Techreport 7950 vs. GTX 660 Ti "Smoothness" videos

[MrK6]

It is obvious that both nvidia's and AMD's solutions suffer from microstutter. Anyone claiming one does not is on a fool's errand and trying to heal their (pathetically) misplaced ego.

Just a quick question and I apologize if its been hashed over the in the previous pages, but can youtube's 30fps (or is it 15?) show the MS issue in full light?

Most certainly, and I think at the least it makes it more difficult to appreciate. Also, YouTube encodes with its own "stabilization" process which I imagine would also affect the fidelity of the end result.

 

[RussianSensation]

Exactly. TR themselves said in their report and in the podcast that they found no difference with 12.8 to 12.11. 12.11 isn't responsible for the results they got.

I personally saw no difference in stutter from the launch beta to the driver I was using 6 months later, but I never tested 12.8 so there is a chance this driver made it better, otherwise its likely from what I saw its been there since launch.

You have 3 monitors and 2x GTX680s, correct?

The truth of the matter is GTX680s also have their fair share of issues as they run out of VRAM/memory bandwidth and SLI scaling is often not ideal for triple monitors. You shared great real world personal experience of using AMD cards and noticing micro-stuttering but gaming on 3 monitors NV cards can be just as awful in many games as well. There is no perfect solution right now. A 60 inch 8K resolution monitor with a GPU 100x more powerful than we have would be a solution

SLI and CF both have major issues in different games. On average with AA, 3xGTX680s are only as fast as 2x HD7970s on 3 monitors.

While GTX680/SLI work well on a single monitor in BF3, it's entirely different on 3 monitors.
http://www.computerbase.de/artikel/grafikkarten/2012/3x-gtx-680-und-3x-hd-7970-mit-i7-3970x/5/

With more frames on the 7970s in BF3, you can at least play around with Radeon Pro and Vsync to try to minimize stutter. With 680s you simply don't even have the GPU power and/or proper SLI scaling working.

There are also games where 680s run out of 2GB of VRAM and it's a slideshow. Would you be happy spending $800-1000 on 680s and then you can't even play games like Metro 2033 on 3 monitors?

Similarly, there are plenty of games where CF doesn't work well.

The best thing to do is research, research, research specifically for your resolution, your games and what framerates you desire in those games. Ask people around what it feels like gaming with Card(s) ABCD in games WXYZ. Read many different reviews and try to get a general idea of which games run better on what hardware. Share these experiences on forums. We know SLI > CF on average but sometimes it's much worse. Same story with HD7970Ghz vs. 680. This is why I find it amusing when people jump on the bandwagon that one brand is "cheating" or "stuttering" its way to winning benchmarks when both companies have these issues.

The way I look you can fix stuttering by either turn down settings or get a faster videocard. The way to get a faster videocard is to keep upgrading. Upgrading for most of us costs money. The best way to minimize upgrade costs is perhaps focusing on price/performance or reselling your cards and reinvesting some of the resale value into the new card. No one videocard will play all games better than the competing brand, so making sure one considers heavily the time spent on specific games he/she plays is pretty important. What makes things worse is often broken or poorly ported console ports like GTA IV or AC3 or excessive performance hit for minimal image quality increase in games like Sleeping Dogs with AA. Unfortunately I feel the only way forward is to keep upgrading. It's only a matter of time before micro-stutter will overwhelm any multi-GPU setup in next gen games, from NV or AMD. I agree that nV has taken better steps to minimizing micro-stuttering with hardware/driver frame metering but it's still a far cry from single GPU smoothness it seems.

 

[BrightCandle]

Since I own and test both I am kind of in unique place in this discussion. The odd thing is I have in general a lot more problems with performance on the 7970s in regards to microstutter. I very rarely have microstutter on the 680s and normally when I notice its actually a low frame rate. But I get lower frame rates on the 680s without a doubt. Benchmarks and play confirm it.

But I get to play the 680s in 5760x1200 a lot more than the 7970s. This is because most of the time microstutter is unplayable at those resolutions on the 7970s, which is not true on the 680s. So while performance is measurably worse actual gameplay experience and graphics settings turn out to be better on the 680s overall.

I have issues with negative scaling on both planet side 2 and mechwarrior online which in both cases have me reducing the graphics settings one notch. On the 7970s however I have multiple games that are unplayable regardless of what settings I use. The best of the 7970s is better than Nvidia but the worst is also worse. Nvidia ends up being better often and its worst is always playable. That to me is what actually matters, will it work when a new game comes out and it always has on nvidia since I got them. That is definitely not true on the 7970s, they were constantly bad enough to cause rage.

When we get better measurements I have no doubt well see frame metering as an essential technology.

 

[RussianSensation]

BrightCandle, I don't think most people will disagree that on average SLI > CF. What happens if you play the same games on 1 monitor on just 1 680 or 7970Ghz? Does your experience remain the same, that for most games you feel that you need higher FPS on the 7970 than 680 to get the same feeling of smoothness on 1 monitor?

 

[Final8ty]

Since I own and test both I am kind of in unique place in this discussion. The odd thing is I have in general a lot more problems with performance on the 7970s in regards to microstutter. I very rarely have microstutter on the 680s and normally when I notice its actually a low frame rate. But I get lower frame rates on the 680s without a doubt. Benchmarks and play confirm it.

But I get to play the 680s in 5760x1200 a lot more than the 7970s. This is because most of the time microstutter is unplayable at those resolutions on the 7970s, which is not true on the 680s. So while performance is measurably worse actual gameplay experience and graphics settings turn out to be better on the 680s overall.

I have issues with negative scaling on both planet side 2 and mechwarrior online which in both cases have me reducing the graphics settings one notch. On the 7970s however I have multiple games that are unplayable regardless of what settings I use. The best of the 7970s is better than Nvidia but the worst is also worse. Nvidia ends up being better often and its worst is always playable. That to me is what actually matters, will it work when a new game comes out and it always has on nvidia since I got them. That is definitely not true on the 7970s, they were constantly bad enough to cause rage.

When we get better measurements I have no doubt well see frame metering as an essential technology.

I'm not going to question your personal experience and neither do i have an issue with it, it's just at times that go on as yours is the definitive even though others also have had both setups and claim the opposite experience to yours.

So at the end on the day there are 3 possibilities, either they will get your experience, the other persons or even end up with something in between.

 

[BrightCandle]

BrightCandle, I don't think most people will disagree that on average SLI > CF. What happens if you play the same games on 1 monitor on just 1 680 or 7970Ghz? Does your experience remain the same, that for most games you feel that you need higher FPS on the 7970 than 680 to get the same feeling of smoothness on 1 monitor?

At normal resolutions there were still multiple games I found were unplayable if the frame rate ever differed from vsync, one fps below 60 was unplayable. That has however never the case on the 680s. They sometimes stutter but I only notice it around 45 fps.

One game that shows as a prime example of my experience is arma 2. The 7970s at the same settings are about 20% faster in fps. However on dual 7970s at normal and triple screens they are unplayable below 60 fps. The nvidia 680s however feel smooth down to about 40 fps. In practice that means I run the settings higher on the 680s and the game still runs better. The average I aim for is still 60 but I don't need to ensure there minimum is also 60 and that has a large impact on graphics quality.

What I found most strange when comparing the two was a single 7970 compared to a single 680. The wife has 1 of the 7970s in an ivy bridge machine so we both have high end rigs although they obviously aren't directly comparable machines. Playing the same game side by side shows clear differences in smoothness. Her monitor isn't dramatically worse than mine and does not show the same thing with nvidia so its quite odd. The frame times captured in fraps sometimes show the problem, they clearly show stutter. What is most odd is that sometimes its not smooth comparatively but the trace looks like we should be happy. So while frame times when bad always seem to correlate with bad experience a clean trace didn't always mean all was well.

My experience was that crossfire is mostly unplayable, 80% of the time it was worse than not having it. But even single cards compared the 680 is smoother in everything we compared. But the frame times only sometimes show that. The 7970 often performs better in fps but fooled us both less for motion. We have no interest in playing below 60 fps and with anything other than sync on so we are playing with a very different type of test to tech report. Our frame times show nice solid jumps to 33ms for the most part because of sync, but they also clearly show oscillations.

What is odd is that while the wife also sees better motion on the 680s she isn't interested in changing cards. She is happy with the 7970, had she not seen the difference she wouldn't have said the 7970 was bad. Since I play competitively in fps games I suspect I am more susceptible to my play being impacted by the effect.

What I want to do is get a high speed camera and see if I can capture the difference effectively. But I also recognise its something you notice more when you make the inputs, which presumably means latency is having some baring on the feeling of motion as well.

 

[BrightCandle]

I'm not going to question your personal experience and neither do i have an issue with it, it's just at times that go on as yours is the definitive even though others also have had both setups and claim the opposite experience to yours.

So at the end on the day there are 3 possibilities, either they will get your experience, the other persons or even end up with something in between.

I think they both have a problem. Something is wrong in the current generation of graphics cards. My current rigs clearly have problems with AMD cards, others have it with NVidia cards. Its clearly a marginal effect or limited to particular setups. I don't think because others don't see the effect they are blind to it, I think they genuinely don't have it.

My issue continues to be how to find and fix it, because I very much would like +20% on my cheaper cards, they just seem to dislike my rigs strongly.

I don't think people should necessarily take my data/experience and know out will happen for them. Even in January I believed it was a bug and had contacted both AMD and ASUS to see if together they could find and repair it. My experience with AMD support was less than stellar so I think that is something everyone would likely experience if they have the displeasure of talking with them. But had I got better support I could be telling a different story. As it is tr seem now to be in the same place I was and unable to work out how to make it work right like it must do on other peoples machines.

 

[Final8ty]

I think they both have a problem. Something is wrong in the current generation of graphics cards. My current rigs clearly have problems with AMD cards, others have it with NVidia cards. Its clearly a marginal effect or limited to particular setups. I don't think because others don't see the effect they are blind to it, I think they genuinely don't have it.

My issue continues to be how to find and fix it, because I very much would like +20% on my cheaper cards, they just seem to dislike my rigs strongly.

Indeed use what's best on your rig.
I know of a guy on Guru3D who has Quadfire 5970 like me and his experience was a complete nightmare.

Even though i don't own any series of this round they both seem to have too many issues for my liking.

 

[Cookie Monster]

PCGames hardware posted new videos of Far Cry 3 to test micro-stutter

-snip-

Wouldn't a 4 GPU setup be more prone to micro stutter? Isn't the comparison abit invalid from the getgo?

 

[MrK6]

Wouldn't a 4 GPU setup be more prone to micro stutter? Isn't the comparison abit invalid from the getgo?

Only if the drivers are a mess for whatever application you're testing. Adding more GPU's generally smooths out microstutter (although it may add just plain old stutter).

 

[SirPauly]

Wouldn't a 4 GPU setup be more prone to micro stutter? Isn't the comparison abit invalid from the getgo?

An older but outstanding article:

http://www.rage3d.com/reviews/video/ati4870x2cf/index.php?p=2

 

[Final8ty]

The default location for the skyrim.ini is: My Documents\My Games\Skyrim.
*edited with correct instructions to remove v-sync
There are 2 .ini files in this folder. You will need to make changes to both .ini files. In the SkrimPrefs.ini iPresentInterval should be located in the bottom of the [Display] section, just above the [Grass] section.
To remove v-sync from Skyrim:
Delete iPresentInterval=1 from SkyrimPrefs.ini and save changes.
Open Skyrim.ini and place iPresentInterval=0 in the bottom of the [Display] section and save changes.
I did not need to edit the .ini, because I left the v-sync at the Skyrim default of 60fps.

I did try that one as well as the other, no affect.
EDIT> Got it, did what was suggested earlier, delete the files and let steam reload them.
Lifting the cap improved game play immensely.
Will tweak the vsync settings and do more testing.

Mission accomplished...
I can now play without vomiting , no stutter, no input lag, just a smooth playing game.

http://www.overclock.net/t/1340733/...0-microstuttering-and-uneven-render-times/370