PCPer on Crossfire problems in the Titan review

[BallaTheFeared]

Maybe the website is lying, I don't see them mentioning a 120Hz screen, that's the only way they could capture output past 60fps.

 

[Imouto]

lol dude their dvi capture gadget is a 60hz thinghie. They're doing their tests based on the amount of screen tearing. Didn't you read their methodology? That's why I said PCPer is awfully wrong about their approaching.

Just how stupid is that?

 

[Annisman*]

Imouto, you seem to have an obsession with Vsync. Enabling it makes all the problems go away ? maybe it does, but there are some people, myself included, that would never, ever, use Vsync. There are a plethora of reasons for that.

So if I went Crossfire next round, and the same problems with their drivers continued, where would I be left ? With a useless second card ?

 

[DiogoDX]

lol dude their dvi capture gadget is a 60hz thinghie. They're doing their tests based on the amount of screen tearing. Didn't you read their methodology? That's why I said PCPer is awfully wrong about their approaching.

Just how stupid is that?

So if the game is running above 60 fps tearing is expected and the frames locations will be all over the place. They test some game running under 60 fps?

 

[VulgarDisplay]

Ok so I'll try to explain what Imouto is talking about.

Vsync attempts to lock your framerate to the refresh rate of your monitor which should in theory result in a solid bar on the left side of this picture basically showing exactly 1 frame. That would be the ideal result. 1 Frame displayed to match up exactly with every 60hz cycle hence the 60fps target of Vsync.

All these graphs are showing is tearing which is the result of these runt frames. The imperceptible frame on Xfire picture will result in an artificially higher frame rate which is technically cheating, but it does not show SLi being smoother it just shows both sides exhibiting tearing which is the enemy of a smooth experience.

By their reasoning this picture of crossfire would technically be considered smoother because more of one individual frame is visible.

Instead of frame times they should be crusading for Vsync if they are so interested in smooth gameplay. Seriously they crusade for a smooth experience, but are still too caught up in the raw framerate metric which is what seems to be the major culprit for these horrible frame times. Turning on Vsync should greatly mitigate all the problems. Use triple buffering or 120hz to get better response.

 

[BallaTheFeared]

Ok so I'll try to explain what Imouto is talking about.

Vsync attempts to lock your framerate to the refresh rate of your monitor which should in theory result in a solid bar on the left side of this picture basically showing exactly 1 frame. That would be the ideal result. 1 Frame displayed to match up exactly with every 60hz cycle hence the 60fps target of Vsync.

All these graphs are showing is tearing which is the result of these runt frames. The imperceptible frame on Xfire picture will result in an artificially higher frame rate which is technically cheating, but it does not show SLi being smoother it just shows both sides exhibiting tearing which is the enemy of a smooth experience.

By their reasoning this picture of crossfire would technically be considered smoother because more of one individual frame is visible.

Instead of frame times they should be crusading for Vsync if they are so interested in smooth gameplay.

I think believe that is correct.

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Part-2-Finding-and-Defining-Stutter

 

[DiogoDX]

This radeon picture seems like how some consoles devs take care of tearing. They put the tearing in the bottom of the screen to give more smooth gameplay in games that run with less the 30 fps and can't have proper vsync.

I can't undestand where is the cheating since crossfire is rendering both frames.

 

[VulgarDisplay]

This radeon picture seems like how some consoles devs take care of tearing. They put the tearing in the bottom of the screen to give more smooth gameplay in games that run with less the 30 fps and can't have proper vsync.

I can't undestand where is the cheating since crossfire is rendering both frames.

That little silver sliver between the two large frames is technically counted as a frame also. Hence why they consider it cheating. I agree that it is cheating for raw framerate numbers, but it has little to no bearing on actual smoothness in this context.

 

[3DVagabond]

I find these results very interesting and does seem to explain [H]'s visual results of crossfire "not appearing as smooth as it should considering it's FPS".

This is why it's important to actually be able to SEE a problem and then try to assign the problem to a measured output. If this turns out to be correct, and it makes sense to me that these "phantom frames" would cause crossfire to appear slower visually than the measured framerates, it shows that frame time graphs are not showing us "microstutter" at all. Although people are lead to believe they are one in the same.

 

[boxleitnerb]

Now TechReport is somehow not to be trusted? They're reporting the opposite with their 7970 CF setup. Better frame latencies for the 7970 CF in most games.

FFS, even you were reporting your own FRAPS figures giving them some credit. Everyone at this forum gave credit to those reviews and FRAPS testing.

The thing is that they're trying to estimate the user experience when no one would use game options that would make their screen tear. Seriously misleading if not totally biased or just plain utter incompetence.

I'd like to see figures using RadeonPro, seriously.

Fraps doesn't necessarily show what you'll get in the end. I find pcpers method more in line with reality since it shows what actually is sent to the monitor, not some approximation of that. Until now, fraps is all we had, therefore it was used extensively. This has nothing to do with trust.

Aside from that, RadeonPro's frame limiter is not a solution it's a band-aid that comes with a large performance cost depending on the game, thus calling into question the use of CF in the first place. Since you could achieve similar results in terms of perceived framerate with a single card@OC.

I enjoyed the part where the exact same thing happened to everyone who bought a 690.

No it doesn't necessarily. What fraps reports and what you actually get may differ. Nvidia themselves said that their frame metering occurs after fraps reads frametime values.

 

[VulgarDisplay]

I have the solution to the ultimate smooth gaming experience. The question is are there any enterprising electrical engineers that want to help me me get something concrete patented? I'll give you 60% if I can take 40%. Your work for my idea.

Any takers? Seriously?

 

[VulgarDisplay]

I think believe that is correct.

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Part-2-Finding-and-Defining-Stutter

Watched that video and my mind exploded. So, basically all this stutter nonsense is just visual tearing and all these review sites are harping on AMD because theirs is worse.

It's 100% due to a lack of using Vsync. All those proponents of this testing are still caught up in the max fps benchmarks which is causing this whole smoothness fiasco. Do away entirely with the Maximum FPS number and use Vsync or cap to your refresh rate and the problem is solved.

 

[BrightCandle]

I wish more games used triple buffering because double buffering is a major cause for performance problems if you are running even remotely near the limit of your machine at 60 fps. Vsync isn't a perfect solution because unfortunately if your machine misses the deadline for the frame it has to wait a whole 16.6ms to delivery that frame and the old one is shown for twice as long. I have plenty of fraps charts with vsync that still show oscillations, from 16.6 to 33.3ms when vsync is on. The input lag however is quite telling when vsync is turned on and that distracts from being able to play well. There is no silver bullet:

No vsync - best latency, tearing, stuttering issues.
Double buffer - complete screens, half frame rate drops if missed, moderate latency
Triple buffer - complete screens, no artificial frame rate drop, really high latency

You can't have perfect frames and low latency, and since the amount of latency in today's graphics pipeline is still very high minimising it is really important.

 

[BallaTheFeared]

Vsync smooth AFR????

 

[boxleitnerb]

^ no input lag, no tearing, but fps drop if you cannot keep a certain level. But still, a very nice improvement if you have the horsepower.

 

[Pandora's Box]

rather than using vsync and having your fps drop to 30fps randomly, just lock your frame rate at your monitors refresh rate. No more microstutter.

 

[ShintaiDK]

rather than using vsync and having your fps drop to 30fps randomly, just lock your frame rate at your monitors refresh rate. No more microstutter.

Microstutter still happens.

 

[SirPauly]

Vsync smooth AFR????

http://nvidia.custhelp.com/app/answers/detail/a_id/3283/~/what-is-smooth-vsync?

 

[Imouto]

Fraps doesn't necessarily show what you'll get in the end. I find pcpers method more in line with reality since it shows what actually is sent to the monitor, not some approximation of that. Until now, fraps is all we had, therefore it was used extensively. This has nothing to do with trust.

Aside from that, RadeonPro's frame limiter is not a solution it's a band-aid that comes with a large performance cost depending on the game, thus calling into question the use of CF in the first place. Since you could achieve similar results in terms of perceived framerate with a single card@OC.,

All I'm saying is how utterly dumb both TR and PCPer testings are and how all of you made a huge fuss about this. Now none of you can support their methodology after revealing itself as totally irrelevant to actual game play experience when their mouth is full of smoothness.

 

[Keysplayr]

All I see in this forum is bias conflict. Nothing will ever get settled and nothing will ever get done. It is a battleground for AMD and Nvidia fans and the VERY FEW neutral folks always get caught in the muck of crossfire (no pun intended).
Shame.

 

[boxleitnerb]

All I'm saying is how utterly dumb both TR and PCPer testings are and how all of you made a huge fuss about this. Now none of you can support their methodology after revealing itself as totally irrelevant to actual game play experience when their mouth is full of smoothness.

So people who don't like vsync are totally irrelevant? Is that what you're saying?

 

[Ferzerp]

All I see in this forum is bias conflict. Nothing will ever get settled and nothing will ever get done. It is a battleground for AMD and Nvidia fans and the VERY FEW neutral folks always get caught in the muck of crossfire (no pun intended).
Shame.

This board has been in need of some heavy handed moderation until the BS stops for a long time.

 

[Imouto]

So people who don't like vsync are totally irrelevant? Is that what you're saying?

This isn't the point but I can deflect like you.

"So people who don't like their screens tearing nonstop are totally irrelevant?"

 

[Annisman*]

This isn't the point but I can deflect like you.

"So people who don't like their screens tearing nonstop are totally irrelevant?"

No, they're not, they can use your magical vsync, that's the point.

 

[wand3r3r]

Do you think the ever observant PCPer or TR caught this? Or did they just bench Titanic and show the potentially skewed results?

Just curious being they are in the pursuit of 'bettering' the information for all.

http://forums.anandtech.com/showpost.php?p=34664254&postcount=54