Far Cry 3 Performance, Benchmarked

[roghero]

[citation][nom]Cleeve[/nom]Probably not in the same area we benchmarked at 1080p.[/citation]

Why would anyone buy a 660Ti and not use anything above 1080p? Don't just assume something like that especially with the kind of graphics and components that are being used these days. Unless I am mistaken than I am pretty sure 1920x1080 is 1080p. So yeah 660Ti, 1920x1080, everything maxed out. Well I guess not everything is maxed out. I have Vsync set to just 1 frame. Maybe my fraps is just reading wrong but it looks pretty smooth to me at a constant 60 unless it's raining than its 55-58.

 

[cleeve]

[citation][nom]roghero[/nom]Why would anyone buy a 660Ti and not use anything above 1080p? [/citation]

I think you misunderstood.

You said:
"I have an MSI GeForce GTX 660Ti PE and I am getting constant 60 fps with everything on high at least according to fraps."

My answer (paraphrased to make it clearer):
"You're probably not benchmarking at 1920x1080 in the same area we used. If you were, your frame rate would be lower."

 

[silverblue]

Also, BTW, what makes you guys suspect that it's limited to (effectively) using just up to 4 CPU threads?

It's been mentioned throughout the thread, but without a low details test, I'm not sure if we could properly measure it. Right now, what we have looks GPU limited.

 

[ojas]

[citation][nom]silverblue[/nom]It's been mentioned throughout the thread, but without a low details test, I'm not sure if we could properly measure it. Right now, what we have looks GPU limited.[/citation]
For me it's mostly been that the FX-8350's performance has been right around where i'd expect an A10 with the same clocks to perform, and the scaling from the pentium to the i3.

But yeah it's too GPU bottlenecked to be sure of what's happening above 4.

 

[ojas]

EVERYONE:

This is possible by altering your registry, but for those less tech savy, this program can do it easily.

This either will have no effect, or a HUGE effect. Ive heard of some amd Phenoms going from 20 fps to 60.
Most newer intels wont see to big of a difference(I hear i5s dont get any, but there are some gains for i3s and i7s), but amd processors could see a huge difference.
Microsoft even admitted that core parking on amd CPUs in windows 7 was down right broken.
This could very well gain you 40 fps, although I expect that if you do get fps, it will be more like 10-30.

Be warned, although Im on a laptop and saw 0 increase in heat, Ive heard of laptops getting extremely hot with this, so be careful. Desktops should be fine.

http://forums.steampowered.com/forums/showthread.php?t=3029182

You think FC3's facing the same problem (quote relates to PS2)? I'm extremely clueless about this so forgive me if MS has already fixed this.

 

[army_ant7]

It's been mentioned throughout the thread, but without a low details test, I'm not sure if we could properly measure it. Right now, what we have looks GPU limited.

That's why I asked why you guys thought it was limited to just 4 threads because you guys mentioned it multiple times thoughout the thread, unless you mean the reason why you guys say so is there. If that's the case, I might've just missed it.

ojas mentioned his/her reason why, though I guess unless we have tests... The FX-8350 did substantially beat (at least with minimum frame rate) the FX-4170 in the medium settings test, and I'm inclined to think that it's not only because of the architectural improvements. If ever, it could just be that it's utilizing the 4 FPU's of the 4 modules of the FX-8350, or not...

Again, I agree with GPU bottleneck, and that a low settings test may be in order.

 

[silverblue]

To be honest, I don't have the slightest idea whether it's limited to four threads or not, though with an i3 performing on the same level as an i5, it might be even fewer. The fact that the heavily overclocked i7 does nothing for framerates definitely points to a GPU limitation, though worryingly, the GPU bottleneck is alleviated at least partially with the 8350, however we do know the FX has lower IPC than SB let alone IB... so a small overclock would probably put that on a par with the others.

 

[ojas]

[citation][nom]silverblue[/nom]To be honest, I don't have the slightest idea whether it's limited to four threads or not, though with an i3 performing on the same level as an i5, it might be even fewer. The fact that the heavily overclocked i7 does nothing for framerates definitely points to a GPU limitation, though worryingly, the GPU bottleneck is alleviated at least partially with the 8350, however we do know the FX has lower IPC than SB let alone IB... so a small overclock would probably put that on a par with the others.[/citation]
What about the pentium?

Actually you know Don actually missed the mark with the CPU test.

There're usually two charts, using the same processor, one shows core or thread scaling, and the other shows scaling with clock rates. Tom's used to do it this way themselves.

Comparing different CPUs will almost always throw up this problem.

examples:

 

[cleeve]

[citation][nom]ojas[/nom]What about the pentium?Actually you know Don actually missed the mark with the CPU test.There're usually two charts, using the same processor, one shows core or thread scaling, and the other shows scaling with clock rates. Tom's used to do it this way themselves.[/citation]

I was the one who first used the clock rate/number of cores method. I stopped using it on purpose.

The problem is, it represented the theoretical strength of the architecture more than it represented what a person could actually buy.

In the end, after acknowledging some concerned forum posters, I agreed it makes sense to compare actual consumer-purchasable hardware than it does to compare synthetic tests with clock rates and configurations that don't necessarily correlate with shipping products. These are, after all, game performance reviews and not CPU architecture comparisons.

Now I test with actual processors, and I think it better represents what buyers can expect from their rig.

 

[silverblue]

FX seems to benefit from clock speed far more than Core in FC3; shame you use a lot more power getting to the same level. As the following link shows, you'd need to overclock an 8350 to 4.5GHz to match a 3770K at 3.5GHz in very high details.

http://www.techspot.com/review/615-far-cry-3-performance/page6.html

Of course, that says nothing for medium or low detail modes...

 

[ojas]

[citation][nom]Cleeve[/nom]I was the one who first used the clock rate/number of cores method. I stopped using it on purpose.The problem is, it represented the theoretical strength of the architecture more than it represented what a person could actually buy.In the end, after acknowledging some concerned forum posters, I agreed it makes sense to compare actual consumer-purchasable hardware than it does to compare synthetic tests with clock rates and configurations that don't necessarily correlate with shipping products. These are, after all, game performance reviews and not CPU architecture comparisons.Now I test with actual processors, and I think it better represents what buyers can expect from their rig.[/citation]
Aha. Ahhhhhhh. Ok, now i see what's happening.

In our heads, we're trying to figure out
A) What's going wrong with the 8350, which leads us to -
B) How many threads FC3 is using, what's affecting FC3 so much, etc.

While you're saying, hey, there's this awesome new game, if you want to play it, here's what you'll need.

Recognizing that your method does indeed make more sense, I'm not going to say "oh noes plz change it!!!"

BUT. But, but. Butt. ok, no. Do you think it's possible that you could do an article on this? Take a few games, mainly CPU bottlenecked (if they're not, create the bottleneck, obviously) and see what's going on. A purely academic-ish article.

Compare Phenom IIs, Core 2, Sandy/Ivy, Bulldozer/Piledriver. Scale cores/threads/modules, all the way to 12, 8 or 4, from 1. See how HT affects a game.
Then scale clock speeds for Piledriver and Ivy only.

You could, if you want test only one game from each engine or something. I don't know if scaling with threads is an engine-specific feature or not, but if it is, you could have stuff like Crysis, Crysis 2, FC3, BF3, some random Unreal Engine III game, etc.

Some of it is pointless and redundant (games for which you've already done it, for example) and i wouldn't want you to do that again, that would just be a waste of time. But games that do scale till 4, i'd like to see how far they go.

Basically i'm suggesting a CPU oriented article that examines how games react to various CPU architectures, in the process understanding more about the games/engines themselves.

Don't know if you'll have the time for this. But seeing the nature of debates sparked by game articles, i think it's needed.

 

[army_ant7]

I was the one who first used the clock rate/number of cores method. I stopped using it on purpose.

The problem is, it represented the theoretical strength of the architecture more than it represented what a person could actually buy.

In the end, after acknowledging some concerned forum posters, I agreed it makes sense to compare actual consumer-purchasable hardware than it does to compare synthetic tests with clock rates and configurations that don't necessarily correlate with shipping products. These are, after all, game performance reviews and not CPU architecture comparisons.

Now I test with actual processors, and I think it better represents what buyers can expect from their rig.

That does make sense... Here's an idea though, if you have more time (and just if) then maybe you could modify the clocks or disable the cores of certain CPU's you have so as to emulate CPU's that you don't have on hand, just to diversify the testing bed more. Like if you have a i5-3570K, you could clock it down to the weakest Ivy Bridge i5 available, so as to let the reader know what if they bought that weaker i5, if it would be still capable. I don't think you'd really need to test any of the i5's in between since we can interpolate how they'd perform based on the clockrate of the weakest and the most powerful. Just and idea though.

BTW ojas, +1. I'd also like an article like that. It sounds like it could make for something very interesting.

 

[Guest]

Most LCD monitors function at 60 Hertz
A hertz is defined as a second at power -1
So a normal LCD monitor can only display 60 images or frames per second
Therefore it is pointless to have your video card produce hundreds or even thousands of images per second
since you will alwasy see 60

 

[Duckhunt]

Its sad that even the most powerful card can't handle all the settings maxed out.

 

[army_ant7]

Most LCD monitors function at 60 Hertz
A hertz is defined as a second at power -1
So a normal LCD monitor can only display 60 images or frames per second
Therefore it is pointless to have your video card produce hundreds or even thousands of images per second
since you will alwasy see 60

Actually, as I know, as long as you don't have V-sync on, you'll see portions of multiple frames if the game is rendering faster than the refresh rate. This is what causes tearing (having portions of multiple frames shown in one refresh). I have heard that having more FPS is still beneficial in some games like first-person shooters, assuming your reflexes are good enough to benefit from it. You can look into it if it interests you.

... How do you edit mistakes?

I bet you commented using the section under the article itself. Just scroll to the first comment on that page, and click on "View comments in the forums" or something like that, which should be right above the first comment.

 

[Duckhunt]

I might have sensative eyes. I am not asking for higher then 200hz. I do feel better at higher refresh rates on CRTs. I have investigated reflections or lighting environment that effect the display. On CRTs my eyes feel good at 85hz for 15 inch ,CRTs 21 inch CRT 100Hz. I get sore eyes on a Dell Laptop but My eyes feel good on 22inch LCD display. I am sure there is rate multiple where i will experience the same problem.

I have an IPS monitor that does 1440p but I am not sure if its runnng at 60hz because it has some techno wizdardy to enhance things to stop motion blur. Either way. I'd like any technology to make it better. Thanks for your help.

 

[BlizzardGamer]

Who actually plays their games with 5760x1080 anyway? You could not see half of the stuff going on the screen anyway.

 

[Duckhunt]

[citation][nom]BlizzardGamer[/nom]Who actually plays their games with 5760x1080 anyway? You could not see half of the stuff going on the screen anyway[/citation]

most of my buddies do. I guess the ghetto dwellers can't see anything great in surround graphics

 

[army_ant7]

@BlizzardGamer
I and my brother do. Actually, it's 6000x1080 because of bezel compensation. When he was playing Dead Space (1 and 2) on that, it was pretty awesome. He used a Logitech controller to be able to sit a bit farther away though so as to not have to turn your head (at least as much).

@Duckhunt
:lol: I'm curious as to what you said that Don (I assume) had to edit it! Hahaha! I'm guessing it was racist or something since even after the edit, it still feels a bit racist to me. :lol:

 

[cleeve]

[citation][nom]army_ant7[/nom]
:lol: I'm curious as to what you said that Don (I assume) had to edit it! Hahaha! I'm guessing it was racist or something since even after the edit, it still feels a bit racist to me. :lol:[/citation]


Just fixed his double post with a citation tag! No censorship here.

 

[army_ant7]

Oh! I see... I see... :lol:

 

[gloomfish]

These benchmarks are not matching up at all with what I'm getting in game. I have a 7870 and an i5-3570k and on Ultra I'm averaging about 55 fps.

 

[cleeve]

[citation][nom]gloomfish[/nom]These benchmarks are not matching up at all with what I'm getting in game. I have a 7870 and an i5-3570k and on Ultra I'm averaging about 55 fps.[/citation]

Then your test method is not at all matching up with ours.

 

[Gauri Madaan]

are u being paid by intel to commend them?

 

[Novuake]

are u being paid by intel to commend them?

Have you compared this review to other sites? Results are similar.