Microstuttering tests on ATI Radeon HD 4870 X2

[Eastcoasthandle]

LOL, madshrimp says the exact opposite about Crysis with their setup.

Source

At 1920x1200 we see a ~5fps drop for both cards, but numbers are still acceptably high. We did notice a lower min. FPS on the HD4870X2, suspecting that this was maybe due to micro-stuttering we manually measured the FPS with FRAPS in-game in the same Alien Spaceship level, but we didn’t notice any stuttering, the 23 fps drop was early on the level.

TestBed from madshrimp review:
Intel Core 2 E8200 @ 2.66Ghz
Cooling Scythe Ninja 2
Mainboard MSI P45 Platinum
Memory 2 * OCZ 1Gb PC2-6400

Testbed computerbase:
Intel Core 2 Extreme QX9770 (overclocked by multiplier to 4 GHz quad-core)
Noctua NH-U12P Noctua NH-U12P
Asus P5E3 Deluxe WiFi-AP main board Intel X38, BIOS-Version: 1104 test for Crossfire systems
XFX nForce 790i Ultra (Nvidia nForce 790i, BIOS-Version: 811N1P01_Beta) für SLI-Systeme
2x 1024 MB G.Skill DDR3-1600 (7-7-7-18)
2x 1024 MB Patriot DDR3-1600 (7-7-7-18)

Clearly 2 different testbeds. Also, we know that stutter can be induced and can be the results of something other then the GPU. Another thing, these "stutter charts" have no legend nor does the creator explain how each plot on the chart is obtained. Making these type of charts vague. Leaving it up to the reader to make assumptions of how it was obtained or, some validating it without any rudimentary information of where the plots on the chart come from.

[zlojack]

CRYSIS!

I don't really care about Crysis. Been there, done that. (Ok, not totally true, I do look at the Crysis benchies, as well.)

Anyway, I too have my doubts about the microstuttering tests posted.

Very vague and by no means conclusive..

I definitely see microstutter on my 3870 X2, so I know it exists. In Age of Conan, for example, the nametags sort of blur when I'm moving but when you look closely you can see they are stuttering.

just run current games at 1920x1200 @ 4aa and your already in the 60s-70s, next gen games will push this into the 40s. Tought the ammount of console ports will increase, so even a snigel 4870/280gtx will be enough.

I can run my games on that res on my 3870 X2 and still get ok frames.

For me right now, I look at how a card does in Age of Conan (which is hell to bench on) and COD4.

Again, everyone looks at things from their point of view and based on their needs.

[cadaveca]

Considering I took this issue on full force, and now it's gone out of hand, and I see now people complaining of other problems, and calling it microstutter, I wish I could kill the beast I created!

[zlojack]

Considering I took this issue on full force, and now it's gone out of hand, and I see now people complaining of other problems, and calling it microstutter, I wish I could kill the beast I created!

I don't know what else to call what I see then. The screen looks "jittery".

Is that not a possible sign of micro-stutter?

[jmke]

LOL, madshrimp says the exact opposite about Crysis with their setup.

yup, was eager to find out if microstutter would show; but several manual runthroughs later... no signs of it.

@zlojack: the stutter is as of the screen freezes a millisecond and you then jump forward in frames quickly, as if some frames were "cut out" of the animation. PCGameshardware has a few movies on these, check at the bottom of this page: http://www.pcgameshardware.de/?article_id=631668

they do miss the ball though on this movie: http://www.pcgameshardware.com/aid,6...on_HD_4870_X2/
the FPS is just way too low for fluent gameplay and they are mixing stutter with FPS dips

[cadaveca]

I don't know what else to call what I see then. The screen looks "jittery".

Is that not a possible sign of micro-stutter?

Nope. not if FPS is below monitor refresh....then the problem is something else.

Microstutter is exclusively:

high-framerates drawn, less displayed. You cannot measure Microstutter below monitor refresh..if your eyes are sensitive enough, you'll see issues at anything below monitor refresh, and it won't be due to the hardware design, it will most likely be software limitations.

people don't understand that the Crysis video I posted was with 60+++FPS, and they also seem to have neglected to read that the problem is @ 1280x1024, while 1680x1050 runs butter-smooth. The problem is evident only when framrate is above monitor refresh, and the system is obviously not gpu bottle-necked.

Like DevilMayCry4.... approx 300FPS, still stutter, @ 2560x1600. This is what happens when the crossfire connection is not capable of feeding rendered frames to the primary gpu...and does not show itself until the bandwidth of those rendered frames exceeds the bandwidth of the interconnect.

Anyone talknig about Microstutter @ 60FPS is just on the hypewagon, because it looked fun, but they do not know where teh wagon is headed...


Provides a good chance to see which reviewers are worth thier salt, and which are not, IMHO. I've yet had anyone contact me from online media asking about the issue...so I'm not too sure that they are even aware of what the problem is, nevermind that most don't have the hardware to properly create the issue.

[Eastcoasthandle]

I don't know what else to call what I see then. The screen looks "jittery".

Is that not a possible sign of micro-stutter?

...In Age of Conan, for example, the nametags sort of blur when I'm moving but when you look closely you can see they are stuttering...

IMO, possible ghosting which is more LCD related then video card related.

[adamsleath]

start a 'jittery' thread

micro-stutter shot jfk.

[zlojack]

Nope. not if FPS is below monitor refresh....then the problem is something else.

Microstutter is exclusively:

high-framerates drawn, less displayed. You cannot measure Microstutter below monitor refresh..if your eyes are sensitive enough, you'll see issues at anything below monitor refresh, and it won't be due to the hardware design, it will most likely be software limitations.

people don't understand that the Crysis video I posted was with 60+++FPS, and they also seem to have neglected to read that the problem is @ 1280x1024, while 1680x1050 runs butter-smooth. The problem is evident only when framrate is above monitor refresh, and the system is obviously not gpu bottle-necked.

Like DevilMayCry4.... approx 300FPS, still stutter, @ 2560x1600. This is what happens when the crossfire connection is not capable of feeding rendered frames to the primary gpu...and does not show itself until the bandwidth of those rendered frames exceeds the bandwidth of the interconnect.

Anyone talknig about Microstutter @ 60FPS is just on the hypewagon, because it looked fun, but they do not know where teh wagon is headed...


Provides a good chance to see which reviewers are worth thier salt, and which are not, IMHO. I've yet had anyone contact me from online media asking about the issue...so I'm not too sure that they are even aware of what the problem is, nevermind that most don't have the hardware to properly create the issue.

Thanks for the comprehensive response.

IMO, possible ghosting which is more LCD related then video card related.

Interesting. I'll have to check that.

start a 'jittery' thread

micro-stutter shot jfk.

I'm jittery this morning from the double espresso! Maybe that's the problem.

I knew I saw some weird framrates on the grassy knoll footage!

[Cooperdale]

Microstuttering is the smallest problem in the history of problems.

[Aberration]

You obviously dont own a SLI/CF solution then.

[jaredpace]

Microstuttering is the smallest problem in the history of problems.

thank you.

[Bo_Fox]

To see what microstuttering really is, just try Test Drive Unlimited with 2x 7900GTX's in SLI like I had for a while back then...

[E.R]

To see what microstuttering really is, just try Test Drive Unlimited with 2x 7900GTX's in SLI like I had for a while back then...

Works fine with 9800Gx2 also to se the same thing, or even better.

[deathman20]

You obviously dont own a SLI/CF solution then.

I own a CF solution and I still have yet to see what all the fuss is about. I've even been purposely looking for it.

[cadaveca]

you won't see much, deathman, because of your vgas. I notice this problem more than most due to running 4 gpus...and the problem is almost completely solved for dual cards...but only because performacne is what it is.


But, to test, you must try singlecard, notice decent performance, and then enable second card, notice performance improvement FPS-wise, but onscreen display is worse than singlecard.


The problem can occur at any FPS number...so looking for it @ less than 40FPS is stupid. Trying to attain such is a waste of time. You will never be able to distinguish problems because of lack of v-sync.

Mcrostutter is LESS FPS DISPLAYED THAN DRAWN. the problem mostly stems from inter-gpu communication, and is why I see this problem far more than anyone else...4 GPUs create far more traffic than 2.

I'll shoot some video of DMC4 and show the problem above 200FPS. Not using FRAPS...which is a waste of time as well, but by setting up camera external to pc. The problem has NOTHING to do with FPS, other than FPS given by FRAPS or ingame counters does not jive with what's displayed onscreen.

[Bo_Fox]

Works fine with 9800Gx2 also to se the same thing, or even better.

Perhaps the max frame rates were enough that microstuttering no longer really noticeable. Say, it runs at 70 fps, but microstuttering makes it more like 40 fps, which is still nice and fluid.

Well, with my 2x 7900GTX in SLI, Test Drive Unlimited ran at 35 fps, but the microstuttering was unacceptable. It actually looked more like 18-20fps. My 8800GTX on the other computer also ran it at around 40 fps, yet it felt solid.
(Now I have 8800GTX on both of my computers which have the same CPUs for equal performance.. I'm a retired SLi veteran.)

[Cooperdale]

Microstuttering is the smallest problem in the history of problems.

You obviously dont own a SLI/CF solution then.

Actually I owned both sli and cf, am about to build a new system with a 4870X2. I really don't think it is/was such a big issue. There are people who say it's there, people who say it's not. People who think it's there because something else is there. I really never noticed much of anything. Just an opinion of course.

[Calmatory]

For anyone who is still unclear what MS is about.
Low FPS = high delay between frames. High fps = low delay between frames.

For single card 60 FPS is 16.6 ms delay between two frames.
For dual card 120 FPS can be three frames during that 16.6 ms, BUT the delay between first and second frame is e.g. 2 ms, and then the delay between second and third frame can be 14.6 ms("fps" is 68.5 between second and third frame). In this case the increase in FPS would be unnoticeable. All that matters is the maximum delay between frames. High delay is called sttutering.

Single card: |----------|----------|---------|-----------|-------|--------|----------|----------|-----------|----------|---------|
Dual cards: |-|--------|--|------|----|--------|-----|---------|-|-------------||------------|--------------||---------------|-|
| presents new frame, and - presents delay between those frames. The longer the delay is, more the image stutters.
Single card has alot smoother delay between frames, whereas dual cards draw more frames but sometimes with higher delay between frames, thus more stuttering. Also differences between frames can be noticed. If you have 70 fps, it jumps to 40, then to 120, then to 30, then 120, then 30 and then 200, you will notice it even if you wouldn't consider 30 fps bad. For some this is bigger problen than for others. Some use v-sync to stabilize the fps to the monitor refresh rate to get rid of FPS jumping, while having lower average FPS.

[ARGH69]

i want to shed more light on micro stuttering. i run an SLI system and have had enough of this micro stuttering garbage so i decided to take footage of it for you guys. micro stuttering negates the higher fps achieved over the single card and makes me feel like this entire multi-gpu solution is a scam!

http://http://www.youtube.com/watch?v=EeZmgb3eOBI

[ILikeCosmosS]

guys is it better to get 3 4870s or 1 4870x2 and 1 4870
i wanted to get 3 because it would add bling to the system but then again costing more!