Jump to content

DirectX 12 Multiadapter Technology ... the future?

 

 

http://wccftech.com/directx-12-multiadapter-technology-discrete-integrated-gpus-work-coherently-demo-shows-big-performance-gains/

 

This would most certainly require the user to be able to customize desired performance output among the GPUs in question, you could f.ex. set a GTX 1080 to render 80% of the tiles and an AMD 380 to 20% as an extremely simplified example. As of right now only AotS supports this feature and is only capable of working with two relative equal GPU´s, too much disparity and performance actually drops. The above mentioned example makes use of the IGPU which in any case would up frame rates even if its just by a small percentage.

 

Microsoft just stated that they are working on this feature and better implementation of it in DX 12, I find the above however hard to believe because rendering different tiles is one thing but both cards need to be able to work together and have the capability of executing the same kind of instructions. At least thats how my little brain figures it without being the big tech nerd.

 

Having one card however executing instructions which alleviate the workload of the master card is quite plausible although I can imagine it to be hard to program.

 

Is any of this a realistic thing or just some nice thoughts? I am mainly referring to using two very different GPUs performance wise.

 

 

Link to post
Share on other sites

The question is how much benefit could you get from it, for the extra effort the developer has to put in to implement it? Is it worth it?

 

I'm thinking it might be of more use on lower end gaming laptops, where the add in GPU might be weak enough that the integrated GPU is a significant contributor.

Gaming system: R7 7800X3D, Asus ROG Strix B650E-F Gaming Wifi, Thermalright Phantom Spirit 120 SE ARGB, Corsair Vengeance 2x 32GB 6000C30, MSI Ventus 3x OC RTX 5070 Ti, MSI MPG A850G, Fractal Design North, Samsung 990 Pro 2TB, Alienware AW3225QF (32" 240 Hz OLED)
Productivity system: i9-7980XE, Asus X299 TUF mark 2, Noctua D15, 64GB ram (mixed), RTX 4070 FE, NZXT E850, GameMax Abyss, Samsung 980 Pro 2TB, iiyama ProLite XU2793QSU-B6 (27" 1440p 100 Hz)
Gaming laptop: Lenovo Legion 5, 5800H, RTX 3070, Kingston DDR4 3200C22 2x16GB 2Rx8, Kingston Fury Renegade 1TB + Crucial P1 1TB SSD, 165 Hz IPS 1080p G-Sync Compatible

Link to post
Share on other sites

8 minutes ago, porina said:

The question is how much benefit could you get from it, for the extra effort the developer has to put in to implement it? Is it worth it?

 

I'm thinking it might be of more use on lower end gaming laptops, where the add in GPU might be weak enough that the integrated GPU is a significant contributor.

I agree with the first part but Im not sure how big of a percentage gaming laptops are of the PC sales.

 

Using the IGP for post processing seems like a good idea because most desktops have one.

But I think most motherboards disable it by default when a dedicated GPU is connected.

 

Supporting any pair of cards is not feasable from a developers perspective. Especially because you may want completely different rendering techniques based on the power of the cards.

The tile rendering mentioned before requires the cards to duplicate data like textures, 3D models and light volumes. This would not work well

Desktop: Intel i9-10850K (R9 3900X died 😢 )| MSI Z490 Tomahawk | RTX 2080 (borrowed from work) - MSI GTX 1080 | 64GB 3600MHz CL16 memory | Corsair H100i (NF-F12 fans) | Samsung 970 EVO 512GB | Intel 665p 2TB | Samsung 830 256GB| 3TB HDD | Corsair 450D | Corsair RM550x | MG279Q

Laptop: Surface Pro 7 (i5, 16GB RAM, 256GB SSD)

Console: PlayStation 4 Pro

Link to post
Share on other sites

4 minutes ago, mathijs727 said:

I agree with the first part but Im not sure how big of a percentage gaming laptops are of the PC sales.

 

Using the IGP for post processing seems like a good idea because most desktops have one.

But I think most motherboards disable it by default when a dedicated GPU is connected.

I was using gaming laptops as an example, but it equally applies to low-mid range desktops where they add a low end discrete GPU. Even if iGPU is disabled by default, in most cases you can force it to be active also.

Gaming system: R7 7800X3D, Asus ROG Strix B650E-F Gaming Wifi, Thermalright Phantom Spirit 120 SE ARGB, Corsair Vengeance 2x 32GB 6000C30, MSI Ventus 3x OC RTX 5070 Ti, MSI MPG A850G, Fractal Design North, Samsung 990 Pro 2TB, Alienware AW3225QF (32" 240 Hz OLED)
Productivity system: i9-7980XE, Asus X299 TUF mark 2, Noctua D15, 64GB ram (mixed), RTX 4070 FE, NZXT E850, GameMax Abyss, Samsung 980 Pro 2TB, iiyama ProLite XU2793QSU-B6 (27" 1440p 100 Hz)
Gaming laptop: Lenovo Legion 5, 5800H, RTX 3070, Kingston DDR4 3200C22 2x16GB 2Rx8, Kingston Fury Renegade 1TB + Crucial P1 1TB SSD, 165 Hz IPS 1080p G-Sync Compatible

Link to post
Share on other sites

Raja From AMD said that's the future and it all depends on the devs. He made the point by comparing it to multi-core processors.

When dual-cores and quad-cores first came out little to no programs were supporting it, it took time and now there is programs that support 16+ cores.

Instead of taking it up by them self's and doing all the work alone now they give all tools to the game devs and its up to them implement . They are working with Microsoft also to make it easier for devs to implement. Its looking good now and once dx12 support takes off that's when well you see it really take over.

We are going down the road of smaller GPU dies because its just to expensive to make big dies when you go under 14nm 10nm,  dual GPUs dies on the same die/interposer  is the way to go in the future and devs need to start learning how to optimize for that.

Slowly...In the hollows of the trees, In the shadow of the leaves, In the space between the waves, In the whispers of the wind,In the bottom of the well, In the darkness of the eaves...

Slowly places that had been silent for who knows how long... Stopped being Silent.

Link to post
Share on other sites

3 hours ago, 3DOSH said:

Raja From AMD said that's the future and it all depends on the devs. He made the point by comparing it to multi-core processors.

When dual-cores and quad-cores first came out little to no programs were supporting it, it took time and now there is programs that support 16+ cores.

Instead of taking it up by them self's and doing all the work alone now they give all tools to the game devs and its up to them implement . They are working with Microsoft also to make it easier for devs to implement. Its looking good now and once dx12 support takes off that's when well you see it really take over.

We are going down the road of smaller GPU dies because its just to expensive to make big dies when you go under 14nm 10nm,  dual GPUs dies on the same die/interposer  is the way to go in the future and devs need to start learning how to optimize for that.

 

I totally agree with you on that. Imagine buying a 1080 and utilizing it side by side with a 970, the 1080 doing 70% of the job and the 970 the other 30%.

Link to post
Share on other sites

Still comes down to developer implementation. Like multi graphics card support now, why build a PC that picks your games for you depending on support? Until every game supports multi graphics card set ups the single graphics card option will always be king.

If anyone asks you never saw me.

Link to post
Share on other sites

So far game developers have done very little to support multiple cards for a good decade. They have had access to tile based rendering and split frame rendering for years and years, and we can count on one hand how many games utilised it. It never stopped being an option. Even with Alternate frame rendering there is a vast amount of things developers could do to improve performance, something Nvidia/AMD often had to do for them in drivers because they were unwilling to do it.

 

So the idea that some how giving them another chance is going to change the game....its not going to happen anytime soon. Giving developers the only keys to multicard scaling is beyond stupid, DX12 is going to be a disaster for dual cards not its saviour and we are seeing that play out with the initial games already. People who think this is going to work haven't been watching this industry for very long.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×