Anandtech - AMD Comments on GPU Stuttering, Driver Roadmap, Benchmarking

M

Meeho

Supreme [H]ardness
Joined
Aug 16, 2010
Messages
5,915
www.anandtech.com/show/6857/amd-stuttering-issues-driver-roadmap-fraps

Although neither strictly a defense nor merely an explanation of what we’ve been seeing over the past year, AMD wanted to sit down and better explain their position. This includes both why AMD’s products have been impacted in the manner they were, and why at the same time (and not unlike NVIDIA) AMD is worried about FRAPS being given more weight than it should be. Ultimately AMD believes that it’s to the benefit of buyers and journalists alike to better understand just what is happening, why it’s happening, and just what the most common tools can and are measuring.
Click to expand...
 
i can agree with what the article is stating and one of the reason i like H's review methods.. In the end its about the User's experience

the conclusion they come out with was a little suprising though, that when AMD put in fixes for the "stuttering" issues they found they actually increased performance (Borderlands 2 13% increase in performance was from the stutter fix)

i would have expected that to be the opposite, kindof makes you think on why Nvidia never really came out and made a big deal about the difference between Nvidia's and AMD's performance in Scott's assessments

in the end though it sounds like AMD's going back adn fixing the issues and putting this as part of their analysis in the future.

to their criticism though, they should have been doing this from the start
 
Although not perfect, I still think using FRAPS frame time measurements is better than using it (or anything else) for FPS

This stuttering isn't just there at the beginning of the pipeline - you can see it with your own eyes, especially if you record it with high speed video.

IMO it is still the best way to portray the end user experience
 
We've been doing it for years

Here's my issue, people keep wanting to apply the scientific method to gaming, and gaming experience/performance. But it just doesn't work, gaming is very "analog" as an analogy. This isn't science class. We aren't mathematicians calculating the fuel ratio to lift orbiters into space. We are gamers, and we demand a good experience when we play our games.

The way we've been evaluating for years reflects this. The entire nature of how we evaluate gameplay performance naturally reflects issues that come up like stuttering. We've mentioned it for years in multi-GPU. Others are only just now catching on. However, I've never, repeat never, associated stuttering to single-GPU, I've never seen "micro stutter" on single-GPUs. You just don't feel it during gaming. Now, what you do feel, are some games that are just coded badly and have game stutter issues. Some games like that are Skyrim, and Far Cry 3. The stuttering in those games is game related, not video card.
 
It bothers me somewhat that Ryan believes that vertical sync somehow mitigates interframe latency. That just isn't accurate at all.
 
No it isn't. You have to physically limit frames as in Afterburner or Radeon Pro. Vertical Sync will get rid of the "runt frame" problem pointed out in the PCPer article.
 
I don't get why the "runt frames" as they call them would be a problem though?

They don't induce stuttering, they don't affect gameplay, the only thing they MIGHT affect, seems to be that FRAPS shows a higher FPS than actually is displayed, which could explain the 100%+ scaling seen in some tests?

Or am I totally off the ball here?
 
wonderfield said:
It bothers me somewhat that Ryan believes that vertical sync somehow mitigates interframe latency. That just isn't accurate at all.
Click to expand...

Actually I think you might be misunderstanding him. To me it sounded like he was saying that having frames sync'ed to display refresh cycles bakes in a sort of latency "buffer" so weird variance you would've seen with vsync off won't be visible with vsync on, since by the nature of vsync you're seeing a even cadence of frames.

In other words, if you would've seen 2ms then 15ms then 2 ms, instead you'd see 16 ms then 16 ms then 16 ms.
 
[G]Ernesto;1039740749 said:
I don't get why the "runt frames" as they call them would be a problem though?

They don't induce stuttering, they don't affect gameplay, the only thing they MIGHT affect, seems to be that FRAPS shows a higher FPS than actually is displayed, which could explain the 100%+ scaling seen in some tests?

Or am I totally off the ball here?
Click to expand...

You spend a bunch of time rendering a frame for no experience benefit.
 
I think the whole issue depends on game engines.. Bioshock infinite for example it's running flawlessly with everything on maximum with two 7950's I didn't need to use radeon pro and even the mouse movement doesn't have imput lag with vsync on..
 
lowlight said:
Although not perfect, I still think using FRAPS frame time measurements is better than using it (or anything else) for FPS

This stuttering isn't just there at the beginning of the pipeline - you can see it with your own eyes, especially if you record it with high speed video.

IMO it is still the best way to portray the end user experience
Click to expand...

I'm not sure about anyone else, but I just watched that techreport high-speed video comparison of the 7950 vs 660 TI. I notice stuttering and tearing on BOTH sides. Yes the 7950 seemed to have it slightly more, but only fractionally more. As in, even at high-speed, I could barely tell the difference.

I know that micro-stuttering is a real issue with crossfire, and I'm glad that AMD has finally talked about the issue. I sincerely hope that they get it fixed with a Driver update sooner or later.
 
BlueStorm said:
In other words, if you would've seen 2ms then 15ms then 2 ms, instead you'd see 16 ms then 16 ms then 16 ms.
Click to expand...
The variance is a result of a variance in frame times, not buffer flips. A frame that takes 2ms to render takes 2ms to render regardless of whether vsync is enabled. With vsync, the only even cadence is in terms of buffer flips.
 
The shortest answer also the bluntest answer: AMD had a stuttering problem because AMD wasn’t looking for a stuttering problem. AMD does a great deal of competitive analysis (read: seeing what NVIDIA is doing) on overall performance, but AMD was never doing competitive analysis for stuttering.

Because stuttering is such a complex issue and AMD had such great knowledge into their drivers, AMD assumed that stuttering was occurring due to the applications and the OS, things that were out of their control. Furthermore because those things were out of their control, AMD assumed that they were happening to NVIDIA and Intel GPUs too. After all, there wasn’t any kind of competitive analysis to scientifically confirm this. AMD never saw that NVIDIA cards weren’t experiencing as much stuttering, and consequently never saw that they did in fact have more control over stuttering than they first thought.
Click to expand...

Making asses of themselves? :D
 
After all, there wasn’t any kind of competitive analysis to scientifically confirm this. AMD never saw that NVIDIA cards weren’t experiencing as much stuttering, and consequently never saw that they did in fact have more control over stuttering than they first thought.

so they never saw it so it didn't exist?
more like we don't care, If it can't be fixed in 5 minutes we will just say its not there.

Why don't they just admit they spent way to much time making bloatware to say weeeee look what you can set and NOT lets make a small package that does it right.
 
New management is my guess.
techreport put the finger on it.
and they are catching up.
 
Kaori said:
I think the whole issue depends on game engines.. Bioshock infinite for example it's running flawlessly with everything on maximum with two 7950's I didn't need to use radeon pro and even the mouse movement doesn't have imput lag with vsync on..
Click to expand...

well keep in mind that the bioshock inf engine is almost 10 years old, so that doesnt impress me much. Im more concerned with the next gen engines like CryEngine and Frostbyte
 
You must log in or register to reply here.
Back
Top