HD5970 Microstuttering tests

[trinibwoy]

You cant claim that all people notice Microstuttering & are ignoring it or lyingly.

Good thing I didn't then.

Also, from what other's have said, there is no control results and no frame rate results.

You can calculate frame rate yourself. It's just 1000/frame_time. And if you want control results you can also generate them yourself if you have fraps and a single GPU card.

[Helmore]

Actually a fraps graph is far more useful evidence than people claiming to see / not see something. Microstuttering is a well understood phenomenon, people just stick their heads in the sand to avoid acknowledging that their multi-GPU setups are pumping out useless numbers.

That FRAPS chart would only be useful if they had a control chart, with for example a 5870 rendering the same game and you can judge the difference. I'm not 100% sure such a graph would look 'smooth' on a 5870, until they prove it to me.

[trinibwoy]

It's your call but I'm not sure why you need someone to prove something to you that you can do yourself. I think there were some other reviews a few months/years ago with single-GPU graphs too so you might have some luck in google.

[JohnZS]

This is why I do not use a MultiGPU system.
Until ATi/nVidia move away from the traditional "AFR" rendering method for multiGPU communication this issue will remain a problem.
Setting Maximum Pre-rendered frames to 0 in the nVidia control pannel does help significantly.
What does annoy me is with each ATi multi GPU release there is always hype about a "true multi GPU link aka multithreading/HT/true dual core" however we are always let down as ATi still use the AFR technique.
Remember the hype re X4870X2 launch?
We got the same hype re: 5970 too. I wish ATi would deliver the goods and show the world how MultiGPU is meant to be played
John

[[XC] hipno650]

well it is finally nice to see some numbers on a large variety of popular games and benches that seem to shed a little more light on the situation. i have used a few Multi GPU setups and some i have noticed mircostuddering one and some i have not. For example i built a system for a buddy of mine using a CI7 920 at 4GHZ and a 4870 crossfire and the mircostuddering was pretty bad. While his system beat my 5870 system in almost all the numbers when benching my system still plays games much smoother overall even with the same or lower FPS. i have also used a 8800gtx SLI and noticed no mircostuddering at all. Hopefully though ATI and Nvidia will make an effort to fix the problem but i doubt it

[SKYMTL]

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....

[Final8ty]

I agree with you 100%

Coming from a heavy SLI user microstuttering is a major problem; Its happened across multiple setups and multiple monitors. I have personally seen it on:

6800 GS
GTX260 SLI
GTX260 192SP SLI
8800GT (G92) SLI
8800GTS (G80) SLI
HD3850 Xfire

Sometimes its noticable other times its not; I personally think its linked to the saturation of the PCI-E bus; not neccessarly on the GFX end of things in terms of lane saturation but data saturation at the southbridge.

The worst games for microstuttering in my personal experience have been MMOs and games that have large open worlds that have to access heavy amounts of CPU/Mem/HDD loads.

Enclosed areas the microstuttering goes entirely away; however the moment you step outside its really pronounced especially as I get near the 40FPS marker.

Given my experiences; Once Fermi is released I plan on shifting my focus from SLI on my main gaming rig to:

1 GFX card
1 PhysX card
1 PCI-E Raid controller

(for the reasons mentioned above)

I have had up-to quad 3870 & only noticed it in Farcry above 2 GPU.
But i do like to play my games with Vsync as i hate tearing or multiples of capped 60/120/180/240 if the option is there, minimal tearing maximum FPS no Vsync lag.

[Particle]

I agree with the notion that AFR is often beneficial. As someone who has actually used XF for an extended period of time on 4K cards, it's obvious beyond belief that for some games it makes a marked improvement in not only framerate but also user perceived fluidity. This isn't always the case, but for most of the gaming I've done it has been.

For those graphs to matter whatsoever, they'd have to show what a single GPU would have done at that same exact moment so a comparison about the differences in timing can be made. If you're going to compare to a single card in your conclusion (eg XF is worse than a single GPU), you MUST have data from a single card or the analysis is forfeit.

Last thought: If you don't like AFR, use supertiling. It may not always provide the best benefit, but you won't suffer from frame AFR frame timing issues. You can select whatever you want to use now.

[Final8ty]

Good thing I didn't then.



You can calculate frame rate yourself. It's just 1000/frame_time. And if you want control results you can also generate them yourself if you have fraps and a single GPU card.

Actually a fraps graph is far more useful evidence than people claiming to see / not see something. Microstuttering is a well understood phenomenon, people just stick their heads in the sand to avoid acknowledging that their multi-GPU setups are pumping out useless numbers.

Who are the few that you are talking about that you know who are just sticking their heads in the sand to avoid acknowledging that their multi-GPU setups pumping out useless numbers.

[KingOfsorroW]

2all I don't know, I'm not blind and definitely saw no microstuttering - everything was smooth like baby's a$$ with that 3 sli setup.
Macro stuttering - constant big fluctuations (the usual stuttering in another words)
microstuttering - small and frequent fluctuations (may happend randomly).
There was a lot of talks about microstuttering on the internet at the time of gtx295 -285 launch - the microstuttering is a myth. Because it is almost impossible to catch it for a human eye. People tend to confuse it with some other troubles like latency or the margin between the min and max fps for example...
But one thing for sure - one fast card is better than 2 or 3 cards alright.

[Final8ty]

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....

And that's all that matters.

[One_Hertz]

I knew it. I've been ing about CF being entirely worthless for a long time now.

I bought two 5870s as soon as they came out. Two weeks later I put one of my 5870s on the shelf because MS completely ruins it for me. I feel that single card is smoother and providers better gameplay for me. Even in a blind test I choose a single 5870. My second 5870 is still on the shelf months later, waiting for lucid hydra to get better.

[Katanai]

This microsluttering issue you speak of is not real.

Come on now, people...
NOTHING TO SEE HERE, MOVE ALONG!

[Manicdan]

as long as the lowest fps of the stuttering stays above 40fps, you probably will not notice it.

i think there is alot of training needed to know how to setup a PC to properly get as much benefit out of multi gpu setups, without having to noticed these kinds of issues.

for the crysis at 1920x1200, at its worse point was bouncing between 70fps and 50, but stuck right around 60. so the user feels 50fps at the WORST, hardly noticeable im sure

for B:AA there is constant jumping between 15 and 20ms and 1000/15 = 66fps and 1000/20 = 50fps, im starting to wonder if they left vsync on, due to how the average is so damn close to 60fps

with dirt 2 i have no idea what went wrong, i benched it with a 5850 everything maxed with 4xaa and same resolution, and averaged better framerates than they show. (im wondering if they were using 9.12 cats) i was expecting sub 10ms numbers, not an average of 40fps

unigine was a constant jump between 5 and 17ms or 200 to 60fps, dont think you can notice that.

i wish i had a second gpu so i could do some of these tests myself, im really curious because theres so many things to look at and everyones opinion of how noticeable it is will vary. i also dont think its worth doing it for 1-2 seconds worth of content. and i cant read the article itself, so i was hoping under each chart is a user opinion about what they saw when doing the test.

[Chickenfeed]

After using a 4870x2 from launch until septemberish and switching to a 5870, even though benchs often show the X2 ahead, the 5870 has felt night and day smoother. Hell when I spent a month with some 260 c216s the single 260 felt just as good if not better than the X2... This is just my own long term personal experiance using these cards. I've tested a lot of other gpus / configs but I hadn't done more then benched them ( eg didn't get a chance to play even a fraction of the games I own ) But I will mention, unless you compared the 2 I likely wouldn't notice it. Does that change the fact that this is a problem? Nope. It just makes me rethink any future move to a multi gpu setup given it is bad enough when you see roughly a 50% speed up ( in numbers ; I'd consider 50% a fair scaling average for a multi gpu system given my experiance) for 2x the cost, its even worse when the afforementioned speed up isn't always relavent.

[zalbard]

ok, AFR is crap, so what would be the "right way" to do multi GPU?

I'd like to know as well. Maybe Hydra does that? But I doubt it... Maybe it's in fact nearly impossible.

[Beefy22]

Perhaps some people can adjust their perception such that, over the course of time, they acclimate to MS?

After using a 4870x2 from launch until septemberish and switching to a 5870, even though benchs often show the X2 ahead, the 5870 has felt night and day smoother.

Is that MS or another phenomena altogether?

[MirageSys]

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....

I have had much the same experience that you have had, but I think some observations have been accurate, or wrongly classified as "microstutter".

My views:

1: Generally microstutter is not an issue. I have only had Cf'ed 4830's and 2 9800GTX in SLI. In most games I do not notice anything, but there are a few exceptions.

2: The majority of the "real" microstutter that people experience is from situations where FPS is either borderline 30 or very high. I feel Vsync definitely has some causation. I have noticed that while playing Battlefield Vietnam with SLI, it feels very sluggish. I have no alternate explanation, and must assume it is "microstutter".

3: I think some of the "microstutter" people do see is caused by CPU limitation as opposed to GPU limitations

[demonkevy666]

I agree with the notion that AFR is often beneficial. As someone who has actually used XF for an extended period of time on 4K cards, it's obvious beyond belief that for some games it makes a marked improvement in not only framerate but also user perceived fluidity. This isn't always the case, but for most of the gaming I've done it has been.

For those graphs to matter whatsoever, they'd have to show what a single GPU would have done at that same exact moment so a comparison about the differences in timing can be made. If you're going to compare to a single card in your conclusion (eg XF is worse than a single GPU), you MUST have data from a single card or the analysis is forfeit.

Last thought: If you don't like AFR, use supertiling. It may not always provide the best benefit, but you won't suffer from frame AFR frame timing issues. You can select whatever you want to use now.

aren't there like 4 other choices for above two card ?

[Rock&Roll]

This is why dual GPU cards should be communicating through an HT-esque bus, sharing memory and NOT using the same rendering ideas from the ATI Rage Fury MAXX of 10 years ago.

Amen

[trinibwoy]

as long as the lowest fps of the stuttering stays above 40fps, you probably will not notice it.

I agree, if it's fluctuating at the high end of the range then it should be very hard to detect. But that's just one part of the complaint. The other is purely academinc in that the reported average fps in reviews etc is misleading. A series that fluctuates between 40 and 80 fps has an average of 60fps but it's not the same as one that fluctuates between 55 and 65 and produces the same 60 average. The latter is far more desirable.

[Particle]

aren't there like 4 other choices for above two card ?

I'm not sure I understand your question. If you're asking if there are more modes than just AFR and supertiling, yes, there are. Supertiling is pretty much the end-game as far as getting rid of visual problems though, so I recommend it.

[STaRGaZeR]

The human eye can't see more than 60FPS anyway

[Rock&Roll]

Out of some nubile curiosity, has anyone tested to see if MS was mitigated when using SSD's? The idea being that if there was a bottleneck getting data from the hard drive to the gfx memory, that SSD would improve the situation.

[MirageSys]

Out of some nubile curiosity, has anyone tested to see if MS was mitigated when using SSD's? The idea being that if there was a bottleneck getting data from the hard drive to the gfx memory, that SSD would improve the situation.

This is totally improbable, as it would affect a single graphics card as well.