Crysis 3 Benchmark - GTX 670 560Ti 465 GTS 450

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810
I did a benchmark of Crysis 3, a 45 second FRAPS run in the 2nd stage, Welcome to the Jungle, at the very beginning right when you setup out of the building into the Dome for the first time. I ran to the end of the grassy area and back, 3 difference runs for each setting High Med Low.

i3 3220/ i7 3770K
4GB DDR3
Intel ITX H67
Corsair CX750M PSU

Settings:
1080P unless noted
16 AF, Lens Flare on, Motion blur off
Textures set to Very High
FXAA on, turning it off only netted 1 FPS for each card


Avg Min Avg Min Avg Min

GTX 560Ti GTX 465 GTS 450 GTX 670
High 34 22 26 19 18 12 38 24
Med 39 28 34 24 23 17 49 36
Low 53 38 39 28 29 25

GTS 450
Low 1080p 29 25
Low 900p 38 32
Low 720p 41 33

GTX 670
Very High 34 22
High 38 24
Med 51 37
Low 49 36

I'm REALLY running into a CPU bottle neck with the GTX 670 and the i3, on the High runs, the GPU was only like 50-60% load, when I cranked MSAA up to x4, the load was 100%, but the Avg FPS was 23... the i3 just can't keep up, it was about 80-90% for all setups...

EDIT:

I bought an i7 (I sold one of my Mustangs ;) and it DEFINITELY improved FPS on all levels, here is my update and a link to my video review:

Avg FPS / Min FPS

GTX 670
__i3 3220____i7 3770K
VH 34 22 ____ 45 39
Hi 38 24 ____ 59 50
Me 51 37 ____ 70 48
Lo 49 36 ____ 78 52


GTX 560 Ti
__i3 3220____i7 3770K
Hi 34 22 ____ 36 28
Me 39 28 ____ 52 43
Lo 53 38 ____ 62 27*

*All the cards with only 1GB of VRAM were VERY sensitive to the Texture quality
Since I ran all my benchmarks with Very High textures, I went back through and
ran them again on some other settings

The 560 Ti was VERY IMPACTED by lowering the texture settings (these are with the i7):

Avg / Min FPS

HIGH/ VH Textures___36 28
HIGH/LOW Textures___38 34 - I ran this 4 times, same result EVERY time
LOW/ VH Textures___62 27***
LOW/MED Textures___65 57 - massive bump in the minimum


*** this was my standard benchmark for low, and it kept chugging, so I ran it a few more times and noticed the chugging was VERY consistent, so since I was planning on testing textures anyway, I turned it down to Medium and LOOK at the result!


GTS 450
__i3 3220____i7 3770K
Hi 18 12 ____ 18 12
Me 23 17 ____ chose not to run this
Lo 29 25 ____ chose not to run this

GTS 450 at Low Specs for different resolutions - Very High Details:
___i3 3220____i7 3770K
1080p 29 25 ____ 33 28
900p 38 32 ____ 45 37
720p 41 33 ____ 65 55


Here is the video link, should be uploaded within the hour:

http://www.youtube.com/watch?v=SeSGegdzxRw


Dry
 

Z1NONLY

Distinguished
I just downloaded a game update yesterday and decided to try "very high" again.

Now I can play the game at it's highest setting (still with low aa) where I couldn't stand how often it would slow down before the update. (When set to "very high".)

I don't know if the update helped, or if the underground scene is just easier on hardware, but the update *may* have changed things.

Once I complete the game, I will go back and check the same mission that brought my system to its knees on "very high".

 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810
I thought it was strange that the 560Ti was only 10% away from a 670, but I think its REALLY because of my i3, so I need to get something better and see if I get a big boost at on all cards.

But the thing is, I lot of people have i3 and are kind of expecting to game on them... :/

So I'd hate to see an i7 get 30-40% better frame rates.

but even when I'm running 37 FPS, the game feels really playable, no chugs or drags...

the GTS 450 'felt' the worst at 1080p, lots of frames would take longer to render and things like that, but on 900p all those issues cleared up, and it felt smooth

The 450 and 465 BOTH had the same kind of 'artifacts' but the 465 was a much smoother experience overall, mainly due to the frame rates not bottoming out as much.

Dry
 

Z1NONLY

Distinguished


Well my two 560's are probably only getting in the 30's on "max" but that's with low aa. If I try to go max AA, it turns into a slide show. I suspect a single 670 will only fare slightly better and still be unplayable at such settings.

I think the video's you are watching are using a rather loose definition of "maxed out".
 


Not at all. Your GTX 560 only has 1 GB VRAM and Crysis 3 uses up to 1,6 GB or more at 1080p. You got a VRAM bottleneck. :)
 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810
Z-

I REALLY think my 670 can do better, when I'm on High, I get 50's with occasional dips into the high 30's, its NEVER unplayable. on Very High, I get about 37-40 FPS for most of the game with, like in hall ways and such, and it dips into the 20's in wide open areas and with lots of explosions.

AND I usually play with TXAAx2 on, its only like a 1-2 FPS hit and it looks a LOT better than FXAA which is zero hit...

The area I chose to do my benchmark is JUST REALLY BRUTAL because each blade of grass is rendered individually AND casts its own shadow. That area is REALLY a worst case scenario.

It would be the same as me benchmarking in a spot with a bunch of explosions.

I wish I had a couple hundred bucks right now to pickup a 3570K and see if I can get a boost, but I just bought a Mustang GT and my insurance is due... :/ ... first world problems...

Dry
 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810



Maybe if I set the texture details to Med or something, we can see if it produces a difference? I did flip flop the textures from Very High down to Med, so if FXAA is off and the textures are lowered, maybe?

Dry
 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810


It says I should be pulling 57ish on High, and I'm pulling 39-40 (1920x1200 is only like a 5-10% hit over 1080p), so I now I'd def like to see if I can get some kind of i5 to see if that gives me a boost!

It might also depending on where their doing their benchmarks, because my MAX FPS were in line with these charts.

Dry
 


It doesn't matter if there is two, as you just mentioned yourself. And what is the point of those benchmarks? You clearly see that GPUs with 1 GB of VRAM are in the bottom? Though that also has something to do with their lower specifications. :)
 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810


I wonder if your AMD 965 is a bottleneck? do you have a monitor to see if your GPU is running at 90-100% load? if its only running like 50-60%, the CPU might not be able to keep up.


a 7870 should have 2GB of VRAM, and that's at least what is needed to run HIGH textures or some type of AA

Dry
 

Z1NONLY

Distinguished



I would like to have more vram just to see what difference it makes.

I'm torn between selling my two 560's while they are still decent (as a pair) and spending a little money on two 660's with more vram...

Or waiting until the next generation of GPU's for a much bigger performance boost, but getting less money for my 560's, because they will be less valuable as time goes on.
 


Get a single GPU? Less micro stutter and better support in games? :)
 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810
yea, according to that website, my 670 is supposed to hit 60+ FPS... yea I'm not seeing that... :/

so I should probably work that out...

the 670s are just starting to nudge $300, especially if you are ok with used or refurb, and I would do that over 2 660s... I've had issues getting SLI to work... it hasn't been to reliable, especially with all the funk I had to do to get it to work in Cry2...

Dry
 


And this doesn't show any AMD 8-core CPU? :lol:

It's a fact that Crysis 3 supports more cores, still doesn't mean FX 8350 is better.
 

DryCreamer

Distinguished
Jan 18, 2012
464
0
18,810



These numbers for the i3 3220 on this benchmark whatever are VERY similar to the numbers I'm getting for mine, so, at least according to this, I'm really gonna need a i5 3570 to get the max potential out of my GTX 670

Dry