Far Cry 3 Benchmark

Far Cry 3 Logo

Far Cry 3 is an open world first-person shooter video game developed by Ubisoft Montreal in conjunction with Ubisoft Massive, Ubisoft Reflections and Ubisoft Shanghai and published by Ubisoft. The game was developed on the Dunia 2 Engine. For Far Cry 3, the Dunia Engine was updated to support DirectX 10/DirectX 11 and other various features like: New Water Technology, Realistic Weather System, New A.I. Technology, New Animation System, Realistic Facial expressions, Motion Capture Technology and Global Illumination.

These are the same developers (Ubisoft Montreal) that gave us Assassin’s Creed 3, the worst optimized game since Grand Theft Auto IV and I’m curious to see if Far Cry 3 suffers from the same CPU optimization issues. If you played Far Cry 2, you know how hardware intensive that game was and how good it looked for its time. Far Cry 3 is no different, because the Dunia engine is a modified version of the CryEngine, the game “borrows” all the strengths and weaknesses of Crysis.

Methodology

The scene that was used to test Far Cry 3 can be seen in the following video, it’s the perfect scenario for this kind of game, 100% repeatability.

The settings used were: Ultra Preset + Enhanced Alpha to Coverage (transparent AA applied on foliage, DX11 only) + No MSAA + SSAO (DX11 only) + 0 Buffered Frames.

FPS was recorded using the Beepa Fraps 3.5.9 for 60s.
CPU usage was recorded using Windows Performance Monitor. Processor(_Total) with a sample interval of 1 second.
GPU Usage was recorded using MSI Afterburner 2.3.0.

Far Cry 3 is updated to 1.01

Test System Specifications
Test Hardware | Far Cry 3 Benchmark
Processor

Intel Core i5-2500K (Sandy Bridge)

3.3 GHz, OC = 4.5 GHz, 6 MB L3 Cache, power-saving settings disabled, Turbo Boost disabled.

Motherboard MSI P67-C43-B3, Intel P67 Chipset
Memory 2 x 2 GB DDR3 1600MHZ
Hard Drives

WD 500 GB SATA III (OS)

Samsung 750 GB Sata II (Game)

Graphics Card Sapphire HD6950 1 GB
Power Supply Corsair TX 650 W
System Software And Drivers
Operating System Windows 7 x32 SP1
Windows 7 x64 SP1
Windows 8 Pro x32 build 9200 (RTM)
Windows 8 Pro x64 build 9200 (RTM)
Driver AMD Catalyst 12.11 Beta 8

Operating Systems Comparison

Far Cry 3 works exactly the same on both Windows 7 (x32/x64) and Windows 8 (x32/x64). Windows 8 is starting to look more like a gaming OS even though it was designed to be nothing like that.

Far Cry 3 Operating Systems Comparison DirectX 9
Far Cry 3 Operating Systems Comparison DirectX 11

CPU and Cores Comparison

In DirectX 9, there’s absolutely no CPU bottlenecking, even when using only 2 cores (the game doesn’t start with 1 core) but if you look at the CPU/GPU Usage Table you can spot a small inconsistency. Even though the CPU Usage is well bellow 100%, the GPU isn’t used at full capacity. It’s not as bad in Assassin’s Creed 3, in fact it’s nothing like AC3 but those few percentages could mean more than a couple of extra frames rendered per second.

Far Cry 3 CPU and Cores Comparison DirectX 9
Far Cry 3 CPU Usage DirectX 9

Not all of you know that Far Cry 3 supports DirectX 11 Multithreading (more info). Unlike nVidia, AMD’s Catalyst Drivers don’t exactly support this feature (maybe because it’s not being used so often by devs, Civ 5 is one of the VERY few games that make use of it) but I still decided to give it a shot, see how “things” work with and without DirectX 11 Multithreading, the results are disappointing…

Basically, DirectX 11 Multithreading doesn’t work with AMD GPUs and drivers, instead of improving performance (creating more than one “rendering thread”) it kills it. You can’t penalize Ubisoft for this because this option is hidden in the xml config file (only available for nVidia GPUs in-game). If you look at Civ 5 performance difference between two AMD and nVidia Cards (similar in performance) you’ll see a huge FPS difference (100% improvement in some cases) in favor of nVidia, that’s what DX11 Multithreading does…

Again, the GPU isn’t fully utilized. On the other hand the CPU needs of Far Cry 3 aren’t very high and that’s excellent news. These CPU Cores tests have nothing to do with how performance scales with more added cores because these tests are not designed to do that. What these tests show is how much “CPU Power” the HD6950 needs for the best performance. In these scenarios the system is GPU bottleneck (GPU bottleneck is good, m’kay!) and performance will not scale with more added cores.

Far Cry 3 CPU and Cores Comparison DirectX 11
Far Cry 3 CPU Usage DirectX 11

Far Cry 3 CPU and Cores Comparison DirectX 11 Multithreading
Far Cry 3 CPU Usage DirectX 11 Multithreading
DirectX 9 DirectX 9 DirectX 11 DirectX 11 DirectX 11 MT DirectX 11 MT
CPU Usage GPU Usage CPU Usage GPU Usage CPU Usage GPU Usage
2 Cores 81% 94% 84% 89% 44% 26%
2 Cores OC 63% 96% 71% 94% 32% 33%
3 Cores 54% 96% 59% 93% 43% 54%
3 Cores OC 43% 97% 46% 96% 28% 47%
4 Cores 41% 96% 46% 93% 54% 93%
4 Cores OC 34% 97% 36% 95% 51% 94%

GPU Overclock

Because the GPU isn’t fully utilized, overclocking it won’t make a big difference but it still makes sense to do it on highly overclockable GPUs, just don’t expect miracles.

Note: Because 12.11 Beta 8 crashes my whole system while running Far Cry 3 (DirectX 9 only) with the GPU overclocked I was forced to use 12.10 WHQL for the DirectX 9 GPU Overclock test.

Far Cry 3 GPU Overclock DirectX 9
Far Cry 3 GPU Overclock DirectX 11

AMD Catalyst Drivers Comparison

Far Cry 3 is a AMD Gaming Evolved title meaning that the developers worked close with AMD’s Catalyst Driver Team to ensure the best performance on the game’s release date. All three tested drivers showed the same performance, like most AMD titles on release date. If you have Crossfire you might want to use the latest CAP (Catalyst Application Profiles 12.11 CAP 1) because it includes a Far Cry 3 profile that should increase CFX Scaling.

Far Cry 3 AMD Catalyst Comparison DirectX 9
Far Cry 3 AMD Catalyst Comparison DirectX 11

Settings Comparison

As you can see, performance differences between using DirectX 9 and DirectX 11 are almost inexistent, especially on higher settings. All the D3D 11 features (SSAO, Alpha to Coverage) don’t seem to affect performance very much. When using lower quality settings, using DirectX 11 is actually faster than DirectX 9, a direct effect of the game’s inefficient usage of the GPU which is more evident on lower settings.

Leaving that aside, Far Cry 3 has a multitude of graphical settings, most of them are addressing foliage, which let me tell you looks amazing. But anything below very high settings makes it look like it’s made out of cardboard, especially trees.

Lacking support for D3D10/10.1 means that all DirectX 10 or 10.1 GPUs will automatically default Far Cry 3 to DirectX 9. Dunia 2 fully supports DX10 but the developers didn’t use it, doesn’t make much of a difference because the game looks about the same in DX9 as in DX11, the difference between the two are in fine details.

Far Cry 3 Quality Settings Comparison DirectX 9
Far Cry 3 Quality Settings Comparison DirectX 11

Image Quality Comparison

DirectX 9

Far Cry 3 IQ Comparison DirectX 9

DirectX 11

Far Cry 3 IQ Comparison DirectX 11

Conclusions

It’s no secret that Far Cry 3 is a console port, a rather good one. The performance issues the tests revealed are nothing but fixable by a patch and a better display driver. Taking a look back at Assassin’s Creed 3, the problems that game has haven’t been fixed yet; though Far Cry is actually playable by decreasing the quality settings whereas AC3 isn’t…

The DirectX 11 addition is a welcomed one, unfortunately it doesn’t make a huge difference in how the game looks because the D3D11 features added aren’t many, as I said earlier, the differences are in small details. How something looks (good,bad) is more of a personal area, extremely hard to quantify, it’s up to you to judge this.

On the other hand I’m happy to report that the drops in FPS are very small, the average difference between minimum FPS and average FPS is of about 20% and this makes the game feel very smooth.

The Good:

  • Works the same on both Windows 7 and Windows 8.
  • Doesn’t require a very powerful CPU, 2 cores of a 2500k are enough.
  • Great Graphics on higher levels of quality; above decent on lower settings
  • Massive variety of quality settings.
  • Addition of DirectX 11 features while still supporting DirectX 9.
  • Small difference between minimum and average FPS

The Bad:

  • Far Cry 3 fails to fully utilize the GPU.
  • Crashes in DX9 with the GPU OC’ed (Catalyst 12.11 Beta only)
  • MSAA is locked to DirectX 11
  • Lack of DirectX 10/10.1
  • Murderklok

    I seen there was a critical day one patch for PC, does that address any of the issues?

    • http://benchmark3d.com/ Johnny 3D

      The 1.01 patch is installed, the changelog says about improving performance and stability but since I haven’t tested the game without the patch I don’t know if it actually fixes anything…

  • John

    My whole system crashed too when i was using DX9 and my 6950 overclocked, happy to see that the problem is not only on my system. Anyway i hope new drivers will come out soon..

  • Nandiman

    Read on G3D that AMD will be releasing a new driver next week for Far Cry 3 (possibly other games too). Any more details on these?
    Also, DX9 does not crash for me with 12.11 beta 8 (oc’d 7870 1100/1250), and it improves fps tremendously over dx11. I found that under dx11 AA has a huge performance hit, even 2xAA can eat up to 20 fps over dx9. (dx9 avg fps 70; dx11 0xAA avg fps 60; dx11 2xAA avg fps 40; dx11 4xAA avg fps 30).

    • John

      Link to the news please?

      • Nandiman

        Highspeed123 already added the link. Hope it’s true, hope it adds a nice perf boost or at least mitigate the impact of the AA because 2xAA should not have such a huge hit.

    • http://benchmark3d.com/ Johnny 3D

      Couldn’t find any info on that but that “supposed” driver should have come at least a week ago. I’ve been seeing Far Cry 3 slides on Catalyst install screens since 12.10 WHQL. Nvidia already released a driver that makes a difference (http://www.geforce.com/whats-new/articles/nvidia-geforce-310-64-beta-drivers-released).

      If you’re going to run advertisments for a game inside your own software, make sure your software is ready for the game :)

  • Jason_Burne

    Hello, did you tested the DX11 while launching the DX_11 dedicated exe included in the game folder ?

    Heard many saying that selecting DX11 doesn’t work, need to laucnh the dedicated exe inside the game folder named : farcry3_d3d11.exe to have the Dx11 inside the game.

    A simple afterburner check will tell you is the game effectivly runs in D3D9 or D3D11 modes

    When i compare your results between DX9 and DX11 it seems really weird to see so small differences, can you please check it again after launching the game using the farcry3_d3d11.exe launcher please? Thx

    Btw i really like this site, great job done ^^

    • http://benchmark3d.com/ Johnny 3D

      Yep, I triple checked in fact:

      - For DX11 I used the DX11 exe, for DX9 I used the other exe.

      - Afterburner confirmation of the API.

      - Fraps confirmation of the API (it shows the FPS on a black background when in DX10 or DX11).

      So yeah, I’m pretty sure I’ve done this right:). The differences between the two APIs is further reduced because I used SSAA instead of HDAO which would have decreased FPS by a few frames.

      I played a bit yesterday and today, had a few crashes at certain missions with DX11 on but with DX9 the game didn’t crash…weird

      • Jason_Burne

        Thx for taking the time to answrer and confirming it, well it’s weird, i will see soon in few hours as I got it downloading right now, for now will test on a HD5870 and monday my new 7970 should be shipped to my home, with next ATI drivers monday i hope to be able to play the game full utlra but with low AA levels at 60 fps, worried about the DX11 mode i heard many have crashes on cut scenes, i saw so many different things said on forums about the game performances, some say it runs fine on ultra with a basic 7870 some have problems on 7950/70 and gtx 680 saying it runs really bad, as we don’t have all the same perception of ” visual lag” and what is acceptable to our eyes so it’s a bit a pain to see clear in all this … 2 hours more to wait the end of the download on the crappy Uplay thing and i wil lsee all this by myslef .

        Again thx a lot for all and keep up the good job you do on this site ^^

  • skr13

    Good Job Johnny!
    I´m glad to see that in DX9 performance is similar to DX11.

  • Nandiman

    New drivers are up people. Get ‘em while they’re hot. http://support.amd.com/us/kbarticles/Pages/AMDCatalyst1211betadriver.aspx

  • Guest

    Of course the GPU is not fully utilited! It is a 1GB graphics card you FUCKTARDS! That GPU will be thrashing due to the memory limitations and not running efficiently! Far Cry 3 goes through up to 2GB (2560×1600 8xAA) and pretty much starts at 1GB with the bare minumum settings! Please redo this dodgy review with sensible hardware and a multitude of different graphics hardware. The only thing this article is useful for considering the unsuitable hardware and the lack of comparison GPUs is wiping my arse on the toilet if I printed it out! No offence but this is worse than the time that W1zzard at TechPowerUp replied to my email about him testing using in i7-920 (piece of crap frankly) saying that he did not use an i7-980X because he believed in using hardware people could realistically afford! Well so much for accurate benchmarks! GPU benchmarks are supposed to be perfectly representative of the GPU’s performance itself and not other hardware. It is up to the user to decide whether his or her CPU is up for the job!

    BTW love Far Cry 3, it makes Crysis 3 with GPU-melting graphics look crap. And besides I do not feel like using Origin! But FC3 is so clean and pretty with vibrant colour whereas Crysis 3 looks dirty and gloomy.

    • http://benchmark3d.com/ Johnny 3D

      Actually the VRAM stays at about 750-850MB usage, so yeah. I don’t know W1zzard at TPU but I really admire him, actually replying to an email written by someone with your language is really something. Next time you have something “smart” to say, use decent language and you might get more from me…

      • Masteron

        hey please ignore him I love the reviews of this site because it targets basically everyone not just those who have got very high end systems. I have got a Dual Core system with a 460gtx graphic card and many people in my country even use either Quad Core or Dual core (I am from South Asia) it’s really hard to find reviews have been done for my GPU/CPU so this is really valuable to me. thanks anyway I enjoyed the reviews of both saints row 3 and this and they helped me a lot to tweak those games. =)

        • Guest

          It is not about whether people have low or high-end systems. It is about measuring the performance of the actual product being reviewed. The results should not reflect the performance of any other component, should they? Because if they did then they would not be accurate. If you have an outdated CPU you then have to assess whether it is capable of keeping this performance or not. And besides, this benchmark has multiple CPUs so that helps your problem out.

          But, whether you have high-end hardware or low-end hardware, the purpose of the review is to benchmark only the performance of the product being reviewed. Since the results are in FPS it is therefore graphics performance being reviewed apart from the section where multiple CPUs are in question. But, is the graphics card adequate for unbiased results? If you used an AMD Radeon HD 5450 then all of the CPUs would yield extremely similar results. So again, here you must use the best hardware available reqardless of whether your audience has high-end or low-end PCs.

      • Guest

        More from you? I didn’t get anything from you in the first place so perhaps you could start by giving at least something before you start being a pain too.

        You have to admit that the point of performance benchmarks are to have comparison hardware and to have it all set up so that the performance measured is only representative of the product or hardware. This review does not satisfy the game or the comparison hardware. Far Cry 3 cannot run properly with only 1GB of VRAM. Nowdays 1GB VRAM is only found on low-end cards. It was two generations of graphics processor ago that 1GB was recent and thus will be insufficient to base a whole review of a demanding game on. Thus we cannot trust that the benchmarks values are accurate.

        As for W1zzard, you cannot have the piece of hardware limited by another component. It would be like benchmarking an SSD by measuring the write speed of a file transferred from a 7200RPM desktop HDD because that is what people can realistically afford. That would completely destroy the validility of the review. That is why good reviews that I have seen have been use a RAM disk to completely eliminate any possible bottleneck in supplying the file. It is the same deal with the inadequate CPU, even if not to the same extent. Do you agree with me?

        Okay.

        I am offering decent language and something smart to say which is constructed around the actual purpose of writing benchmark reviews and what is required to fultil this purpose adequately. What do you have to offer so far?

        • Guest

          Apparently nothing to offer which means when added to the usefulness of the review it is still zero.

  • http://www.facebook.com/Drejeck Emil Johan Majani

    Hi, I have a 6950 2GB, my stable overclock is 880mhz for the core and 1325mhz for ram.
    Some games doesn’t like when I push to 900mhz, but FPS games which usually are very well made got no problem @ 900mhz core and 1375mhz ram.
    Unfortunately @ 915mhz gets unstable, since my Sapphire Dual Fan HD6950 2GB has locked voltage at 1,100.
    The way I got easily to this overclock was to enter the registry key for the 6970 and then CCC and MSI Afterburner started with the card set at 6970 values (880mhz core 1375mhz ram).
    This is 110% overclock both core and ram. Should yield from 8 to 10% fps more.

    Remember to do not run 64bit OS with 4 or less GB of ram, it’s good to use a ramdisk, which Dataram Ramdisk allows to use up to 4GB for free, to get the full amount of ram at disposal.

    Upgrade as soon as you can to 8 GB at least, if you want to get serious with SSD also, and crush all bottlenecks go to 16GB and open a 8GB Ramdisk drive, I would never have another rig without it (Asrock motherboards got a Ramdisk software for free).

    I see you use 1600mhz memories, i’d suggest low timings like 8 8 8, IPC gets very similar to 1866mhz 9 9 9, costing less. Crucial Ballistix Tactical cost less and it’s way better because of the low profile, instead of those uselessly tall Corsairs Vengeance. I got 4x4GB Patriot Viper Xtreme 2000 @ 1866mhz 9 11 9. This are low profile due to seriously made copper heatsink instead of aluminum, which allows me to use a NH D14 on a i7 2600k @4ghz. I don’t like to get over 45°C on CPU and over 55° on GPU, but 10°C more could be ok for a really short time like benchmarking.

    Any way a 100% GPU load is very unlikely to happen in the real world, especially because no one plays without Vsync, but it gets hot faster without probably give any FPS. It’s good that you pointed it out, but not so relevant. I get a 100% load in Street Fighter X Tekken benchmark without Vsync 0 MSAA but over 360fps for 5 seconds, maybe less. And my GPU fans are set like this: from 1 to 60°C goes from 1 to 60% fan speed. From 61°C to 70°C goes from 61% to 100%. Furmark 15minutes stops at 67°C.
    Beta drivers are really bad to try overclock, always go for the stable release.
    For what I see, DX 11 yielded more FPS probably because of optimization instead of loading with effects. Gamespot awarded the game with “Technically Proficent Graphics”.

    One day when I was trying to OC the 6950 I got stuck on 725mhz for the core. I had to reboot.

    • Guest

      10% overclock means adding 10%. 110% would be like running a 1000MHz GPU at 2100MHz. Good job though with the registry tweaks! And may I recommend Sapphire TriXXX with its overclocking and custom fan controls? Awesome program!

  • Guest

    This is not accurate! How do you think my AMD Radeon HD 7970 3GB 1200/1750 works at 100% usage in the game with an Intel Core i3 530 @ 4.12 GHz. [1080p, ultra preset and 4xAA]. So people, overclocking your GPU will help A LOT even with a poor CPU.