Metro Last Light early benchmarks on gtx 670

[toyota]

The charts at TPU didn't have PhysX enabled. And yes, drivers and OC are enough to account for that. That's exactly my point, that those charts aren't the gospel Toyota thinks they are.

AGAIN I have ran the same benchmark on my gtx660ti and gtx670 and know that they are basically in line. what the hell he gets on an AMD card thats oced has nothing to do with the fact that Kepler is still getting those basic scores in that game.

 

[2is]

again genius I have ran the same benchmark on my gtx660ti and gtx670.

And I've played the game on my 680 with a fraps counter running

 

[toyota]

And I've played the game on my 680 with a fraps counter running

and? I have spent more time testing actual gameplay then you ever have and know damn well its not smooth on max settings with on oced 670. in fact I found it irritatingly sluggish at times.

 

[2is]

and? I have spent more time testing actual gameplay then you ever have and know damn well its not smooth on max settings on a on oced 670.

Well I can only conclude you don't know as much as you'd like to believe. Seeing I didn't get the performance you think I should have gotten.

 

[toyota]

Well I can only conclude you don't know as much as you'd like to believe. Seeing I didn't get the performance you think I should have gotten.

lol yeah I got the exact performance I was supposed to. only difference is I am not delusional enough to claim it is perfectly smooth on max settings.

 

[2is]

My claim was that I get higher performance than the TPU charts suggest and that "smooth" is subjective.

 

[toyota]

My claim was that I get higher performance than the TPU charts suggest and that "smooth" is subjective.

and you have nothing to back up your claim. I AGAIN do because I have tested 660ti and 670 with various drivers and know for a fact that the numbers are right in line with what tpu shows. I have also tested the hell out of the actual gameplay. Metro is a game where the average does not mean much as it has very low and sluggish minimum framerates at times on anywhere close to max settings.

 

[2is]

You provided nothing but what you've observed. I provided nothing but what I've observed. Yet I have nothing to back me up and you do? You deserve some sort of award for that statement... Here, I found one.

http://bbsimg.ngfiles.com/1/6711000/ngbbs434d72e6e4337.gif

 

[bystander36]

As long as you don't use DoF and PhysX, Metro 2033 ran quite well. You also have to realize the built in benchmark is far more demanding than any place in game. Even the area the benchmark is done in, as there are no where near as many people to fight as the benchmark has.

Some people will still consider the game maxed without those two features, though I think people should mention it.

 

[Headfoot]

My claim was that I get higher performance than the TPU charts suggest and that "smooth" is subjective.

blackened23 and Toyota, i really feel sorry for both of you. Your life seems to revolve around tangible numbers, i came in here saying the game ran smooth and you both ridiculed me and blasted me. Really sad that the forum has trolls like u guys around. You guys are really just.... nerds with no life it seems. You do realize you're getting all riled up about the FPS in a 3 year old video game? but hey all the best to both of you. Now if you'll excuse me i'm hitting the gym, then meeting some friends for dinner. Yes, people do such things.

PS i graduated high school 11 years ago.

Minimums below 25 fps is not smooth to me. Hell you could even lower the bar and put the cut-off at 20 fps minimums. These cut off points are subjective, yes. But smoothness itself is not. Smoothness is quantifiable.

Smooth is NOT subjective. For example, look at one of the frame-rate graphs you generate from running the in-game benchmark, or look at the one previously shown in this thread. Smoothness can be shown using the delta between the peaks, in many fields the lack of 'smoothness' is called jitter and it is a very important metric to know. Perfectly smooth would be a flat line across. You could take the perfectly flat line at the average FPS and compute the integral both above and below that line. If either the positive or negative integral is a value other than 0, it is quantifiably less smooth than perfectly smooth. Just because it takes more than algebra to quantify smoothness does not mean it cannot be quantified...

Poohbear,
I love how you digress to the classic "nerds with no life" bit. Really? Are you kidding me? You are posting in ANANDTECH's GPU FORUM at this very moment. You somehow think that you are on some different level than anyone else here? Rather Ad Hominem, imo

The truth of the matter is that when asked for objective data to substantiate your claim, you instead deflected and tried to move the goalposts so the most important metric is suddenly the most subjective metric (from your perspective).

 

[Headfoot]

Here is a diagram to better explain the point:

If the absolute sum of the integrals is greater than 0 then the game is not perfectly smooth. This is mathematically true

Practically, how do you get perfect smoothness? Be able to run the game at minimum frame-rates higher than your monitor's refresh rate (ie 60 FPS minimums) with vsync on

Theoretically, if everyone (or reviewers specifically) compiled results like the absolute integral sum, we could directly compare smoothness. So it is quantifiable. As we've seen with FCAT measurements (see the Anandtech article or any recent PCPer review) though, FPS doesn't tell the whole story. So any method to objectively quantify smoothness must be much more robust than the simple diagram above. Frame-time will also affect perceived smoothness, and so will each individual's ability to perceive small differences in the delivered frame-times. It is not subjective, rather it is an objective metric where the individuals perceiving the frames also have objectively definable differences in perception performance. 'Subjective' means it is a matter of taste, such as art or music. Smoothness is instead objective with a whole lot of variables

If you really want to go crazy with it, we could perform a double-blind test of a large enough sample size of people and have them watch a series of benchmarks where each FPS metric decreases by, say, 5%. We could then have the participant report at which benchmark do they notice that it is no longer "smooth" and we could statistically determine what the human minimum, average, and maximum minimum frame rate(s) and standard deviations in between. If its a normal distribution (which it probably will be, though I can't say for sure) we could then say that any FPS number reported as 'not smooth' below the 3rd standard deviation is therefore not smooth from 99.85% of the human population. With this approach we would be averaging out the perceptive differences between people and arrive at a FPS that is smooth for 99.85% of the population and a FPS that is not smooth for 99.85% of the population

 

[toyota]

You provided nothing but what you've observed. I provided nothing but what I've observed. Yet I have nothing to back me up and you do? You deserve some sort of award for that statement... Here, I found one.

http://bbsimg.ngfiles.com/1/6711000/ngbbs434d72e6e4337.gif

no genius I have actually ran the benchmark many many times including tonight. so while you sit there and just pull crap out of your butt, I know damn well that the numbers at tpu are spot on for 660ti, 670 and 680. so no you dont have anything to come back with.

 

[toyota]

Minimums below 25 fps is not smooth to me. Hell you could even lower the bar and put the cut-off at 20 fps minimums. These cut off points are subjective, yes. But smoothness itself is not. Smoothness is quantifiable.

Smooth is NOT subjective. For example, look at one of the frame-rate graphs you generate from running the in-game benchmark, or look at the one previously shown in this thread. Smoothness can be shown using the delta between the peaks, in many fields the lack of 'smoothness' is called jitter and it is a very important metric to know. Perfectly smooth would be a flat line across. You could take the perfectly flat line at the average FPS and compute the integral both above and below that line. If either the positive or negative integral is a value other than 0, it is quantifiably less smooth than perfectly smooth. Just because it takes more than algebra to quantify smoothness does not mean it cannot be quantified...

Poohbear,
I love how you digress to the classic "nerds with no life" bit. Really? Are you kidding me? You are posting in ANANDTECH's GPU FORUM at this very moment. You somehow think that you are on some different level than anyone else here?

The truth of the matter is that when asked for objective data to substantiate your claim, you instead deflected and tried to move the goalposts so the most important metric is suddenly the most subjective metric (from your perspective).

but he is about to head to the gym and then meet friends for dinner. maybe I should post a picture of me flexing to show I have actually been to the gym. lol

 

[2is]

Well I can only conclude you don't know as much as you'd like to believe. Seeing I didn't get the performance you think I should have gotten.

We'll just have to leave it at that.

The fact you think I'm full of shit doesn't bother me anywhere near as much as my claim has bothered you.

 

[blastingcap]

Edit: never mind, I misread

 

[blackened23]

The fact that these guys still haven't uploaded a screenshot from the built in benchmarking application, and one of them claimed "Metro 2033 is a role playing game".

These facts are extremely telling. Neither of them own the game in all likelihood; of course metro is smooth at 20 fps. So normally, I see nonsense here all the time and I don't care - but this time, it's time to call someone out on their egregious exaggerated claim. And when we call them out on their BS and being FOS, you know, it's our fault. We're nerds for calling them out. We don't have lives. And they're off to the gym to hang out with their friends and booze with movie stars. Sure they are.

I highly doubt either of these guys own the game. Period. Especially anyone who claims "Metro 2033 is a role playing game.". And i'll stick by that statement since neither, STILL, have uploaded a benchmark that takes 25 seconds to run. They don't own the game. Talking out of their rear. It's pretty simple. And when confronted about it, we'll just get a circular argument about "I don't care what you think, i'm not uploading a benchmark". There's really nothing left to discuss about this.

Anyway, like I said the game is a great game. The engine is highly questionable, but I can overlook these things. In the meantime, don't make exaggerated claims about the game and performance maxed out.

 

[NIGELG]

I have not seen an enjoyable thread here for a while....until this one.

 

[SirPauly]

I didn't say 30 fps, and smooth is subjective. Some people perceive smoothness at 30 fps, others need 60 while others claim they need 100. Perhaps you don't know what the word subjective means?

Agreed! Subjective is key based on tastes, tolerances and thresholds. Ideally, would desire 60FPS consistent but usually settle for 40-45 for sustained minimum lows and if the immersion factor is dramatically raised will settle for sustained minimum lows in the 30's based on single GPU's.

AFR rules were different to me and really needed 60 FPS consistent and if the immersion factor was dramatically raised would settle for 40-45 for sustained minimum lows and needed more care and tweaking.

The key with titles with a lot of features is one can find the right balance for their subjective tastes, tolerance, thresholds and platform of choice.

 

[Black Octagon]

Well, isn't this thread one massive snore-fest? I can't believe some of the rubbish being argued over here.

1) Does anyone have anything extra to contribute about Metro LAST LIGHT, or need we continue arguing about the performance of 2033 until proper Last Light reviews and benchmarks are out?

2) Opinions aside, the world's fastest and most expensive single-GPU card only averages 39fps at 1600p when "maxing out" 2033. Conclusion = game runs like a dog. Discussion over, seriously.

3) Why gives a monkey's turd whether 2033 is an 'RPG' or not? Yes, I think it's a ridiculous choice of genre to confer to a game like this too. That doesn't make me arguing about it any less off-topic.

These boards seem oscillate between amusing and excruciating each time I log in.

 

[Black Octagon]

Practically, how do you get perfect smoothness? Be able to run the game at minimum frame-rates higher than your monitor's refresh rate (ie 60 FPS minimums) with vsync on

I like the rest of your post, but tell this quoted bit to people running 120Hz monitors. Within days of moving to 120Hz 90fps became the new acceptable minimum for me. 120fps is better. 60fps upsets me. It is far from perfectly smooth after comparing it to 120fps.

 

[Jaydip]

I am not sure what we are arguing about here?

 

[bystander36]

2) Opinions aside, the world's fastest and most expensive single-GPU card only averages 39fps at 1600p when "maxing out" 2033. Conclusion = game runs like a dog. Discussion over, seriously.

I personally don't agree with this, as it is 1 setting that makes this true. One experimental settings; advanced DoF. It's like The Witcher 2 and Ubersampling. Without it, the games run good, with it, they don't. If you rephrased that to: "Conclusion = game runs like a dog completely maxed out. Discussion over, seriously." I'd agree, butI just don't believe a game should be judged on one feature.

 

[blackened23]

I personally don't agree with this, as it is 1 setting that makes this true. One experimental settings; advanced DoF. It's like The Witcher 2 and Ubersampling. Without it, the games run good, with it, they don't.

And I don't agree with this unless you want to bust out the benchmarks to prove it. I can post some benchmarks with 680 SLI to prove my point, if you wish.

Both physx and DOF incur an outrageous performance hit in Metro 2033, but this is besides the point. This entire discussion was about their claim - Their claim was that Metro 2033 is "perfectly smooth" maxed out on a single GPU. That is not a true statement - they can argue until they're blue in the face about subjective-ness, but smoothness is quantifiable, not subjective. 20 fps can be felt in Metro 2033 - panning the mouse around at 20 fps produces noticeable stutter.

This is about their claim. Their claim was clearly not true. On top of this, hilariously, we have poohbear stating that Metro 2033 is a role playing game. If you've played metro 2033 you would understand how absolutely ridiculous and hilarious this statement is at the same time. Obviously poohbear doesn't own the game and is talking out of his rectum. And he can't produce a benchmark because he doesn't own the game.

Also, I realize this is somewhat of a tangent because the main discussion is about the sequel. Yet it may be semi relevant because Metro Last Light is using the same 4A engine. I hope they were able to fix the shortcomings in the engine, I guess we'll see soon enough.

 

[cmdrdredd]

Remember when crysis 1 ran at 1280x1024 at 27fps and it didn't have issues with gameplay?

 

[bystander36]

And I don't agree with this unless you want to bust out the benchmarks to prove it. I can post some benchmarks with 680 SLI to prove my point, if you wish.

Both physx and DOF incur an outrageous performance hit in Metro 2033, but this is besides the point. This entire discussion was about their claim - Their claim was that Metro 2033 is "perfectly smooth" maxed out on a single GPU. That is not a true statement - they can argue until they're blue in the face about subjective-ness, but smoothness is quantifiable, not subjective. 20 fps can be felt in Metro 2033 - panning the mouse around at 20 fps produces noticeable stutter.

This is about their claim. Their claim was clearly not true. On top of this, hilariously, we have poohbear stating that Metro 2033 is a role playing game. If you've played metro 2033 you would understand how absolutely ridiculous and hilarious this statement is at the same time. Obviously poohbear doesn't own the game and is talking out of his rectum. And he can't produce a benchmark because he doesn't own the game.

I added a little to the last message to hopefully make you understand.

My point is, the game doesn't run like a dog. The game runs like a dog if "completely maxed out". Much like The Witcher 2, you just don't use the setting(s) that are unreasonable and the game is fine. Don't use Advanced DoF or PhysX, which most review sites don't even test anyways, unless they wanted to push the limits. You also can't use that benchmark as a judge of the in game play, as there is no place in the game that compares to the action in the benchmark.