Is 4gb VRAM on the Fury X going to be a problem on Crossfire?

[drakioned]

I'm considering moving up from 2x GTX 780s to either 2x Fury X or 2x GTX 980 Ti.

I'm running at 4K and regularly see that VRAM usage max out on my 780s.

Given that Fury X only has 4gb RAM and doubling up on the GPU doesn't increase the amount of RAM available, would it make sense to go with the 980 Ti instead?

Is there a 8gb Fury X on the horizon in anywhere near a sensible timeframe?

 

[ChrisD.]

Team green. Maybe I'm biased, but had nothing but issues with my 290X's in Crossfire. YMMV but I won't be getting another AMD card for a long time for my main rig.

 

[africanos23]

Of what has been going around we will probably see 8GB next year. Your going to have to wait and see if the 4GB HBM is enough for 4K . (personally I doubt it)

Reviews will be out today and hopefully they push the settings to see how HBM copes.

 

[RavenXXX2]

I would go SLI 980 ti for more memory and the better SLI drivers, crossfire has been meh of late.

8GB fury looks to be way off.

 

[JediFragger]

Team green. Maybe I'm biased, but had nothing but issues with my 290X's in Crossfire. YMMV but I won't be getting another AMD card for a long time for my main rig.

Fortunately DX12 natively supports mGPU setups, so once it's bedded in then it'll be worth getting AMD cards again

 

[Poneros]

As much as it pains me too see it, I'd definitely say that a 980ti is looking like a better proposition than the Fury X, and more so for SLI vs CF.

 

[baldmonk]

As much as it pains me too see it, I'd definitely say that a 980ti is looking like a better proposition than the Fury X, and more so for SLI vs CF.

this ^

 

[TheMorningStar]

The GFX RAM is independent under DX12 from what I understand, this is probably not the same as pooled RAM I'm guessing, so they don't share RAM but they can optimise for each core which may provide some benefit. I don't think they could really pool the ram unless they put 2 figi cores on one substrate/package whatever and have the ram between them, that actually seems super interesting but I have no idea if any RAM can share it's data with two hardware devices at the same time like this, maybe HBM can? This would have been something I would have been looking at if I was in bed with Samsung developing it. X2 cards would be even smaller if it's technically possible to pull anything like that off, power would be down slightly more I imagine also.

From what I understand about the promised HBM optimisations, which apparently equate to notably less memory for the same tasks that's down to the game developers themselves so that could be implemented at any time between now and never, hope I'm wrong about that one. Maybe it's just part of the DX12 API or just the driver level? Can anyone enlighten me?

 

[Born_2_Kill_83]

All crossfire is good for is looking at. I went 18 months spending £900 in total and only using £450's worth of card, Sli isn't perfect either but it's better. That's why I always go for the most powerful card solution available and a G-Sync/Freesync monitor for added fluidity. Cuts out all multi GPU known problems as well.