http://www.overclock.net/t/1594675/pcper-amds-raja-koduri-talks-moving-past-crossfire-smaller-gpu-dies-hbm2-and-more
Good. You're better off putting multiple gpus on the same card so the system kernel will treat it like one.
multicard setups were always a headache
LMAO only 4gb max limitation
>Polaris will have a limit of 4GB VRAM
>no more multi-GPU support
AMD IS FINISHED
M
D
I
S
F
I
N
I
S
H
E
D
>>53531832
please delete this
>>53531832
It sounds more as though they want to replace crossfire
I loved my crossfire and SLi setups when they were working, I'm SLi two 980s right now, I really hope both companies move past this shit because when it doesn't work it's such a fucking headache.
Also, how many fucking years has it been and we still have micro stutter
>>53531832
Did you even read what it said?
Are you retarded?
They are saying they are working on IMPROVING multi gpu and that current methods are not good enough.
>>53532452
This is another nvidia shill thread, you're wasting your time.
>>53532452
>current methods aren't good enough
>crossfire still leaps and bounds above sli
what are nvidia even doing? did they just give up on the idea of making multi-gpu decent in 2005?
![d8kbno2su575wvtp8alf[1].png](https://i.imgur.com/6ut8wwI.png "d8kbno2su575wvtp8alf[1].png")
>>53531949
it wasn't a problem from fury x in 4k why would it be a problem for polaris? RotTR uses ~7gb of vram at 4k the fury x and still performs as good as a 980 ti which has 6gb of vram.
https://www.techpowerup.com/reviews/Performance_Analysis/Rise_of_the_Tomb_Raider/4.html
>NVidiots on suicide watch!
>>53532508
>RotTR uses ~7gb of vram at 4k and the fury x still performs as good as a 980 ti which has 6gb of vram.*
The equivalent of two separate graphics card in a single board, those cards better come with a fire extinguisher.
Also, AMD will increase the price of its cards due to this, barely giving and advantage in performance but with a price equal or higher than novodio.
>>53532556
>barely giving and advantage in performance
the entire point of the thread was how he was saying crossfire sucks and dx12/vulkan will allow for better scaling.
not to mention crossfire scaling is already much better than sli
Just give me a GPU solution that runs 1440p at 100+fps on a single card that isn't running at 95c. If you can't do that yet, then this is a pointless conversation as the only people buying multi-gpu setups have a 1440p or 2160p panel which no single card solution is really going to tackle quite yet. I doubt this coming gen is going to change that.
it sounds like what they're actually saying is because dx12 and vulkan and all of that crap is capable of automatically splicing cards together - crossfire is pointless to have.
the OS does this shit already.
>>53532475
When retards keep buying your cards anyway why bother? This is what happens when you get a monopoly. Amazing to me that Nvidia fanboys want AMD to fail like it helps them.
>>53532621
if it weren't for amd we'd have been on dx9 with a few games here and there on dx11 for the next two decades
>>53532508
Nice 20~ FPS, human eye doesnt see more, huh?
Console 1 : 0 PC
>>53532646
you aren't on /v/ you know.
Aren't dual GPU cards bottlenecked by the PCI bus on your average motherboard?
>>53532632
If it weren't for competition PC gaming would be fucked.
If AMD didn't exist it would be Nvidia vs. the fucking Intel GMA bullshit and they would have literally no reason to be competitive because they would have the entire market on lockdown.
The average Nvidia fan is too retarded to understand the concept that monopolies are bad for the consumer though, so we'll keep seeing threads like this one forever.
>>53532556
>vastly improve your multi-GPU interface so it's better than the competition
>build graphics API that works better than anything else on the market for multi-GPU and scaling (then pawn it off on the people that actually make those)
>start designing smaller, power efficient dies
>sell 2 smaller, vastly cheaper dies with much lower rate of failure in manufacturing on one board with all this new tech coming out that you basically invented anyways
>more dies from a single platter means you can sell those dies cheaper
>sell cheaper, faster, and more powerful cards than your opponent's cards
Smells like AMD has just walked into the lunchroom, walked over, slapped Nvidia in the face, then took their lunch and shit on their face.
>>53532692
No. Even PCIe 2.1 won't bottleneck high end cards at x8 each.
Maybe with ultra high end HBM2 cards 2.1 might start being bottlenecked at x16 but it's unlikely.
![Boris_Yeltsin[2].jpg](https://i.imgur.com/cbbjORS.jpg "Boris_Yeltsin[2].jpg")
Who polaris 11 here? Never paid more than 100$ for gpu ;^).
>>53532815
>Yeltsin
>>53532815
>using the smiley with a carat nose
AMD HAS NO DRIVERS
>>53532721
You call?
AMD poorfags on suicide watch
Nothing here implies multi-GPU is going away in any form. What they market it as is irrelevant.
>With changes in Moore’s Law and the realities of process technology and processor construction, multi-GPU is going to be more important for the entire product stack, not just the extreme enthusiast crowd. Why? Because realities are dictating that GPU vendors build smaller, more power efficient GPUs, and to scale performance overall, multi-GPU solutions need to be efficient and plentiful. The “economics of the smaller die” are much better for AMD (and we assume NVIDIA) and by 2017-2019, this is the reality and will be how graphics performance will scale.
Article says they're planning on moving from one large monolithic GPU to several smaller ones because smaller GPUs are more efficient, smaller dies are easier and cheaper to produce without defects.
Didn't DX12 and Vulcan make crossfire obsolete?
>>53536021
In mantle/vulkan/dx12 the computer sees that there's additional resources available and it puts them to work. Its even supposed to allow mixing different GPU manufacturers.
>Polaris is 4gb
AMD is kill.
>>53535743
in a reddit AMA a while back I remember seeing that one of the guys at AMD was thinking about how great it would be if they could have modular GPUs working in concert across a matrix of interposers, so that end users or OEMs could simply plug in another GPU to the interposer matrix to improve performance rather than replace the whole PCB.
Sometimes it really seems like AMD doesn't want to make money.
>>53531850
that's not how it works faggot
