new AMD Catalyst driver from Alpha Micro stuttering

[csbin]

http://www.computerbase.de/news/2013-04/amd-mit-neuem-alpha-catalyst-treiber-gegen-mikroruckler/

AMD is preparing to launch frames pacing technology (frame pacing) to solve this problem, they intend to add a new option in the Catalyst driver panel, the user can adjust the output frame interval, the simple answer is to turn this option, the graphics rendering is completeframe does not immediately output to the monitor, but will wait for a cycle to the next frame when the output, so that the generation time of each frame may become increasing (the mean number of frames will be reduced), but the output of each frame intervalmore evenly, Caton phenomenon will be improved.
Look at a concrete example, the following figure is the first to test the HD 7990 13.5 beta2 driver frame delay performance fluctuations than most of the GTX 690.

 

[SPBHM]

yes, AMD's software frame metering it seems to work well

the video is pretty good to show the huge difference in smoothness

https://www.youtube.com/watch?v=RIFYyDV3DKs

http://www.pcper.com/reviews/Graphi...-Improves-CrossFire-Prototype-Driver/Crysis-3


still, it's hard to understand how so many users have been happy with CF without this for so long...

 

[Lonyo]

There's no distinct reason that frame pacing would reduce the FPS, it just changes how frequently the frames are displayed, whether it's when ready, or when allowed.

It might result in some frames being cut from being displayed, although presumably they would still be rendered, as these frames would get to the end of the process, then get cut, but it all depends on how they set up their frame delays to smooth out the display of the frames.

 

[3DVagabond]

still, it's hard to understand how so many users have been happy with CF without this for so long...

It's not present in all games. At least to a point of being noticeable. If you read [H]'s reviews, for example, they will point out when a particular game seems to be running slower than the FPS would indicate. They've been doing this far longer than FCAT has been around.

Many people use vsync which fixes the problem in most all games. The ones that it doesn't you can use a FPS cap.

 

[SPBHM]

It's not present in all games. At least to a point of being noticeable. If you read [H]'s reviews, for example, they will point out when a particular game seems to be running slower than the FPS would indicate. They've been doing this far longer than FCAT has been around.

Many people use vsync which fixes the problem in most all games. The ones that it doesn't you can use a FPS cap.

from the PCper tests is clear that the problem is present in most games, any game GPU limited seems to suffer from it,

as for vsync, it should work as long as the CF could easily keep it stable at over the desired rate (like 16.6ms) I think, but anyway, a huge portion of users is not going to try anything other than the most default settings,

sure there has been reports of micro stuttering over the years, but I'm amazed by how many people always defended CF as a solution with no additional problems,

 

[Leadbox]

from the PCper tests is clear that the problem is present in most games, any game GPU limited seems to suffer from it,

as for vsync, it should work as long as the CF could easily keep it stable at over the desired rate (like 16.6ms) I think, but anyway, a huge portion of users is not going to try anything other than the most default settings,

sure there has been reports of micro stuttering over the years, but I'm amazed by how many people always defended CF as a solution with no additional problems,

You're fixating on the graphs and the numbers a tad too much. They represent what the tools and the associated software can detect which is far removed from what a person's eyes might notice. Not being able to see something or not being bothered by it is NOT defending CF

 

[Face2Face]

yes, AMD's software frame metering it seems to work well

the video is pretty good to show the huge difference in smoothness

https://www.youtube.com/watch?v=RIFYyDV3DKs

http://www.pcper.com/reviews/Graphi...-Improves-CrossFire-Prototype-Driver/Crysis-3


still, it's hard to understand how so many users have been happy with CF without this for so long...

Wow.. That video of Crysis 3 is a pretty huge boost to smoothness. This is good news because I plan on going with another 7950 in the coming months.

 

[JAG87]

Wow check it out! It does almost what vsync has been doing for the past four decades but you still get tearing! Awesome technology.

I can't believe people actually play games without triple buffered vsync.

 

[DarkKnightDude]

Granted, judging smoothness in a Youtube video might not be the best idea, but 13.5 seemed really choppy as hell compared with Prototype driver. If they keep up with the fixes, might considering picking up another 7970 for the hell of it.

 

[SPBHM]

You're fixating on the graphs and the numbers a tad too much. They represent what the tools and the associated software can detect which is far removed from what a person's eyes might notice. Not being able to see something or not being bothered by it is NOT defending CF

it's not simply about graphs and numbers, watch the videos (download the better files from their website if you can)

 

[Stuka87]

it's not simply about graphs and numbers, watch the videos (download the better files from their website if you can)

Unless the videos are being recorded and played back at the exact same framerate, its not really valid.

 

[SPBHM]

Unless the videos are being recorded and played back at the exact same framerate, its not really valid.

both videos were recorded at 60FPS from the DVI output, it's all the same hardware, just with a different driver,

if you are talking about the rendering frame rate, it looks close in both cases

read the article, it's all very easy to understand.

 

[railven]

Wow check it out! It does almost what vsync has been doing for the past four decades but you still get tearing! Awesome technology.

I can't believe people actually play games without triple buffered vsync.

I've been doing this for years. Perhaps this is why I haven't experience the stuttering issues as severely as been reported.

There is still only one game where I get stutter, but that is true for both my rigs (AMD rig is worse though) and I can't for the life of me figure it out.

World of Warcraft: Mist of Panderia

 

[VulgarDisplay]

Does anyone know if the lower the ms on the graph = lower input lag? If that's the case does that mean that both gtx690 and 7990 are better than Titan when dual gpu drivers are working?

Another question. Are the slightly higher variations in the prototype drivers on the 7990 due to higher FPS in general?

Anyone else think Far Cry 3 should be thrown out of all benchmarks because of how terrible it runs no matter what hardware you have?

 

[psolord]

yes, AMD's software frame metering it seems to work well

http://www.pcper.com/files/imagecache/article_max_width/review/2013-04-22/Crysis3_2560x1440_PLOT.png

the video is pretty good to show the huge difference in smoothness

https://www.youtube.com/watch?v=RIFYyDV3DKs

http://www.pcper.com/reviews/Graphi...-Improves-CrossFire-Prototype-Driver/Crysis-3


still, it's hard to understand how so many users have been happy with CF without this for so long...

Some of us have been using vsync for ages. Personally it's the only setting with which that I can perceive total smoothness.

Even in single gpu on my Nvidia system, disabled vsync cannot possibly provide me with the desired result, even when framerate is very high. Vsync does provide me with that.

In that regard, I can totally understand that excessive higher framerate does not mean better smoothness (80fps is not smoother than 60fps vsynced), but what people seem to have forgotten all of a sudden, is what dual gpu really offers.

It may bring rant frames, frame time inconsistency or whatever, but what really matters is that it essentially almost doubles your available gpu processing processing pool which in turns does wonders in real life gaming, if that expanded processing pool comes to be needed.

For example, if a game runs at 40fps on a single card, enabling the second one could go up to 75-80fps. The game may still not be smooth due to the frame time inconsistency but if you vsync the sucker, it will run at a perfect smooth 60fps and that exactly is the profit of dual gpu. With single gpu you could never do that if such a gpu limit came into play.

A faster gpu could easily go up to 65fps for example, thus providing a great vsynced experience, but it will still fall short on the next demanding game, where the dual gpu expanded pool will still be able to provide good playability.

Of course I have price to performance ratio as my highest point of intereset. If everyone could afford Titans for example, sure go ahead.

 

[Jacky60]

Well the PC perspective vid certainly convinces me. Lets hope it's released asap^_^

 

[blastingcap]

http://www.computerbase.de/news/2013-04/amd-mit-neuem-alpha-catalyst-treiber-gegen-mikroruckler/

AMD is preparing to launch frames pacing technology (frame pacing) to solve this problem, they intend to add a new option in the Catalyst driver panel, the user can adjust the output frame interval, the simple answer is to turn this option, the graphics rendering is completeframe does not immediately output to the monitor, but will wait for a cycle to the next frame when the output, so that the generation time of each frame may become increasing (the mean number of frames will be reduced), but the output of each frame intervalmore evenly, Caton phenomenon will be improved.
Look at a concrete example, the following figure is the first to test the HD 7990 13.5 beta2 driver frame delay performance fluctuations than most of the GTX 690.

How will input lag be affected?

 

[Hitman928]

How will input lag be affected?

Input lag will go up, just like with any frame metering. That's why AMD said they will put in a slider to give the user choice as to how much smoothness versus input lag they want.

 

[Fx1]

vsync 59 fps cap is the best middle ground of smoothness and input lag. Fact is that you dont get any problems if you use that setup with CF or SLI

 

[thilanliyan]

vsync 59 fps cap is the best middle ground of smoothness and input lag. Fact is that you dont get any problems if you use that setup with CF or SLI

Any disadvantages to using this method?

 

[Fx1]

Technically there is some input lag but its barely there. Personally i cant tell.

Vsync does add lag but the 59 cap seems to do away with 75% of any lag that is felt.

It is the best solution i have ever seen so far. its something to do with 59hz v 60fps vsync

 

[SPBHM]

Some of us have been using vsync for ages. Personally it's the only setting with which that I can perceive total smoothness.

vsync is not always a good solution, if for some reason the PC can't keep the same performance the effects are obvious and unpleasant,
if you look at the graph I posted from Crysis 3, it's running at 35-40FPS on the 7990, Vsync doesn't sound like a good way of improving the experience, frame rate metering does.

looking at the 60FPS videos from this comparison
http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Visual-Effects-Vsync-Gaming-Animation

it's clear that even when you would think you are pretty close to having 60FPS all the time, some problems can occur, the animation on the non vsync at 60fps looks smoother than vsync at 60fps.


still, I think an input latency test, with single GPU vs multi gpu with and without frame rate metering, and vsync would be quite interesting...

 

[JAG87]

Any disadvantages to using this method?

Yes, a tiny hitch every 1/60th of a second. You can't pull up 59 frames into 60 hz, so the monitor has to display one frame twice once every second. It's unnoticeable vs the horrendous stuttering mess that vsync off is, and definetely much better than dealing with input lag. However, for games that do not rely on mouse input (i.e. controller games), you should cap at 60 since input lag is irrelevent. Then you have a perfect display feed.

You can set profile based caps with Nvidia Inspector. No idea how to do it on AMD, maybe with RadeonPro?

 

[cmdrdredd]

Wow check it out! It does almost what vsync has been doing for the past four decades but you still get tearing! Awesome technology.

I can't believe people actually play games without triple buffered vsync.

Some people want their aim to be precise and avoid added input lag.

 

[Hitman928]

Wow check it out! It does almost what vsync has been doing for the past four decades but you still get tearing! Awesome technology.

I can't believe people actually play games without triple buffered vsync.

Some people want their aim to be precise and avoid added input lag.

I'm one of them. I've tried using triple buffered vsync several times and every time I have to turn it off because the input lag bothers me more than the hitches in smoothness.

still, I think an input latency test, with single GPU vs multi gpu with and without frame rate metering, and vsync would be quite interesting...

I agree. Personally, like I said above, input lag bothers me more than uneven frames. While I can appreciate smoother frame rates, added input lag just makes me feel like my character is moving through jello the whole time and it is more disruptive to me than even stuttering . . . unless you get to really extreme stutters that disrupt my ability to play of course.