Interesting data on Crossfire vs SLI

wonderfield said:
No, I don't believe the answer is that simple.
Click to expand...

Do you recall the very fist Xbox gpu was Nvidia? What a mess it was when the 360 switched to a AMD gpu. Backward emulation compatibility was possible with certain titles chosen by Microsoft because of the royalties Nvidia wanted for each title chosen.

Why do you believe Sony, Microsoft and Nintendo all decided to go with AMD?
 
It really looks like with the AMD Crossfire setup, Crossfire is inefficient transferring data fast enough between multiple cards. On the other hand, it looks like it's "brute-force" rendering to get a frame processed/rendered fast enough regardless. That ping-pong action between frame numbers looks like each card is waiting for the other to output data and/or process it before the next frame can be rendered onto the screen.

It looks like:
... Radeon card #1: "Here's a completed frame 1, sending it to screen, send me the other half of frame 2"

... Radeon card #2: "Data for frame 1 is still there, clear it and I'll process half of frame 2"

... Radeon card #1: "Data cleared, processing other half of frame 2, sending to my memory"

... Radeon card #2: "Retrieving your frame data, assembling completed frame, sending to your memory to output"

... Radeon card #1: "Received completed frame 2, sending it to screen... "
​
(Something like that... :p )

Memory speed/bandwidth/technology issue?
Data transfer issue between cards?
Issue with how the GPU renders each frame when in Crossfire?
 
crossfire definitely has issues, i can only vouch for SLI, i didnt even want to bother getting into ATI's drivers last year when i had some high end cards... i just said screw it back to NV
 
Regardless, I just went with CF'd 7970s. Despite this "issue," there seem to be workarounds, and I believe that AMD will respond accordingly and we will see an official fix in the future. I have better faith in their hardware, especially in recent months with all of their driver updates, and I just couldn't bite on 680's knowing that Titan was their true flagship card all along, and they released it months down the road at such a ridiculous premium.

Maybe I'm just an AMD fanboi :-x
 
if I saw this a year ago i'd be rage'n now. I got 5 7970s at launch.. 2 to crossfire in my HTPC and 3 for my main box.

I replaced mine the HTPC a few ago because of audio issues.. and my 3 way crossfire (overlooking the "fake" fraps numbers...) most games are slower with 3 way than 2.. *sigh*

but titans are here... SHIP ALREADY so I can get this crap crossfire out of my system.
 
Hmm, this changes everything almost bought 2 7970s earlier today. Maybe ill just get the one titan and add another later this year.
 
What exactly are these "runt" frames? Are they something that only manifests itself with vsync disabled? I.E. are they real, whole frames with vsync enabled or an artifact AMD created to fudge benchmarks?

As for that graph showing no performance improvement with Crossfire, if I find a game that plays with vsync enabled on one card at 32fps, and fraps shows it at 32fps, if I then add my second card and fraps shows it playing at 60fps, I then record the screen with my camera the camera will still only show 32 distinct frames being displayed? I may test this.
 
Fooshnik said:
What exactly are these "runt" frames? Are they something that only manifests itself with vsync disabled? I.E. are they real, whole frames with vsync enabled or an artifact AMD created to fudge benchmarks?

As for that graph showing no performance improvement with Crossfire, if I find a game that plays with vsync enabled on one card at 32fps, and fraps shows it at 32fps, if I then add my second card and fraps shows it playing at 60fps, I then record the screen with my camera the camera will still only show 32 distinct frames being displayed? I may test this.
Click to expand...

Test it. In fact, test it with and without the frame limiter set to your average FPS in that game.
 
When I am running one 7970, I can not hit 60 fps on max in Crysis 3. When adding the other 7970, I can easily hit 60 fps on max.

I can tell the difference with my own eyes without using fraps or OSD - Ok, so the frames are not finished being rendered in real time based off third party software? So what, this does not effect my real time gaming experience and in the end, that's all that really counts, not third party statistics that are not visible to the naked eye.
 
solidnickel said:
When I am running one 7970, I can not hit 60 fps on max in Crysis 3. When adding the other 7970, I can easily hit 60 fps on max.

I can tell the difference with my own eyes without using fraps or OSD - Ok, so the frames are not finished being rendered in real time based off third party software? So what, this does not effect my real time gaming experience and in the end, that's all that really counts, not third party statistics that are not visible to the naked eye.
Click to expand...

Glad to hear it. What resolution are you running?
 
solidnickel said:
Do you recall the very fist Xbox gpu was Nvidia? What a mess it was when the 360 switched to a AMD gpu. Backward emulation compatibility was possible with certain titles chosen by Microsoft because of the royalties Nvidia wanted for each title chosen.

Why do you believe Sony, Microsoft and Nintendo all decided to go with AMD?
Click to expand...

Yep, had nothing to do with the fact that they completely changed the core system architecture on the Xbox from x86 to PPC. Nope, nothing to do with that.
 
Pinski said:
Yep, had nothing to do with the fact that they completely changed the core system architecture on the Xbox from x86 to PPC. Nope, nothing to do with that.
Click to expand...

I know this! why I said it was emulated...

The question was why do you think all the next gen consoles chose AMD over Nvidia?
 
not going to read the whole thread but i have to say that this is 'micro benchmarking' at its best..

i'm so glad that [H] goes with realworld experiences and how the game plays.. we could micro analyze ourselves to death and then we'd never play any games.


why hasn't 3dmark needs to come out with FrameTimeMark2013 and lets be done with it.
 
H-street said:
not going to read the whole thread but i have to say that this is 'micro benchmarking' at its best..

i'm so glad that [H] goes with realworld experiences and how the game plays.. we could micro analyze ourselves to death and then we'd never play any games.
Click to expand...

Apparently you didn't even read the article, much less the thread.

This "micro analyzing" is helping us find out why we have better "real world" experiences.

Enjoying your game? Ever wonder how things get fixed or improved? It's by finding the problem. You can't find the problem by just saying "oh - it kind of looks worse/better". It gets fixed by finding a scientific way to explain why we see what we see. Frankly, the new frame rating system developed by PCPer is an incredible achievement in benchmarking in my opinion.

As far as I know - they are the first to explain why even though CFX has higher framerates, Nvidia SLI has a visually smoother presentation.

At the end of the day - this "micro anazlying ourselves to death" really benefits everybody. It helps even AMD to attempt to fix the CFX issue and helps to finally explain a phenomena the industry has experienced for so long but could not be explained.
 
three4seven said:
It helps even AMD to attempt to fix the CFX issue and helps to finally explain a phenomena the industry has experienced for so long but could not be explained.
Click to expand...

Again they're not explaining what these 0ms frames are. So they only show as two rows of pixels with vsync off. So what? All that demonstrates is that if you play with vsync off some of your frames are only two pixels tall. Who plays like that? If enabling vsync and Afterburner's frame server delays those frames so they're displayed in a homogenous manner, then stating there's no performance benefit with Crossfire is disingenuous at best, an outright lie at worse. If that were the case, anyone who bought a second card because their game was only running at 30fps, and now fraps is telling them they're at 60fps, was still playing at 30fps and didn't notice. Also, that would mean the whole time you're playing the second card was generating 125W of heat to do absolutely nothing. If that were true, I'd think someone nvidia would've made it known that AMD was scamming people.

Until they publicly state that these 0ms frames are, in fact, non-frames, this research isn't telling anybody anything they didn't already know.
 
Fooshnik said:
Until they publicly state that these 0ms frames are, in fact, non-frames, this research isn't telling anybody anything they didn't already know.
Click to expand...

I respectfully disagree. Just because you play with V-sync on and frame limited to 60, doesn't mean that's true for everybody. This research also tells us that long accepted industry standard of FRAPS is actually flawed, and actually attempts to give us an objective reason why. Even Kyle commends them for this. How can you say the research has shown nothing new? Can you say you had the answer, or even how to come to one? This research, while you may claim to be incomplete, is still new and helpful in finding out what exactly is the "smoothness" difference. You can't dismiss the research altogether just because it doesn't answer a specific question for you.

I wanted to add, they never claimed that there is zero performance benefit with CFX. They didn't make a claim and then tried to prove it with data. They had data, and made an objective interpretation. How you construe their findings is your own interpretation. Even their most damning statement "Removing the tiny slivers gives us a "perceived frame rate" that differs quite a bit - CrossFire doesn't look faster than a single card" is just a statement objectively interpreting the data.

Finally, they also state their research is still ongoing and the article was simply a preview.
 
Last edited:
People have been measuring Crossfire frame times for years, like here, and demonstrating the extreme inconsistency. What's new with this?

Also interesting is frame times with RadeonPro.

Regardless, I'm not questioning any of that. I'm questioning their implication that there's no benefit with Crossfire.
 
Last edited:
EPOQ said:
CF'ed 7970s for 1080p, dayum son. You must have foreseen the Crysis 3 requirements for obtaining 60fps.
Click to expand...

Actually, on my new build; I experimented with various GPU's including the GTX 680, GTX 670 Power Edition and the 7970...

All in sli and xfire; the 7970 was more appealing to me because of the extra vram and the ability to run all my games at max especially Sleeping dogs! This game truly looks next gen when cranked to max. Next gen is already here with this game and I don't see next gen systems such as the ps4 looking much better then this game. Love the fact square optimized it for the pc, its a technical marvel IMO.

The 7970 xfire was the only combination to run this game at max from all the cards I experimented with btw.
 
Fooshnik said:
People have been measuring Crossfire frame times for years, like here, and demonstrating the extreme inconsistency. What's new with this?
Click to expand...

The Tomshardware article identifies that the quality of each frame is unequal, but still uses FPS as a metric to explain microstuttering. What PCPer is doing is rating the quality of frame itself, not just using FPS as a metric.
 
three4seven said:
What PCPer is doing is rating the quality of frame itself
Click to expand...
What is a "low quality" frame? I've never heard that phrase before outside the context of IQ. All I'm seeing are instances where two frames appear to hit the display driver in rapid succession. They're dropping the first because the second arrived 1ms later instead of 14. Why? Is it not a full frame with vsync enabled? I would be truly stunned if their DVI capture card recorded only 30 full frames with vsync enabled while fraps was showing 60. Unless that's the case, dropping the runt frames has no relevance to actual gameplay unless you're playing with vsync off.

This was already partially addressed by the author:

http://www.pcper.com/reviews/Graphi...Frame-Rating-Update/Frame-Rat?page=1#comments

February 22 said:
Seriously? You're still doing this? Can you please give me one reason, just ONE reason, as to why I would want to play a game without Vsync and Triple Buffering? Because that is the only, and I repeat, the ONLY situation in which your "novel" and "state-of-the-art" method of benchmarking VGAs would be relevant to an actual real-life scenario. I'd nominate you for a Nobel prize if you do.
February 22 said:
I kind of agree on Vsync, but frame buffering, probably not:

"1. If it is not properly supported by the game in question, it can cause visual glitches. Just as tearing is a visual glitch caused by information being transferred too fast in the buffers for the monitor to keep up, so too in theory, can triple buffering cause visual anomalies, due to game timing issues for example.

2. It uses additional Video RAM, and hence can result in problems for those with less VRAM onboard their graphics card. This is particularly true for people who also want to use very high resolutions with high quality textures and additional effects like Antialiasing and Anisotropic Filtering, since this takes up even more VRAM for each frame. Enabling Triple Buffering on a card without sufficient VRAM results in things like additional hitching (slight pauses) when new textures are being swapped into and out of VRAM as you move into new areas of a game. You may even get an overall performance drop due to the extra processing on the graphics card for the extra Tertiary buffer.

3. It can introduce control lag. This manifests itself as a noticeable lag between when you issue a command to your PC and the effects of it being shown on screen. This may be primarily due to the nature of VSync itself and/or some systems being low on Video RAM due to the extra memory overhead of Triple Buffering."
Click to expand...
Click to expand...

I have to agree as well with triple-buffering, in fact it's not even an option on many recent games, and there are better ways of addressing frame latency than triple buffering. That said, I see no relevance of these efforts to actual gameplay unless you're playing at 120Hz, with vsync disabled, on Crossfire. Outside that context their statement that "removing the tiny slivers gives us a "perceived frame rate" that differs quite a bit - CrossFire doesn't look faster than a single card" is false and going to be the source of great FUD among the misinformed for years to come.
 
Fooshnik said:
Snip.
Click to expand...

You've went off on the deep end on your own tangent. I don't agree or disagree with you on those points. All I'm saying is that I respect PCPer for trying something new to improve or more accurately assess our current technology. You can't dismiss all their research just because it doesn't effect you necessarily.

If you're content with your experience, more power to you. You're talking to me like I'm the one who wrote the article. If you have beef with it, bring it up there.
 
EPOQ said:
I wouldn't call it "damning" by any means. Limit the frames to 59FPS and you have a different story, no? What's the point of the 80-100 frames on a monitor with a 60hz referesh rate, anyway?

Yes there would be greater disparity when you are gaming at 120hz but we are talking about high resolution or multi-monitor setups. People arent going to be trying for 120fps in games like BF3 on maxed settings, but if you are for some reason I guess this article is telling you to go Nvidia.
Click to expand...


People are trying for 120 FPS games like bf3-- I am because I have 120 Hz monitor and I run it with lightboost/0 motion blur. So it has to hit 120 fps plus deal with 3d overhead.
 
three4seven said:
You've went off on the deep end on your own tangent. I don't agree or disagree with you on those points. All I'm saying is that I respect PCPer for trying something new to improve or more accurately assess our current technology. You can't dismiss all their research just because it doesn't effect you necessarily.

If you're content with your experience, more power to you. You're talking to me like I'm the one who wrote the article. If you have beef with it, bring it up there.
Click to expand...

Oh. I didn't know it was something new and unexplained since people have been quantifying unequal frame times for years.

Anyway, I'll move this over there, instead of this thread, which was started to discuss the article.
 
Last edited:
Fooshnik said:
What is a "low quality" frame? I've never heard that phrase before outside the context of IQ. All I'm seeing are instances where two frames appear to hit the display driver in rapid succession. They're dropping the first because the second arrived 1ms later instead of 14. Why? Is it not a full frame with vsync enabled? I would be truly stunned if their DVI capture card recorded only 30 full frames with vsync enabled while fraps was showing 60. Unless that's the case, dropping the runt frames has no relevance to actual gameplay unless you're playing with vsync off.
Click to expand...

I think you greatly underestimate the number of people who game with Vsync off.
 
Fooshnik said:
Oh. I didn't know it was something and unexplained since people have been quantifying unequal frame times for years.

Anyway, I'll move this over there, instead of this thread, which was started to discuss the article.
Click to expand...

I agreed that people have been quantifying equal frames in the past. They're approaching their research in a new manner, which to my point again, I respect.

It seems you need validation that you made the right decision in purchasing your 7870 Mysts recently - which you apparently will be playing with V-sync on at 60fps.
 
Forceman said:
I think you greatly underestimate the number of people who game with Vsync off.
Click to expand...

No I don't. But if someone is playing with vsync off then screen tearing isn't an issue for them, so frame height probably isn't going to be either.
 
solidnickel said:
Do you recall the very fist Xbox gpu was Nvidia? What a mess it was when the 360 switched to a AMD gpu. Backward emulation compatibility was possible with certain titles chosen by Microsoft because of the royalties Nvidia wanted for each title chosen.

Why do you believe Sony, Microsoft and Nintendo all decided to go with AMD?
Click to expand...

Yes it was and Nvida tried to screw MS over.... because of that they got canned on the next two designs and lost out....
 
three4seven said:
I agreed that people have been quantifying equal frames in the past. They're approaching their research in a new manner, which to my point again, I respect.
Click to expand...
Whether you use a new stopwatch or old, you're still measuring time. You respect them for that, I don't care.

three4seven said:
It seems you need validation that you made the right decision in purchasing your 7870 Mysts recently - which you apparently will be playing with V-sync on at 60fps.
Click to expand...
This is exactly it. I never comment on misleading articles unless I have a vested interest. Everyone knows that once it's discovered AMD's been lying about their framerates this crap'll be impossible to sell, not to mention the soul-crushing misery of having to live with an ill-conceived decision. Now I'm going to take a break from seeking affirmation to go read through your post history.
 
Fooshnik said:
This is exactly it. I never comment on misleading articles unless I have a vested interest. Everyone knows that once it's discovered AMD's been lying about their framerates this crap'll be impossible to sell, not to mention the soul-crushing misery of having to live with an ill-conceived decision. Now I'm going to take a break from seeking affirmation to go read through your post history.
Click to expand...

lol. Sarcasm eh? You seemed overly vested in this discussion, so I took a look. Low and behold, you recently purchased 7870 Mysts (more than one it would seem). All of a sudden this discussion seems more pertinent to you.

You can call it coincidence. We can let the readers of this fine board decide.
 
Lemme end the suspense: They don't care. This thread is about PCPer's paper. If one of the two major video processor designers is lying about their framerates or fudging the numbers that's an enormous issue. Anyone who buys products from either manufacturer has a vested interest. There are tens of thousands of people using Crossfire or considering it, I'm sure they'd appreciate knowing if AMD is falsifying their data. That's not the case unless PCPer is withholding significantly more compelling information.

I've been using Nvidia/ATI for years, and 3Dfx SLI before that. That you'd go through someone's post history when that has nothing to do with anything is frankly a little bizarre and obsessive.
 
You must log in or register to reply here.
Back
Top