At the same time, this is a much more interesting solution as it will allow legacy cards to still be used. If DX12 is supported by Southern and Northern Islands AMD cards as rumored (Or anything with GCN 1.0 it seems). Then those could be combined with the upcoming 300 series for even more performance.

It’s potentially revolutionary.

I mean performance increase per additional GPU in an sli or crossfire setup.
A 30-50% gain in frame rate per additional GPU would be amazing considering right now you are lucky if you get an extra 5-10%.

It could certainly be better than 30-50% but I’m thinking cautious here. Going from say 60fps to 90 by adding a GPU would be better than what we get right now.

Even though DX12 can do it doesnt mean nvidia will allow it in the drivers. At the minimum DX12 would just use the GPU to boost framerates but physx for example would most likely be disabled when an AMD GPU is detected in the system

Anyway have you seen the Toejam and Earl kickstarter? i backed it. the first one was so good and this one is true to that

http://www.kickstarter.com/projects/1578116861/toejam-and-earl-back-in-the-groove

30-50% gains would mean incredibly bad scaling and much worse than today. More like close to 100% gains or it’s not worth buying additional gpu’s or do you mean performance increase per generation??

If your talking per generation it still won’t happen as its down to how much power they increase the gpu’s by per generation and knowing NVIDIA they’ll use this to hold performance back per generation even more

With DX12 even with low level on multi gpu setups boosting frame rates 4K setups will get a pretty big performance bump for free but the “big gpu” mode will mean memory issues thing of the past.

Hopefully such a thing will be possible. Even if it isn’t ill likely go AMD anyways. Another great thing will be when it comes time to upgrade your GPU. Today when you do that you sell or toss your old GPU because its useless in a multi GPU setup as the system runs at the level of the worst GPU. In this new setup your old GPU can just be set as slave to the new one and even if its only doing 20% of the total rendering workload…thats better performance. Or you could even dedicate a single GPU to physics tasks or whatever else a game or the api supports.

Exciting stuff in the pc space coming with windows 10.

The api handles it all. All the discrete GPUs do is render what direct x tells them like render this half of the screen or this frame or these pixels. It will also allow each card to have unique data in memory unlike today where memory is cloned. That means 2 4 gig cards will mean 8 gigs of vram instead of just 4 gigs duplicated.

Could finally allow the true potential..the potential nvidia promised years ago to be unleashed. No more 5-10% gains in frame rate. We may now see 30-50% gains with each GPU added