Micro-Stuttering And GPU Scaling In CrossFire And SLI

[Final8ty]

We've received many emails from readers asking about the phenomenon known as micro-stuttering and what it means to multi-GPU setups in CrossFire and SLI. After running plenty of benchmarks, we're ready to weigh in on what turns out to be a real issue.

Single GPU or multi-card setup? That's a question we face every time we start a System Builder Marathon series or evaluate the worth of a flagship GPU.

Do you want a just one high-strung racehorse or a pair, trio, or quartet of draft horses? Can a team of inexpensive cards perform the work of a pricier one and still come in at a lower cumulative cost?

Welcome to Groundhog Day. Due to recurring forum questions and direct requests by our readers, we decided that it's time to go beyond the usual performance-oriented benchmarks of CrossFire- and SLI-based systems, and shed some light on the underlying principles. Frame rates in and of themselves do not necessarily translate into a high-quality experience.

Yes, we're going to tackle the issue of micro-stuttering, which seems to keep so many sensitive eyes from investing in multiple cards running in concert. We'll also look at the scaling of two, three, and four GPUs. Where is the benefit? And at what point is actual added value really realized, or is churning out high (but ultimately useless) frame rates a self-defeating exercise? As we're sure you can imagine, at some point, the pursuit of performance can become a money pit and a power hog. At the same time, we've seen multi-card setups yield incredible value.

http://www.tomshardware.co.uk/radeon...iew-32256.html

[BababooeyHTJ]

Let’s revisit that PowerColor Radeon HD 6870 X2 for a moment. This card places two Barts GPUs on a single PCB and connects them through the use of a bridge chip from Lucidlogix.

Sort of skews the results imo.

[eXa]

Now, let's compare the behaviour of the Radeon HD 6870 X2 with a pair of regular Radeon HD 6870s connected with a CrossFire bridge. It turns out that the average frame rates are comparable. However, the setup without Lucidlogix's bridge chip demonstrates slightly smaller performance dips.

Didnt read the article did you?

[mAJORD]

It would seem to me the conclusion to that article is if you want higher FPS for the purposes of smoother gameplay (and not for getting off on FPS numbers) then crossfire/sli is a waste of time if a single card will consistantly give you say 50 +FPS

[eXa]

Or tri-xfire >_> weirdly enough.

[Manicdan]

Or tri-xfire >_> weirdly enough.

exactly, i hope they contact AMD and try to get some insight on why this is.

[bamtan2]

good visualizations:





but they still didn't hit the mark describing those measurements. it is like they come close to covering the whole story, but bounce off a bit.

they need to measure instantaneous frame rates (measured in ms, like the pie chart), and then graph the whole benchmark run. fraps can measure this. do some math to calculate how long each frame takes to render. compare that to the 30fps-equivalent (33ms?) and score down whoever goes under it the most often, or by the largest total amount, etc. or create a microstutter index by taking the average or minimum fps of a benchmark, then measure how many frames are rendered slower than that fps-equivalent. this would show how misleading the fps number was by showing the per-frame rendering speed variation.

instantaneous frame rate is what matters here. if you render one frame at 20fps-equivalent, then another at 40fps-equivalent, the average is 30fps, but that is irrelevant. you never hit 30fps. and you look like 20fps.

[NapalmV5]

lol smoother gameplay @ single gpu ?? hahaha i never ever had smooth gameplay on single gpu!

whoever started this bs about micro stutter/stutter/everything bad about multi gpus needs their arses whooped hard

its simple really.. cards get properly power supplied you got nothing to worry about

if it still stutters its the games or driverss fault

single gpu cards are not immune from this

[BababooeyHTJ]

exactly, i hope they contact AMD and try to get some insight on why this is.

Rage3d came to the same conclusion in their article from a couple of years ago along with an explanation as to why. It really is a good read.

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

Didnt read the article did you?

No, thats kind of where I stopped reading. Thanks for pointing that out to me.

[Sesto Sento]

exactly, i hope they contact AMD and try to get some insight on why this is.

Source: the comments on the review :P

alangeering 24/08/2011 12:19
Hide
-0+

I'd like to first thank Igor and Greg for a very insightful article and for

discussing the not often talked about phenomenon of stuttering.

There's one thing I'd like to expand upon.

A few times in the article the observation is made that while dual GPU scaling is

good, the stuttering effect is bad.
No real point is made that when scaling is poor, stuttering is less pronounced.

It's precisely because three cards aren’t as efficient that stuttering is reduced.

Bear with me and I'll explain.


For the following thought experiment I've used the data from the Call of Juarez

graph on the page called "Step 2: Crossfire with three GPUs"

Three situations:
A: 1 card @ 70 fps average
B: 2 cards @ 135 fps average
c: 3 cards @ 160 fps average

In other words:
A: The card takes an average of 14.3 ms to produce the frame.
B: Each card has 14.8 ms to produce the frame to maintain the average.
C: Each card has 18.8 ms to produce the frame to maintain the average.

Look again at the data from Call of Juarez.
The lowest frame rate recorded for the single card is 60fps or 16.7 ms per frame.

This is well below the 14.8 ms required to not delay/stutter the pipeline in

situation B but...
This is well within the 18.8 ms time frame for the 3 card set up in situation C.

As frames are now arriving in time for use, the evidence of stuttering is reduced.

So efficiency is good; but inefficiency in scaling allows each card a little

longer to provide its frame, and the eventual combined frame rate is less

variable.

A quote from the article:
"This phenomenon manifests itself even more seriously in CoJ. While CrossFire

scales well under load, it becomes even more susceptible to micro-stuttering."

And another:
"For some reason, the third GPU almost always eliminates micro stuttering and has

a less-pronounced effect on performance."

You got so close; it just needed another jump of statistical thinking. Efficiency

correlates with stuttering (NVIDIA and AMD) and there is a logical reason why.
alangeering 24/08/2011 12:37
Hide
-0+

The above post isn't trying to explain why microstuttering occurs - only why it's more pronounced as multi-gpu scaling increases. (and less so as scaling efficiency decreases)



P.S. The explanation of this guy, kinda makes sense to me :P

[STEvil]

lol smoother gameplay @ single gpu ?? hahaha i never ever had smooth gameplay on single gpu!

whoever started this bs about micro stutter/stutter/everything bad about multi gpus needs their arses whooped hard

its simple really.. cards get properly power supplied you got nothing to worry about

if it still stutters its the games or driverss fault

single gpu cards are not immune from this

I agree, single GPU's are not immune.

edit

For the record i've seen the phenomenon induced by things outside of the GPU relation, as in chipset drivers and pagefile size.

[saaya]

wow, i thought this was an old thread somebody revived...
cant believe this is still such a big issue...

shocking to see that amd is having so many probs with multi gpu considering their entire strategy is to have 2 gpus as a highend part while nvidia has both, a super highend single gpu and dual gpu and their dual gpu is an alternative solution...

[Manicdan]

if amd is having problems, then why is 3gpus the best solution in price and performance and reduced stuttering?
and nvidia only did well because the scaling seemed to be reduced to match the minimum that stuttering would have shown

both companies still have alot of work to do here

[eXa]

Manicdan:
Sesto Sentos quote seemes quite plausible to me and if true would explain your question.

[Manicdan]

Manicdan:
Sesto Sentos quote seemes quite plausible to me and if true would explain your question.

yeah i saw that, i was basically telling saaya that amd cards are not behind nvidias with multi gpu configurations

[eXa]

Oh, i see

[BababooeyHTJ]

yeah i saw that, i was basically telling saaya that amd cards are not behind nvidias with multi gpu configurations

Well, that was only one game its hard to draw too many conclusions from that.

[Manicdan]

Well, that was only one game its hard to draw too many conclusions from that.

they said:
"We repeated this same test using a number of games yielding different average frame rates. So long as the average is above 30 FPS, the resulting output is almost stutter-free. Minor differences in when frames hit the screen are not apparent, and so they don't bother us at all. The experience is similar to using a single GPU.

After analysing the dual-card configuration, this is certainly not what we were expecting."

[BababooeyHTJ]

they said:
"We repeated this same test using a number of games yielding different average frame rates. So long as the average is above 30 FPS, the resulting output is almost stutter-free. Minor differences in when frames hit the screen are not apparent, and so they don't bother us at all. The experience is similar to using a single GPU.

After analysing the dual-card configuration, this is certainly not what we were expecting."

I disagree with that. I saw microstutter on a 6950 crossfire rig at a little over 40fps at times, don't get me wrong that wasn't the norm but it did happen occasionally.

I have a problem with this statement.

We repeated this same test using a number of games yielding different average frame rates. So long as the average is above 30 FPS, the resulting output is almost stutter-free.

Microstutter is not a stutter, at least not a noticeable stutter. The game just feels more sluggish than it should at a given framerate.

[Mungri]

lol smoother gameplay @ single gpu ?? hahaha i never ever had smooth gameplay on single gpu!

whoever started this bs about micro stutter/stutter/everything bad about multi gpus needs their arses whooped hard

its simple really.. cards get properly power supplied you got nothing to worry about

if it still stutters its the games or driverss fault

single gpu cards are not immune from this

+1.

I have owned:

SLI Geforce 6800s
Xfire 3850s
Xfire 4850s
Xfire 4870s
Xfire 5770s
SLI GTX 460s
SLI GTX 560s.

I have 100% perfect vision and have never ever seen microstutter in a single situation. Most people who whine about microstutter have never even used a single multi GPU setup, or if they did it was badly configured (They prolly forgot to get a better PSU, activate Vsync, overclock their CPU from 2.6 Ghz to 4 Ghz+, or use a 2ms response time monitor. Maybe they only had 2 Gb ram).

Or they simply have eyeball lag, since that would be the only difference between me and another user playing games on my dual GFX cards builds.

I mainly use my dual cards to enable 4x SSAA in any games that it can be forced into. If I try turning off SLI / Xfire and using SSAA on a single card, I get a slideshow, so yes SLI + Xfire on mid range cards works very well for me

I have yet to be convinced to buy a single high end card if it gets smoked by two mid range ones for the same cost. I havnt found any situation where even a single 3 Gb GTX 580 would be any better than my SLI GTX 560 ti.

[BababooeyHTJ]

I completely believe that you have never seen microstutter but just because you have never noticed microstutter does not mean that it isn't there.

It seems like most people that don't notice the gamebyro stutter in Oblivion, Fallout 3, Divinity 2, and New Vegas but it is present with any 60hz display. The gamebyro stutter annoyed the crap out of me.

I also picked up an HP LP2465 a while back for a really good price. This was a pretty popular display on the forums due to the price. I found the input lag to be awful but most people claimed not to notice it.

The first thing that I noticed on my Acer GD234hz was input lag. Most users on the forums claim not to notice it.

Just because you don't notice it does not mean that it isn't there.

[DeltZ]

not to mention the effects of playing with Vsync are probably worse than micro stutter itself

[zerazax]

I wonder how many people who notice stuttering are also people who are on the look-out for it. It's like when someone says "watch out for xyz" they start noticing "xyz" everywhere

[Mungri]

I completely believe that you have never seen microstutter but just because you have never noticed microstutter does not mean that it isn't there.

It seems like most people that don't notice the gamebyro stutter in Oblivion, Fallout 3, Divinity 2, and New Vegas but it is present with any 60hz display. The gamebyro stutter annoyed the crap out of me.

Oblivion only lags when I have it heavily modded and its running at 17-22 FPS? I've never seen any kind of reliable method that shows that 'microstutter' is a real issue, and even if it is that its a fault of multi GPU setups. From what I can tell, 'microstutter' is due to sharp lag spikes causing the FPS to rapidly drop. I dont believe that this is due to having multiple GPUs as it can happen on a single GPU too.

not to mention the effects of playing with Vsync are probably worse than micro stutter itself

Why is this? I find that playing with Vsync leads to noticeably less lag spikes. In any case that I do notice lag spikes, they are in situations where a single GTX 580 would be unplayable anyway, so I cant place the blame of SLI or Crossfire.

I wonder how many people who notice stuttering are also people who are on the look-out for it. It's like when someone says "watch out for xyz" they start noticing "xyz" everywhere

I've had lots of debates over multi GPU and Vram before, and during them have sat with my eyes wide open during multiple Metro 2033 benchmarks with DOF and everything else up to full.

The only time I could see any lag was literally when the FPS was falling into the 20s.

Just because you don't notice it does not mean that it isn't there.

Ok, but even if microstutter does exist, where is the evidence that multiple GPUs is the only cause, and that this issue is completely absent on single GPU systems?

[Lanek]

The real question is why use thoses games ? and not Dirt 2-3, crysis or whatever ..

Metro2003 is horrible in the term of constant framerate .. it allready drop permanently ... ( run the benchmark and see the result, with one or 2 .. it is allready ridicule ). Stalker and specially the benchmark give allready some high lock down on some part without reason, specially cause as most of thoses benchmarks, they have never been patched. but game have been patched... ..... mafia, huum... Call of Juarez.. i don't know ..

Seriously there's 2 pc for gaming here.. 1 is with a GTX460 1gb the other is the one in my sig with 2x5870 ... we have try all games and put the monitors side by side with my gf... ask peoples, nobody can see a difference in smoothness ( microstutter ).. we have test with different monitor size ( 24-22" ), exchange them, use a TV LCD 40+ for see if something appears ... nobody can see a difference.... the only thing, i can push a lot more the graphism and quality is ofc a bit better. ( depend the game ofc )