HD5970 Microstuttering tests

[annihilat0r]

I'm not sure if Vsync solves MS (well, theoretically it should) but it makes most games unplayable for me. I can't play games that don't have snappy controls, that's why the console ports suck extra for me, they usually have sucky controls. In Bioshock at 150 FPS, controls are much less accurate than Crysis at 20 FPS. And even when the game does have snappy mouse controls, Vsync ruins it.

I absolutely can't understand why most FPS's like Bioshock or Far Cry 2 for example have controls that are somewhat laggy and not precise. Damn, Quake 2 is an 13 year old game and its controls were extremely awesome.

[mak1skav]

I am using a 4870x2 Asus card at stock clocks and the only game that gave me MS problems was Lord Of The Rings Online. When i was moving the camera at open places i had mini pauses and the character was looking like walking instead of running but my frames were above 140-150 all the time.

When i installed Windows 7 ultimate 64bit and reinstalled the game all the ms problems were disappeared with a magic way. My guess was that the problem solved from ATI's 64bit drivers or from Windows or maybe the problem solved when i turn on DirectX 10 use in Lord Of The Rings options.

[RaZz!]

I'm not sure if Vsync solves MS (well, theoretically it should) but it makes most games unplayable for me. I can't play games that don't have snappy controls, that's why the console ports suck extra for me, they usually have sucky controls. In Bioshock at 150 FPS, controls are much less accurate than Crysis at 20 FPS. And even when the game does have snappy mouse controls, Vsync ruins it.

I absolutely can't understand why most FPS's like Bioshock or Far Cry 2 for example have controls that are somewhat laggy and not precise. Damn, Quake 2 is an 13 year old game and its controls were extremely awesome.

i couldn't agree more. however, controls in far cry 2 aren't that bad imo. it may appear to be unprecise due to far cry 2's crap fov, provided you have a widescreen monitor.

furthermore, if the refresh rate is high enough vsync doesn't affect the mouse control att all - or atleast only by an insignificant amount. back when i still had a crt and was able to set the refresh rate to 100hz i used vsync (even with triple buffering) in almost every game. using vsync on a tft which only has 60hz in most cases pretty much sucks. mouselag + only 60fps, possibly the worst that could ever happen to someone who likes playing fast paced fps

[annihilat0r]

i couldn't agree more. however, controls in far cry 2 aren't that bad imo. it may appear to be unprecise due to far cry 2's crap fov, provided you have a widescreen monitor.

furthermore, if the refresh rate is high enough vsync doesn't affect the mouse control att all - or atleast only by an insignificant amount. back when i still had a crt and was able to set the refresh rate to 100hz i used vsync (even with triple buffering) in almost every game. using vsync on a tft which only has 60hz in most cases pretty much sucks. mouselag + only 60fps, possibly the worst that could ever happen to someone who likes playing fast paced fps

well dunno man, there's a certain "console-like" type of control where it's obvious that it was designed for gamepads, and to me FC2 looks like one of them. Do you think FC2's controls are as snappy as Quake 2, 3 or 4? And I don't think it's an issue with FPS either, like I said Crysis provided snappy controls even at 20FPS and I play FC2 with around 70 FPS.

[Musho]

Whether there is any microstuttering or not, I am completely clueless how the site thinks they can prove there is microstuttering by showing those graphs, without showing any microstuttering-free graphs (AKA, single GPU setups).

Those spikes in framerate could be just about anything! The game having to read from the HDD, the CPU bottlenecking, because it's doing some physics calculation or maybe some frames are just harder to render. Think about a lightning flash, increasing the complexity of the lightning of a scene for 1-2 frames.

[Xoulz]

I don't know if you can call this a "huge" problem since many people just don't see "microstutter" unless they are looking at one of those graphs....myself included. We all know the problem exists but it doesn't impact upon my gameplay experience in any way, shape or form. But that's just me....

Hey bro, it's just that microstutter gives us more to notice... with more opportunities to ruin gameplay. Becoming frustrating & reducing immersion.

Also, MAX anything, just doesn't matter when benchmarking. I'm steadfast on this, (again) the ONLY metric that matters when benching or comparing GPU's, is how often a card hits minimum frame rates..!

Not to many websites focus on this aspect, either. But microstutter is a no-no and places people if difficult situations.

[Xoulz]

Whether there is any microstuttering or not, I am completely clueless how the site thinks they can prove there is microstuttering by showing those graphs, without showing any microstuttering-free graphs (AKA, single GPU setups).

Those spikes in framerate could be just about anything! The game having to read from the HDD, the CPU bottlenecking, because it's doing some physics calculation or maybe some frames are just harder to render. Think about a lightning flash, increasing the complexity of the lightning of a scene for 1-2 frames.

Because there is no AFR used... thus, microstutter cannot exist..

[eRazorzEDGE]

I just learned something while reading this whole thread over the past day and searching google... ATI uses Supertiling by default. It's in the registry, search for the term "MVPUmode". 31 is for AFR, 32 is for SFR, 33 is for Supertiling... which is what is set by default (at least on my rig).

So everyone is talking crap on AFR and they own ATI cards, yet ATI doesn't use AFR by default... so how can you talk crap about it?

The only game I can notice any stuttering in, to any degree, is Crysis. It's more pronounced with AA enabled, however, whenever stuttering occurs, out of the corner of my eye, I see the HDD light blinking like a mother trucker. I have two SSD's in RAID 0 and more than enough memory to run the game with even high res textures, so I'm not sure exactly why it's chuggin from the HDD so much. Anyone with thoughts on this?

I've also done a little test by enabling AFR while playing Crysis and running benches in it. While the benches turned out basically the same (using AFR and Supertiling, withing 1 FPS of each other), playing the game seemed like there was more stuttering with Supertiling enabled. Not only that, but disabling Crossfire and only using one 5870 still seemed to give the same "stuttering", but I don't have any empirical evidence to show at this time to support what I "noticed". I should try the same tests as the article, since I haven't read it yet, I'll see if they wrote down their testing methodology. :p

[annihilat0r]

I just learned something while reading this whole thread over the past day and searching google... ATI uses Supertiling by default. It's in the registry, search for the term "MVPUmode". 31 is for AFR, 32 is for SFR, 33 is for Supertiling... which is what is set by default (at least on my rig).

So everyone is talking crap on AFR and they own ATI cards, yet ATI doesn't use AFR by default... so how can you talk crap about it?

:p

Wherever you got this info is wrong. All games with a profile are defaulted to AFR in both SLI and Crossfire.

[zalbard]

With vsync on - no, it doesn't.

In fact that's a good point, isn't VSync supposed to help with MS?

[herderien]

well i have a 4870x2 and never noticed such a thing, it might exist thou but i just never noticed, i played a lot moderne warfare 2 in 1920x1080.
the only thing i see is slower then should be fps with one particular particle effect in one particular dunjon on WoW.

[Xoulz]

I just learned something while reading this whole thread over the past day and searching google... ATI uses Supertiling by default. It's in the registry, search for the term "MVPUmode". 31 is for AFR, 32 is for SFR, 33 is for Supertiling... which is what is set by default (at least on my rig).

So everyone is talking crap on AFR and they own ATI cards, yet ATI doesn't use AFR by default... so how can you talk crap about it?

The only game I can notice any stuttering in, to any degree, is Crysis. It's more pronounced with AA enabled, however, whenever stuttering occurs, out of the corner of my eye, I see the HDD light blinking like a mother trucker. I have two SSD's in RAID 0 and more than enough memory to run the game with even high res textures, so I'm not sure exactly why it's chuggin from the HDD so much. Anyone with thoughts on this?

I've also done a little test by enabling AFR while playing Crysis and running benches in it. While the benches turned out basically the same (using AFR and Supertiling, withing 1 FPS of each other), playing the game seemed like there was more stuttering with Supertiling enabled. Not only that, but disabling Crossfire and only using one 5870 still seemed to give the same "stuttering", but I don't have any empirical evidence to show at this time to support what I "noticed". I should try the same tests as the article, since I haven't read it yet, I'll see if they wrote down their testing methodology. :p

If it is crossfire/xfire mode, wouldn't it have to be AFR..? Though, if it defaults to one card... then microstutter is moot. Correct?

[RaZz!]

In fact that's a good point, isn't VSync supposed to help with MS?

considering the original purpose of vsync (which is to prevent tearing on the monitor) it's, imo, very likely to influence microstuttering as well.
vsync synchronizes the fps rendered by the gpu with the refreshrate of the monitor. while gaming on a 60hz panel the gpu synchronizes the fps to 60hz, which results in 60fps. to put it simple, vsync waits for every frame to be completely shown on the screen and only then starts to send the 2nd frame to the monitor etc.
if you now look at the cause of microstuttering (variance of time between each rendered frame) i think vsync might help to reduce the effect of ms, but doesn't solve the initial problem.

however, i'm only speculating. i neither tested sli/cf with vsync nor do i even own a sli/cf setup

[KingOfsorroW]

In fact that's a good point, isn't VSync supposed to help with MS?

Well, it eliminates it completely )

[zalbard]

Well, it eliminates it completely )

Would be interesting to see the tests with and without VSync.
I'm sure someone has 3 cards and can find out!

[eRazorzEDGE]

Wherever you got this info is wrong. All games with a profile are defaulted to AFR in both SLI and Crossfire.

You're absolutely right. I come in here stating false truths to get a reply.

If you have an ATI card, or cards, look in the registry. Search for MVPUmode... default is 33... 33 IS EQUAL TO SUPERTILING.

It's all over google, just search for "mvpumode" or "mvpu".

If the game itself overrides the default driver preferences, then I'm wrong. However, I was just playing Crysis with the MVPUmode set to 32 (which is Split Frame Rendering) and boy howdy can you see a difference between the top and the bottom halves of the screen when moving around... it's like image tearing, only it does it below the display frequency of my monitor. It definitely didn't do that when I had AFR (31) or the default Supertiling (33) on.

@ Xoulz... no, it doesn't have to be anything. It uses whatever you tell it to use. Look at the program by AgentGOD, ATI Proflies or something (I know I'm butchering the name, I'm sorry), it allows you to force different rendering techniques.

Also, everyone is just in love with the word "microstutter" and its generally accepted meaning has spread like wild fire throughout the internet to people who like to talk about things when, in fact, they know nothing of these things.

If you notice from the articles graphs, they got virtually no stutter in 1680x1050, but once they went to 1920x1200 it was all over the place... seems peculiar to me. If you notice when the hard drive is being accessed a lot while playing a game, it slows and un-even's the flow of the frames to the screen. Now since we don't have a baseline control example, there is no definitive way to conclude that the results are simply cause by multiple gpu's.

There's a thing called a scientific method, if it's not followed, any resulting data cannot be have a conclusion drawn from it with absolute certainty.

[KingOfsorroW]

Would be interesting to see the tests with and without VSync.
I'm sure someone has 3 cards and can find out!

That was done many times, but yeah anyone with a multi gpu setup can verify it by there own run fraps and read the record from the log file with and without the vsync
Just like they did it on this forum: http://www.evga.com/forums/tm.aspx?m=121152

[eRazorzEDGE]

I really can't make heads or tails of the translation, does anyone know what testing methodology they used?

[cowie]

i'v been useing cf/sli for 6yrs and it has never bothered me
my take on it is that ms is over-diagnosed 99.9999% of the time.
after articles like this even some on single cored cards complain about ms
so if it happens you cant tell so who cares?
\ i read in that report that nv has helped the ms somewhat with drivers????? i always thuoght that drivers could never help ms?
or was the "ms" because of not so optimized drivers in the first place?

[Final8ty]

You're absolutely right. I come in here stating false truths to get a reply.

If you have an ATI card, or cards, look in the registry. Search for MVPUmode... default is 33... 33 IS EQUAL TO SUPERTILING.

It's all over google, just search for "mvpumode" or "mvpu".

If the game itself overrides the default driver preferences, then I'm wrong. However, I was just playing Crysis with the MVPUmode set to 32 (which is Split Frame Rendering) and boy howdy can you see a difference between the top and the bottom halves of the screen when moving around... it's like image tearing, only it does it below the display frequency of my monitor. It definitely didn't do that when I had AFR (31) or the default Supertiling (33) on.

@ Xoulz... no, it doesn't have to be anything. It uses whatever you tell it to use. Look at the program by AgentGOD, ATI Proflies or something (I know I'm butchering the name, I'm sorry), it allows you to force different rendering techniques.

Also, everyone is just in love with the word "microstutter" and it's generally accepting meaning has spread like wild fire throughout the internet to people who like to talk about things when, in fact, they know nothing of these things.

If you notice from the articles graphs, they got virtually no stutter in 1680x1050, but once they went to 1920x1200 it was all over the place... seems peculiar to me. If you notice when the hard drive is being accessed a lot while playing a game, it slows and un-even's the flow of the frames to the screen. Now since we don't have a baseline control example, there is no definitive way to conclude that the results are simply cause by multiple gpu's.

There's a thing called a scientific method, if it's not followed, any resulting data cannot be have a conclusion drawn from it with absolute certainty.

I have seen at least 3 posts now on another foum by people who claim to have MS on a single GPU now.

[eRazorzEDGE]

Well, this is no definitive review, but I just completed a couple runs on my system... and learned how to use spreadsheets and how to use formulas in said spreadsheets. So here's the details:

Crysis 64bit DX10 1920x1200 4xAA running on an AMD X4 965 @ 4.4GHz Core, 2.88GHz NB, 2.88 GHz HTT Link, 1600MHz DDR3 RAM. I have two ATI 5870's at 950/1250. I'm running Windows 7 64bit.

I ran the Crysis benchmark tool for 3 loops in one run, for both Dual card and Single card setups. I set FRAPS to capture the frame times and stop after 10 seconds of recording. I started recording the frame times on the 3RD loop right when the black truck exits the scene (the camera moves past it) to the right. This is to hopefully be as repeatable a test as possible.

Dual Card setup...

This one is with the Default ATI driver Supertiling technique:



This one is with the AFR technique:



and the very much needed control setup... Single card...




OMGWTFBBQ?!?!

Could it be there is the same damned thing happening on a single card as a dual card? Could it be that everyone is a little more influenced by what others tell them then by what they can actually prove?

On a side note, if this variation in frame output from one frame to the next is responsible for the so-called microstutter, then that would mean everyone on here can see the difference from one frame to the next while being shot at with 30-60 frames per second. For a test on that, in the span of ONE second, count from 1 to 60 and tell me how you do.

[eRazorzEDGE]

And now a CoD4 test. The game is default locked at 91 FPS, so I didn't change that for this test. Same setup, maximum visual settings in game.

Dual card...



Single Card...

[Stukov]

Well, this is no definitive review, but I just completed a couple runs on my system... and learned how to use spreadsheets and how to use formulas in said spreadsheets. So here's the details:

Crysis 64bit DX10 1920x1200 4xAA running on an AMD X4 965 @ 4.4GHz Core, 2.88GHz NB, 2.88 GHz HTT Link, 1600MHz DDR3 RAM. I have two ATI 5870's at 950/1250. I'm running Windows 7 64bit. I also changed the registry for MVPUmode to 31, to enable AFR from the default of 33, which is Supertiling. I plan to run another test with Supertiling later.

I ran the Crysis benchmark tool for 3 loops in one run, for both Dual card and Single card setups. I set FRAPS to capture the frame times and stop after 10 seconds of recording. I started recording the frame times on the 3RD loop right when the black truck exits the scene (the camera moves past it) to the right. This is to hopefully be as repeatable a test as possible.

Dual Card setup...

OMGWTFBBQ?!?!

Could it be there is the same damned thing happening on a single card as a dual card? Could it be that everyone is a little more influenced by what others tell them then by what they can actually prove?

On a side note, if this variation in frame output from one frame to the next is responsible for the so-called microstutter, then that would mean everyone on here can see the difference from one frame to the next while being shot at with 30-60 frames per second. For a test on that, in the span of ONE second, count from 1 to 60 and tell me how you do.

Very well done. So the question is, is the person who did the article stupid or do they have an agenda?

[Leeghoofd]

has anyone ever considered FPS lag via HD access etc ? Is the fraps thing 100% bulletproof ? Is it all reproduceable, noticeable when playing ? Isn't it all overhyped ?

Who runs a crossfire setup at 1920 x 1200 anyway ? if A single 5870/5970 can't handle that , then it's a poorly coded game engine or bad GPU driver for that particular game... I think it's time to kick the game encoders in the n*tts, then the hardware guys...

[eRazorzEDGE]

LMAO Leghoofd

No offense, but yes, several people have stated other reasons for the lag... which were all discussed in this thread that you should have read. Again, no offense.

There is no 5890, but who the hell goes to the moon? There's no real practical reason other than it's never been done before. I run what I run for the same reason a man pisses out a car window while driving down the freeway... sure you can hold it or pee in a cup, but it's more fun to hang your wang out of a moving vehicle. Just like it's more fun to get 9000 FPS than it is to only get 4500 FPS.