[Crossfire vs SLI] 780Ti/780/290X/290 x Dual/Triple/Quad

[DGLee]

Hello guys. I'm a newbie on this forum. Glad to post this thread here.

I tested four flagship single-GPU graphics cards scale one to four respectively, along with AMD/NVIDIA's existing 2-GPU SKUs such as 7990, GTX 690 and ASUS 760 MARS. Test presets are as follows : 1920 x 1080 without AA, 1920 x 1080 with 4x AA, 2560 x 1600 without AA, 2560 x 1600 with 4x AA, 3840 x 2160 without AA and 3840 x 2160 with 4x AA. Other hardware configurations and softwares used are as follow image:

(Note that I'm going to upload results under Ultra HD (3840 x 2160) only since there're tons of imgs and results under Full HD (1920 x 1080) and Quad HD (2560 x 1600) are nearly meaningless in terms of CF/SLI scalability. You can read the full text here: http://udteam.tistory.com/585)

Here's the results:

1. Aliens vs Predator (without AA / with 4x AA)

2. Batman : Arkham City (without AA / with 4x AA)

3. Battlefield 3 (without AA / with 4x AA)

 

[DGLee]

4. Battlefield 4 (without AA / with 4x AA)

5. Bioshock : Infinite (without DDOF / with DDOF)

6. Call of Duty : Ghosts (without AA / with 4x AA)

7. Company of Heroes 2 (without AA / with AA : Medium)

8. Crysis : Warhead (without AA / with 4x AA)

 

[DGLee]

9. Crysis 2 (without AA / with 4x AA)

10. Crysis 3 (without AA / with 4x AA)

11. DiRT : Showdown (without AA / with 4x AA)

12. Far Cry 3 (without AA / with 4x AA)

13. Hitman : Absolution (without AA / with 4x AA)

 

[DGLee]

14. Just Cause 2 (without AA / with 4x AA)

15. Metro 2033 (with AAA / with 4x AA)

16. Metro : Last Light (without AA / with SSAA)

17. Sleeping Dogs (without AA / with AA : Extreme)

18. Splinter Cell : Black List (without AA / with 4x AA)

 

[DGLee]

19. The Witcher 2 (without UberSampling / with UberSampling)

20. Tomb Raider : Reboot (without AA / with 4x AA)

Now we've covered 20 game titles. Here are two summaries for absolute framerate and CF/SLI scalability.

- UHD Performance Summary (without AA / with 4x AA)

- UHD Scalability Summary (without AA / with 4x AA)

Okay. That's all.
Thanks for reading my thread!

 

[Silverforce11]

OMG that is some lovely data presented, do you have a review website?

Nvm: saw the link now. Great work.

 

[wand3r3r]

Wow there's a lot of charts. Now to go analyze them a bit.

 

[3DVagabond]

Thanks for the effort.

Did you warm the cards up until their clocks stabilized? How was the micro stutter with the 7990? It was obviously faster than the nVidia cards, but was the gameplay as good?

 

[BallaTheFeared]

Is there a reason the summery section lacks the games AMD had no results in or poor scaling such as FC3 or JC2 but includes titles where Nvidia has no support such as Dirt:Showdown?

 

[Shmee]

Welcome to Anandtech and thanks for the great effort in testing!

 

[blackened23]

I question the legitimacy of these tests because the GTX 780 does not support quad SLI.

To be clear Quad SLI with GTX 780 DOES NOT WORK. 780ti does, Titan does, but 780 GTX does not work and never has worked in a quad sli config. So what's up with that, OP?

 

[Silverforce11]

Read it carefully, modded drivers to enable 780 quad sli support. There's no reason its disabled besides NV saying "NO! No candy for you!"

 

[blackened23]

Modded drivers. 320.00. Sounds legit...

 

[Silverforce11]

Modded drivers. 320.00. That sounds legit

Feel free to ignore the 780 results then.

 

[Ventanni]

I was hoping to see a chart showing the power consumption too, just for kicks.

 

[3DVagabond]

I was hoping to see a chart showing the power consumption too, just for kicks.

Lots!

 

[guskline]

Thank you for all of your effort.

 

[Spidre]

Wacky results. Thanks Lee!

 

[desprado]

Any one can 1000s of graph and chart with different performance.In farcry 3 and some other games R9290X and R9 290 4X is no where where as Battlefield 4 same performance whether it is single or 4x in GTXs cards

 

[desprado]

Feel free to ignore the 780 results then.

Fell free to ignore farcry and some other games give 0 with R9 290X and R9 290 4x.

 

[Face2Face]

DG's Nerdy story... where have I seen that before? Is this the guy that leaks card info before they are released? Anyway, nice job, this must have taken a crap load of time to do. I would like to have seen some boost clocks on the cards, but can't complain much. Thanks for sharing.

 

[SlowSpyder]

Pretty good showing for the R9 290.

 

[el etro]

I question the legitimacy of these tests because the GTX 780 does not support quad SLI.

To be clear Quad SLI with GTX 780 DOES NOT WORK. 780ti does, Titan does, but 780 GTX does not work and never has worked in a quad sli config. So what's up with that, OP?

That's why it haves an note in every result it makes. Was added only for information purposes and not to compete with 4xCF of 290/290x.

Source: http://www.youtube.com/watch?v=o5KdSwZJibQ
The guy used GTX 780 launch driver with a custom inf file(this last has made modding the inf file of a posterior Forceware driver.)

Is there a reason the summery section lacks the games AMD had no results in or poor scaling such as FC3 or JC2 but includes titles where Nvidia has no support such as Dirt:Showdown?

CF generally scales better than SLI at 3x-4x GPU configurations(Since 6xxx series), that's the truth. Maybe can lose to Nvidias in 99%FPS or frametimes matter.
I think he avoided only the Zeroes in the performance Summary.

 

[Fastx]

Wow nice work and thanks!

 

[3DVagabond]

Fell free to ignore farcry and some other games give 0 with R9 290X and R9 290 4x.

That's valuable info for quadfire purchasers (The 1/2 of 1 % ). I don't know why you'd want anyone to ignore it?

Pretty good showing for the R9 290.

Agreed! Even @ $500 ea. they aren't bad value, but I wouldn't buy them until supply increases and they come back down to reasonable prices. Especially if I was going to buy more than one (unless I was going to mine, of course.). I also wouldn't run 3 or 4 cards on air, but that's another topic.