You saying so doesn't make it so.
Especially when what you say is against the evidence. You think Highly OC 980TI SLI will make this suddenly good?
And it isn't vram limited without MSAA, that's running maxed + FXAA.
2 Highly OC GTX 980 Ti SLI are enough but Not 4X Fury CF due to lack of Vram.
What we know now, based on review evidence:
1 GPU is not enough to even run 4K at semi-max settings.
2 GPUs are not enough to run 4K at maxed settings with 4x MSAA where vram matters.
Note this is FXAA, not MSAA, performance would tank massively to a slideshow.
This isn't even maxed, it's using post AA and not MSAA, vram doesn't matter in Witcher 3 as well, its all GPU processing power.
4 GPUs would be required to truly max and have 4x MSAA in some games.
What's the major logical assumption moving forward? Future games will be more GPU demanding, not less.
So in the hypothetical scenario, 2 years from now, applying current-gen Fury X/980Ti scenario:
1 GPU is not enough to run 4K at medium/high.
2 GPU is not enough to run 4K at semi-max settings without MSAA.
4 GPU is probably enough to run 4K at maxed without MSAA.
--------------------------
I used to believe that 4GB vram was not enough at 4K but certainly after reading Russian's posts and the evidence presented, it's very hard to argue that case anymore. Games will increasingly demand more GPU grunt and texture quality is already 4K for some, 4GB will be enough until games start pushing 8K textures, at which point, these gen GPUs will just tank due to lack of performance in general, so it doesn't matter.
Fair enough. But someone like me who plays at 2560x1440 on a gtx 970 I'm already pushing close to 4gb on many games, I can't see myself upgrading to a card 40 to 50% faster with the same vram. Maybe your right about these cards arent strong enough to push 4k, other resolutions like 1440/1600p are a different story.
Just becareful how your interpret vram usage when you use OSDs, games like COD: AW will dynamically use up to 8GB vram just because it can. Likewise for GTA V, people on Titan X show it can use up to 7-8GB but there's no difference in performance compared to GPUs with 4GB (ie. CF/SLI R390X/980 pwns Titan X).
Vram matters more at higher resolution than lower, so if it isn't a problem at 4K, its absolutely not an issue at 1080/1440p.
You can check [H] too, where 980Ti is 25%-50% faster than Fury X.
But feel free to make it seem like whatever % you want. I trust sites like TPU, computerbase, sweclockers that run a lot of games, neutral, AMD GE and NV GW combined, and offer a performance summary chart.
They don't even run Civ BE in Mantle mode, where the gains are huge:
If those sites actually did, the 1440p gap will shrink from 5-10% to even less.
What we know now, based on review evidence:
1 GPU is not enough to even run 4K at semi-max settings.
2 GPUs are not enough to run 4K at maxed settings with 4x MSAA where vram matters.
Note this is FXAA, not MSAA, performance would tank massively to a slideshow.
but using CF would give enough horsepower to make those setting playable but not enough vRAM?
Looks like CF just wrecks SLI. In dual-GPU configs as we've seen, it beats reference Titan X & 980Ti easily and matches OC 980Ti models.
In Quad-GPU setups, it blows NV away. XDMA scaling awesome at 4K.
Since no single GPU is enough for 4K, I would have to conclude that multi-GPU is required and as such, Fury X wins the 4K contest. This would also apply to 1440p with DSR/VSR from 4K.
https://www.youtube.com/watch?v=fFpy2L3B8lk
Reference Titan X have a tendency to throttle due to temps in multi-GPU configs. Gotta max that fanspeed!
FC4 is benched with NV GW features enabled. Fury X doesn't care!
Looks like CF just wrecks SLI. In dual-GPU configs as we've seen, it beats reference Titan X & 980Ti easily and matches OC 980Ti models.
In Quad-GPU setups, it blows NV away. XDMA scaling awesome at 4K.
Since no single GPU is enough for 4K, I would have to conclude that multi-GPU is required and as such, Fury X wins the 4K contest. This would also apply to 1440p with DSR/VSR from 4K.
https://www.youtube.com/watch?v=fFpy2L3B8lk
Reference Titan X have a tendency to throttle due to temps in multi-GPU configs. Gotta max that fanspeed!
FC4 is benched with NV GW features enabled. Fury X doesn't care!
Lets be honest here. Reference 980 Ti is off the table for anyone intelligent buying 980 ti SLI. And I would hazard a guess that for someone like yourself, who is very knowledgeable about GPUs, was NEVER on the table to begin with. You would go aftermarket right off the bat.
Lots of people said not to compare with reference Hawaii; now we are getting comparisons to reference 980 Ti.
My prior multi-GPU setup was put on water. I didn't like open air squashed next to each other, the top card gets massively warmer & louder. The ambient inside a closed case rises, causing my CPU temps to go 10C hotter.
So multi-GPU, for cards of ~250W range, works best with reference blowers or the best with water. Fury X definitely has a major advantage being already under water with full warranty unlike DIY solutions. I give a nod to the EVGA Hybrid 980Ti, a great solution and one I would seriously recommend to others who go multi-GPU.
My CF 7970's beg to differ. I got better performance / temps when I ditched both reference coolers and with with custom coolers.
History repeats itself I see:
http://forums.anandtech.com/showpost.php?p=37522993&postcount=85
http://forums.anandtech.com/showpost.php?p=37523104&postcount=92
Silverforce11 never provided any data to backup the claim in that thread either, as far as I can tell.
It may very well be true (in general at least, there are of course different levels of quality for both blowers and open air coolers), but it would be nice to see some actual data to confirm it.
History repeats itself I see:
http://forums.anandtech.com/showpost.php?p=37522993&postcount=85
http://forums.anandtech.com/showpost.php?p=37523104&postcount=92
Silverforce11 never provided any data to backup the claim in that thread either, as far as I can tell.
It may very well be true (in general at least, there are of course different levels of quality for both blowers and open air coolers), but it would be nice to see some actual data to confirm it.
It's because I have much higher ambient than most of you here. Summers are brutal here in Australia.
I have tried to cool open air cards in a case before, it requires a lot of airflow and fans, at which point the entire system is noisy. Not saying it cannot be done, lots of people do it. It just wasn't a good trade-off that I like. Two blowers (7970 reference at 40% fanspeed was enough) and the rest of the system airflow can be cut down (less system fans) due to the blowers acting as both GPU & system exhaust.
The R290s blowers were crap and required 55% fan speed which got really loud, so I put them on water. Again the rad acts as both GPU & system exhaust.
It would be nice for a review site to do an article on it, but what I do know is open air setups with two 250-300W GPUs need a lot of airflow, even then, the case ambient is higher, so CPU & other component get hotter.
Most reviews you read, sites go with open bench so its completely unrepresentative of what a closed case situation is. There's very little study on this aspect.
Edit: Here's a back to history thread on AT forums, discussing merits of blowers/open air cards: http://forums.anandtech.com/showthread.php?t=2269142
All the usual suspects are there, including me, with my mITX SG05 build. TLDR: you need a ton of airflow/pressure to ensure CF/SLI 600W heat dumping in your case. 7970 isn't as hot as R290/X, 780Ti or 980Ti class, gaming load for the original 7970 is usually around 180W each.
Yes, it's called professional reviews, performed for the last 20 years. Also, if you read the thread from start to finish, the Tri-SLI and Quad-SLI/CF situations have already been discussed. Not only are they irrelevant to the OP, but these 2 niche situations are exactly the few exceptions where you'll need to strongly consider blowers. However, even then, it's not conclusive. You would need to do research and make a judgement call based on responses from gamers/your own experience for specific SKUs you want to Tri-/Quad-stack. You can easily have a situation where 4x open air cooled 250W TDP GPUs will still be quieter, cooler and overclock better than 4 blowers.
Perfect example is HD7970 MSI Lightning x 3-4 vs. HD7970 x 3-4 Reference.
HD7970 Lightning Tri-fire
GTX780 ACX Tri-SLI
GTX770 Tri-SLI
GTX970 Tri-SLI - all open-air cooled SKUs
Current specs:
Intel 5820K @ 4.6
Asus Rampage V Extreme
GSkill 4X4GB DDR4 2400
(3) Gigabyte GTX 980 G1 Gaming 3-Way SLI @ 1525
EVGA Supernova 1300G2
NZXT Switch 810
Corsair H100i
http://forum.overclock3d.net/showthread.php?t=67691
Do you think you'll be able to get GTX980 Tri-SLI reference blower cards stable with 1525mhz Boost clocks on all 3 at quiet noise levels? Not happening considering a single GTX980 reference blower at 1500mhz hits 50 dBA! ^_^
that is a 10% bump from just drivers! huge!:thumbsup:@railven
My 2x R290 was fine for clocks at 50% fanspeed in every game, mining required 70% fan which was rdiculous. 50% fan is already very noisy beyond tolerable levels imo. Review sites definitely should do some testing on 2-4 GPU setups, blower vs open air. There's a real lack of info in this arena and its all left to end users who have different tolerances for heat, noise etc.
Anyway. Seems AMD have fixed the crap sub 4K performance with the 15.7 drivers.
~20% gains in FC4 @ 1440p => http://www.overclock.net/t/1547314/...nano-x-x2-fiji-owners-club/2430#post_24147663
Some decent gains in other games: http://www.overclock.net/t/1547314/...nano-x-x2-fiji-owners-club/2460#post_24149806
I noticed about a 10 fps gain in min fps in World of Tanks in city map, that game is horrendous for being 1 thread & major CPU bottlenecked. Witcher 3 about 10 fps avg in cities (my default setting is x16 tessellation override), less in the wilderness. Really awesome driver.
that is a 10% bump from just drivers! huge!:thumbsup: