 100%? Also so many other questions unanswered, give us some tips.

File: bbb.png (33KB, 834x894px) Image search: [Google]

33KB, 834x894px

>>58650747
IOMMU was disabled in genkernel

>>58650773
So now

dmesg | grep DMAR

shows a bunch of stuff, right?

File: 2017-01-25 13_37_39-PCI passthrough via OVMF - ArchWiki.png (39KB, 2347x264px) Image search: [Google]

39KB, 2347x264px

>>58650702
I'm sure you're right

>>58650691
Gonna try it soon, hopefully it all goes well.
Again thanks again, papa bless.

>>58649156
ahahahahahah I know how to fix this cus it happened to me too.

All you have to do is turn your monitor off and on by pulling the plug if it doesnt have a switch.

bam, there's your VM

>>58650789
Haven't finished it yet, I'm actually disabling a bunch of shit because I was too lazy to do that when I installed this machine

I'll report later if nothing goes wrong

>>58650791
>I'm sure you're right
I am. Yes, both exists, but this is because PCI lanes directly to the CPU is typically gen 3 and intended for GPUs and stuff, whereas from the chipset is typically intended for gen 2 and is there for legacy reasons. This is typically for server architectures though, not so common in consumer CPUs.

The slot doesn't matter though.

>>58647080
i only have one gpu

>>58650830
So as has been said all a-fucking-long in this thread.

Not all hardware will work, and even if your motherboard, CPU, etc support passthrough, there are situations where you rmotherboard simply wont allow the isolation you need for passthrough to work in your setup.

The solution is to buy a new motherboard, but that's not really a solution is it?

>>58647144
apparently, last time i seriously looked into it was in 2009, and you needed to navigate quite a number of hoops and cross the fingers of everyone in your vicinity

>>58650853
It will work, again the device itself doesn't matter and motherboards combining lanes like this will simply drop down to gen 2, but it will still fucking work just not as gen 3.

Again, this is an IOMMU thing, it has nothing to do with devices or motherboards or lanes. PCIe passthrough works because you're just virtualising the bus addresses and pass virtual IO addresses to the device driver. It's analogous to virtual memory on x86.

File: ultimate GPU passthrough VM tree of maximum efficience.png (56KB, 800x600px) Image search: [Google]

56KB, 800x600px

I always wondered if something like this was even remotely possible. Tried to draw the idea up for you real quick.

This way you wouldn't have to use one increasingly bloated gaming / production VM. Different software you install separately wouldn't be able to gather information on what else is installed on your system either.

Easy to keep every VM updated and clean, if this kind of branching is somehow achievable with VMs.

How impossible is this, /g/?
No one would need to install Windows on bare metal ever again.

>>58650888
Not it wont fucking work, just read the picture I posted.

Unless you are fine with passing through ALL devices on that root, which in some cases you dont want to fucking do, you're screwed.


Or you're doing deep level changes like ACS patch
http://vfio.blogspot.com/2014/08/iommu-groups-inside-and-out.html

>>58650916
>Unless you are fine with passing through ALL devices on that root, which in some cases you dont want to fucking do, you're screwed.
This is vastly different than "this doesn't work at all", which is what I am arguing against.

Anyway, again, this is an issue if you don't have enough CPU bound lanes. What consumer motherboard released the last 5 years doesn't have at least x16 to the CPU?

>>58650898
It's possible with:

- Containers: Windows doesn't support that though.
- VMs: Overlay images are supported with qemu/KVM. Changes in the base image would break all other images though.

So just keep all your data on a share and snapshot those with btrfs or something like that.

>>58650916
>>58650951
Also, you can always drop down to the number of lanes that actually do go to the CPU. I doubt that many computer games or whatever you do with your GPU fully utilise x16 all the time anyway.

>>58647080
I would do that if my spare GPU didn't have some issues resetting in pci passthrough. it resulted in the hypervisor freezing occasionally, and the host requiring a reboot.
now i just use two computers and synergy.
i notice no latency with synergy over GbE (connected to the same switch) even with a wireless mouse.

Wine 2.0 came out yesterday. Needing VM or dual boot will soon be obsolete.

>>58650969
In the various benchmarks of games vs. PCIe bandwidth, the only ones that seem to even notice changes are the ones heavily leveraging MegaTexturing.

Probably 90%+ of games can go from PCIe 3.0 x16 (16 GB/s) to PCIe 1.0 x4 (1 GB/s) and not lose more than 10-15% performance.

>>58650965
>Changes in the base image would break all other images though.
But in most cases there would be no good reason for that, would there be?
Let's say you update some drivers on the Windows base VM. The Steam base VM branching from that only contains the original changes Steam made to the old Windows base (the 'overlay', if that's the correct term here). So the old Steam overlay should work without issues if it was using the updated Windows base. Since nothing really was changed that directly affects it.

But best to play around with it to figure it out, I guess.
>snapshot those with btrfs
I'll have to read up on btrfs, might lead me somewhere. Thanks.

>>58651147
>the only ones that seem to even notice changes are the ones heavily leveraging MegaTexturing.
Makes sense because these are probably so big you need to move them in and out of system RAM, whereas most games just do rendering which is a compute intensive task with so little data dependency it fits on the on-board memory.

File: ebinbacon.jpg (239KB, 1280x850px) Image search: [Google]

239KB, 1280x850px

>>58651133
Thanks, i needed new /g/ humor material

>>58651133
Can I play witcher 3 on wine?

>>58651133
What's so special about Wine 2.0?

>having dedicated desktops
One up'd ya, I have a linux laptop and my desktop boots straight to Steam Big picture.

>>58651531
It runs Office 2013. :^)

how to set up without recompiling my kernel?

>>58651430
witcher 3 was supposed to get a linux release guess what never happened

I just built a computer for it but haven't set it up yet. Anyone have a link for a good guide? I'm running Debian Jessie

>>58651846
https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF

http://vfio.blogspot.com/

>>58651531
>>58651678
It also supports 64 bit on macOS and improved direct3d 11 support, and new release schedule for faster releases.
https://www.winehq.org/announce/2.0

>>58648774
Do we know if this is going to trickle down to other distros, like *buntus or is it bound to Fedora?

>>58651846
This guy uses debian unstalbe https://forum.level1techs.com/t/play-games-in-windows-on-linux-pci-passthrough-quick-guide/108981

although you should be able to make it work with just >>58651874

>>58651733
That was just steam fucking up everyone, it wasn't even announced by cd project red, it was probably a stupid error from the people in charge of updating the page.

>>58651966
I'm upgrading to Stretch right now. I tried it last week and audio through HDMI didn't work, but I didn't troubleshoot it too much. If anything I'll just reinstall Jessie and go from there.

>>58651992
Don't upgrade to stretch yet. It's transitioning from testing to stable and updates are going to be few starting early February.

>>58651992
You shouldn't need stretch to make it work, maybe adding the testing repos is a good idea though.

As for the audio, I've never tried the audio through hdmi, I either passthrough the integrated audio or just emulate AC97.

>>58652008
Well shit, I've already got half the packages downloaded

>>58652025
I've been wanting Stretch anyway for the updated kernel with newer Btrfs support, but it sounds like I should stick with Jessie for now. I tried installing the backports 4.8 kernel and just got a bunch of dependency errors

Is it possible to configure PCI passthrough that way so I can still use my dedicated GPU to play Linux games sometimes?

In other words, when I'm not using passthrough I want access to my dedicated card on Linux, and when I want to use Windows I could just swap my Xorg to integrated

>>58652026
Stretch entered soft freeze early january, meaning no new packages from sid. Early February it'll enter full freeze so no updates to existing packages unless necessary and once deemed stable enough stretch will become stable and jessie oldstable.

Do I have IOMMU? There is no such option in BIOS
Gigabyte H81M-HD3 rev 1

$ dmesg | grep DMAR
[    0.000000] ACPI: DMAR 0x00000000DD7DA3C8 000080 (v01 INTEL  HSW      00000001 INTL 00000001)
[    0.022779] DMAR: Host address width 39
[    0.022780] DMAR: DRHD base: 0x000000fed90000 flags: 0x1
[    0.022787] DMAR: dmar0: reg_base_addr fed90000 ver 1:0 cap d2008c20660462 ecap f010da
[    0.022787] DMAR: RMRR base: 0x000000df683000 end: 0x000000df691fff
[    0.022789] DMAR-IR: IOAPIC id 8 under DRHD base  0xfed90000 IOMMU 0
[    0.022789] DMAR-IR: HPET id 0 under DRHD base 0xfed90000
[    0.022790] DMAR-IR: Queued invalidation will be enabled to support x2apic and Intr-remapping.
[    0.022958] DMAR-IR: Enabled IRQ remapping in x2apic mode

>>58652056
you can detach your gpu from vfio and reatach it to it's drivers, look for rmmod and modprobe

Here are a couple of scripts, the downside is that you have to kill x

look for "Melvin Vermeeren scripts qemu" on google, the guy uploaded some scripts, you can adapt them for your setup, I can't link because for some reason 4chan think it's spam

>>58652345
that's pretty neat, I honestly didn't even hope that's possible

what about this? >>58652126
fucking gigabyte I cannot find any information about IOMMU on this mobo in internet, and there is no bios option so it propably won't work
but then there is this command and output is not empty

>>58652126
>>58652388
What is your CPU?

>>58652407
i5-4460
it supports vt-d 100%, there is an option in bios too
but vt-d is not iommu right?

>>58652436
read
>>58650478

and it will all make sense

>>58652436
from https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF#Setting_up_IOMMU

>Ensure that AMD-VI/VT-d is enabled in your BIOS settings. Both normally show up alongside other CPU features (meaning they could be in an overclocking-related menu) either with their actual names ("Vt-d" or "AMD-VI"), legacy names ("Vanderpool" for Vt-x, "Pacifica" for AMD-V) or in more ambiguous terms such as "Virtualization technology", which may or may not be explained in the manual.


so yeah, it is, just enable vt-d in your bios and add the line to your kernel bootloader options.

>>58652474
holy shit life is beatiful
will try later, thank you

>>58650336
No he's not. You are. For PCI passthrough to work the device to be passed through has to be the only device in its iommu group.

>>58649203
so tired of these prison-trans with bad hair

>>58648015
>all those shit monitors
No, not really.

>>58647080
you mean my Intel HD Graphics 4400 ?

>>58647080
Sure, but that's complicated. Rebooting into Windows takes like 15 seconds with an SSD

>>58655154
It's like you faggots don't actually want the best of both worlds, at the same time.
Anyway, Fedora 26 has in its roadmap full integration of GPU pass-through in Boxes, for any lazy fucks still stuck in the last century.

>>58656362
Except you don't have the best of both worlds

>>58656403
I guess my current setup must be imaginary.
Thanks for clarifying.

>>58652819
On my AMD 990X + 8320, I can pass through any number of the PCIe devices to a VM. I can pass the USB controllers, WiFi, and GPU to the VM with full performance.

>>58656485
Post IOMMU groups:

#!/bin/sh

# List the devices in each IOMMU group

for iommu_group in $(find /sys/kernel/iommu_groups/ -maxdepth 1 -mindepth 1 -type d); do 
    echo "IOMMU group $(basename "$iommu_group")"; 
    for device in $(ls -1 "$iommu_group"/devices/); do 
        echo -n $'\t'; 
        lspci -nns "$device"; 
    done; 
done

>>58647080
You totally can't tell where the cut & paste happens.

Video is super legit.

>>58656577
It is

https://www.youtube.com/watch?v=37D2bRsthfI

>>58656683
Are you retarded? That is the same edited video except OP converted it to webm format to shit up /g/.

File: 1431590733407.gif (1MB, 258x193px) Image search: [Google]

1MB, 258x193px

>>58656723
you're right, that setup, all the setups people are already running, and all the info on dedicated wiki such as arch are fake and designed to shit up /g/
thanks for your high quality post

>>58656723
Exactly

>>58656740
>implying I am bitching about the ``technology''
>comments directed towards the shitty fake video

You don't think some shit tier ``enthusiast'' on /g/ would fake his ``epic'' gaming computer after reading your fabulous wiki?

It is a shame you can't discern exactly where the edits are in the video.

>>58647080
>he taints his install

>>58656577
>>58656951
is fake

look carefully at the pixels
and the fact that the sun obviously travels very far in an instant

So what's the best distro for doing this, if I don't want to use Arch or Gentoo?

i tried getting gpu passthrough to work but for qemu/kvm it was really slow and when i fixed one of the problems causing bad performance (disc type needs to be scsi with some redhat driver) the sound turned into pure static cackling which apparently is DPI latency and there's no fix for it

for xen it wasnt possible to compile on arch nor did the one in the AUR work at all, when i switched to debian to try it the xen kernel would boot to a black screen


there are barely any guides on doing it either

also apparently barely any nvidia gpus are supported for doing this?


i'd really like to do it since i got the hardware for it but there's just a bunch of issues that i don't feel like attempting to troubleshoot again. for anyone who wants to bother don't use qemu/kvm as the hypervisor, its for servers and it seems to have really shitty performance

Don't you need multiple GPUs for this?

>>58657340


you can use onboard gpu for linux

>>58657359
I don't have an onboard GPU. I remember seeing this years ago when I bought my motherboard and made sure to get a compatible one, a 990FXA which was one of a very select handful, but it doesn't have an integrated GPU.

>>58657191
Arch

>>58657420
your CPU should have one

File: [Ajin2.com] Ajin Season 2 Episode 13 [End] [720p] [Subbed] _0001_00:19:44_[2017_01_25_17].jpg (96KB, 1280x720px) Image search: [Google]

96KB, 1280x720px

>>58647080
Well, my current hardware doesn't support it, but Zen will probably will. I do dual boot and use Windows exclusively for some gaymes since Linux does have a few native games and emulation. My questions are the following:
How is the latency between the KVM and the host?
Can I use my main GPU on Linux and then switch to a crappy GPU on the fly without rebooting when I want to use the Windows KVM?
Is there ever going to be supported with one GPU?
Will Fedora make it one button click retard proof?

>>58657311
>for anyone who wants to bother don't use qemu/kvm as the hypervisor, its for servers and it seems to have really shitty performance
Um.
Xen is the server one AFAIK

Qemu is literally designed to be handy to use

>>58657471
Socket AM3+ CPUs don't have iGPUs.

File: Screenshot 2016-11-26 14.14.13.png (51KB, 891x593px) Image search: [Google]

51KB, 891x593px

My CPU doesn't support VT-d, otherwise I'd jump to Linux+PCI Passthrough in a heartbeat.

File: lain_computer.jpg (6KB, 280x267px) Image search: [Google]

6KB, 280x267px

>>58648015
Is this the future they spoke of?

>>58657706
Does the ms toolkit work with Windows 10? I've been on Linux for three years now, with a W7 spare SSD that I'd like to sidegrade. I was thinking 8.1, but I might as well go full botnet and go with 10 for shenanigans and familiarity for the future VM.

>>58657706
i7-4770K supports VT-x tho?

https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF#Prerequisites

http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Processor-8M-Cache-up-to-3_90-GHz

>>58657815
Yes, but if I understand correctly you must have both, VT-d and VT-x to get it working as intended.

>>58657764
Don't know. I don't have disabled/removed anything related to telemetry. I stopped giving a fuck I guess.

>>58657815
https://wiki.debian.org/VGAPassthrough
>Your CPU should support virtualization and IOMMU (not supported by K variant of Intel CPUs).
lmao doesn't the -K suffix mean unlocked? As in, it's supposed to be "un-nerfed edition"?

>>58657871
Intel are massive cunts. I really hope Zen follows previous iterations of having all of the instructions.

>>58657871
Intel being jews as always.

>>58657871
My 5820k says otherwise.

>>58650069
Zoinks

>>58657730
>monitors scattered around the room
>requires her to move her neck in order to see anything on another screen

shitty art directors REEEEEEEEEEEE

>>58647080
op is a fag. take your shitty thread to /v/. /g/ has had this for about 6yrs now.

>>58657871
>>58657889
>>58657896
https://en.wikipedia.org/wiki/List_of_IOMMU-supporting_hardware
according to this, K cpus support that. At least my 4790K is on that list.

I want to play what games are on linux only on linux and Im pretty sure you can only pass a GPU to a VM and back seamlessly on certain graphics cards that I dont own
and I dont own two graphics cards or strong enough hardware to give the vm decent specs without probably crashing everything

>>58647080
>using a VM instead of installing and running Linux natively

>>58647194
this

>>58658715
OP's saying to run Linux natively.

>>58658730
Then why would I need GPU passthrough for gaymen?

Linux runs all my gaymes natively.

My processor architecture doesn't support GPU passthrough or I would

>>58658715
OP is running linux natively retard, Windows is the VM

>>58649203
Fucking traps

>>58658767
>Linux runs all my gaymes natively.
Good for you, but many games don't. In my example I like playing simulator games (racing and flying) which will never work either natively or through wine

>>58649203
>more since
>pic shows you lack sense

>>58656403
Runs fine on my machine. I switched to Manjaro last year around Dec 15'/Jan 16'. AMD GPU Intel IGP. Ran an extra cable. Made the switch because Win10 Insider was even more of a Win8 aids edition. Installed my new 500GB Evo drives that were $100 bucks and off I went. Just as good as Windows on my hardware.

You should refrain from commenting unless you have actual experience. Thankfully you're somewhat anonymous because you end up looking like an ass.

>>58658632
4790k and 4690k are devil's canyon parts and support passthrough. He has the older Haswell part so it won't work per the guides AFAIK. I have a couple DC Intel CPU's and it works.

bump for this quality thread

>>58647080
Would you recommend Kvm or wine on a laptop for photoshop and other Adobe bullshit.
I have i5 6200, 8 gb, ssd. no strong gpu.

>>58649156
Git gud

Tried it.

My HD4600 on Xubuntu was just crap. I wouldn't be able to watch youtube videos at 1080p. It was also slower to load windows that way than it was to just dual boot.

>>58647080
my 2500k doesn't support it

>>58647114
Only what you do inside w10

Since OP is a useless fuck, I'll try to answer some of the questions you guys have
>>58657608
>How is the latency between the KVM and the host?
If you are referring to mouse/keyboard input latency, it depends on your setup - if you feel synergy doesn't do a good enough job (since it is routing packages over the network), you can always set up a software mouse/keyboard switch, like so: https://rokups.github.io/blog/#!pages/full-software-kvm-switch.md
>Can I use my main GPU on Linux and then switch to a crappy GPU on the fly without rebooting when I want to use the Windows KVM?
Yes, however it is hardware dependent - so far nvidia cards had more success with this. See: https://rokups.github.io/blog/#!pages/second-xserver-second-gpu.md
>Is there ever going to be supported with one GPU?
Possibly, so far it looks like there is support for Intel's vGPU technology, see: https://www.kraxel.org/blog/2017/01/virtual-gpu-support-landing-upstream/
>Will Fedora make it one button click retard proof?
At this point, it's already trivial to automate the setup if all your hardware is supported - so I don't see why they won't.

>>58657191
Fedora, it has very good support for latest vitalization packages: https://fedoraproject.org/wiki/Virtualization_Preview_Repository

>>58657311
Xen is deprecated for this kind of setup, that's why you don't see anyone recommending it anymore.
Also your assertion that somehow a piece of software can't be used in whatever environment you have a need for it, is quite presumptuous.
DPI latency is commonly fixed by enabling MSI support for your hardware: http://vfio.blogspot.com/2014/09/vfio-interrupts-and-how-to-coax-windows.html

>>58660481
Most laptops are shit out of luck for GPU passthrough, since even if they have two GPUs available, they are usually hardwired to HDMI/DP/VGA outputs and share a framebuffer, or they are even locked in an Optimus configuration (thanks w540), making individual assignment impossible.

>Fell for the Intel K meme back when Ivy Bridge came out

>Don't even overclock, have stock heatsink
>Desperately want to Linux host/ Windows VM with passthrough but Intel disabled VT-d on K processors because PAY UP FOR THAT EXTREME EDITION GOYIM

I can't justify buying a new mobo and CPU, and buying used parts in this 3rd world hell hole is a shit show of tough RNG rolls for quality and not getting shafted by a scammer.

>>58660719
Now that's complete bullshit, the HD4600 on my 4790k has no issues with any desktop usage, including 4k60fps youtube.

>>58661046
My man, my 280x can't handle 4k60fps youtube. The bullshitter is none other than you

>>58647080
I wish this worked with VMware on Windows.

>>58661015
I feel you, I was in the same boat having originally bought a i5-4670K. I traded that shit in at the first opportunity for a haswell refresh chip and lucked out on the OC too.

>>58661075
Why would you ever need another OS if you already use Windows?

>>58661103
To switch between a clean Windows and one for gayman (pirated shit).

>>58647080
How is GPU passthrough meant to help me in FL Studio?

>>58661147
At least you'd get decent graphics.
Also, I believe you can passthrough other devices as well, like soundcards.

File: Screenshot from 2017-01-26 13-26-25ed.png (902KB, 1889x901px) Image search: [Google]

902KB, 1889x901px

>>58661062
Turns out you're right, it is dropping frames at 4k60fps.
But can we agree that there shouldn't be an issue with fucking 1080p, or do you need a screen for that too?>>58660719

From what I read, it appears two GPUs are needed. Since my CPU doesn't have an integrated GPU (and I am extremely grateful for that), I just can't.

File: 751685.jpg (8KB, 251x221px) Image search: [Google]

8KB, 251x221px

let's see, this stunt requires hardware with VT-d support, an integrated or second dedicated GPU, and requires you being okay with setting it all up, the possibility that this might be slower than dual booting (>>58660719), as well as the inherent overhead and potential performance and compatibility issues.

the only good reason i can think of that anyone would want to do this is to run Windows and Linux on one machine at the same time. is it because you're running a server or something else on Linux that has to stay up 24/7? or is it to be able to play games and shitpost on /g/ at the same time?

Seems to be a lot of effort to play garbage videojaimes

File: 5.jpg (4MB, 4032x3024px) Image search: [Google]

4MB, 4032x3024px

>>58647080
>Dual booting
>ever.

Son I just run two identical computers.

Does the host OS use your GPU when you aren't running the guest OS?

How is the guest OS compatibility with USB peripherals?

>>58661991
What happened to the velcro? Oh thats right, you decided taking a blade to those zip ties was too much of a risk.

>>58661991
why would you put 60 pounds of weight on something for no reason

>>58647805
Not OP but the performance is close to native, because VT-d there is near no CPU overhead.

>>58651911
Seeing the nature of open source, I'm sure it will. Fedora is the pushing force in GTK/Gnome so most projects will start there.

>>58650898
qemu can have separate base images and deltas with their qcow2 image format.

>>58649717
the VM I pass through to isn't UEFI, I don't even know what that GPU bios thing is about or why it matters, but I never messed with that.

>>58650489
3-5 monitors is the sweet spot for 3D animation, enough room for viewports, your timeline, separate motion editors, and asset/entity lists.

>>58650516
Normally it depends on the CPU not the motherboard, what do you have?

>>58650728
No, you just wont be able to do things on your host for obvious reasons.

>>58661911
If you want to talk about practical reasons, from the top of my head, at least in my case, it's faster than dual-booting, also it's annoying to reboot or physically go from one PC to another when all you want to do is play something for an hour or so.

Depending on the person, someone may have to use windows for one or two programs they need.

It's also more secure, you can take snapshots in case something fucks up with your windows install. Microsoft has been tending to fuck up the something in every update.

>>58657954
>implying lain even NEEDS monitors, at all
>not connecting to the wired by schumann resonance alone
ishiggity giggity

>>58647080

that's so cool...

>>58662601
Pomodoro timer to git pumpd
>>58662564
There is no velcro, you stupid fucking faggot.

>>58661062
4k60fps VP9 decodes fine on my cpu (fx8320)

didn't find any AVC 4k60fps youtube videos, do your cards decode VP9, or are there some AVC 4k60 yt videos?

Is it possible for the Windows guest to access content on the host's ext drives?

>>58663454
use network file shares

>>58662864
I might try it again sometime, but I wish there was an idiot-proof guide...

>>58663454
theres 2 ways

-you make sure thr ext partition isn't mounted and you add it to the VM, and then boot said VM, you'll need a software called ext2fsd ( or something like that, I can't recall the name) for windows to recognize the partition

-Or just make a samba share

>>58663466
>>58663488
>network share
That's a pretty great idea actually.

>>58663424
I don't think any video web content is hardware decoded by the GPUs, so it's stupid in the first place to imply that you can't watch youtube due to your HD4600.
Also youtube has been pushing away from AVC/h.264, so there might not be any 60fps AVC content out there.

My main point in all of this is that there is no practical reason why you wouldn't be perfectly fine with just the integrated Intel GPU on Linux - heck it even has the best free drivers.

>>58663497
My one suggestion regarding this, is that if you're going to the trouble of running Linux as your main OS, you might not want to let Windows have access to your Linux partitions - instead create a share on the Windows VM, and mount that on Linux for moving/sharing files.

Is it possible to have a virtual OS with multiple partitions coming from multiple vdi files? I've only made relatively small virtual OSes that use one vdi file. I guess I want the virtual OS to have the vdi that would be the C: drive be located on my SSD and then a separate bigger vdi file for the D: drive on a separate HDD.

>>58663724
yes
I do it on daily basis with all my data from main linux rig to my vm-s for school

>>58663724
an image per partition would be a pretty odd setup, but you can just use multiple whole-disk images in a vm

File: f63.jpg (22KB, 396x382px) Image search: [Google]

22KB, 396x382px

>>58661991

>>58663774
Basically what I mean is I want to recreate what I do right now in Windows where I run some games off of C: (SSD) and then other games off of D: (HDD). I want to separate them physically so I'm not using up all the space on my SSD. SSD for games with very long load times and HDD for the rest.

>>58663814
Again, yes, there is no problem assigning multiple virtual disks to a VM, each with its own .qcow2 image file stored on a separate physical drive. You can even assign an entire disk if you'd like to eliminate IO overhead from writing/reading from an image file. Or, if your motherboard has multiple storage controllers, you can even pass-through an entire controller to the VM, giving Window direct access to the attached disks.

>>58663814
just use two full-disk images, put a partition on each

>>58663929
>pass-through an entire controller
This is getting pretty insane. So many different possibilities.

>>58650728
>does this setup require dual monitors

>>58663002
>No, you just wont be able to do things on your host for obvious reasons.

so there's no way to redirect the guest's GPU framebuffer output directly back to the host's compositor/GPU?

>>58663989
Adding a Spice/VNC Server to the VM would allow you to do this, and with the amount of bandwidth available from host to VM, it could run ok for even video. However it does introduce significant overhead by moving all that data through software, and it does add latency/compression artifacts, ultimately defeating the purpose of a gaming VM.

>>58664057
do spice and/or vnc servers support the avc/hevc offloading on newer nvidia/amd cards, or would it be realistic to just use the built-in ShadowPlay/ReLive services to stream the content to the host?

File: Screenshot from 2017-01-26 19-48-38.png (3MB, 1600x937px) Image search: [Google]

3MB, 1600x937px

>>58664399
I guess you're actually thinking about Steam streaming with AMD VCE/Nvidia ShadowPlay hardware encoding and Intel hardware decoding. On one hand you have a lot more bandwidth (I think I was hitting 20Gbps with iperf3 testing) and stability over the virtio network interface that in any physical setup, so it does run quite reliable. However on the other hand, you need to consider if you are ok with the image quality reduction resulting from performing GPU hardware compression.
That being said, I have to admit, I do use Steam streaming if I'm only casually playing something while also browsing other stuff - having the game in a scale-able window is nice - pic related.

Spice/VNC is done all on the CPU, no hardware acceleration as far as I know.

Does this still require having dual peripherals for each "machine?"

>>58664618
I don't see why it ever did. We've had Synergy for ages, and there are other solutions as well posted itt.

>>58663799
>One guy who has a stick up his ass about reddit posting everywhere.

Thanks for subscribing to me, senpai. sry mine got more upvotes than yours

>>58650139
There is a lot of tutorials out there about coreboot (and disabling the intel me in thinkpads) and libreboot. I can elaborate if you want to.