Author : Daeguen Lee (DGLee)

(Any action violating either CCL policy or copyright laws is strictly prohibited. All rights reserved)

 

※ You may see "Flagship Gang Fight" here -> http://iyd.kr/585

크파 vs SLI / 280X 크파 vs GTX 770 SLI / 7950 크파 vs GTX 760 SLI / 270 크파 vs GTX 660 SLI

 


Introduction


About a month ago, I posted the article <Crossfire vs SLI : Flagship Gang Fight> which tested 4 flagship VGAs from two major manufacturer, AMD and NVIDIA - R9 290X, R9 290, GTX 780 Ti and GTX 780 with single, dual, triple and quad Crossfire / SLI configurations. If we could ignore the issues come along with CF such as microstuttering OR if the issues were ignoreably small, then there is a clear way forward us : CFing Radeons benefits more than SLIing GeForces not only in terms of scalability but also in terms of absolute frame per second. 4x R9 290 actually faster than 4 x GTX 780 Ti in several cases I tested, indeed, even 3 x R9 290X bests more numerous configuration of its rivals more than a couple of fights.

 

At this moment, what we should remind is, however, digits in our bank account. It's obvious that a R9 290 is very attractive combination of sexy performance and affordability yet we are already aware of the absolute fact that everything is relative. R9 290's affordability could only be derived by comparing it with GTX 780 or GTX 780 Ti, not R9 280X nor GTX 770. How about CF/SLIing a pair of R9 280X or GTX 770 then? Could they be more affordable way to taste TITAN-class performance?

 

Okay. Let's have a look for materials & methods.



Test Setup & Methodology


 

 

Note that Radeon HD 7950 Boost is highly anticipated that it is to be re-branded as R9 280, a SKU between R9 280X and R9 270X. So comparison group could also be translated as "R9 280X/280/270X".

 

Every technique in benchmark is exactly the same as before.

 

I want to mention that I adopted two different way to analyze the results. One is "framerate accumulation", another is "framerate normalization". There are simple formulas for them. Let Game_n(vga_k) means framerate measured in game "n" on VGA "k". Then:

 

- Framerate accumulation : Game_1(vga_k) + Game_2(vga_k) + ... + Game_n(vga_k)

- Framerate normalization : [{Game_1(vga_k)/Game_1(vga_fixed)} + ... + {...}] / n

 

Now you understand what a "block" in this graphs below implies.

 

 

All right. Let's get drowned in graphs!

 


Result - 1. 3DMark 11


 

 

 

 

2. 3DMark Fire Strike


 

 

 

3. Alan Wake


 

 

 

 

 

4. Aliens vs Predator


 

 

 

 

 

5. Batman : Arkham City


 

 

 

 

 

6. Battlefield 3


 

 

 

 

 

7. Battlefield 4


 

 

 

 

 

8. Bioshock : Infinite


 

 

 

 

 

9. Borderland 2


 

 

 

 

 

10. Call of Duty : Ghosts


 

 

 

 

 

11. Crysis : Warhead


 

 

 

 

 

12. Crysis 2


 

 

 

 

 

13. Crysis 3


 

 

 

 

 

14. DiRT : Showdown


 

 

 

 

 

15. Far Cry 3


 

 

 

 

 

16. Hitman : Absolution


 

 

 

 

 

17. Just Cause 2


 

 

 

 

 

18. Metro 2033


 

 

 

 

 

19. Metro : Last Light


 

 

 

 

 

20. Sleeping Dogs


 

 

 

 

 

21. Splinter Call : Black List


 

 

 

 

 

22. Tomb Raider : Reboot


 

 

 

 

Congratulations! You just escaped from all the graphs............... except those in follow chapter:



Performance Summary - 1 : By accumulation


 

 

 

 

 

Performance Summary - 2 : By normalization


 

 

 

 

 

Multi-GPU Scaling Summary


 

 

 

 

 

...The article is finally over.

 

Thanks for reading (or attempt to read) this. I'll be very happy if you leave your opinion here.

Have a nice day :-)

 

※ You may see "Flagship Gang Fight" here -> http://iyd.kr/585

 

//

 

(아래 위젯은 티스토리의 크라우드펀딩 시스템인 '밀어주기' 위젯입니다. 100원부터 3000원까지의 범위 내에서 글쓴이에게 소액 기부가 가능합니다. 사견으로는 이러한 형태의 펀딩이야말로, 성공적으로 정착될 경우 이해관계자로부터 독립된 벤치마크가 지속가능해지는 원동력이 될 것이라 생각합니다. 제가 작성한 글이 후원할만한 가치가 있다고 여기신다면 밀어주기를 통한 후원을 부탁드립니다. 물론 글을 '가치있게' 쓰는 것은 오롯이 저의 몫이며, 설령 제 글이 '후원할 만큼 가치있게' 여겨지지는 못해 결과적으로 후원을 받지 못하더라도 그것이 독자 여러분의 잘못이 아니란 건 너무 당연해 굳이 언급할 필요도 없겠습니다. 저는 후원 여부와 관계없이 제 글을 읽어주시는 모든 독자분께 감사합니다.)