Will 2018 be the year of multi GPUs?

sblantipodi

2[H]4U
Joined
Aug 29, 2010
Messages
3,759
What do you think about multi GPUs?
SLI and CrossFire will die along with DX11, will DX12 multi GPUs convince more people to get a multi GPUs system?

Do you think that next games like Far Cry 5, AC Origins, Wolfenstein II and the other will support DX12 mGPU?
 
Probably not. It's fading, but maybe if developers code support into their engines it can work nice.
 
I don't think 2018, but unless they find a better medium to use than silicon soon, then GPU's are going to hit a similar plateau to CPU's.

I'm betting some time around the mass adoption of 8k gaming, the industry will be forced into multi-GPU support and optimization.

Though I hear AMD's latest GPU architecture is built with scaling in mind. Will have to see how that pans out.
 
Do you think that next games like Far Cry 5, AC Origins, Wolfenstein II and the other will support DX12 mGPU?

Not unless AMD or Nvidia or Microsoft do it for them. mGPU should be part of the driver, not DX-level, nor in the game.
 
mGPU should be part of the driver, not DX-level, nor in the game.
No necessarily. Game developers would understand their engine better, and likely be able to provide the optimum performance possible, *if* they spend the effort to program it.
 
No necessarily. Game developers would understand their engine better, and likely be able to provide the optimum performance possible, *if* they spend the effort to program it.

The only way i can possibly see this happening is if Nvidia and or AMD pays them to make the effort. Developers are always pushed to get there software out as early as possible in order to maximize profits. Coding something like mGPU for at the most 1% of the gaming people out there would not be a good use of there time. Not money wise anyway.
 
Multi-GPU looks alive and well to me ...

6-gpu-ethereum-mining-rig-running-amd-rx-480-gpus.png


just not for gaming :p
 
Last edited:
While I like the idea of going multi-GPU for those who want to push their hardware to the limit, so far it has just not been the most practical. Added latency, often support for it coming months after the release of a game or it gets broken on and off, not utilizing the rest of the GPUs fully.

I'm happy that single GPUs have become so beefy that GPUs in a year or two will run 4K at high framerates without much trouble.
 
It's not dead yet. AMD just release there new crossfire driver. But it's not called crossfire anymore. I been running triple monitors for years now. So I do use multi gpu's. And if they made a 65" monitor, not those cheap ass tv's but real monitor, but not costing a arm and leg and first and last child. I'd buy it and would go to a single gpu.
 
There may not be more than a handful of games in 2018 that will support more than 1 GPU.

Its a technology on life support for gaming. The patient is already dead.

You can already see dedicated circuits disappearing from the bottom up and being replaced with fully compute oriented solutions top down.
 
I personally don't see developers going through the trouble of supporting this, at least not many of them, and even less so doing a good job of it. I was running SLI for about 2 years with a pair of 680's. The first year was good, the second was not so good. It was my first time dabbling with multi GPU setups and in all likelihood my last.
 
I think the fact that most games play pretty well @4K on a 1080TI has just about hammered the nail into the coffin of MGPU. Back a few years ago if you ran a 2560x1600 monitor you needed SLI/CF to hit 60FPS.
 
I had nothing but issues with my 780 Ti's in SLI. The 1080 I have now has been rock solid since the day I got it, not to mention uses significantly less power. I still have no trouble maxing out settings for most all games at 3440x1440 with 60+ FPS.
 
the short answer, no... the long answer, stick a nail in it mgpu's dead and neither AMD nor Nvidia is sad about it, lol.

I wouldn't say that AMD and NVIDIA couldn't care less. The more cards they sell the better.

And mGPU turns over multi GPU support over to developers. Traditionally a CF/SLI enabled game usually required custom profile driver tweeks from their respective company. This is no longer necessary as it's baked into code before the driver level.

In other words, AMD and NVIDIA are no longer responsible for helping getting multiple GPU's to play nicely together. So it's a win/win for them.

For the developers, it gives them more power to play with. But it's like 1% of the population. And what percentage of people who label themselves as gamers have more than one GPU any way? So what game company is going to devote those kind of resources for such a small target audience (AoTS being the exception)
 
The only way i can possibly see this happening is if Nvidia and or AMD pays them to make the effort. Developers are always pushed to get there software out as early as possible in order to maximize profits. Coding something like mGPU for at the most 1% of the gaming people out there would not be a good use of there time. Not money wise anyway.

This guy gets it.

If anything, graphical fidelity is no longer a priority like a few years ago. We have reached a point where it's good enough. Call us lazy or complacent but it genuinely is at a point where it's just diminishing returns on effort.
 
I wouldn't say that AMD and NVIDIA couldn't care less. The more cards they sell the better.

And mGPU turns over multi GPU support over to developers. Traditionally a CF/SLI enabled game usually required custom profile driver tweeks from their respective company. This is no longer necessary as it's baked into code before the driver level.

In other words, AMD and NVIDIA are no longer responsible for helping getting multiple GPU's to play nicely together. So it's a win/win for them.

For the developers, it gives them more power to play with. But it's like 1% of the population. And what percentage of people who label themselves as gamers have more than one GPU any way? So what game company is going to devote those kind of resources for such a small target audience (AoTS being the exception)

Exactly. It being in the game dev's hands is basically a death sentence IMO.
 
No necessarily. Game developers would understand their engine better, and likely be able to provide the optimum performance possible, *if* they spend the effort to program it.

Wouldn't the game engine will haves issues once GPU manufacturer moves into a vastly different architecture? But I guess at that point, GPU will be powerful enough to run on 1 GPU or the game is already dead that it wouldn't matter anyway.
 
Wouldn't the game engine will haves issues once GPU manufacturer moves into a vastly different architecture? But I guess at that point, GPU will be powerful enough to run on 1 GPU or the game is already dead that it wouldn't matter anyway.
Ideally, no. The graphics API is supposed to abstract away the specifics of the hardware so it should be seamless. Same way you can play old DirectX9 games with modern DX12 GPUs and it still mostly works.
 
It’s dead Jim.

In all honesty I don’t see mgpu making any sort of comeback. The market is to small to invest that much money into a game that people mostly get 8 hours out of...
 
It’s dead Jim.

In all honesty I don’t see mgpu making any sort of comeback. The market is to small to invest that much money into a game that people mostly get 8 hours out of...

It'll come back if we hit a manufacturing process wall.

That could take years more, even if 5nm were the end of the road.
 
E
It'll come back if we hit a manufacturing process wall.

That could take years more, even if 5nm were the end of the road.
ven then I am not 100% sure it will come back.
Minus the few games SLI support from the vendor was not the greatest nor is it now when comes from game developers.

For SLI to work it would need to just “work” all the time every time.
 
Back
Top