who here AMD Processor + Nvidia Video Card?

>>61722655
I'd say a good 20% of us

>>61722655
Gaymers BTFO
>>>/pcbg/

>>61722655
I would be very interested in doing this and a combination as a mega-super-turbo server+data+gaming box that I used headless 80% of the time. Ryzen would be great for iterated simulations and CUDA has would save me personal labor. However, residential upload speeds and inconsistent dynamic address bullshit mean the entire idea isn't practical (now). My area is getting fiber this autumn.

Reminder that ryzen can use ecc udimms.

Has nVidia fixed their fucking retarded drivers yet? Or do you STILL get lower performance with Ryzen?

>>61722958
Proofs? Cause I might get a ryzen to go with my GPU, since it's the 1000 gtx series

>>61722958
>>61723077
AFAIK this only happens in DX12 games and nobody runs DX12 versions on Nvidia anyway.

>>61723197
>dx12
Sounds about right, nobody uses that shit anyways

>want to run a GeForce card in a Windows VM using GPU passthrough
>the drivers detect the OS is running in a VM and refuse to install
>complain to nVidia, get told to buy a Quadro
>want to use a GeForce card and a Radeon card in the same system for different things
>the nVidia drivers detect there is a Radeon card in the system and disable PhysX
>only way to get PhysX back is to physically remove the Radeon card
>want to use Gefore Experience for easier driver installation and Shadowplay
>need to sign in with your email
NVIDIA: Never Again.

>>61722655
Ryzen 7 1700 + gtx970 reporting

>>61722655
I have an 1800x and a gtx 1080, it was a little be rough in the beginning, but its been smooth sailing for the past 2 months

6300 with a 950 wishing someone would kill me
>going 8350/1060 soon

>>61722655
1600x and GTX 1070 is my rig

>>61722655
>AMD Processor + Nvidia Video Card
What's the difference between this and being a cuck?

I wanted to go r5 1500x + gtx 1060, any issues with this combo?

>>61722655
8320 @ 4.4 + 970.

it's alright I guess, plays all the games I have great except h1z1.

eventually I'll go to r5 1600/X + 1070 or 1080. anything I should worry about going to r5 1600/x?

I play on 1080p but have a 144hz monitor. I just want most games to be minimum of 120fps on most games on high.

>>61722655
During the TNT/3Dfx era, that was an ideal combo.
You either used AMD CPU+Nvidia GPU or Intel CPU+3Dfx GPU.
The Nvidia drivers used the 3Dnow for the T&L, while the 3Dfx drivers relied on the CPU FPU, that was better on the Pentiums.

File: 324864.jpg (9KB, 299x168px) Image search: [Google]

9KB, 299x168px

i5-2500 + HD 7870 + Realtek RTL8111E LAN

to

Ryzen 1600 + GTX 1060 + Intel I211AT LAN
AMD + Nvidia + Intel

who /comfy/ here

What's the equivalent 1080TI card for AMD?

>>61722655
I'll have it the other way around soon. (waiting for Vega) Kinda wishing I had waited to get Ryzen instead of this i7-4790K however. This will likely be my last Intel chip.

>>61723531
The Vega 64 will be.

>>61723552
Did they say the price?

>>61723318
Yeah, I've had different issues on my end. You use Nvidia with Linux, you have to kill 20% of your perf just to reliably fix screen tearing. Took them multiple years to publicly acknowledge that the issue was even there.

Fuck Nvidia and their stupid broken bullshit. Should've never tried switching teams.

>>61723552
>>61723564
Probably won't hit 1080 Ti perf, but it will likely get near 1080. Price should float around $500.

>>61723564
499 for air and 699 for water

>>61722655
me

>>61723576
>>61723585
Well fuck. I'll wait for a 1080ti equivalent. The price point could be lower if its likely near a 1080.

>tfw RADICAL COMPUTER CENTRALISM

>>61722655
>who here cpu bottleneck

Not me

File: 1499378554100.jpg (2KB, 125x125px) Image search: [Google]

2KB, 125x125px

Nah mate i was a Piledriver 8320e+ Maxwell GTX960 4gb then i went Skylake i5 6600k+ Polaris Rx480 8gb and iv'e realized i was using the wrong CPU and GPU brand all around. Ryzen Seems a bit tempting but unless my i5 does not do well in the games i mostly play for me 4 cores are fine for now, until some AAA game start to really using more cores i will be considering Ryzen 1700. Until then i am good.

>>61723694
So you have a Ryzen?

>>61723737
It doesn't surprise me you would come to that conclusion

>>61723708
The only reason AMD's behind is because Intel wouldn't stop paying companies not to buy any AMD stuff. They've been pulling that for like 20-30 years now. Intel's shit won't work anymore.

File: poozen gaming.jpg (547KB, 2531x1359px) Image search: [Google]

547KB, 2531x1359px

>>61723737
No

Sempron 3850 + GT 710
Gonna upgrade soon to A8 7600 + GTX 760

>>61723753
>>61723782
By the price those CPUs do cost, you should be getting twice the frame rate of the ryzen parts, not marginally better.

>>61723799
The 7700k is almost half the price of the 1800x you moron

>>61723782
Date of benchmarks?
>>61723799
bigger price=/=bigger gaming perf
Ryzen destroys as a workstation card and after the updates it does just fine with games. Intel shills just like to hang on to the old benchmark results and leave dates out.

scared about inevitable driver gimp

>>61723843
>date of benchmarks
2 weeks old

https://www.youtube.com/watch?v=Rnf5QzoLjG4&t=132s

>>61723796
wew anon

nevertheless, you're going to get cpu bottlenecked with that gpu

>>61723843
>old benchmarks
>2 weeks old
Time to move the goalpost somewhere else

>>61723782
damn, didn't know 133 is half of 157, nor that 197 is half of 211

>>61723918
You do realize the 1800x is alot more expensive than the 7700k right?

Also you must be new to technology as a whole if you think price scales linearly with performance

>>61723612
>I'll wait for a 1080ti equivalent
Navi is in 2019

>>61723945
Christ what the fuck. I'm going to save my shekels

>>61723929
kill yourself and take whoever was the retard that made those graphs with you

8350 and a 1080 ftw2 icx

File: GPU-Growth-Large.jpg (162KB, 1335x732px) Image search: [Google]

162KB, 1335x732px

>>61723945
They have something planned for 2018 that competes with Titan Xp. Notice how Vega stack extends all the way up

>>61723332
r7 1700x with 2 gtx970
>h-hey my internet penis is 10% bigger than y-yours

>>61723782
>i7-7700K @ 5GHz
How many died in the fire?

>>61724006
By 2018 the Volta will be out and blow anything AMD puts out to shreds so long as it is based on the Vega arch

>>61723782
what ram is he using

>>61722655
1800x with a 760 here, waiting for vega™ to hit the shelves

miners leave my cards alone REEEEEEE

>>61723534
Got 4790k over a year ago.

Only kinda regret the gtx970.

Especially now that I'm switching to linux and want gpu passthrough.

I may keep the gtx970 for the host, and if I run a windows VM I'll passthrough an AMD card.

Though I'm seeing less and less need for Windows to game really. Star Citizen will be ported to Linux when its finished.. and I can always just dual boot instead for gaming if I really want to, and avoid gpu pcie passthrough...

In which case I can sell my 970 and get a Vega.

>>61723585
Aint 1080 also $499? So they're really matching there not undercutting

And for the water version thats 1080ti price... unless the water cooling allows for a higher clock meaning it obtains 1080ti performance..which would be big news if so... but even then they match nvidia's price

>>61724208
Wait, why can't you do passthrough? Unless your motherboard doesn't support it, you should be all set. Devil's Canyon has that feature enabled.

>>61724208
>Star Citizen will be ported to Linux when its finished
Hahahah

Not that guy, but I don't care if it takes them another 10 years to finish it. It's already pretty beast and these guys obviously put a lot of attention to detail.

>>61724236
I just hear its annoying with nvidia GPU you have to run through more hoops to trick it to think its not in a VM, whereas with AMD you don't

>>61724239
;___; when its finished

Seriously though, I'm thinking they release the beta in 2 years. Been keeping track and really all this foundational work is the hardest. Once they transition to just making content (items, planets) and not frameworks (systems for persistence tracking, tools to build solar system), things should move quicker.

>>61724257
Oh yeah, now that you mention it, I do recall that stupid Nvidia bullshit when reading guides. It's shit like this that's got me ready to buy a Vega card despite knowing that it'll probably be buggier due to drivers.

Why are they making the Vega $500 if it matches the 1080? 1080 often sometimes below $480 or $450 maybe below $450 if you can snipe that shit.

>>61724279
They haven't even got a concept of how it all fits together. They have all these individual modules but they have no idea the system which ties them all together and the groundworks which will make the basis of that system.

>>61724304
They probably figure the cards will disappear due to miners so why price them much lower?

Shits annoying because I would normally be OK with buying a used GPU, but now knowing what kinda shit they may be run through wih mining...yeah probably wont buy a used amd

>>61724304
Miners. They don't have to undercut Nvidia because their stock will sell out regardless if it is good or not.

>>61724327
Actually the 3.0 alpha release on sept 8 will tie things together a lot more, they're expanding the missions, ship cargo will be persistant, their will be tracking of ship's legal owner, etc. the Universe is coming together nicely.

Squadron 42 is essentially its own stand alone single player game, and Arena Commander is basically an arcade mode.

>>61723972
Is that how you deal with evidence that conflicts with what you believe?

>>61723318
>>want to run a GeForce card in a Windows VM using GPU passthrough
>>the drivers detect the OS is running in a VM and refuse to install
it sucks but there is a simple fix for it if you are using a QEMU script to launch the VM.

# Basic CPU settings.(hyperv tweak + Nvidia workaround)
OPTS="$OPTS -cpu host,kvm=off,hv_spinlocks=0x1fff,hv_relaxed,hv_time,hv_vendor_id=Nvidia43FIX"

>>61724672
Quick rundown on qemu scripts..

Is that something that applies to VM player? Vbox?

>>61723379
Run well? This is what I want... I'm 8350 and r9 270 for now...

>>61724698
Not a precise technical explanation so take it with a rock of salt.
KVM is kernel based virtual machine which is already included in the linux kernel, you just have to enable it.
Qemu is a hardware emulation layer you'd have to install it, you use it to start up the virtual machine and set up and all the devices you want to passthrough, audio, video, network connection, etc.

In layman it is like a configuration file for your VM, you just write a script from a template and execute it to run the VM.
There are GUIs like virtmanager that facilitate this process but you have less control since you tweak it with what is available in the GUI, and through the messy XML file if there are missing features in the GUI. Qemu is easier once you get the jist of things.

have a look at this guide it explains the basics, its what I use as a reference to set it up. Arch wiki is more thorough but this is easier to grasp for a beginner.
>https://www.evonide.com/non-root-gpu-passthrough-setup/

>>61724784
>Arch wiki
Thats main reason I'm considering arch right now. I've only done very basic work with gcc and command line in centos, trivial stuff I'm sure to most.

Was considering NixOS because I like the idea of easily being able to format and reinstall everything (documents offloaded to a nas).

>>61724016
It's probably still behind the 1800x in power consumption, but nice meme

me
>>61722958
never really had a problem since I don't game much and when I do I don't care about getting 200 fps

>>61724856
Looks like the 1800x uses about 3-4 watts more on stock settings, but it has twice as many cores.

Your point?

>>61724784
QEMU/KVM gets more performance than VMware/Virtualbox?

>>61724951
yes, pretty sure it has the best performance for GPU passthrough but you can only use it with Linux as the host OS.

>>61724984
Right, gotcha

One thing I cannot find is a comparison between: just running the VM as normal, no gpu pcie passthrough, so both the host and guest use the same gpu; versus gpu passthrough performance. Is it leaps and bounds better? Because I imagine the host OS isnt demanding a lot from the GPU when its not passed through.

>>61725005
Generally, that configuration will give a virtual GPU to the guest OS. It's effectively software rendering. If we could do it that way, passthrough wouldn't be considered nearly as frequently.

>>61725005
hardware emulation for 3D acceleration is pretty poor currently, probably due to the overhead in passing the display output from VM to host OS.
You get leaps and bounds more performance by passing the physical GPU to the VM since it uses the device directly.

I hope newer GPUs start including features that enable them to run one or multiple VMs in the host OS simultaneously on a single card in windowed/borderless-fullscreen mode. Just like emulating the video except close to the performance you'd get with GPU passthrough. It would be amazing

>>61724784
>>61725005
>One thing I cannot find is a comparison between: just running the VM as normal, no gpu pcie passthrough, so both the host and guest use the same gpu
on that topic, if vega rx supports SR-IOV, does QEMU/KVM support it or will it need some sort of special update?

>>61725078
I imagine with GPU's getting more RAM space this should help part of the problem of using a single GPU for both host/guest?

>>61725078
>>61725057
Its fairly simple to swap between dedicated gpu for host-linux, no vm & igpu for host linux w/ gpu passed to vm?
Or is it annoying enough that most people get two high power gpu's to run for both host/guess so they dont have to always game on the vm?

File: 1487841880472.jpg (37KB, 640x360px) Image search: [Google]

37KB, 640x360px

>>61722655
Intel + Nvidia masterrace
>tfw your PC powers your gaymes but also ISRAEL
>tfw your PC is exterminating muslims at every active second

File: 1a651f96a851.png (391KB, 1920x1031px) Image search: [Google]

391KB, 1920x1031px

ha ha ha ... w

>>61725140
>Its fairly simple to swap between dedicated gpu for host-linux, no vm & igpu for host linux w/ gpu passed to vm?
>Or is it annoying enough that most people get two high power gpu's to run for both host/guess so they dont have to always game on the vm?
Funny you ask that, I am struggling with it.

I managed to get it to work but it is such a hassle, so far I have had limited success.
I have two solutions, one partially works, and the other definitely works but takes more time than writing a single command in terminal.

>method 1
I wrote 2 xorg config files, one for iGPU and another for the dGPU.
I wrote a command that switches between the two GPUs but it automatically logs you out of your session and logs you back in to load the relevant Xorg.conf file.

The xorg file for the dGPU uses intel modesetting to set the iGPU as inactive. This method currently does not seem to work for nvidia it is really strange.
running lspci -k show that my GPU is actually using the nvidia driver.
However, running glxinfo | grep renderer shows that the dedicated GPU is using Gallium 0.4 llvm instead.
The Gpu is displaying and everything but I cannot run 3D games in steam.

>method 2
I reboot then go to bios and disable vt-d and the iGPU.
I use a xorg.conf file I have previously generated using nvidia-xconf command.
This method works, its basically just running linux with the dedicated GPU as the primary GPU right from boot.

I am frustrated that I cannot figure out how to get method one to work properly and use openGL4.5 like method 2. I am sooooooo fucking close its infuriating.
Method 1 is very quick, just open terminal type one command and bob's your uncle. It automagically logs out and back in with the GPU switched.

It might be an issue with modesetting or nvidia not supporting GPU offloading, might work with an AMD card haven't really tried.
Two powerful dGPUs is wasteful power consumption and cost wise. It is like using a sledgehammer to tackle a small nail.

>>61725362
I agree and I cant afford another dgpu right now.

Whats the "best way" people stream their games these days? Not OBS, but like Steam game streaming? So to stream the game (or whole OS?) back to the host OS. I figure this will have a performance hit, but not as much as not doing gpu passthrough.

>>61725401
streaming from steam is a pain, I managed to get it working at 720p on a gpd win from my desktop reasonably well. But it takes a while to tweak and relies heavily upon how solid your local network is.
Quality takes a hit.
Don't think its worthwhile on a Linux desktop even unless perhaps you are sitting next to the router connected via Ethernet. Even then quality will not be as good as running straight from dGPU.

>>61725401
it might be possible to use prime to switch GPUs, I haven't really dug into it on arch wiki.
Tried it with the nvidia GUI, the GPU switching works as intended but I cannot unbind and passthrough nvidia to vfio because it becomes the primary GPU.

It did work on previous Xubuntu setup before by some freak incident, I was able to swap Using Nvidia's GUI between iGPU and dGPU and still was able to passthrough when I have iGPU as main.
So I know IT IS possible, there must be something I am doing wrong.

File: 4593948789_cae251caa5.jpg (50KB, 446x361px) Image search: [Google]

50KB, 446x361px

>>61725164

>>61725475
Cant you stream to yourself the same way you can for example, change the port for ssh from 22 to 2222 on a vm, then connect to the vm from the host on localhost:2222? Or something like that?

Does that not skip the router? Or am I retarded and it does use the router it just uses your internal Ip?

>>61725590
haven't really dabbled with it much so I am not sure if you can bypass your router. If you figure it out I'd like to hear from you.

>>61725610
Will do. I'll make a thread when I play around, just got a seperate 850 evo to put linux on; want to leave my windows install alone on its own drive so if I bork something on linux drive im not fucked

I use 750Ti along with 8150 because some programs need CUDA.

File: 1477346282273.jpg (67KB, 800x600px) Image search: [Google]

67KB, 800x600px

>>61722655
>tfw fx-4300 + gtx 660

>>61725629
yeah its what I've done as well. Linux on SSD and windows in its own 3TB HDD.
If you are worried about borking linux you can get a cheap ass HDD, I have a 500GB 5400 RPM from some old crummy laptop that I use for scheduled backup snapshots with rsync.
What I do is unmount the 500GB HDD and have it only mount when the automated scheduled backup script runs once per day.
That way even if something nasty like rm -rf / is triggered your backup HDD in theory should be untouched since it is not mounted.

>>61725729
Well I figure while I play around it should be easy to reformat the whole thing and just restart worst case, still need to put in the 850 evo, but I figure I should be able to easily reformat the 850 from a usb boot or something worst case.

I was considering NixOS but I dont know if with playing with VM's and pcie passthrough if havin arch's bleeding edge is more important

>>61725164
>tfw you hate jews but also hate muslims and not sure who you wanna stick it to more

>>61723796
At least get a Ryzen 3.

>>61722655
I am, would have bought rx580 8G, but fucking miners... All the cards were out of stock...
So got myself 1050ti...

>>61722655
FX9590 Housefire
GTX 970 4gb OC

Never a problem

FX 8320 w/H100i & STRIX GTX 1070

Bottlenecked to fuck but at least it renders video nicely.

>>61722655
absolute heresy

Me in 2018

File: 1501458960889.jpg (966KB, 2661x1638px) Image search: [Google]

966KB, 2661x1638px

Mfw i5 4690 390x

I need a new pc

>>61728293
Isn't that as powerful as a 1060-6GB? How is that needing to be replaced badly?

>>61728332
Some games it's 1070 level some it's 1060

Mainly my 100% all the time tho

>>61729511
100% what, CPU usage? Because I can see that happening with an old i5.

>>61723318
Nvidia cards are more reliable than AMD ones for passthrough though. if you get around Nvidia's dickery everything just werks after that, while AMD cards will hang or even crash more frequently.

I'm Intel processor and AMD GPU

Just a reminder that Ryzen sucks for PCI-e passthrough. Despite Wendell shilling hard for AMD pci-e passthrough is trash on Ryzen compared to Intel, I know because I've used both. IOMMU groupings are fixed, yes. The real issue is nested paging tables though. By deafult NPT is on, which makes Windows run great but GPU performance is halved in a lot of games. It doesn't affect 100% of games, no one's sure why some work and some don't, but you will most likely not have a good gaming experience. Turn NPT and your GPU can be utilized fully, but your CPU performance will tank due to slow I/O speeds. The only sort of workaround is by installing Windows directly to a hard drive and not using the Red Hat virtio drivers, but then you lose a lot of the advantages of running a VM in the first place, and you're still not getting good VM performance.

In short, I would STRONGLY discourage you from getting Ryzen solely for setting up a Windows VM with GPU passthrough right now. It is completely unknown if or when AMD will fix this.

>>61722655
Yeah, me. I fell for the 970 meme at release. Never again.

>>61723424
Being a cuck is when you let other people fuck your girl. Having amd cpu and nvidia gpu is when you fuck other people's girl. Like yours.

>>61731501
how is nine hundred and seventy a meme?

>>61732441
(YOU) (YOU)
(YOU) (you)

>>61732519
now I am a meme? you are pulling quite the strawman there.

>>61722655
I guess it doesn't really count, but I used to have an AMD processor + AMD video card + NVidia chipset (nForce 2)

for a real tour de force, use an nVidia chipset with an Intel CPU and an AMD video card!

File: wut.jpg (44KB, 584x517px) Image search: [Google]

44KB, 584x517px

>>61722655
I literally never thought I would say this, but nVidia + AMD masterrace

>>61724024
Maybe, but the price/performance will likely be skewed. They're pushing a psuedoshrink and increasing overall die size meaning their yields will probably be pretty bad. Then there is the compute capacity which they probably will continue the trend of disuse in consumer cars. With RPM coming to fruition this could cause a lot of issues on nVidia's side where Ryzen will shine though it's difficult to speculate on the performance capabilities of either since we don't have any hard specs for consumer level cards.

>>61723354
Why 8350?

>>61732955
The next race is the first to MCM GPUs. AMD knows it, NVidia knows it.