Far Cry 3 Video Card Performance and IQ Review @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,601
Far Cry 3 Video Card Performance and IQ Review - Far Cry 3 is here, and we take a full look at performance and image quality comparing several video cards. We will find out what it takes to run this beast of a game with each video card, how those compare, and look at performance with certain graphics features, and we compare image quality with those features on and off.
 
Question.

If you like SSAO better than HDAO (and I agree it looks better), then why use the later for your testing? And why did you use HBAO rather than SSAO with nvidia?
 
On the "GTX 670 vs. HD 7950 Boost vs. GTX 660 Ti" ..

Even though the GTX 660 Ti has the same clock speeds as the GTX 670, its lesser CUDA core count means we had to back off on Alpha to Coverage...
Click to expand...

Both the GTX 660 TI and the GTX 670 have the same CUDA core count. Their memory bus is what has changed. Unless I missed something and there is a new GTX 660 TI sku?
 
Stoly said:
Question.

If you like SSAO better than HDAO (and I agree it looks better), then why use the later for your testing? And why did you use HBAO rather than SSAO with nvidia?
Click to expand...

I was looking to find the highest playable settings. HBAO is closer to SSAO performance, it was the next quality down option to use, to gain performance in those scenarios. I needed to find what was necessary to get it playable, so that's what I've informed the reader of. AO is subjective, naturally, technically HDAO is the better technology, but my opinion was SSAO looks better, others may agree or disagree. When showing you what is playable, I have to take all settings into consideration and try to maintain the highest settings. So I showed you both ways, I showed performance of all AO methods, and let you decide what to try and given my suggestions. I always like to use the phrase "as the developer intended" and the developer intended for the game to be seen with its highest in-game quality settings.
 
martinmsj said:
On the "GTX 670 vs. HD 7950 Boost vs. GTX 660 Ti" ..



Both the GTX 660 TI and the GTX 670 have the same CUDA core count. Their memory bus is what has changed. Unless I missed something and there is a new GTX 660 TI sku?
Click to expand...

Actually the difference is the ROP count and memory bus, 24 ROPs on 660 Ti, 32 on 670, and different bus widths. I think my brain took vacation when I was typing it out and I used cuda cores, let me fix that.

/edit Fixed
 
Nice review.

I was surprised of the demands this game put on my cards.

Neither my GTX 670 SLI nor my HD 7970 CF can run this game in Surround/Eyefinity.
I have to play this on on a single screen, something I haven't had to do since trying Metro 2033 on CF 5870s in EyeFinity when the game was first released.

So I either have to go three way or wait until next generation to play in Eyefinity.:(
 
Nvidia has proved once again that when it comes to multi GPU they are the best. AMD has always done well with single GPU performance. its multi GPU performance which has been and continues to be a problem.

I think the developer could have done a much better job with texture quality in the PC version. some of the places in the game show poor texture quality. BF3, Max Payne 3 had some of the best texture quality seen in PC games. Also HDAO resembles real life lighting most in that even shadowed regions have clear visibility. In the first screenshots the logs of wood in the HDAO screenshot are visible whereas the SSAO is almost pitch black. HBAO has little bit more visibility. But HDAO is the best and thats how it normally is in real life.

here is an interesting comparison of Crysis vs Farcry 3

http://www.youtube.com/watch?v=E3ohqbEn1v8

the original Crysis is still the benchmark of how much a developer can push the limits of PC gaming graphics quality. for a 2007 game it still looks damn good.
 
Last edited:
Brent_Justice said:
Actually the difference is the ROP count and memory bus, 24 ROPs on 660 Ti, 32 on 670, and different bus widths. I think my brain took vacation when I was typing it out and I used cuda cores, let me fix that.

/edit Fixed
Click to expand...

GJ Brent! I honestly thought they did something to the 660 line. I haven't been keeping up with video cards as much. Thanks for the info. I had it in my head the width was the only difference. Good to know!
 
from the conclusion page

"The GeForce GTX 680 and Radeon HD 7950 with Boost video cards seemed to be on par in this game"

i'm pretty sure that is suppose to say 670.

anyway.. i'm actually rather surprised the game is this hard on GPU's. Didn't think i would see a game stress out these cards in 2012.
 
raghu78 said:
Also HDAO resembles real life lighting most in that even shadowed regions have clear visibility. In the first screenshots the logs of wood in the HDAO screenshot are visible whereas the SSAO is almost pitch black. HBAO has little bit more visibility. But HDAO is the best and thats how it normally is in real life.
Click to expand...

So I guess Brent did right.

I like SSAO better cause I feel it looks more natural.
 
raghu78 said:
Nvidia has proved once again that when it comes to multi GPU they are the best. AMD has always done well with single GPU performance. its multi GPU performance which has been and continues to be a problem.

I think the developer could have done a much better job with texture quality in the PC version. some of the places in the game show poor texture quality. BF3, Max Payne 3 had some of the best texture quality seen in PC games. Also HDAO resembles real life lighting most in that even shadowed regions have clear visibility. In the first screenshots the logs of wood in the HDAO screenshot are visible whereas the SSAO is almost pitch black. HBAO has little bit more visibility. But HDAO is the best and thats how it normally is in real life.

here is an interesting comparison of Crysis vs Farcry 3

http://www.youtube.com/watch?v=E3ohqbEn1v8

the original Crysis is still the benchmark of how much a developer can push the limits of PC gaming graphics quality. for a 2007 game it still looks damn good.
Click to expand...

Not going to lie, but Crysis looks far better in comparison while having better performance.
 
Wow! So you have to have a GTX670 or better to play this on max settings at 1920x1080! That is awesome! Means there are some developers out there pushing the graphics envelope but best of all it means I finally have a reason to upgrade! :D
 
I play on Quad-CF 5870 2GB Matrix and have little problems with micro-stutters.

Some how Tri-CF helped me a lot with micro stutters, and now i got couple of months a go for 300 NOK ($55) a 4th Matrix for Quad-CF, to bridge me over till the 8970s come out. (tho the last one dose not have a waterblok ;-)

When i got my CF 5870s, the micro stutters ware so bad i almost put my card up for sale, the i was reading a post from a guy on the WSGF forum that for him Tri-CF fixed the problem, so i also gave it a try, thinking i could always send the card back, but for me it also fixed most of the problem, certainly to a degree it was really playable again.

With Tri-CF the stutters ware a lot less, with Quad-CF now, they are almost totally gone, and its about the same as on my friends PC that has 4GB 680 SLI setup.

Have you ever tried to test the micro stutter difference between CF, Tri-CF and Quad-CF?

From the consensus what most people that have Tri or Quad-CF is that the extra card(s) give the buffer more time to be ready to display the next frame.

This is the theory behind it is that every card got more time to get ready for the next frame.

Game @ 50 fps (taken 50 as it a round Nr), then the card have 20ms time to display a frame

Whit 2 cards every card got 40ms to display a frame.
Whit 3 cards every card got 60ms to display a frame.
Whit 4 cards every card got 80ms to display a frame.

The extra card(s) give every card more time to get ready, and so lessen the stutters.

I ask Vega how has a kick-ass OCed 3960X + Quad 7970s, and he also has no stutters, and when he gets them, just like me he can uselay get rid of them by playing with the settings.

player-x said:
One question Vega.

Do you have problems with micro stutters with Quad-CF?

[...]

Vega said:
In BF3, everything is smooth. In GW2 though I've noticed some settings introduce some stutter. I can't remember if it was shadow quality, or shaders or similar. Sometimes you have to play around with settings as one "thing" can introduce stutter.
Click to expand...
Click to expand...


Wonder if [H]ard|OCP is willing to test it out, how Tri and Quad-CF/SLI will help the game.
coool.gif
 
2x 7970's here, and I am getting a much smoother and more playable experience when I disable crossfire. Even with the new AMD driver + CAP. Fail!
 
considering an article that came out on the web that was saying even single amd cards stuttered, i am glad to hear it is not so and will only partake in hardocp articles from now on.

great review... worries me on how much i'll have to turn the graphics down one i get to playing farcry xD
 
Thank you so much for doing this! You really sold me when you said this was the pc game you had been waiting for. I can't wait to try this out on my 60" 240hz led tv.
 
Didn't think I saw this in the article, but settings Post-Process to "Low" will net you upwards of 7fps and gets rid of the over-the-top blurring.

Despite many many tweaks / latest beta drivers I'm still getting stuttering with my 6870x2 @ 1920x1200 no matter the gfx settings. Some sites have suggested using a 3rd party program to force Triple Buffering (without v-sync), and while this helped, it caused graphical issues.

Also, there are some manual ini tweaks you can add to up the graphics fidelity another step further, BUT the game apparently updates to ini randomly while playing, and will reset them if you don't set it to read-only.
 
I am very curious about Ram usage during the tests. The recommended specs say 8GB. Does that mean it uses a 64 bit executable? Is it possible my 6GB won't be cutting it anymore?
 
lilbabycat said:
Didn't think I saw this in the article, but settings Post-Process to "Low" will net you upwards of 7fps and gets rid of the over-the-top blurring.

Despite many many tweaks / latest beta drivers I'm still getting stuttering with my 6870x2 @ 1920x1200 no matter the gfx settings. Some sites have suggested using a 3rd party program to force Triple Buffering (without v-sync), and while this helped, it caused graphical issues.

Also, there are some manual ini tweaks you can add to up the graphics fidelity another step further, BUT the game apparently updates to ini randomly while playing, and will reset them if you don't set it to read-only.
Click to expand...

I hate some of the post-processing effects used in games. For Nexuiz, as an example, I left all of that crap off. I'm not looking for "cinematic" graphics - I just want pleasant textures, specular, HDR, and the like. Image "enhancements" can really get on my nerves sometimes. Glad it saves some FPS. :)
 
Outstanding review, Brent's gone to the wall for us with this one. Liked his explanations on the special considerations page too but some of the screenshots came up a bit small for me to see the difference. Was considering an 660Ti upgrade for my folding farm ( I game just a l'il too ;) ) but now it looks like it'll need to be 670. Can't wait to play it!

Kyle : get Brent a round of beer and tamales for us :D
 
wow the 680gtx doesnt seem that powerful!! cant even play this game in max settings 60fps at 1080p reso
 
Weird here's my system

I5 2500k at 4.4ghz
Gigabyte OC 660TI with a 105mhz memory OC boost over the manufacturer boost

FC3 settings

1080p
Frame at 1
VSYCH OFF
MSAA : 2x
ALPHA to Coverage : Enhanced
Setting at ULTRA PRESET

Playtime with FRAPS

Frames, Time (ms), Min, Max, Avg
38809, 857725, 13, 58, 45.246

The MIN FPS is the cut loading time i had 2 times, and my estimate is more 38FPS than 45 since i played 2 hours before that with FRAP on top just for fun and the game over between 32 and 48 fps most of the time. Vehicle travel seems to eat most FPS

In that playtrought, i drove vehicle, liberated an outpost, hunting mission, just running in the jungle etc..

Since i come from a console background anything over 30 FPS is playable (for me) and 60FPS is bonus..
 
Strange that Alpha To has such a big hit on the AMD cards. The HDAO hit on Nvidia seems reasonable since the compute performance is so much worse than AMD, but Alpha To doesn't seem like it should be such a big deal.
 
magoo said:
Neither my GTX 670 SLI nor my HD 7970 CF can run this game in Surround/Eyefinity.(
Click to expand...

Are you using Cat 12.11 beta and CAP2 12.11 ? They give a huge performance boost to my 7970 CF setup. I have been able to hit ~45FPS on 6000x1080 with everything maxed and MSAA turned off.
 
Right now, the GTX 680 certainly has a leg up on Alpha to Coverage performance in this game.
Click to expand...

7970 has the same performance with Alpha to Coverage as 680 without it, so why would the latter have a leg up? Couldn't you say that the 7970 has the leg up on non-AtC performance?
 
WTF!!! No testing with FXAA? I use a single 680GTX and am willing to bet that with solely FXAA this game screams.

FXAA is all i've used since it was introduced. It barely has any performance hit and looks terrific. I can't believe it was not mentioned once!

Please tell me the game is compatible with it, I will holding off on this until FXAA is working if that is the case.

Great review BTW! Just like the "old days" :D
 
Ok, well I looked around the web and it appears that the game is compatible with FXAA, which covers both alpha and edges, takes little to no performance hit and does look great. It will spare you 10-20 FPS along with going with SSAO, another 10+ FPS boost, the game should run like butter on most current gen cards.

I still find it odd that it was not mentioned in the review, but oh well. ;)
 
I play Crysis Wars mostly, and have seen this stuttering issue for a while; since 4870CF.

Certain settings make it seem to 'skip' a frame, and it stutters as a result.

Going to three cards instead of two makes sense, to cure a missed frame.

Lowering the overall settings works to eliminate it; IDK what specifically works.

I'm waiting for my newegg xmas present to play FC3. :)
 
they mention using "16x AF", this is forced through the card driver/control panel yeah? im using nvidia and was wondering if i should be turning that on?
 
Your review bemoans the lack of FXAA or MLAA options in the game, but why not run benchmarks with these options enabled from the Control Panel? By all accounts MLAA2.0 is a significant improvement, especially in performance on the HD7970. And I don't think I've seen any of the reviews on here test it as an option. It may not have as high image quality as FXAA or in-game AA, but it would be nice to see if the performance boost is worth it.

Other than that, GREAT review, very thorough!
 
my sli gtx 670 set up has some hiccups now and then at 1920 x 1080 with high settings , not ultra , is this normal ? 16gb ddr3 , ssd , i5 3570k @ 4.2 ghz.
 
One thing that's a bit flawed in this review/shoot-out... comparing a pre-overclocked 7970 against a stock non-overclocked GTX680. The 7970 is also $30 more expensive in which case you could have got a pre-overclocked GTX680 for the same money - which would have made for a better comparison.
 
eiren said:
One thing that's a bit flawed in this review/shoot-out... comparing a pre-overclocked 7970 against a stock non-overclocked GTX680. The 7970 is also $30 more expensive in which case you could have got a pre-overclocked GTX680 for the same money - which would have made for a better comparison.
Click to expand...

7970 GHz edition cards essentially have a "stock" clock of 1,050 MHz. 1,000 of it is "base" clock and the rest is "boost" which is only enabled when the thermals and power allow it (read: all the time).

The 680 dynamically overclocks itself (rather than the static boost that the AMD card offers), and the samples we use in our reviews tend to boost to nearly 1,200 MHz in game.

Overall, I'd call it a fair stock vs stock fight...
 
eiren said:
One thing that's a bit flawed in this review/shoot-out... comparing a pre-overclocked 7970 against a stock non-overclocked GTX680. The 7970 is also $30 more expensive in which case you could have got a pre-overclocked GTX680 for the same money - which would have made for a better comparison.
Click to expand...

all of the gpus are at stock clocks. the 7970 stock clock is 1ghz with a boost clock of 1050mhz.

where are you getting a gtx 680 for $30 less then at 7970?
 
I found that SSAO tended to get a tad over-zealous on the shading effect. I kept seeing black halos around objects that shouldn't be shaded, and the effect was so distracting that I ended up turning off ambient occlusion altogether. If only my 5850 could handle HDAO...
 
Ugh, FXAA looks terrible. I still don't get what you people see in it. Blur-o-vision.
 
You must log in or register to reply here.
Back
Top