HD5970 Microstuttering tests

[Pyr0]

yes, i know where they should be, i've searched the registry already
as I have already stated, there are no entries for MVPUmode in mine

you mentioned using agent god's crossfire extension app, and I think that is where the MVPUmode entry comes from

[eRazorzEDGE]

Sorry, I was just posting it so you could compare it to yours. If it's not there, add it, or try the app from AgentGOD and see if it adds it.

[jaredpace]

Unfortunately this doesn't offer anything constructive because the frame rates were omitted.

I don't think framerates need be known, since those are available on dozens of reviews, and a quick google search will tell you what framerates those cards get.

The point of the video is for you to watch AFR/Scissor/Supertiling in action, and afterwards, directly compare it to the fastest single gpu rendering that exact same scene. You already know a 5970 gets more FPS than a 5870. You then make an attempt to notice any visual differences. The difference in "smoothness" is painfully obvious to me. Especially when the character is turning on the horizontal yaw.

If you can't find that constructive, your missing the point, or unaware of what to look for.

[cegras]

Microstutter in action:

- Time: 1:12-1:16 on 5970 screen watch carefully as he rounds the corner around the pole, then rewind and watch how much smoother it looks on the 5870 (even though 5870 has less fps).

- Time: 1:21-1:26 on 5970 screen watch carefully at the Camouflage netting on the top of the FOV, and again watch the screen underneath it and see how smoother it is on the 5870.

http://www.youtube.com/watch?v=emG7ZNIsxw8&fmt=22

Single GPU rendering is less likely to show this type of stuttering compared to multi-GPU.

Whoever recorded that video is an idiot. The first cardinal sin is that they didn't use the same video for all four video cards. The second cardinal sin is that for the supposed microstutter, they purposely swept left and right on the crossfire rig to expose rendering artifacts.

How can you even claim the ride around the pole was choppy? Almost all of that choppyness was entirely due to the fact that whoever was using the mouse has to lift it up and reposition it; he couldn't even manage to smoothly move his mouse in a circle.

[BababooeyHTJ]

I don't think framerates need be known, since those are available on dozens of reviews, and a quick google search will tell you what framerates those cards get.

The point of the video is for you to watch AFR/Scissor/Supertiling in action, and afterwards, directly compare it to the fastest single gpu rendering that exact same scene. You already know a 5970 gets more FPS than a 5870. You then make an attempt to notice any visual differences. The difference in "smoothness" is painfully obvious to me. Especially when the character is turning on the horizontal yaw.

If you can't find that constructive, your missing the point, or unaware of what to look for.

I think that the smoothness thing is due to the larger swings in framerates that multi-gpu setups see. That always bothered me. I have never noticed "MS" in cf or sli, personally.

[jaredpace]

I never noticed it on my SLI 260's as well. But I can see it when you compare multi-gpu to single gpu, like in these youtube vids. Yes I agree that reivewer is purposely sweeping on the horizontal axis left to right, to expose stuttering on multi-gpu, but that isn't any different to what a player would do 'in-game'. Ive seen MS on grid, clearsky, & vantage. PCGH does a lot of videos on it. Here is another one of vantage playing with an irregular framerate on multi-gpu - http://www.youtube.com/watch?v=GbcGy...B2778B&index=4

[jaredpace]

How can you even claim the ride around the pole was choppy?

So which one looks smoother to you, the 5870, or the 5970? How about the entire last 15 seconds of the video when hes going through the fog & trees.

The 5970 and 5850cf and GTX295 all look choppier than the 5870. That is my opinion, maybe someone agrees?

[Final8ty]

I never noticed it on my SLI 260's as well. But I can see it when you compare multi-gpu to single gpu, like in these youtube vids. Yes I agree that reivewer is purposely sweeping on the horizontal axis left to right, to expose stuttering on multi-gpu, but that isn't any different to what a player would do 'in-game'. Ive seen MS on grid, clearsky, & vantage. PCGH does a lot of videos on it. Here is another one of vantage playing with an irregular framerate on multi-gpu - http://www.youtube.com/watch?v=GbcGy...B2778B&index=4

Well that MS in both vids are awful & i have never had that to be honest & that Farycry comment i made was a completely different issue in comparison.

Maybe because i have always stuck with AMD mobos for my CF is why i have not had MS.

[Rock&Roll]

I'm not sure that's completely MS in those stalker videos...I think the xray engine is goofy anytime you are near an environment with fire. Been like that on every graphics card I've ever tested with stalker.

Just saying, I would be interested in seeing more irrefutable MS evidence.

[BababooeyHTJ]

I never noticed it on my SLI 260's as well. But I can see it when you compare multi-gpu to single gpu, like in these youtube vids. Yes I agree that reivewer is purposely sweeping on the horizontal axis left to right, to expose stuttering on multi-gpu, but that isn't any different to what a player would do 'in-game'. Ive seen MS on grid, clearsky, & vantage. PCGH does a lot of videos on it. Here is another one of vantage playing with an irregular framerate on multi-gpu - http://www.youtube.com/watch?v=GbcGy...B2778B&index=4

Honestly that just looks like a normal stutter to me. Nothing different than I saw on my 280 while playing The Witcher. Maybe more like that weird Far Cry 2 stutter. It dosen't surprise me that a multi-gpu setup will see more irregularities than a single gpu setup. I'm also not surprised by seeing "MS" in Vantage. We all know how much ATI cards love Vantage.

[rge]

I tried same experiment on my gtx 295.... FRAPS on crysis benchmark, 3d6, cod mw, I dont see any bigger swings in FPS or in frametimes (plotted) with 2 gpus enabled versus 1. In fact, comparing same runs on 3d6, 295gtx with both gpus enabled had smaller swings in frametimes than with just 1 enabled, but neither was very much.


I can tool around in a game quickly to rapidly change sceneries and record varying FPS, then disable sli and watch it look exactly the same or worse if FPS gets too low.

I was actually impressed at the consistency of frametimes for both sli and single gpu.

[annihilat0r]

Indeed. Sorry for this, guys, I never thought a single GPU system could have so much microstuttering. I did a test myself with Crysis and can confirm similar results. Should have done this BEFORE opening the thread.

[saaya]

I never noticed it on my SLI 260's as well. But I can see it when you compare multi-gpu to single gpu, like in these youtube vids. Yes I agree that reivewer is purposely sweeping on the horizontal axis left to right, to expose stuttering on multi-gpu, but that isn't any different to what a player would do 'in-game'. Ive seen MS on grid, clearsky, & vantage. PCGH does a lot of videos on it. Here is another one of vantage playing with an irregular framerate on multi-gpu - http://www.youtube.com/watch?v=GbcGy...B2778B&index=4

to experience the awesomeness of microstuttering on an 260SLI rig fire up FC2 and hop on a jungle boat and cruise around... usually doesnt take long until you notice stuttering, especially when the river makes a turn. it takes some time until you notice it, especially if you dont know how fluid it looks and feels without microstuttering on a single 260, but once you played FC2 on a single 260, going SLI is really annoying and the stuttering and lags arent worth the extra fps... especially since its really fast enough on a single 260 even at 1920x1200 with max details and 2aa or even 4aa if your rig is oced...

[saaya]

I tried same experiment on my gtx 295.... FRAPS on crysis benchmark, 3d6, cod mw, I dont see any bigger swings in FPS or in frametimes (plotted) with 2 gpus enabled versus 1. In fact, comparing same runs on 3d6, 295gtx with both gpus enabled had smaller swings in frametimes than with just 1 enabled, but neither was very much.


I can tool around in a game quickly to rapidly change sceneries and record varying FPS, then disable sli and watch it look exactly the same or worse if FPS gets too low.

I was actually impressed at the consistency of frametimes for both sli and single gpu.

too much data... can you stretch those graphs a bit? (x axis)

Nvidia:12 Force : ATI

no nubs *cough*

ATI

noNvidia pftt

lolly pops availble the ATI counter (US) who? :p

\0/ [f u]

wasted talent:

serious

ati NOW ^&

ELITE players onboard - "the best of the best"

your move, prepared to be owned.

mmmmkay.... you might wanna run a quick virus scan on your synapses m8...

[annihilat0r]

some stupid graphs

how long can I work in US from UK, I need sunshine ;p

judging from your posts, I'd say moonshine

[BababooeyHTJ]

to experience the awesomeness of microstuttering on an 260SLI rig fire up FC2 and hop on a jungle boat and cruise around... usually doesnt take long until you notice stuttering, especially when the river makes a turn. it takes some time until you notice it, especially if you dont know how fluid it looks and feels without microstuttering on a single 260, but once you played FC2 on a single 260, going SLI is really annoying and the stuttering and lags arent worth the extra fps... especially since its really fast enough on a single 260 even at 1920x1200 with max details and 2aa or even 4aa if your rig is oced...

That is an issue with FC2. Do you want me to post a video of that exact same stuttering on my single 280? I have to cap my framerate to play that game with a single gpu, it's a common issue. Once again that is not "microstuttering". Are you really surprised that you are more likely to see irregularities on an sli/cf setup than with a single gpu?

[SocketMan]

Maybe it's the saturated FSB bottlenecking,not the graphics ?
All that travelling to the chip-set and back might be tiring.


I've used:

7600gt SLI
8800gt SLI
GTX260 SLI
3850+3870 CF
3x4850 CF /3x4850+4890 CF
5770 CF
(No FSB)
Never seen a "stutter",even with the mixed cards.
I don't know why DIRT 2 FPS is so jumpy on that graph,FRAPS shows a constant number for me (2x5770/win7x64/DFI M2RS/Phenom710@3.5/8gb DDR2),no fluctuations.Certain levels give much higher FPS,but it never jumps/stutters.
As for dual gpu's on a single PCIe x16 -who knows maybe the street is too small....

Don't forget 2 cards get x16 lanes each,no sharing.

[BababooeyHTJ]

Never seen a "stutter",even with the mixed cards.

Yeah, drivers are always 100% optimized for you. I love these claims.

[SocketMan]

Yeah, drivers are always 100% optimized for you. I love these claims.

I didn't say it - you did

I've seen low fps (SLI/CF) and negative scaling using CF ( often with early drivers), or
stuttering from low frame buffer/network issues (on MP games) ,but
not this "famous" "micro stutter" - that's all.


These are all different issues,but for many it's a lot more convenient to blame the MS rather then tweak/diagnose their systems.
It's always a waiting game with drivers,I couldn't agree more,but again a separate issue.


Also to be fair - I do only play a few select titles,thus
reducing the chances of getting MS or any other abnormalities present in one title or the other (but not all).

I do speak from personal, hands on experience only,so what the other have
or have not seen/read about is up to them.
I know I still trust my eyesight more then some web site out there,what ever
they claim.
Thanks to the eRazorzEDGE for running the tests !
You might gain a few extra FPS by using the ganged mode/unganged
memory mode -depending on the game.64bit should do better
with ganged - as it uses the 64 bit access mode (as supposed to 32 for unganged).

[Eastcoasthandle]

I don't think framerates need be known, since those are available on dozens of reviews, and a quick google search will tell you what framerates those cards get.

The point of the video is for you to watch AFR/Scissor/Supertiling in action, and afterwards, directly compare it to the fastest single gpu rendering that exact same scene. You already know a 5970 gets more FPS than a 5870. You then make an attempt to notice any visual differences. The difference in "smoothness" is painfully obvious to me. Especially when the character is turning on the horizontal yaw.

If you can't find that constructive, your missing the point, or unaware of what to look for.

The highlighted portion is really the gist of your reply. The problem with your reply to me (& the gist of what is MS is claimed to be in this thread) is that it's based on assumption(s). You want me to assume what I believe to be is an illusion based on the omission of frame rates. Because I am aware of other variables that can restrict frame rates which can cause stuttering, etc I cannot agree at this time.

[rge]

too much data... can you stretch those graphs a bit? (x axis)

It wont look any different if less compressed, at least not to me, just see same only stretched out. The previous run was slightly out of sync by 100 frames from my slight variance in hitting benchmark key, but this one is in perfect sync. If I graphed a bunch more, basically you see similar variability in frametimes of both 1 gpu and sli, but overall sli just has less on frames I captured.

After hearing all the hoopla about microstuttering, I was against getting sli card...but then thought I would give it a try, and atleast see what all the hoopla was about. And I am still waiting. I guess instead of playing games I need to search the net for claims of ms, go buy the game, then try and find the ms, then test it on 1 vs 2 gpu and maybe if I search enough ms claims, I can finally see at least 1 for myself caused by sli and can even graph it out...but that sounds like a lot of work.

And I have farcry 2, mine did stutter in places on my 8800 ultra single card. Maybe I will get urge to load it again on win 7 and try and find the ms, then graph 1 gpu vs 2, but that means I will have to play it...now if it was farcry original I would be up for that. The few times I tried to duplicate claims of ms in games I have, it always looks same or worse with only 1 gpu...but now I guess I can graph it as well.

[eRazorzEDGE]

I'm glad everyone is trying this for themselves, comparing Single GPU's vs. Multiple GPU's. It's really great to see the graphs guys. Great work.

[kadozer]

+1000 best thread on Microstutter so far. spread the word to the lesser noob forums.

[zerazax]

Microstuttering = one of the most over-hyped, mis-understood, and carelessly dangerous terms ever

I'm glad people here see why there's a need for a scientific BASELINE.... ironic isn't it when the single GPU sometimes is worse than the multi-GPU?

It's too bad that the term micro-stuttering is now being thrown around so carelessly

[Katanai]

I have a question: what if this is a problem with the motherboards and not the video cards themselves? That would explain why for some people it works and for others it doesn't. Has anyone done a CF or SLI benchmark on multiple motherboards investigating fluctuations in the framerates?