[H] Yet again say SLI is smoother than crossfire

[The Alias]

Is that what you're going to interpret my comments as? Have at it. You'll have fun conversing alone.
I'm saying EXACTLY what my words say. As I usually do.

ok let's review and see exactly what you said :

Then why didn't we hear about all this "smoothness" talk back then?
That makes more sense than your comment. Doesn't it?

which was a response to :

That makes no sense because crossfire had more stutter long before 12.11 was released.

So he said crossfire had more stutter long before catalyst 12.11, then you asked why didn't we hear about that before catalyst 12.11 .

I'm pretty sure that's how 99% of the forum populus would view it too, which would then scream bs because radeon cards have had more micro stutter for generations

 

[Keysplayr]

ok let's review and see exactly what you said :



which was a response to :


So he said crossfire had more stutter long before catalyst 12.11, then you asked why didn't we hear about that before catalyst 12.11 .

I'm pretty sure that's how 99% of the forum populus would view it too, which would then scream bs because radeon cards have had more micro stutter for generations

I said why didn't we hear about this "smoothness" back then? Right. Only recently have we heard the mention of Nvidia, single or multicard configurations offer a "smoother" gameplay experience than AMD. We now hear that what "feels smooth" on Nvidia SLI at about 40 fps necessitates 60 to 70fps on a CFX setup.
Again I'll ask. WHY haven't we heard this mention of "smoothness" further back in time? If you have seen this mentioned, I'd like you, or anyone, to provide a link, because I don't know of any.

 

[blastingcap]

Fraps will actually record the frame time of every frame. As long as it is getting the true frame time, that's enough.

NO NO NO NO NO.

You need to capture the end-user experience, and there are elements of the graphics chain that you are not taking into account when you do what you suggested. I suggest that you read this:

http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/11

"The slide above shows the frame production pipeline, from the game engine through to the display, and it's a useful refresher in the context of this discussion. Things begin with the game engine, which has its own internal timing and tracks a host of variables, from its internal physics simulation to graphics and user input. When a frame is ready for rendering, the graphics engine hands it off to the DirectX API. According to Petersen, it's at this point that Fraps records a timestamp for each frame. Next, DirectX translates high-level API calls and shader programs into lower-level DirectX instructions and sends those to the GPU driver. The graphics driver then compiles DirectX instructions into machine-level instructions for the GPU, and the GPU renders the frame. Finally, the completed frame is displayed onscreen."

Once again. NO. You guys bickering back and forth for pages and pages in this thread need to face the fact that you will need to take into account whatever black magic NV and AMD do that is not captured by FRAPS. The only real way to do that is to literally point a high-speed camera at the monitor and compare. Use multiple monitor types if you are afraid that the results are monitor-specific. That is the ONLY way to capture end-user experience objectively. Not FRAPS frametimes, not subjective user experiences.

 

[VulgarDisplay]

I said why didn't we hear about this "smoothness" back then? Right. Only recently have we heard the mention of Nvidia, single or multicard configurations offer a "smoother" gameplay experience than AMD. We now hear that what "feels smooth" on Nvidia SLI at about 40 fps necessitates 60 to 70fps on a CFX setup.
Again I'll ask. WHY haven't we heard this mention of "smoothness" further back in time? If you have seen this mentioned, I'd like you, or anyone, to provide a link, because I don't know of any.

The whole point is that we have. I've heard about microstutter in crossfire for at least 5 years, and that's about the entirety of the time I have actually payed attention to PC hardware . It's why I have never had a dual GPU solution from Nvidia or AMD because I don't want to even bother wasting the money on something I may not like.

Interestingly enough, Crossfire is now the only dual/etc. card option that has very little to zero microstutter thanks to RadeonPro. Don't worry though because it's a software solution and I'm sure someone will come up with something similar for Nvidia. It just needs to get some attention from review sites like Anandtech or Techreport.

Some of you are also saying that you shouldn't have to use 3rd party software to get your enthusiast set ups running properly, and that makes some sense. Honestly though, how much headache is installed a program on your computer in comparison to installed a dual GPU solution anyways. Not a big deal at all.

 

[Jacky60]

I just installed radeon pro and made my flip queue 2 and I think Arma 2 now subjectively 'feels' smoother. TBH I really haven't a clue whether it does or not. :\

 

[supremor]

This thread only serves to remind me how glad I am to be so "blind" as to be pretty much oblivious to microstutter. There is no doubt that it a problem and one made worse by multiple GPU's but it's pointless to argue without a reliable way to measure it because peoples perception of it is subjective. I'm sure there are many happy CF users out there even if their setup is supposedly inferior in terms of MS because it appears smooth enough for them.

I for one almost never noticed it in my life and it sure as hell didn't bother me too much the few times I did. I would find it hard to go back to single monitor gaming after having done Surround.

 

[thilanliyan]

The whole point is that we have. I've heard about microstutter in crossfire for at least 5 years,

I've heard about microstutter in ALL dual-GPU setups for many years now. It is a consequence of the AFR method from what I understand. Something like supertiling or SFR wouldn't have microstutter (possibly, at least the multi GPU version of it) IIRC since both GPUs would be working on the same frame...why has there not been a bigger push into SFR and/or supertiling?

 

[Keysplayr]

I've heard about microstutter in ALL dual-GPU setups for many years now. It is a consequence of the AFR method from what I understand. Something like supertiling or SFR wouldn't have microstutter (possibly, at least the multi GPU version of it) IIRC since both GPUs would be working on the same frame...why has there not been a bigger push into SFR and/or supertiling?

I agree, we have been talking about microstutter for years.
Now I'm not trying to be nasty or mean or anything, but folks DO try and keep up with this question.
In all the years talking about this, when does anyone remember so much talk about Nvidia being smoother than AMD? Serious question.

 

[blackened23]

I'll throw my .02 into this massive flame fest, I've used both 7970 xfire and 680 sli and never noticed a difference in terms of microstutter -- any MS with the 7970xfire is completely eliminated by turning vsync on. I have noticed MS on older graphics cards in dual gpu configs, but not on recent heardware. I understand that hardocp is testing with vsync off (as they should for benchmarking purposes) -- I don't know if that has a discernible effect on stuttering but I personally never play with vsync off. When someone mentions that CF is full of microstutter, I really don't think so; with vsync on I never had any such issues.

Anyways, i'm running away from this thread. Running far away, continue the flame wars.

 

[schmuckley]

I prefer a powerful single card.
People want SLI to be like this guy
http://www.youtube.com/watch?v=Ej0tItNqItc

 

[The Alias]

I said why didn't we hear about this "smoothness" back then? Right. Only recently have we heard the mention of Nvidia, single or multicard configurations offer a "smoother" gameplay experience than AMD. We now hear that what "feels smooth" on Nvidia SLI at about 40 fps necessitates 60 to 70fps on a CFX setup.
Again I'll ask. WHY haven't we heard this mention of "smoothness" further back in time? If you have seen this mentioned, I'd like you, or anyone, to provide a link, because I don't know of any.

http://www.hardocp.com/article/2012/03/28/nvidia_kepler_geforce_gtx_680_sli_video_card_review/9
published march 28

 

[Ferzerp]

I prefer a powerful single card.

There is no one who doesn't, but neither brand makes a card that is good enough for 2560x1600 yet.

 

[Silverforce11]

Radeon Pro testing, single card. Civilization V. One turn gameplay after initial map loading ~30s to stabilize.

vsync off ingame

vsync off ingame, with radeon pro dynamic frame rate control, set at 55 fps.

vsync on ingame

With vsync off, there's a hiccup near the end, that WAS noticeable.

With vsync on, did not notice any stutter.

With vsync off and radeon pro, did not notice any stutter. But had screen tearing. Tried again with vsync force through radeon pro at 60 fps, with dynamic frame rate at 59 fps. No stutter noticeable, no tearing.

Looking at the frame time in fraps alone, it does seem to have an effect. All the gameplay frames are locked in at ~18ms constantly no spikes up or down, like normal with small fluctuations. The blips to 30-40ms for all the charts is due to changing screen to manage cities.

 

[Jacky60]

I too prefer a single card but the gamesI play run like crap on any single card, in fact single cards only really perform on console ports or older games adequately. With today's better more graphically intensive games (and some from more than 3 years ago e.gArma2/Crysis) no single card can really produce a seamless consistent 60fps at maximum settings. I have just learned however that turning on VS even at 23fps on a 60hz monitor DOES remove ms pretty much completely. I thought if fps below refresh rate the experience would be worse. That 23fps is typical flying an A10 over Cherno at low altitude in Arma 2 with video settings far from maxxed (OK SSAA) which indicates how much further we have to go.

 

[Lepton87]

I guess 4x7970GHz with radeon pro provides the smoothies experience in games. I would like to see it compared to 2x690 with a high speed camera with V-sync on, the way that I actually play games. Obviously that's much better than any single card, which is not enough for new games at 1080p not to mention normal desktop resolution and not some TV standard. But mid-range cards(GTX680) were never meant for 2560x1600 and it shows. Yet nv fanboys are perfectly happy with an overclocked mid-range card serving as a high-end, with real high-end reserved entirely for the professional market. (GF110 is already out as a tesla card, too bad it's not out as a quadro so we could actually test its performance in games, but then again drivers could be sub-optimal

Jacky60 the more GPUs you have rendering the game the less micro-stutter you have. So four gpus getting the same frame-rates as two gpus should feel smoother.

 

[Jacky60]

Reviewers should test high end multi-gpu setups with vertical sync on. If we're going 16xAF, 4/8AA MSAA or SSAA then vs on is blindingly obvious. I was blind about it until 2 hrs ago.

 

[Silverforce11]

Reviewers should test high end multi-gpu setups with vertical sync on. If we're going 16xAS, 4/8AA MSAA or SSAA then vs on is blindingly obvious. I was blind about it until 2 hrs ago.

First thing i did in old days while gaming was enable vsync. Night n day difference in smoothness. Especially with tripple buffering.

 

[Keysplayr]

http://www.hardocp.com/article/2012/03/28/nvidia_kepler_geforce_gtx_680_sli_video_card_review/9
published march 28

Thanks. And that was with 12.3 drivers. So After the 12.11's It must be getting more pronounced if now TR is reporting on it. I wonder who else will examine this phenomena.

 

[Lepton87]

Reviewers should test high end multi-gpu setups with vertical sync on. If we're going 16xAF, 4/8AA MSAA or SSAA then vs on is blindingly obvious. I was blind about it until 2 hrs ago.

It would hide any performance differences between the set-ups. They should test without v-sync and then turn on v-sync and see if the smoothness problem is gone. If it's not gone only then should they bitch about it.

 

[Zanovar]

NO NO NO NO NO.

You need to capture the end-user experience, and there are elements of the graphics chain that you are not taking into account when you do what you suggested. I suggest that you read this:

http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/11

"The slide above shows the frame production pipeline, from the game engine through to the display, and it's a useful refresher in the context of this discussion. Things begin with the game engine, which has its own internal timing and tracks a host of variables, from its internal physics simulation to graphics and user input. When a frame is ready for rendering, the graphics engine hands it off to the DirectX API. According to Petersen, it's at this point that Fraps records a timestamp for each frame. Next, DirectX translates high-level API calls and shader programs into lower-level DirectX instructions and sends those to the GPU driver. The graphics driver then compiles DirectX instructions into machine-level instructions for the GPU, and the GPU renders the frame. Finally, the completed frame is displayed onscreen."

Once again. NO. You guys bickering back and forth for pages and pages in this thread need to face the fact that you will need to take into account whatever black magic NV and AMD do that is not captured by FRAPS. The only real way to do that is to literally point a high-speed camera at the monitor and compare. Use multiple monitor types if you are afraid that the results are monitor-specific. That is the ONLY way to capture end-user experience objectively. Not FRAPS frametimes, not subjective user experiences.

So why havent any of these review sites used these then? incompetent,lazy or unable to afford them?maybe other variables.maybe we all should chip in in for one of these high speed cameras and pass them around to these reveiw sites.just a thought

 

[BrightCandle]

High speed cameras can not capture the age of the frame. What they allow reviewers to do is see the ms. So you can subjectively measure that their is some from the screen shots slowed down from normal but not get a comparable number.

 

[supremor]

Are we really going back to that mid range part nonsense? I guess then you're implying that Nvidia's so called mid-range is on par with AMD's high-end parts since the 680 and 7970 GE aren't very far apart?

I never understood why people even argued this point to begin with. And no, I have no brand loyalty but this particular point just irritates me.

 

[thilanliyan]

I agree, we have been talking about microstutter for years.
Now I'm not trying to be nasty or mean or anything, but folks DO try and keep up with this question.
In all the years talking about this, when does anyone remember so much talk about Nvidia being smoother than AMD? Serious question.

I think for a shorter time than we have been talking about microstutter in general, but it could also be that a quantitative method of measuring the difference it was not thought of until more recently?

When (if they have done) did nVidia introduce some sort of frame time limiter to their drivers? Maybe that is when people started noticing that SLI was smoother than CF.

High speed cameras can not capture the age of the frame. What they allow reviewers to do is see the ms. So you can subjectively measure that their is some from the screen shots slowed down from normal but not get a comparable number.

This is what I was thinking of as well...since the camera is decoupled from the computer, you can't really say you have X ms at frame Y and compare between setups?

 

[cmdrdredd]

Per the vsync question...for some games and players, input lag is worse than a second of screen tearing here or there.

 

[Lepton87]

Are we really going back to that mid range part nonsense? I guess then you're implying that Nvidia's so called mid-range is on par with AMD's high-end parts since the 680 and 7970 GE aren't very far apart?

I never understood why people even argued this point to begin with. And no, I have no brand loyalty but this particular point just irritates me.

Exactly, check 560ti and 5870.
http://www.anandtech.com/bench/Product/511?vs=547 If GF100 failed NV could increase GF104(114)clocks further and make some turbo shenanigans and have a very competitive product performance-wise from their MID-RANGE LINE. They would be competitive until Cayman release. There're further parallels between those chips, both have seriously crippled compute performance, because they are gaming chips, never meant to be their highest tier solution. If 7970 was 80-90% faster then 6970 instead of 45% with release drivers I'm pretty confident that we could now buy GK110 based GeForce cards. AMD was too conservative with 7970 because even 300mm2 NV mid-range chip could beat it at release.
There isn't a big performance gap between 5870 and 560ti. GTX460 was released with very low clocks and its shaders were cut-down. 560Ti shows us what GF104(114) can do when fully unleashed. Locked voltage control, cheap PCB it all screams 300$ card. Do you really think that NV planned their next high-end card to have the same memory bandwidth then its predecessor? Apart from that, NV already released their high-end chip in the form of GK110. So arguing that GK104 was always meant to be their high-end flagship card is from my point of view stupid.

UPDATE:

I guess then you're implying that Nvidia's so called mid-range is on par with AMD's high-end parts since the 680 and 7970 GE aren't very far apart?

Yes, that's exactly what I'm saying but it only pertains to gaming performance, let's leave compute performance out of this.
560ti and 5870 there you have it, NV mid-range card pretty match performs on par with AMD high-end card. It already happened in the past so why do you think it's so improbable?

560ti almost ties 6950 which is a tiny little bit faster then 5870. So that's another example of NV mid-range chip performing on par with AMD high-end, the other example is 680 vs 7970. Although with 7970 GHz edition AMD managed to eke out a small win, but that's nothing major. With current drivers vanilla 7970 pretty much ties with 680. Even at release GTX680 was only about 5-8% on average faster then 7970 at 2560x1600 resolution. After OCing both cards 7970 was faster.