AMD Radeon HD 7990 Graphics Card Reviews

[STEvil]

Baba is right though. The original German term only covers AFR stuttering - and from that micro stutter was derived. I'm pretty sure this term was conceived as early as 2007 when tombman, an Austrian hardware enthusiast (RIP) was on of the first if not THE first to discover the phenomenon and made it public at 3dcenter:
http://www.forum-3dcenter.org/vbulle...d.php?t=371844

The issue's been around since the X1000's (hint: R520/X1800, in 2005) in its "current" form. Let alone good old 3dfx SLI vs nVidia TNT2 debates and 3dfx's "30 vs 60" demo... and then "stepdaughter/PGP" from Metabyte..

If you really want to go back to where it all originated I can point out to you that we've been having "Intel vs AMD" smoothness threads for you since the P2/3 era, which, surprisingly enough, was GPU indipendant.

Really, it blows my mind how uninformed the current generation is on the issue.

[BababooeyHTJ]

Yeah, ok. Throw out insults and get all condescending when you don't have a reply.

The fact of the matter is the term microsutter which is a different issue than you are describing didn't exist that far back.

Also did you read that techreport article yet? Or are you above that?

[jimba86]

is there a source which shows how bad the mricostutter is on these dual GPU cards such as this 7990?

I read stuff on forums like this but never see any hard facts.

[STEvil]

Yeah, ok. Throw out insults and get all condescending when you don't have a reply.

The fact of the matter is the term microsutter which is a different issue than you are describing didn't exist that far back.

Also did you read that techreport article yet? Or are you above that?

Its the same as it was a year ago, and the same as it was last night, why would I read it a third time?

[STEvil]

is there a source which shows how bad the mricostutter is on these dual GPU cards such as this 7990?

I read stuff on forums like this but never see any hard facts.

There are SOME but things change, and system setup greatly affects microstutter sometimes, so while it may not be perceptible in a certain game with a dual-gpu card, it could become perceptible in that same game with a single gpu card on a different system.

Again, the issue on the single GPU card would be something system related while the dual-gpu could be AFR timing or system timing. Microstutter encompasses many issues to create the problem. AFR stutter is a single issue.

[boxleitnerb]

The issue's been around since the X1000's (hint: R520/X1800, in 2005) in its "current" form. Let alone good old 3dfx SLI vs nVidia TNT2 debates and 3dfx's "30 vs 60" demo... and then "stepdaughter/PGP" from Metabyte..

If you really want to go back to where it all originated I can point out to you that we've been having "Intel vs AMD" smoothness threads for you since the P2/3 era, which, surprisingly enough, was GPU indipendant.

Really, it blows my mind how uninformed the current generation is on the issue.

Then please point out a forum discussion/article not related to AFR where the term "microstutter" was used before 2006.
I don't deny that there are many types of stuttering that have occured long long ago. I only deny that the term "microstutter" was used back then for phenomena such as these as a generally accepted term (and not an anomaly that was mentioned maybe twice in 10 years). Prove me wrong if you can.

[AMDDeathstar]

Not a word on microstuttering - only half a review. Seriously, what is up with that?

Is it so difficult to record some frametimes AND put 2-3 people in front of the keyboard, let them play a couple of games for a few minutes and have them write down their experiences in terms of smoothness in 1-2 sentences for each game? It's not like multi-GPU is tested on a daily basis. So if you do it 2-3 times a year, at least do it right and as thorougly as possible. Geez, the continued lack of coverage on microstuttering/frametime consistency is becoming really annoying. It is an integral part of AFR, so test it!

That is absolutely true. Even more - it is a nature law: Whenever there are two frequencies, they always create 3rd frequency as differential. It is particularly seen when say, monitor is 60Hz refresh rate and card is outputting frames at few more or less, e.g. 57Hz. It would create 3Hz. This could create a tearing, or stutter. And note: this happens even when GPU outputs stable framerate.

This effect particularly well seen when used FRAPS. Or when converting video in any combination of widely used in movies FPS': 24 , 25 and 30 (unless any special technique used, e.g telecine / inverse telecine, multiplying frame rates)

From the Techreport article

Fraps measures the same degree of jitter we saw initially, but try as I might, I can't see the problem. We may need to spend more time with (ugh) faster TN panels, rather than our prettier and slower IPS displays, in order to get a better feel for the stuttering issue.

Microstuttering might be happening but your monitor may cancel it out. Or the otherside is there is no mircostutter but but because of oddball frame rate vs refresh rate it looks like microstuttering.

The Fraps redender times don't refect what is is being put out on the screen. A monitor makes a draw call every 16.7msec(@60Hz refresh rate), depending on the render time some frames get doubled up and/or some get dropped.

Then throw in input lag and pixel reponse time, you could have worse cards for microstuttering but you would never know it just because of your monitor.

Don't forget game engines, CPU/motherboards and DirectX versions (DX9 wasn't made to work with multiply GPU's) plays a part in this too.

Question. If one reveiwer overclocks their system and see microstuttering and a second reveiwer with same system but stock speeds doesn't see mircostuttering. Whose review is more valid and is it up AMD/nVidia to fix it?

AMD and nVidia by themselves can not fix this problem it would require MSFT(DirectX),Kronos(OpenGL),Intel and AMD(CPU/motherboards), monitor manufactures and game developers working together.

The only way to get rid of mircostuttering is if everyone agrees that rendering time matches refresh rate and come to agreement on some type of methodlgy to acheive that.
Good luck with that.

[boxleitnerb]

When you test SLI/CF on the same system with the same display and the same software, you can very isolate AFR stuttering and isolate it. I also have never heard that overclocking has an influence on microstuttering (if both GPUs have the same OC clocks of course). And I have S-IPS panels since 2002 and notice microstuttering just fine. These are all excuses not to cover the topic and keep the same inadequate testing methodology. Almost all reviewers aren't even making an effort, and that is just sad. The only ones that do are techreport, HardOCP and on the German side PCGH, HT4U and Computerbase.

Anywhere else, the pitfalls of AFR are not mentioned with a single word. People reading those reviews don't even get told that their SLI/CF uses AFR and what that means - even though it is basic stuff to know. You are told that card A has 50fps and card B 40fps, but what does that mean, how relevant is this if playability and smoothness doesn't correspond to those fps rates between those two AFR-cards? And it is certainly not comparable to single GPUs that are always listed as a comparison.

[BababooeyHTJ]

There are SOME but things change, and system setup greatly affects microstutter sometimes, so while it may not be perceptible in a certain game with a dual-gpu card, it could become perceptible in that same game with a single gpu card on a different system.

Again, the issue on the single GPU card would be something system related while the dual-gpu could be AFR timing or system timing. Microstutter encompasses many issues to create the problem. AFR stutter is a single issue.

How would you describe the term microstutter and can you provide some links to a discussion or article that backs up that description?

I think that techreport article does a good job at showing off these inconsistent frametimes on a single gpu. At its worst it still happens differently than the "AFR microstutter" that happens every other frame very consistently for obvious reasons. I don't think that the two will look the same. Even the cpu bottleneck in BF3 that was posted earlier looks far different than the very consistent microstutter graphs that we have all seen. I would imagine that looked more like a regular stutter.

Here is a good example mentioned in the techreport GTX690 review.

Multi-GPU schemes generally divide the work by asking a pair of GPUs to render frames in alternating fashion—frame 1 to GPU 0, frame 2 to GPU 1, frame 3 to GPU 0, and so on. The trouble is, the two GPUs aren't always in sync with one another. Instead of producing a series of relatively consistent frame delivery times, a pair of GPUs using alternate frame rendering will sometimes oscillate between low-latency frames and high-latency frames.

To illustrate, we can zoom in on a very small chunk of one of our test runs for this review. First, here's how the frame times look on a single-GPU solution:



Although frame times vary slightly on the single-GPU setup, the differences are pretty small during this short window of time. Meanwhile, look what happens on a CrossFire setup using two of the same GPU:



You can see that alternating pattern, with a short frame time followed by a long one. That's micro-stuttering, and it's a potentially serious performance issue. If you were simply to measure this solution's performance in average frames per second, of course, it would look pretty good. Lots of frames are being produced. However, our sense is that the smoothness of the game's animation will be limited by those longer frame times. In this short window, adding a second GPU appears to reduce long-latency frames from about 29 ms to about 23 ms. Although the FPS average might be nearly doubled by the presence of all of those low-latency frames, the real, perceived impact of adding a second card would be much less than a doubling of performance.

This problem affects both SLI and CrossFire, including multi-GPU graphics cards like the GTX 690. How much micro-stuttering you find can vary from one moment to the next. In this example, we can see a little bit of jitter from the GTX 690, but it's fairly minimal.

[STEvil]

Then please point out a forum discussion/article not related to AFR where the term "microstutter" was used before 2006.
I don't deny that there are many types of stuttering that have occured long long ago. I only deny that the term "microstutter" was used back then for phenomena such as these as a generally accepted term (and not an anomaly that was mentioned maybe twice in 10 years). Prove me wrong if you can.

http://forum.doom9.org/showthread.php?t=90158 2005, discussion on video playback. Not even gaming related. HMMMMMMMMMMMMMMM........

When you test SLI/CF on the same system with the same display and the same software, you can very isolate AFR stuttering and isolate it. I also have never heard that overclocking has an influence on microstuttering (if both GPUs have the same OC clocks of course). And I have S-IPS panels since 2002 and notice microstuttering just fine. These are all excuses not to cover the topic and keep the same inadequate testing methodology. Almost all reviewers aren't even making an effort, and that is just sad. The only ones that do are techreport, HardOCP and on the German side PCGH, HT4U and Computerbase.

They are not exuses, they are alternate issues which cause microstutter. Microstutter can be caused by anything in the system or the program it is running.

Anywhere else, the pitfalls of AFR are not mentioned with a single word. People reading those reviews don't even get told that their SLI/CF uses AFR and what that means - even though it is basic stuff to know. You are told that card A has 50fps and card B 40fps, but what does that mean, how relevant is this if playability and smoothness doesn't correspond to those fps rates between those two AFR-cards? And it is certainly not comparable to single GPUs that are always listed as a comparison.

There are some reviews that cover this and some very well done threads on this forum (CallsignVega has a few iirc). Yes, there are better ways to do things that I would greatly enjoy seeing done and there are issues that ARE NOT BEING ADDRESSED BY THE INDUSTRY...

How would you describe the term microstutter and can you provide some links to a discussion or article that backs up that description?

A short stutter. Likely less than 100ms worth, but i'm sure there's some wiggle room depending on the person.
2005 (Video playback) http://forum.doom9.org/showthread.php?t=90158
2005 (Flight Simulator 2004) http://www.tomshardware.com/forum/18...l-texture-size
2005 (Creative PCI latency) http://forums.creative.com/showthread.php?t=131285
2005 (Creative PCI latency) http://forums.creative.com/showthread.php?t=23568

I think that techreport article does a good job at showing off these inconsistent frametimes on a single gpu. At its worst it still happens differently than the "AFR microstutter" that happens every other frame very consistently for obvious reasons. I don't think that the two will look the same. Even the cpu bottleneck in BF3 that was posted earlier looks far different than the very consistent microstutter graphs that we have all seen. I would imagine that looked more like a regular stutter.

Here is a good example mentioned in the techreport GTX690 review.

It does a decent job, but does not attempt to look deeper into the issue. They test stock settings and do not test for settings in the drivers or games or system that could influence the results..

[boxleitnerb]

http://forum.doom9.org/showthread.php?t=90158 2005, discussion on video playback. Not even gaming related. HMMMMMMMMMMMMMMM........

One down, 99 to go. I'm pretty sure you cannot prove that this term was widely/generally used for anything aside from AFR stuttering.

They are not exuses, they are alternate issues which cause microstutter. Microstutter can be caused by anything in the system or the program it is running.

There are some reviews that cover this and some very well done threads on this forum (CallsignVega has a few iirc). Yes, there are better ways to do things that I would greatly enjoy seeing done and there are issues that ARE NOT BEING ADDRESSED BY THE INDUSTRY...

When my games are running smooth without SLI/CF (at lower fps) and stutter once I turn them on, it is pretty clear...
The large sites like Anandtech, Tomshardware, Techpowerup don't mention anything in their review articles. Toms did a special microstutter article a while back, but that was pretty flawed and much raw frametime data was missing. When you read a SLI/CF review, in 98% of the cases you will not read the terms microstutter, profiles, AFR. They test these setups as if they were comparable to single GPUs - they are not.

[BababooeyHTJ]

It does a decent job, but does not attempt to look deeper into the issue. They test stock settings and do not test for settings in the drivers or games or system that could influence the results..

Have you personally experienced microsutter and by microstutter I mean the "afr stutter" as you call it? When you post things like a thread titled "Please help me CREATIVE !!! Audigy 2 ZS causes hiccup/stutter/freeze during game play" and label that as a microsutter it makes me wonder. The two are a completely different sensation. Microstutter does not look like a stutter. It normally just looks like a lower framerate than something like fraps will report. Its not a normal stutter and not everyone is all that sensitive to it.

[Manicdan]

It normally just looks like a lower framerate than something like fraps will report. Its not a normal stutter and not everyone is all that sensitive to it.

this.
many people call lots of other problems micro stutter and were dead wrong. micro stutter will generally only be noticeable to users if the framerate is low.
the causes of it however are far and wide. we have seen weird problems where weaker cpus or certain northbridges were the culprit instead of blaming just the gpu and drivers. weve also seen how adding a third card gives only a smaller boost in performance, while fixing the problem completely. and weve also been told by nvidia that how its measured by fraps is wrong since it only calculates the time between certain points of the process and not the whole thing.

i had only 1 setup with dual gpus and it ended up being rather annoying, so i stick to single stronger cards now.

[BababooeyHTJ]

Yeah, I plan on picking up a third card just because I've always heard that it really helps to alleviate microsutter. When you think about it, it really makes sense why it would.

Didn't you use 3870 crossfire? I never used that combo personally but I did own a 4870x2 for a bit and sold it due to microsutter. Actually, I didn't know that was microstutter at the time but in a lot of games despite the higher framerate it didn't feel any smoother than my GTX280. I didn't realize until later that was microsutter. When I later tried 6950 crossfire it was much less of an issue for me. Don't get me wrong I did notice it in a few games. Anyways, I find it to be less of an issue than it was in the past.

[Andrea deluxe]

most of the people have a 60hz monitor with a poor response time.... and never see stutter because of it

people like me with 120hz monitor are able to catch more framerate problem!

but when the game run fine im play like a god!

[boxleitnerb]

Microstutter can indeed manifest itself as stuttering. In Mirrors Edge for example 50fps with my GTX580 SLI feel stuttery. If they were to feel like 30 single GPU fps, it would be "okay", but in fact, 30 single GPU fps feel better. But I have to add that this was with a custom SLI profile that increased performance by 15-20%. It may very well be due to that and another profile would feel better.
In Skyrim I feel microstutter below 40-45fps, and again I feel them as real stutter. It's like a hiccup, very small but noticeable pauses. With a single GPU, all is smooth.

[BababooeyHTJ]

The both of those could be an actual stutter introduced by the sli profile though.

[boxleitnerb]

With Mirrors Edge I'm not sure, could be. Skyrim on the other hand is the typical AFR stutter. Maximum GPU load, uneven frametimes. Introduce a CPU bottleneck (ugrids = 11 and/or disabling SGSSAA) and it is smooth with SLI at similar fps. I'm going to get some GK110 cards when they hit and I'm very curious how the hardware based frame-metering is working out on Kepler.

[BababooeyHTJ]

Going by the techreport GTX690 review I'm not really sure if the frame metering is also in effect with the single gpu cards. The frametimes seem to be consistently higher on the 680s in sli. I've never seen any solid confirmation on that.

[STEvil]

One down, 99 to go. I'm pretty sure you cannot prove that this term was widely/generally used for anything aside from AFR stuttering.

You only asked for one. One instance of it being used before the german article is enough. The terminology is not locked to AFR stutter, thats all there is to it. AFR stutter can be microstutter, but maybe we should call it millistutter to better reflect that its happening in milliseconds?

Here's a lesson in Etymology for you: Just because the current definition of a word has a meaning doesnt mean it was never used before nor that the current use is the "correct" use of the word.

When my games are running smooth without SLI/CF (at lower fps) and stutter once I turn them on, it is pretty clear...
The large sites like Anandtech, Tomshardware, Techpowerup don't mention anything in their review articles. Toms did a special microstutter article a while back, but that was pretty flawed and much raw frametime data was missing. When you read a SLI/CF review, in 98% of the cases you will not read the terms microstutter, profiles, AFR. They test these setups as if they were comparable to single GPUs - they are not.

See previous reply which you dont seem to have read in full.

Have you personally experienced microsutter and by microstutter I mean the "afr stutter" as you call it? When you post things like a thread titled "Please help me CREATIVE !!! Audigy 2 ZS causes hiccup/stutter/freeze during game play" and label that as a microsutter it makes me wonder. The two are a completely different sensation. Microstutter does not look like a stutter. It normally just looks like a lower framerate than something like fraps will report. Its not a normal stutter and not everyone is all that sensitive to it.

I've owned SLI Voodoo 2's, Crossfire X1600XT's, SLI 8800GTS's, a 4870X2, Crossfire 3870x2's, and am now running SLI GTX670's. Yeah, i've experienced microstutter. I've also owned creative sound cards and done PCI latency testing, which surprisingly enough, caused hitching and stuttering quite similar to what AFR stuttering looks like, and sometimes was much worse, to the point I wouldnt call it microstutter.

Yeah, I plan on picking up a third card just because I've always heard that it really helps to alleviate microsutter. When you think about it, it really makes sense why it would.

Care to explain? I'm sure ATi gave us the same line with the MAXX and double/triple buffering if I remember my "old days" correctly.

[boxleitnerb]

You only asked for one. One instance of it being used before the german article is enough. The terminology is not locked to AFR stutter, thats all there is to it. AFR stutter can be microstutter, but maybe we should call it millistutter to better reflect that its happening in milliseconds?

Here's a lesson in Etymology for you: Just because the current definition of a word has a meaning doesnt mean it was never used before nor that the current use is the "correct" use of the word.

I said "generally accepted term". Today when we talk about microstutter, in 99% of all cases we mean AFR stutter, so that is the generally correct use of the word. Anyway, this discussion is beside the point. The point is, that reviewers have a very very long way to go to properly test MGPU. I would take 680 SLI over 7970 CF any day of the week - and not for fps reasons, which is the only thing most reviews show.

See previous reply which you dont seem to have read in full.

Yes I have. If the only thing that changes from "smooth" to "stutter" is activating SLI/CF, the other potential issues are irrelevant. If there were issues of this sort that you described, they would occur with one GPU, too. If they don't (they never ever have on my systems), the conclusion is crystal clear.

[STEvil]

I said "generally accepted term". Today when we talk about microstutter, in 99% of all cases we mean AFR stutter, so that is the generally correct use of the word. Anyway, this discussion is beside the point. The point is, that reviewers have a very very long way to go to properly test MGPU. I would take 680 SLI over 7970 CF any day of the week - and not for fps reasons, which is the only thing most reviews show.

It is A use of the word, but not the ONLY use of the word, let alone the FIRST use of the word. Same goes for gay, :banana::banana::banana:, and hundreds of other words. Just because you only have an issue with AFR stutter and have never had another issue does not mean the word has never been used before to describe other issues.

Today when you talk about microstutter and only link it to a single problem (AFR stutter) you are ignoring the rest of the issue. You are doing what all those reviewers you have a problem with are doing and only scraping the surface.

Yes I have. If the only thing that changes from "smooth" to "stutter" is activating SLI/CF, the other potential issues are irrelevant. If there were issues of this sort that you described, they would occur with one GPU, too. If they don't (they never ever have on my systems), the conclusion is crystal clear.

No, you did not read it because I am pointing out other issues, not AFR only issues. If you fix your issue by turning SLI/Crossfire off, then great, that was your issue and it did not relate to the other causes of microstutter.

Would you be so kind as to point out where in my posts I have pointed out your problems with AFR are more than one thing being combined?

[boxleitnerb]

It is A use of the word, but not the ONLY use of the word, let alone the FIRST use of the word. Same goes for gay, :banana::banana::banana:, and hundreds of other words. Just because you only have an issue with AFR stutter and have never had another issue does not mean the word has never been used before to describe other issues.

Today when you talk about microstutter and only link it to a single problem (AFR stutter) you are ignoring the rest of the issue. You are doing what all those reviewers you have a problem with are doing and only scraping the surface.



No, you did not read it because I am pointing out other issues, not AFR only issues. If you fix your issue by turning SLI/Crossfire off, then great, that was your issue and it did not relate to the other causes of microstutter.

We are talking about MGPU and the 7990 and the reviews of MGPU (see thread name). Please observe the context. How relevant are those other issues then, really? It is widely accepted that AFR has stuttering issues compared to single GPU. And it is also widely accepted that the term for this is microstutter. The term may also apply to other issues, but its main applications - especially in the discussion in THIS thread - is AFR stutter.

No, you don't (want to?) understand what I'm saying:
If one has SLI/CF and experiences stutter, it is very very likely that AFR is the cause. That doesn't mean that other issues don't exist at all, but in this context, they are highly unlikely, especially if no stuttering occurs with SLI/CF off, but when it does with it on. Simple cause and effect. If this is observable in many cases (which it is), why would one think the problem is somewhere else (engine, IRQ sharing and what not)?

Would you be so kind as to point out where in my posts I have pointed out your problems with AFR are more than one thing being combined?

You seem to forget that you indicated, microstutter (AFR) could not be observed/investigated independently from other causes:

Microstutter can be caused by anything in the system or the program it is running.

I already said that that is highly unlikely. Look at the thousands of user reviews/experience reports with SLI/CF and the videos and the few review that do investigate/comment on this matter - do you really think that with all these sources, there was something else that was causing the stuttering besides AFR? In dozens of games, with many generations of cards and many many different setups? The only common element is...AFR.

[STEvil]

We are talking about MGPU and the 7990 and the reviews of MGPU (see thread name). Please observe the context. How relevant are those other issues then, really? It is widely accepted that AFR has stuttering issues compared to single GPU. And it is also widely accepted that the term for this is microstutter. The term may also apply to other issues, but its main applications - especially in the discussion in THIS thread - is AFR stutter.

The topic of discussion is the term microstutter. The thread was about 7990 originally, then moved to microstutter as affected by AFR and shortly after moved to additional causes of microstutter. If you cannot follow that then you should not be here.

No, you don't (want to?) understand what I'm saying:
If one has SLI/CF and experiences stutter, it is very very likely that AFR is the cause. That doesn't mean that other issues don't exist at all, but in this context, they are highly unlikely, especially if no stuttering occurs with SLI/CF off, but when it does with it on. Simple cause and effect. If this is observable in many cases (which it is), why would one think the problem is somewhere else (engine, IRQ sharing and what not)?

Please point out where I have said that SLI/Crossfire cannot be the sole cause of microstutter and is not the likely culprit in a system which has a SLI or Crossfire setup or that the likiness of some other cause being the culprit is larger than SLI or Crossfire causing the issue.

You seem to forget that you indicated, microstutter (AFR) could not be observed/investigated independently from other causes:

The quote quite flatly states that there are other causes to be aware of, that is all. It does not state that they always intermix. How could it be read any different?

I already said that that is highly unlikely. Look at the thousands of user reviews/experience reports with SLI/CF and the videos and the few review that do investigate/comment on this matter - do you really think that with all these sources, there was something else that was causing the stuttering besides AFR? In dozens of games, with many generations of cards and many many different setups? The only common element is...AFR.

You know what else was common in other instances? Flatout: Ultimate Carnage (data streaming for physics items). Creative sound cards. VIA northbridge/southbridge (PCI latency). DiRT with PhysX on GPU enabled (single or multi-card). Half-Life 2 GCF fragmentation (BSP streaming). Far Cry texture/data streaming collisions (high resolution induced). Trackmania texture/data streaming (resolution/player count induced) collisions. Operation Flashpoint texture/data streaming collisions (resolution induced, usually cause mouse input lag but not always). GTA3 texture/data streaming collisions. 3dfx Voodoo 5 5500 "AGP2x" bug (usually fixed by testing other AGP driving values).

[boxleitnerb]

Don't get insulting! You don't have the right to tell someone that he should not partake in the discussion.
You sidelined the topic with your additional stutter-causes that are simply not relevant in this thread about AFR-based setups (in this case the 7990).

As for your bold paragraph:
You implied it several times, always bringing other unlikely causes into the discussion when the discussion itself really is about AFR. You said your part, okay, but that doesn't change the fact that you cannot test MGPU as it is tested right now.

And your last paragraph:
Your examples consist of a miniscule amount of games and problems that existed years and years ago, and I'm pretty sure every single one of these issues didn't exist on every system or even on a majority of systems.
On the other hand, AFR induced microstutter can be shown in almost every game on every system. It's just a question of GPU load and framerate.