Corsair BF4 Memory Scaling - Fake or Fact ???

[UaVaj]

http://www.corsair.com/en-us/blog/2013/october/battlefield-4-loves-high-speed-memory


i7-4770k w/ 780 SLI with 32GB RAM.



1920x1200 Ultra

DDR3 1600
72 fps min
92 fps avg

DDR3 2400
79 fps min
113 fps avg



5760x1200 Ultra

DDR3 1600
35 fps min
47 fps avg

DDR3 2400
43 fps min
54 fps avg

 

[Dasa2]

memory can scale like that if the game is cpu bottlnecked but thats is very rare especially at 5760x1200
that was done back in the beta the final game responds much different since its been optimized

so while they may not be correct i dont think they were fake

i have looked into memory performance scaling with arma which is a cpu bottlnecked game and the gains are decent
http://forums.bistudio.com/showthre...rmance-comparison-1600-2133-up-to-15-FPS-gain

most games are gpu bottlnecked even with a 4 year old cpu so they see no gain at all from faster ram or cpu overclocking

i think bf4 mp is currently a bit of a blend so if you have a 144hz monitor and say gtx980 sli its possible that faster ram and cpu together could see noticeable gains

 

[PPB]

With terrain and mesh quality on ultra, in 64 MP there is a certain memory bandwith bottleneck that needs to be passed to get vastly better minimums. 1333 cl9 to 1600 cl9 did it for me

 

[UaVaj]

:thumbsup::thumbsup:

arma 3 stratis shows a significant improvement going from 1600 to 2133.
min frame rate got an 18% boost. (44 -> 52)

arma 3 altis shows a significant improvement going from 1600 to 2133.
min frame rate got an 19% boost. (37 -> 44)

one more thing. what was the 1600 timing?




back to BF4

greatly appreciate if you or some else can do a run on BF4 final
with 1600 as the base line and 2133+ for comparison.

 

[Dasa2]

in those first few tests 2X4G SAMSUNG 2133 (10-10-10-24-1T used for all tests)
further down that post i updated it with some timings tested at a later date but i didnt triple check the results that time so there is under 1fps variance in each test which sometimes meant a lower speed would match the performance of a higher speed
the minimum fps in arma had a tendency to be fairly random i wouldnt focus on them so much as the average
where as in gpu limited games the minimum fps may be where the difference is most obvious in a cpu bottlnecked game like arma the fps increase is right across the board

bf4 isnt a easy one to test
a lot of reviews just use a bit of sp gameplay where the game is on rails so so that they get consistent results and they managed to get 95fps with a i3\fx8350 then a 4790k only manages 107fps
http://www.anandtech.com/bench/CPU/1073
so its a very gpu limited test

but there is a few mp maps that are far more demanding of cpu performance however results in mp can be anything but consistent

 

[UaVaj]

thanks for clarification.

2133 CL10
vs
1600 CL10

given that timing was kept the same.
this explains why there was such a significant discrepancy in arma3.
corsair probably did the same when testing bf4.



your later testing with 1600 CL8 does solidified that performance difference was minimal when timing is taken into consideration.