Has there been any benchmarks using 2 video cards in direct12 or vulcan where no sli or crossfire is needed?

[mcduke]

From what I''ve read games can now use direct X12 or Vulcan and support the use of 2 (or more?) video cards without the need to use sli or crossfire. It also can take advantage of the memory on both video cards, where in sli only the first cards memory is used and the second card just provides extra gpu power. I don't know if it's the same for crossfire. I've read that you can also use different model of cards (at least in sli you needed the same model of card ). Once again I don't know if it's true for crossfire.
Has there been any benchmarks done on the use of 2 video cards on this new multi card support?

 

[sizzling]

Yes but I believe only 2 games actually support it. So while DX12 has this feature the game has to be designed to use it. Ashes of a singularity is one of the games that supports it but I cannot remember how well it works.

 

[sizzling]

Here, found this. Doesn't look great http://www.guru3d.com/articles_pages/ashes_of_singularity_directx_12_benchmark_ii_review,8.html

 

[mcduke]

Well, it did show that the 2 gtx cards worked well together. They should have run a comparison against them in sli and see if Explicit Multi-GPU works better then SLI. I would think there'd be an advantage in being able to use full memory of both cards in the explicit multi-gpu support which you only get use of one card memory in SLI.

 

[sizzling]

I wouldn't say they worked well but it did work. The % gains were poor in my view.

 

[renz496]

Well, it did show that the 2 gtx cards worked well together. They should have run a comparison against them in sli and see if Explicit Multi-GPU works better then SLI. I would think there'd be an advantage in being able to use full memory of both cards in the explicit multi-gpu support which you only get use of one card memory in SLI.

it is not like that. in CF or SLI each GPU need their own pool of memory. imagine you have two GTX980 4GB in SLI. when the game for example need 2GB of VRAM then GPU 1 will use 2GB from it's own pool and GPU 2 will use 2GB for from it's own pool as well. that's why in traditional multi GPU adding more GPU does not mean there is more VRAM available to your game. now with DX12 there are claim to overcome this limitation. each GPU no longer need it's own memory pool. because using this method the game will only see one GPU instead of two and all the VRAM available being pool together. if you have two 980 4GB then the game will have 8GB VRAM to work with. if you have 3 then you will have 12GB. but to be honest i'm not sure if this really work in games even in Ashes of Singularity. this thing is very complicated that to this date only Ashes developer are really experimenting with this stuff.

though for game developer supporting even the most basic multi GPU is not an encouraging matter.