AMD, say it ain't so!

[BiggieShady]

Apparently, frame times logged by FRAPS differ from what you get using hardware capture.
It seems that FRAPS data, while showing micro stuttering issues with CrossFire, was actually doing AMD a favor.
The truth seems even worse:

FRAPS frame times:



Using hardware capture:



The article is at http://www.pcper.com/reviews/Graphic...formance-Tools

Does anyone has any more info on this? Why would FRAPS data differ so much for CrossFire?

[silkstone]

Urgh not this again. I was unable to notice any difference when watching the videos at all. Someone needs to post a couple of "blind" videos along with a non stutter comparison and get people to vote on the one they think is stuttering.

For example: A direct comparison between a 680 and 7850, then the 680 vs. 680 and 7850 vs. 7850 but having all of them unlabeled. Label them something like card 1,2,3,4,5,6 but do 1 and 2 side by side etc. and set up a vote to see if most people can actually notice this stutter.

Edit - Then do a second blind crossfire test. I won't believe that it is an issue until it's verified in an unbiased way.

[the54thvoid]

I'm lost for words. I play BF3 (on crossfired 7970) and get 80-120 fps on MP depending on scene/players etc. To my eyes it runs perfectly smoothly. I know it does because it was choppy on Far Cry 3 (embarrassingly an AMD sponsored title).

Given that i know what choppy frame rates look like to me, I can hand on heart say BF3 runs as smooth as any game I've played on any gfx card (And I've had more Nvidia in my life - first AMD card was a 5850, then I got a GTX 580, then i went for 7970's).

There may be more at play from a very neurophysiological level that people are also ignorant off. Our eyes and brain cobble together the images we see. We cannot register everything that flickers up in front of us so our brain makes the image tangible to us.

From the OP's source post the crossfire looks like it should be abyssmal but it is not on my system. Perhaps the more people delve into frame latency the more we get transfixed by what the computer can see and not what humans beings are capable of seeing.

I cannot deny the findings of the source post. I'm sure it is not fabricated but I'm pretty sure there will be a raft of crossfire owners on here saying how well it works for them and they do not see any on screen indication of poor performance.

On the same note, I'm running around Crysis 3 land transfixed by how good it all looks and how smooth it is. Frames might drop occasionally to 30's but generally it's good for me. Certainly no stutter (like the Skyrimgate latency posts from last few months).

Think about it. It doesn't matter what anything says about the latency if we can't see it. An invisible problem is not a problem if the visual fidelity is what matters.

Perhaps someone can do a poll and record how many folk use crossfire in BF3 and how many have stutter or issues that would tie in with the OP source.

Again - I'm not denying crossfire is NOT as good as sli but in the instance of BF3 - it works butter smooth for me.

[Anusha]

Originally Posted by silkstone

Urgh not this again. I was unable to notice any difference when watching the videos at all. Someone needs to post a couple of "blind" videos along with a non stutter comparison and get people to vote on the one they think is stuttering.

For example: A direct comparison between a 680 and 7850, then the 680 vs. 680 and 7850 vs. 7850 but having all of them unlabeled. Label them something like card 1,2,3,4,5,6 but do 1 and 2 side by side etc. and set up a vote to see if most people can actually notice this stutter.

Edit - Then do a second blind crossfire test. I won't believe that it is an issue until it's verified in an unbiased way.

you will notice the impact of MS when you are playing as a sniper. drove me insane when i played as a sniper in BF3 with a couple of HD6950's. trying to lock on a moving object made my head explode!

[silkstone]

Still, I'd like to see some scientific method in these tests. So far it's just showing me a bunch of graphs that have no meaning. I have seen the direct comparison between the cards on Skyrim, but i noticed nothing.

I'd like to have a bunch of comparisons from different games and different situations and have some real evidence that people do actually notice what the graphs are showing us and not that people just notice something because they are told that there is a stutter.

[claylomax]

Originally Posted by BiggieShady

Apparently, frame times logged by FRAPS differ from what you get using hardware capture.
It seems that FRAPS data, while showing micro stuttering issues with CrossFire, was actually doing AMD a favor.
The truth seems even worse:

FRAPS frame times:

http://www.pcper.com/files/review/2013-02-22/fr-5.png

Using hardware capture:

http://www.pcper.com/files/review/2013-02-22/fr-2.png

The article is at http://www.pcper.com/reviews/Graphic...formance-Tools

Does anyone has any more info on this? Why would FRAPS data differ so much for CrossFire?

This is the original article, I think.
http://techreport.com/review/21516/i...e-benchmarking

[BiggieShady]

Originally Posted by silkstone

bunch of graphs that have no meaning

You should notice on fraps graph minimum is around 5 ms and on the frame capture graph minimum is almost 0 ms. That means some frames are only couple of pixels high (with frame times near zero) so there is even larger impact on boosting the average fps metric. They calculated the perceived fps which is around what a single card gets.

That is what they say, but I also hear different experiences by people who use CF.

I do agree, we need a scientific deconstruction of the experiment.

Originally Posted by claylomax

This is the original article, I think.
http://techreport.com/review/21516/i...e-benchmarking

Yeah, they initiated whole frame latency thing. I like the colored bars pcper has that show "runt" frames, though their latency graphs suck (tr's are much better imo)

[Kaynar]

My guess is that FRAPS is a software therefore it is a process in windows and it is subject to alot of "interference". Maybe setting it to be high priority could help?

To that you need to add random system components (including PSU quality) that produce random differences every time and how much the game was optimised, particularly when talking crossfire. Maybe the "hardware capture" measures the frames at a very early stage before they are "broadcasted" to the screen and there is time until then for the whole thing to become unstable?

That said, AMD 7970 and 7950 owners KNOW that the random frametime review from that random biased website is overly exaggerated. I personally play Skyrim and have not a single issue. Oh well, maybe the release version had instabilities like those but not the recent versions! Playing the game with tons of HD addons and nearly 3GB vRam loaded and the game is perfectly smooth! Same goes for my smooth BF3 and Hitman recent experiences. On the other hand, FarCry 3 performance was bad, as described above, with frequent choppy frames and the effect of "bad frametimes" being obvious most of the time while the game is "AMD" special, as said above.

[silkstone]

I just read the linked article. I'd also like to know what hardware they used so i could look up the information about the tolerance on the measurements. +/-5% would be what I would guess, but I couldn't be sure without looking it up.

Anyone have any idea on the typical tolerance of these measuring systems?

I am always skeptical of any kind of 'experiment' that does not talk about any kind of errors or faults in the process and declares itself to be definitive proof.
It's kind of similar to me saying that i am psychic because I predicted 100 results from coin tosses correctly (without telling you the total number of coin tosses).

[theoneandonlymrk]

Ive used crossfire on and off for the last 5-7 years first 3870s now old 58##s and I have seen very few issues that I couldn't sort or a games patch didn't fix .
Win8 made a great many issues occur for me inc piss poor game performance gfx oc ability and stuttering .
Im back on 7 and im getting 15- 30 percent more performance and no issues. Odd but true

[cadaveca]

Originally Posted by silkstone

I just read the linked article. I'd also like to know what hardware they used so i could look up the information about the tolerance on the measurements. +/-5% would be what I would guess, but I couldn't be sure without looking it up.

Anyone have any idea on the typical tolerance of these measuring systems?

I am always skeptical of any kind of 'experiment' that does not talk about any kind of errors or faults in the process and declares itself to be definitive proof.
It's kind of similar to me saying that i am psychic because I predicted 100 results from coin tosses correctly (without telling you the total number of coin tosses).

3DMark also captures this problem pretty well. frame times are a nice thin line in single card, but a thick line in crossfire. If you have Crossfire, and use any sort of FPS counter, even ingame ones, you can easily see how the framerate is far less consistent with Crossfire, and take a 10-15 FPS range every second, while one card might be 3-8FPS.

AMD's memory management hit bottlenecks, these bottlenecks cause frames to be rendered slower, and it IS perceivable. AMD has already admitted memory management is broken, so I don't see any reason to be so up in arms about people relating issues.

If you were local, silkstone, you could come over to my house, sit at my rig, and you'd see the problems. I have every platform under the sun, and they all have these problems, so I feel this can be fixed by driver. But perhaps not.

Originally Posted by theoneandonlymrk

Ive used crossfire on and off for the last 5-7 years first 3870s now old 58##s and I have seen very few issues that I couldn't sort or a games patch didn't fix .
Win8 made a great many issues occur for me inc piss poor game performance gfx oc ability and stuttering .
Im back on 7 and im getting 15- 30 percent more performance and no issues. Odd but true

I tend to get slightly higher benchmark scores in Win8. I've had a few people ask me why my 3dmark scores with single 7950 are so high..., matching and beating 7970's...I can only attribute it to Win8.. and yes, Win8 intrioduces some problems, but only because it is more sensitive to OC. get rid of OC, and the problem miraculously go away.

[Xzibit]

Didnt AMD say they were going to release a new memory manager for GNC cards ?

Is it in any of the Betas yet ?

[cadaveca]

Originally Posted by Xzibit

Didnt AMD say they were going to release a new memory manager for GNC cards ?

Is it in any of the Betas yet ?

No, there are per-app fixes for stutter in specific apps, most of which hare DX9-based fixes. AMD related that DX10 and DX11 would need the memory management fix before being 100%, and have yet to mention these fixes in release notes or anything. They did also way that it would be a bit of time before it was ready, and since they have had over a year already to fix this, I trust it's not a small problem.

[radrok]

Originally Posted by the54thvoid

I'm lost for words. I play BF3 (on crossfired 7970) and get 80-120 fps on MP depending on scene/players etc. To my eyes it runs perfectly smoothly. I know it does because it was choppy on Far Cry 3 (embarrassingly an AMD sponsored title).

Given that i know what choppy frame rates look like to me, I can hand on heart say BF3 runs as smooth as any game I've played on any gfx card (And I've had more Nvidia in my life - first AMD card was a 5850, then I got a GTX 580, then i went for 7970's).

There may be more at play from a very neurophysiological level that people are also ignorant off. Our eyes and brain cobble together the images we see. We cannot register everything that flickers up in front of us so our brain makes the image tangible to us.

From the OP's source post the crossfire looks like it should be abyssmal but it is not on my system. Perhaps the more people delve into frame latency the more we get transfixed by what the computer can see and not what humans beings are capable of seeing.

I cannot deny the findings of the source post. I'm sure it is not fabricated but I'm pretty sure there will be a raft of crossfire owners on here saying how well it works for them and they do not see any on screen indication of poor performance.

On the same note, I'm running around Crysis 3 land transfixed by how good it all looks and how smooth it is. Frames might drop occasionally to 30's but generally it's good for me. Certainly no stutter (like the Skyrimgate latency posts from last few months).

Think about it. It doesn't matter what anything says about the latency if we can't see it. An invisible problem is not a problem if the visual fidelity is what matters.

Perhaps someone can do a poll and record how many folk use crossfire in BF3 and how many have stutter or issues that would tie in with the OP source.

Again - I'm not denying crossfire is NOT as good as sli but in the instance of BF3 - it works butter smooth for me.

Agreed with you, sir.

I have no issues whatsoever on BF3, that title runs like a champ on ATI hardware.

Still when you drop below your monitor refresh rate with your FPS you'll notice stuttering when running multiple GPUs vs a single one, I've noticed this so many times in multiple games, if I actually disabled CFX I'd have lower framerates but a SMOOTHER experience.

If CFX/SLI is implemented properly and it is easily codeable for games, like BF3 for example, it's an awesome resource but when it doesn't work it makes you forget the times it works because it becomes frustrating.

Still, guys, trust your eyes, not these reports.

[silkstone]

Originally Posted by cadaveca

3DMark also captures this problem pretty well. frame times are a nice thin line in single card, but a thick line in crossfire. If you have Crossfire, and use any sort of FPS counter, even ingame ones, you can easily see how the framerate is far less consistent with Crossfire, and take a 10-15 FPS range every second, while one card might be 3-8FPS.

AMD's memory management hit bottlenecks, these bottlenecks cause frames to be rendered slower, and it IS perceivable. AMD has already admitted memory management is broken, so I don't see any reason to be so up in arms about people relating issues.

If you were local, silkstone, you could come over to my house, sit at my rig, and you'd see the problems. I have every platform under the sun, and they all have these problems, so I feel this can be fixed by driver. But perhaps not.

Maybe, if you get time, do you think it would be possible to post a few runs of 3d Mark with different cards but not label them?

I have heard about various issues with Xfire and would not be surprised if it were more noticeable in that kind of set up. But, I'd really like to see some comparisons. They don't have to be edited to be side by side or anything like that. People could just open the videos in separate windows.

Maybe like 2 runs with a 680 and 2 with a 680 SLI then 2 runs of each 7950 (or card equivalent to the 680) and xfire.

[cadaveca]

Originally Posted by silkstone

Maybe, if you get time, do you think it would be possible to post a few runs of 3d Mark with different cards but not label them?

I have heard about various issues with Xfire and would not be surprised if it were more noticeable in that kind of set up. But, I'd really like to see some comparisons. They don't have to be edited to be side by side or anything like that. People could just open the videos in separate windows.

Maybe like 2 runs with a 680 and 2 with a 680 SLI then 2 runs of each 7950 (or card equivalent to the 680) and xfire.

I don't have any Nvidia cards, AMD only.

I've run Crossfire since it came out with the X800's. Sometimes it works, and sometimes it doesn't. With 7900-series, it scales worse than it should. AMD has acknowledged the stutter issues and such, so that's that. It's not like this has come out, and AMD denies anything.. they agree it's an issue. Running benchmarks to confirm the problem is rather useless at this point. If you don't have such a config, and don't have problems, then don't worry about it. If you do, AMD is aware, confirmed it, and is working on fixing it. There's no "drama" to be found here.

[radrok]

Originally Posted by cadaveca

I don't have any Nvidia cards, AMD only.

You should bribe W1zzard for a couple of Titans

[erocker]

I'm not really convinced. Things seem to have gotten much better with these newer drivers. I can't tell the difference between AMD and Nvidia without opening the case and seeing what card is physically in the system.

[radrok]

Originally Posted by erocker

I'm not really convinced. Things seem to have gotten much better with these newer drivers. I can't tell the difference between AMD and Nvidia without opening the case and seeing what card is physically in the system.

Actually AMD/ATI colors look more vivid if you compare stock vs stock systems.

Of course nothing that a five minute NV control panel tweak can't overcome.

[theoneandonlymrk]

Originally Posted by cadaveca

3DMark also captures this problem pretty well. frame times are a nice thin line in single card, but a thick line in crossfire. If you have Crossfire, and use any sort of FPS counter, even ingame ones, you can easily see how the framerate is far less consistent with Crossfire, and take a 10-15 FPS range every second, while one card might be 3-8FPS.

AMD's memory management hit bottlenecks, these bottlenecks cause frames to be rendered slower, and it IS perceivable. AMD has already admitted memory management is broken, so I don't see any reason to be so up in arms about people relating issues.

If you were local, silkstone, you could come over to my house, sit at my rig, and you'd see the problems. I have every platform under the sun, and they all have these problems, so I feel this can be fixed by driver. But perhaps not.



I tend to get slightly higher benchmark scores in Win8. I've had a few people ask me why my 3dmark scores with single 7950 are so high..., matching and beating 7970's...I can only attribute it to Win8.. and yes, Win8 intrioduces some problems, but only because it is more sensitive to OC. get rid of OC, and the problem miraculously go away.

Yeh well I dont think o got my point across sorry, I wasnt trying to blame win8 but succinctly.
Its down to individual hard and software setups in my example something had clearly gone wrong with win8 , but it was simply part of the issue ,but its far easier to blame amds drivers then fresh install or actually find the issue and some setups probably make stuttering more an issue

[the54thvoid]

Originally Posted by erocker

I'm not really convinced. Things seem to have gotten much better with these newer drivers. I can't tell the difference between AMD and Nvidia without opening the case and seeing what card is physically in the system.

Absolutely the case. I've got two friends regularly playing BF3 with me (both kick my ass mind) but they have Nvidia cards (580/680) and they have driver crashes- or at least game crashes far more often than I do and I run crossfire and they're single gpu.
Finners is on the TPU forum occasionally so he might be able to shed more light but I genuinely haven't crashed in BF3 for a good few months, if not half a year. Though I've not played yet using 13.2 (6) beta's so maybe I'll come back and edit this post, lol.

But so i don't come across all "AMD is great - Nvidia sux", MOH for me was an absolute joke. Crossfire didn't even work for me at all. Then Far Cry 3 was choppy. But Crysis 3 is good candy and plays nice.

[TRWOV]

Is this an issue if you play with VSYNC? I've never experienced any kind of stuttering.

The only issue that comes to mind is the BF3 memory leak.

[DayKnight]

Originally Posted by erocker

I'm not really convinced. Things seem to have gotten much better with these newer drivers. I can't tell the difference between AMD and Nvidia without opening the case and seeing what card is physically in the system.

... or you can just use GPU-Z.

[silkstone]

Originally Posted by cadaveca

I don't have any Nvidia cards, AMD only.

Running benchmarks to confirm the problem is rather useless at this point. If you don't have such a config, and don't have problems, then don't worry about it. If you do, AMD is aware, confirmed it, and is working on fixing it. There's no "drama" to be found here.

It's not that i don;t believe that the problem exists. I know that crossfire has issues with it. It's more that I don't believe it is an accurate way to 'benchmark' cards like stated in the original article.

Also the OP states that the 'issue' is worse than we thought. Surely it is only as bad as it is perceived when playing a game with the issue? I just don't see the point of graphs without any kind of demonstration.

[cadaveca]

Originally Posted by silkstone

It's not that i don;t believe that the problem exists. I know that crossfire has issues with it. It's more that I don't believe it is an accurate way to 'benchmark' cards like stated in the original article.

Also the OP states that the 'issue' is worse than we thought. Surely it is only as bad as it is perceived when playing a game with the issue? I just don't see the point of graphs without any kind of demonstration.

I kind of agree, but since I have many AMD cards here, I do not need any demonstration. I have been complaining about these problems for years, literally.


If you understand what the graphs show(an increase in frame-to-frame render latency), that should be enough, really. Visually it amounts to a momentary pause.

When you add a second card, because this doubles that delay, the pause is even longer, and becomes even more noticeable, to the point that singlecard performance is perceptibly smoother, even though the framerate is significantly lower.

AMD has said, that with DX10 and DX11, this is due to memory management. With DX9, performance is adequate enough that it can be dealt with easily, as far as I can tell, but I'm not expert with that sort of stuff at all.

So, imagine playing at 60+FPS, and it's nice and smooth. FPS drops below 60 with some action, and it becomes less smooth. Staying above 60 FPS will keep that smoothness. Adding a second card, and increasing performance above 60 FPS should keep the same smoothness, right?

Unfortunately, what happens is that those stutters from going to less than 60 FPS overall still apply, but PER CARD, so that now to maintain a smooth experience, you need 120 FPS, negating the usefulness of the second card. Add in that the added driver overhead from managing both cards, and overall you end up with a worse experience than with one card.