Microstuttering tests on ATI Radeon HD 4870 X2

[shiznit93]

That is not true... With respects to multi gpu solutions and my own findings. V-sync, without triple buffering, will not automatically go from 58 fps to 30 fps, or anything like that. V-sync without TB can give framerates between 31 and 59 fps perfectly.

well im gonna make a video and we will see. any way to make the fraps counter show in the video?

in the meantime read this: http://www.blazingpc.com/forum/showt...379/index.html http://www.hardforum.com/showthread.php?t=928593

if the refresh rate is 60hz with double buffer vsync, the only possible framerates are 60/N, where N is a positive integer. Sorry to spoil it for you, but vsync w/o triple buffer is pretty much worthless, unless you like your framerate capped at 30 like the console kids.

[eleeter]

Having trouble reading a graph?

Is anyone really surprised?

[Hu1kamania]

There is no microstutter advantage when you compair a 4870X2 vs. a pair of single 4870's in cf....they are basically on par...so why are you busting shintai's balls? It just goes to show that the miscrostutter issue was not solved on the hardware end with the 4870X2.

[WhiskeyAlpha]

Do you not know how vsync works? Without triple buffering, if your framerate ever falls below the refresh rate even by 1 fps, it will be immediately cut into half of the refresh rate. So people who play on 60hz LCDs and want to avoid the nasty tearing that comes with such a crappy refresh rate (majority of PC gamers) will be stuck bouncing back and forth between 60 and 30 frames per second. I don't know about you but that drives me crazy, which is why in games where I can afford to use Vsync (where input lag won't get me killed), I always force triple buffer with D3DOverrider to avoid the 60-30-60-30 bull and allow the framerate to fluctuate freely. Last night I found out that AFR and triple buffer can't be used together (I have no clue why people have been buying SLI and Xfire setups all this time and giving up triple buffer), so until something can be done multi-gpu is a no-no for me I guess.

Off topic, but I have exactly this 30/60 fps problem with Bioshock in DX10 mode, when V-Sync is enabled (tripple buffering is enabled in the driver panel). I don't have this problem in any other game as far as I'm aware.

With respect, do you have a source for this information or a link to somewhere where I could get a little more info? I did search recently but found nothing interesting.

EDIT:

Nevermind, just found this

Essentially this means that with double-buffered VSync, the framerate can only be equal to a discrete set of values equal to Refresh / N where N is some positive integer. That means if you're talking about 60Hz refresh rate, the only framerates you can get are 60, 30, 20, 15, 12, 10, etc etc. You can see the big gap between 60 and 30 there. Any framerate between 60 and 30 your video card would normally put out would get dropped to 30.

[annihilat0r]

That's why I suggested turning the settings down. I played crysis when it came out and had the same problem...thats why I am waiting for a rig that can play it in all its glory. So far there is no such animal. Don't get me wrong...thats not the only reason for building a new rig

A pair of R700's might do the trick...only time will tell. Even then the microstuttering debate will linger on.

In regards to Vsync, I have found my experience to be similar to what Seraphiel mentioned.

With a non-overclocked 4870 it's about 30-35FPS where SLI was 40, but it's really and I mean REALLY playable and enjoyable.

Crysis - High/Very High mixture - 16xAF - 1680x1050

Highly playable with a non-overclocked 4870! Looks bloody good and feels right. I think with some serious overclocking I might just try to crank everything to Very High, as I found even 25FPS to be playable with the 4870. Not that I would play the entire game at 25FPS, but if it's just some really intense combats at 25 and higher in the rest of the game, I'll definitely go for it.

[emoners]

subscribed...wanna see how this would turn out

[Seraphiel]

I was referring to CF users. Do you have those results?

Also there was a well known driver hack that would enable SLI for v1.0. I am referring to test results of Grid in this review without added driver tweaks.

I thought you mentioned multi gpu as something that implied both CF / SLI, or else I would have assumed you would just say CF instead. My fault then, and no I do not have such results, but I haven't applied that driver hack either to my SLI setup

well im gonna make a video and we will see. any way to make the fraps counter show in the video?

in the meantime read this: http://www.blazingpc.com/forum/showt...379/index.html http://www.hardforum.com/showthread.php?t=928593

if the refresh rate is 60hz with double buffer vsync, the only possible framerates are 60/N, where N is a positive integer. Sorry to spoil it for you, but vsync w/o triple buffer is pretty much worthless, unless you like your framerate capped at 30 like the console kids.

Do you have a mult gpu setup? Because I have always suspected, that it will show otherwise (which it does).

I can use fraps and have eyes, just as well as you do. There is no 60 / n(INT) performance with my setup using v-sync. There is a max frame cap, of course, but nothing else.

I know what you're talking about, and I have observed it myself years ago. But since I have used SLI, I have never ever experienced it again.

I have no reason to lie, nor do my eyes or observations lack anything that will give errored results regarding this matter.

Oh, and yes, it is possible to show fraps fps while capturing

[Solus Corvus]

There is no microstutter advantage when you compair a 4870X2 vs. a pair of single 4870's in cf....they are basically on par...so why are you busting shintai's balls? It just goes to show that the miscrostutter issue was not solved on the hardware end with the 4870X2.

Oh, perhaps not in relation to 4870 CF. But nether the 4870 CF or X2 appear to have microstuttering, at least in that game so that observation is essentially useless. The X2 (and CF) do appear to have an advantage in microstuttering tough - when compared to other cards. Why would one make a comment that eliminating MS (In the tests we have seen so far) isn't an advantage?

[shiznit93]

Do you have a mult gpu setup? Because I have always suspected, that it will show otherwise (which it does).

I can use fraps and have eyes, just as well as you do. There is no 60 / n(INT) performance with my setup using v-sync. There is a max frame cap, of course, but nothing else.

I know what you're talking about, and I have observed it myself years ago. But since I have used SLI, I have never ever experienced it again.

I have no reason to lie, nor do my eyes or observations lack anything that will give errored results regarding this matter.

Oh, and yes, it is possible to show fraps fps while capturing

I don't have multi-gpu, been waiting for years for all the issues to get sorted out but looks like I will have to wait longer. So you are saying that in SLI and Xfire the framerate can fluctuate freely without triple buffer when vsync is on? Or is that just what the FPS counter says? I ask because this microstutter business has shown us all that in multi-gpu the FPS counter doesn't mean $hit.

[Hu1kamania]

Oh, perhaps not in relation to 4870 CF. But nether the 4870 CF or X2 appear to have microstuttering, at least in that game so that observation is essentially useless. The X2 (and CF) do appear to have an advantage in microstuttering tough - when compared to other cards. Why would one make a comment that eliminating MS (In the tests we have seen so far) isn't an advantage?

Eliminating microstuttering IS an advantage, like you said....but its not an advantage that the R700 holds exclusively. I would be interested to see how the GTX200 series does in relation, as it is a different generation than the 9800X2, similarly how the 3870X2 is a different generation from the 4800 series.

You can't dispute that the improvments are astounding and it makes miscrostuttering something to be that much less fearful of.

[Seraphiel]

I don't have multi-gpu, been waiting for years for all the issues to get sorted out but looks like I will have to wait longer. So you are saying that in SLI and Xfire the framerate can fluctuate freely without triple buffer when vsync is on? Or is that just what the FPS counter says? I ask because this microstutter business has shown us all that in multi-gpu the FPS counter doesn't mean $hit.

Yes, that is my observations. And a drop from 60 -> 30 fps or further to -> 15 fps is not that hard to notice. No TB and with v-sync, of course. We are talking about the same here.

But I will limit this to my own SLI setup, which have sometimes behaved odd or differently than others have. I will not claim that multi gpu in general does this, nor would I say SLI does this. It does with my setup, and I can claim no more truth than just that. It may or may not apply in general, but I do not know, as I haven't the money to buy so many setups or have that kinda access to hardware - even though I may have the time.

No fps counters or fraps may not be very reliable, and I sincerely doubt they really are. But it is the best tool, aside from our eyes, which can be very subjective, to measure performance.

[grimREEFER]

Hi,

I'm working with full article of microstuttering issue including tests in Crysis and Race Driver: GRID with ATI Radeon HD 4870 X2, ATI Radeon HD 4870 CrossFireX, ATI Radeon HD 3870 X2 and GeForce 9800 GX2.

can you underclock and overclock the pcie clock, and see if it's the bus bandwith has any effect on it?
my theory right now is that microstuttering inversely correlates with the amount of bandwith the cards have to share with each other.
it shouldnt have any effect on the the dual gpu cards, but would be interesting to see what effect that has on 4870 crossfire.

[Xoote]

nice test

[Devious]

subscribed *

[hieuhef]

i cannot believe that i never read anything about this until after i sli my 8800GTs..

i thought something was off when my FPS went up in COD4 but my kills went down dramatically as if my aim had gone awry. subscribed as well.

[safan80]

@ Sampsa
try testing 2560x1600.

[GoldiClocks]

@ Sampsa
try testing 2560x1600.

As far as I have understood, he doesn't have a monitor capable of that resolution in arms reach.

[Seraphiel]

@ Sampsa
try testing 2560x1600.

I am going through alot of tests at the moment, this resolution included. Grid is weird to say the least, but I test other games as well. Unfortunately, I don't have a R700

But I have two XFX 8800 Ultra XXX 685M, and perhaps they can give some perspective to all of this. Microstutter is a weird phenomenon, and not always present - or at least not as bad it is sometimes in some games.

[safan80]

I am going through alot of tests at the moment, this resolution included. Grid is weird to say the least, but I test other games as well. Unfortunately, I don't have a R700

But I have two XFX 8800 Ultra XXX 685M, and perhaps they can give some perspective to all of this. Microstutter is a weird phenomenon, and not always present - or at least not as bad it is sometimes in some games.

I had mircostutter on my 9800gx2 at 2560x1600. I'd like to see if the R700 has it.

[awdrifter]

According to the graph, there's the same amount of microstutter on an CFX 4870 system and the 4870x2, so what solved (or at least minimized) the problem is not the PLX chip but something else then?

[L7R]

There is no microstutter advantage when you compair a 4870X2 vs. a pair of single 4870's in cf....they are basically on par...so why are you busting shintai's balls? It just goes to show that the miscrostutter issue was not solved on the hardware end with the 4870X2.

You misundersood this. There is NO microstuttering in Sampsa's 4870 crossfire system or 4870x2 system. That's why this thread was made in the first place and is somewhat groundbreaking news if it's true. What is the most intresting how Crossfire has been fixed and is it really. We'll wait and see. For sure, before it has been a well-known problem for Crossfire and SLI.

Eliminating microstuttering IS an advantage, like you said....but its not an advantage that the R700 holds exclusively. I would be interested to see how the GTX200 series does in relation, as it is a different generation than the 9800X2, similarly how the 3870X2 is a different generation from the 4800 series.

I think you should think before you write... People here are first people in the world witnessing this splendid, spectacular news, that for the first time in the history of multi-GPU hardware, one product seems to be free without microstuttering. You should really watch what you write before posting, pal. No offense. We are just saw a hint that this could be true, but we have to wait and see. If you have GTX280 or 260 SLI, sure go ahead and post your findings. But do not speculate too much, this is just a begining.

[NapalmV5]

for the 3rd time..

is there ms @ 2 x 4870x2 ?

[safan80]

for the 3rd time..

is there ms @ 2 x 4870x2 ?

patience

[NapalmV5]

yeh but they got a bunch of 4870x2 there to test so if ms is fixed @ 4870x2 thanks to the plx.. how does it play @ 2x

2 x 4870: not surprising if ms still occurs
2 x 4870x2: good chance it doesnt occur

btw guys, ms goes back to the 3dfx days.. reason why i stayed away from multi gpus for so long

[Mumid]

Tbh, I havnt seen any micro stuttering in the setup in my rig, results so far look pretty accurate to me...