Far Cry 3: Benchmarks and Review

Article Index:   

GPU Performance

We benchmarked Far Cry 3 on a Core i7-3770K with 8GB of DDR3-1600 RAM. We used AMD's Catalyst 12.11 beta drivers for measuring Team Red's performance, while Team Green was tested using the 306.97 driver set. We tested the game at "High" and "Ultra" presets at 1920x1080 with 4x MSAA, Enhanced Alpha to Coverage, and HDAO enabled.

Our benchmark was a lengthy drive around the island, with pauses to hop out and fight various enemies that spawned along the way. Far Cry 3 randomly generates encounters as you explore; so we repeated the test multiple times on each card to ensure our frame rates were representative of the game's performance.



Far Cry 3 tolerates a low frame rate exceptionally well; the GTX 660's 28 fps is absolutely playable. The 7950 is significantly faster than the GTX 660, but the GTX 670 is the fastest card at High Detail. The Radeon 7950 seems to suffer from more texture pop-in than the GeForce cards, but that could be a problem on the software end. High detail levels with AA, AF, and DX11 are no problem.



At ultra-high detail, the GTX 660 starts to stagger a bit. This is most easily remedied by turning down water and shadow rendering a notch. Our benchmark path takes us past the open ocean and across a beach.

Given the way ambient occlusion is handled in game, we opted to test performance with that feature disabled via profile editing. Both the 7950 and the GTX 670 pick up a few frames here as a result.



Turning AO off improves game performance enough that the option to play without it really ought to be included. While we're on the topic of questionable design decisions, can we talk about the fact that buying the game on Steam forces you to install Uplay?

If I wanted to use Ubisoft's online store, I'd have bought the game through Ubisoft. Steam is nothing but a launch tool, it literally can't detect if the game executable is running, unless UPlay is. Talk about annoying.
 

Related content

Comments

blog comments powered by Disqus