AMD Crossfire a scam - Almost no benefit over single card

Status
Not open for further replies.
I digress on my comment about the newbies not knowing the lineage of dual video cards in PC gaming..... I was gaming back when Cirrus Logic was the big dog in the video card world so I have seen it all.... However, I could see others being more concerned with Dual cards issues than I would be. Single Cards are the way to go.
 
Meh, I gave up on multi-gpu's a long time ago. When I went SLI, I was expecting a game to go from medium settings to ultra or high settings at least. But instead it went from medium to medium high. From that day forth, I decided money is better spent by selling the one you have and upgrading to a new card rather than trying to add another card.

I've got to locate this article but are they testing with the flagships? Are they testing them on PCIe 3.0 boards? Are the CPU supporting PCIe 3.0?

I do believe that multiple flagship GPU's are saturating the PCIe 2.0 bandwidth finally aren't they?

Everytime i am tempted to add a second card for Sli or X-Fire.. i always end up just buying the next gen card that is faster anyway... So its a moot point for me...
 
Not taking into account the article at PCPer, with what experience I have had with dual GPU configs (last being the 5850 Crossfire so I have not experienced recent GPUs); it has been my experience that 2 GPUs makes the experience effectively worse. Microstutter makes 70fps worse than 40 FPS (on a single card) in terms of playability.
 
Depends on the game engine, drivers, etc. Been running 6970s in Crossfire for nearly 2 years (I think) now, and while there have been issues along the way, that gen seems to have the drivers nailed down. While the FPS numbers may not tell the whole story, they never have. There is a perceivable performance increase when a second card is added, in many games it is substantial.

TLDR; while perhaps the stated FPS gains are not as high as claimed by benchmarks like FRAPs, there is still a significant advantage gained by Crossfire in many games.
 
They are essentially giving measurable data as to what microstutter is and opening the possibility to have amd/nvidia mitigate it to the best of their ability, this is exciting for me since i've used dual/tri cards for the last 10 years or so having owned multiple pairs/tris of cards from each vendor. This has long been a thorn in my side and it is really meaningful that we can finally move in a positive direction for the hordes (and their are) of us that employ multi gpu setups.
 
So Dual Card issues.. that have been going on since the VOODOO2 days are all of a Sudden NEW issues? some of you guys must have just started gaming on the PC with WOW or something

I think in the past, when SLI/CF had more problems, and didn't have anywhere near "perfect scaling", it really made sense to just get the best single GPU. But now I feel like dual-GPU is advertised or referred to others as literally having "double the performance of a single GPU". Though the frame rates may agree with that, the perception of a doubled frame rate won't be anywhere near that.

I had a similar experience to the OP using unlocked HD 6950s in CF briefly. Compared to a single HD 6950, of course there was a very large frame rate boost, and it felt better overall. But something felt off with the displayed frame rate - it always seemed too high. At some point I realized I was experiencing micro-stutter. And once I knew, it was too late. It just completely destroyed multi-GPU for me and I was able to finally understand why others complained about it (despite so many people saying MS wasn't a big deal anymore - WRONG). It's not something you can literally see (if you see stutter it's no longer "micro-") but you definitely feel it - but only if you have a feel for what a certain frame rate should feel like otherwise (if that makes any sense at all).

Anyway, then I got a GTX 670, and when overclocked still put up less frames than the 6950 unlocked CF, but yet the 670 felt so much better. Then and there I came to the conclusion that many others are starting to now realize. I started ignoring CF/SLI benchmarks from there on out, and only focused on single GPU.

These new frame time reviews aren't the end-all conclusion, as it can be argued to be highly subjective. However, I honestly don't believe that anyone is immune to witnessing MS. It's just that they don't know. I didn't know for a little while when using CF 6950s... and then one day I just... "saw it".
 
I had a similar experience to the OP using unlocked HD 6950s in CF briefly.

I actually had that exact setup in one of my rigs i built a little while back. It was replacing a trifire 5970+5870 that didn't pan out performance wise in games (benched like a champ though!)

Again, you left with a sour taste in your mouth feeling deceived by what was being advertised by ATI/AMD at the time. Since they couldn't fix your issue, as there wasn't someone to call them out on it in a measurable/repeatable way, you simply decided not to do business with them again, which is why i hope they choose to remedy what they have control over since i like having a healthy choice in gpu horsepower.
 
Not quite sure why focus on AMD while this has been the issue since Day 1 for all Multi-GPU.

For some title, GTX 680 SLi indeed feels smoother, and 7970CF feels stutter, but that only happened to limited title of games, and sometime it happened the other way around.

I have been playing SLi and CF for ages, I have to say that both have it's own issue, stutter happened across both brand, but happened on AMD more often lately for some reason.
Both of them also have a horrible input delay, which is the main thing that push me off to Single GPU.

PS: Didn't AMD were coming up something called "sunspot" or something to fix all these? I wonder whats the development of it now.
 
I've been suffering from stuttering for years, everyone said I was crazy and googling for solutions always came up with nothing. I assumed I was the only person in the world with the problem. I even showed friends that came over to my place and either they couldn't see it or if they did see it they thought I was being picky.

It drove me nuts! I discovered a way to fix it by turning my high precision timer off in BIOS. Ran with it off for over a year, few months ago I turned it back on to see what would happen and discovered that Nvidia had actually fixed the issue and everything ran really smooth again. Guess it went from stutter to micro-stuttering when I turned the HPT off on a single GPU.

After reading the original post on Anandtech and seeing AMD was saying things along the lines of it's not the drivers causing it, it's windows or the application, so it's not their problem so they're not fixing it really pissed me off. They knew there was a problem but since they convinced themselves that it wasn't their issue they didn't have to fix it -- until someone exposed them and they realized they might lose sales. They have a black eye and now it's time for damage control.

It's disconcerting that both AMD and Nvidia lost the point of people paying $500 or more for video cards: the end user experience. We don't do it so we have the most FPS possible we do it trying to get the best game play possible. I don't care if the game is running 60 fps or 160 fps, I want it to be smooth as silk. I understand the tech behind it all is very complex and there are things out of their control but they failed in understanding the whole point of what they are making. Screw FPS, I want smooth game play, end of story.
 
I actually had that exact setup in one of my rigs i built a little while back. It was replacing a trifire 5970+5870 that didn't pan out performance wise in games (benched like a champ though!)

Again, you left with a sour taste in your mouth feeling deceived by what was being advertised by ATI/AMD at the time. Since they couldn't fix your issue, as there wasn't someone to call them out on it in a measurable/repeatable way, you simply decided not to do business with them again, which is why i hope they choose to remedy what they have control over since i like having a healthy choice in gpu horsepower.

I hope they do as well. It actually turned me off from multi-GPU setups entirely, even though Nvidia is better.

I'm not expecting any miracles with the current generation of cards, as maybe they'd have to adjust something on the hardware level to fix this, but now they've at least acknowledged the problem, a slightly faster HD 8000 refresh with stuttering issues fixed would certainly make for a worthy refresh.
 
Gee that's funny OP, because FRAPS and my own eyes tell a different story. I guess I am just hallucinating when I see the difference Crossfire makes in a game like BF3.

edit: Guess calling out a troll post is frowned upon by moderators, I am sorry :rolleyes:
 
Last edited:
man use a framelimiter like the one built into radeonpro, you will not get choppy gameplay and a LOT more performance (varies by title obv.)
 
After reading the original post on Anandtech and seeing AMD was saying things along the lines of it's not the drivers causing it, it's windows or the application, so it's not their problem so they're not fixing it really pissed me off. They knew there was a problem but since they convinced themselves that it wasn't their issue they didn't have to fix it -- until someone exposed them and they realized they might lose sales. They have a black eye and now it's time for damage control.

It's disconcerting that both AMD and Nvidia lost the point of people paying $500 or more for video cards: the end user experience. We don't do it so we have the most FPS possible we do it trying to get the best game play possible. I don't care if the game is running 60 fps or 160 fps, I want it to be smooth as silk. I understand the tech behind it all is very complex and there are things out of their control but they failed in understanding the whole point of what they are making. Screw FPS, I want smooth game play, end of story.

You completely nail it right here.

For those starting to flame, this thread is a reactionary discussion to the article from pcperspective i linked to in the first post, please familiarize yourself with the data if you are going to make an argument. While I appreciate all opinions, my own experiences have aligned virtually 100% with every measurement and issue that is raised in that article. If you can disprove the factual data they have presented i would appreciate it. Or if you have other methods that can correct the issue that you have created data sets for and can repeat the fixes please share them. Having someone tell you the second card you bought and for a lot of people saved up money for over time is actually providing a worse experience can be hard to swallow. If we didn't get upset with data like this why would AMD ever bother to work towards a fix?
 
Crossfire looks like shit in a lot of games unless you cap the framerate. I'm glad that sites are finally looking for a way to test for smoothness.

Yes, capping framerates is a nice workaround but its not a great solution in some circumstances. 120hz displays come to mind.


Yeah, that thread title is asking for the thread to be closed.

man use a framelimiter like the one built into radeonpro, you will not get choppy gameplay and a LOT more performance (varies by title obv.)

Thats not such a great option for 120hz displays or eyefinity and still only works unless you have framerates over the cap. Yeah, it works well with a 60hz display but microstutter is really something that AMD finally needs to attempt to address.
 
Last edited by a moderator:
Frame capping seems pointless to me. If your dual GPUs are running 80fps but they "feel like" 60fps on a single card, so you frame cap at 60fps so it "feels like" 60fps - what's the point? :p
 
80fps on a 60hz display is going to feel like 60fps anyways but less smooth than it would with a framerate cap. I would use a cap with Nvidia cards too if I had a 60hz display. It makes things really smooth.
 
However, I honestly don't believe that anyone is immune to witnessing MS. It's just that they don't know. I didn't know for a little while when using CF 6950s... and then one day I just... "saw it".

I completely agree with you there.

I actually sold my 4870x2 because in some games like crysis it didn't feel any smoother than my GTX280. I didn't realize until later that was microstutter.
 
the problem with GCN based cards is not dropped or runt frames as some point out, the major issue with the 7k series by and large seem to be the memory controller not feeding as it should, so it simply is not processing efficiently. This has been a known issue for quite awhile, cat 12.11 did some work on it, 12.12 a massive amount but most of the "fixes" are being polished off.

Yes it sux to not have your stuff working as one wants, and AMD is guilty of things such as advertising a FX as a "quad-6-eight core" and it really sux not to have great performance all the time, but folks also have to face some facts, Nvidia focuses on graphics development, more or less always have, they simply have far more $ to throw behind development in hardware as well as software compared to AMD, and Intel well LOL they are massive compared, but even they do not always do things perfectly.

If there are ways we can help them, that's what we should do, it benefits us all in the long run. It is a known issue, has been for many months, scaling as I worded and so many folks have pointed out over the years, historically Nvidia has more support(at the moment) AMD/ATI have great scaling in the games that play "nicely" with it. something like 86% and 92% respectively for dual cards and AMD overall by far has more performance gained in quad configs then Nvidia has, there is always exceptions or provables as with anything else.

GCN is "new" and it will take time to work the kinks out of it, even Nvidia 600 series have their own issues apparently, be patient, get a more powerfull single card solution, make your own profile/fixes or something.

The only thing I do not like with my 4870-6870-7870 is some games simply put didn't behave so nicely, such as BC2-BF3 from texture flickering and the like, but they are mostly all sorted out now, and the random driver freezing sux, been like that for years from both companies, but hey, overall I am happy, nothing is perfect, even a well designed car will have unforeseen issues sometimes :)
 
I don't know about that. I really think that memory controller comment was an excuse. Cayman and Cypress used to do far worse than GCN ever did as far as frametimes go in older techreport reviews. The driver wasn't optimized for those cards either?
 
single 7970 FTW, but seriously are we looking at a new age when the tools allow us to peak inside what is really going on. Single GPU's seem to still be the best solution and multiple GPU's just some elaborate hoax when it comes to acutal performance. AMD will suffer some backlash for the data coming out but we all may benefit in the long run.
 
Good to see this article posted again for the fifth or sixth time. I was never able to master the search function either.

Of course if you're having stuttering with Crossfire you could always fix it.
 
Crossfire looks like shit in a lot of games unless you cap the framerate. I'm glad that sites are finally looking for a way to test for smoothness.

Yes, capping framerates is a nice workaround but its not a great solution in some circumstances. 120hz displays come to mind.



Yeah, that thread title is asking for the thread to be closed.



Thats not such a great option for 120hz displays or eyefinity and still only works unless you have framerates over the cap. Yeah, it works well with a 60hz display but microstutter is really something that AMD finally needs to attempt to address.

Yeah this is a good point. I haven't had issues using CFX and my 72 Hz display in most games except crysis 3 (which I framelimit to 50fps and it still feels pretty smooth...the animations are good in this game or something idk). yes, your minimum framerate needs to be above the limit that you set. Still, most games can be run max @ 60-70fps (rarely if ever dipping below) which I don't really think can be said for single cards, avg fps maybe but I doubt you'd see minimums at 70 for some of the more demanding titles. It is not a perfect solution at all I agree with you--AMD needs some work, but I do not think that the second card is "useless".
 
Yeah this is a good point. I haven't had issues using CFX and my 72 Hz display in most games except crysis 3 (which I framelimit to 50fps and it still feels pretty smooth...the animations are good in this game or something idk). yes, your minimum framerate needs to be above the limit that you set. Still, most games can be run max @ 60-70fps (rarely if ever dipping below) which I don't really think can be said for single cards, avg fps maybe but I doubt you'd see minimums at 70 for some of the more demanding titles. It is not a perfect solution at all I agree with you--AMD needs some work, but I do not think that the second card is "useless".

Some games are not playable at a high resolution settings (ultra) without a second card. 3d especially also comes to mind because its very taxing. A question to the OP, do you feel scammed by AMD because of the articles you read or from self experimentation?

I rely on my own experience and honestly can tell a big difference when running xfire on high demanding games. Real time gaming @ 60fps on ultra is all I really care about.
 
Hrmmm this is strange....

It would seem I must be imagining double the performance when running crossfire mode on my 6990 in games? Must be a dream. Pinch me.
 
no the memory controller is not an excuse, it has been proven to not feed appropriately(as was designed) this is the mem controller used specifically on GCN based cards, the fix will be implemented in hardware for GCN V2 based cards e.g the real HD 8k series whatever they decide to call them, magic bullet no, but it will def help according to internal and their party test by a magnitude.

I do not remember the whole spiel, though it very much had to do with it(memory controller) not being "stable" in its processing even though the speed(bandwidth) was there, it was termed an errata, this is specific to GCN based products and part of the reason why the 7700 showed quite poor performance (at launch) more then AMD had wanted.

It took them(and the community-3d party) quite awhile to work out why, as mem controller simply was not acting as it should, and yes it also had a bearing on crossfire performance(to a point), I wish I could remember where I had seen the article describing it, it was around same time they started talking about the whole runt frame thingy, AMD investigated, did their internal testing and talked to major 3rd party testers/devs to figure it out, Cat 12 end of year major performance boost from starting to sort it out 12.11 if I recall, 13 range will finish it, apparently was not a major problem, but a bunch of small ones adding up to major "quirks"

Nvidia "tuned" its controller for the 600 series, AMD really did not even though GCN based was a new arch and also a redesign of mem controller among other things, bug in the architecture I do not know, am not a coder or chip designer, just a crap ton of reading :p
 
i game at 60hz, played tomb raider last week and beat the game, at 1440p maxed out with FXAA & 16xAF, as long as I kept vsync on it was much smoother than 1 card

i've been playing sleeping dogs at 1440p and there's definite hitching even though MSI OSD says 59.9fps at all times (like when you rotate the camera all around you), this is with vsync on and off

so yeah, the micro stuttering exists, this isn't new information
 
Yeah, sleeping dogs was not smooth at all in crossfire without a hard cap.

@dragonstongue So why were frame latency so bad on Cayman and Cyrpress? The driver wasn't tuned for those either? Yeah, I'm not buying that.
 
more and more games are supporting crossfire....I don't think CF is going away any time soon...especially with the cf with the new apu's.
 
Good to see this article posted again for the fifth or sixth time. I was never able to master the search function either.
fix it.

This is in reaction to an article posted yesterday on pcper, as linked to in the op, if you can't even read or understand what a thread is discussing, why post?
 
Some games are not playable at a high resolution settings (ultra) without a second card. 3d especially also comes to mind because its very taxing. A question to the OP, do you feel scammed by AMD because of the articles you read or from self experimentation?

I rely on my own experience and honestly can tell a big difference when running xfire on high demanding games. Real time gaming @ 60fps on ultra is all I really care about.

This is why i have multiple cards. I game at high res - 2560x1600 or 2560x1440 depending on what monitor I'm using, and prefer gaming at 60fps avg if possible.
 
i game at 60hz, played tomb raider last week and beat the game, at 1440p maxed out with FXAA & 16xAF, as long as I kept vsync on it was much smoother than 1 card

i've been playing sleeping dogs at 1440p and there's definite hitching even though MSI OSD says 59.9fps at all times (like when you rotate the camera all around you), this is with vsync on and off

so yeah, the micro stuttering exists, this isn't new information

Seriously? Why xfire at that resolution? Also, is the micro stuttering constant?
You should be able to play on a single card at that resolution.
 
I have never had a single problem with SLI nor Crossfire. My k/d ratio in online FPS has always been around 2.5 and above. My perfect framerate is 60 to 80 on a Dell 3007WFP-HC 2560x1600.

If I do not use v-sync or frame limiter both setups are not giving my a smooth gameplay..
 
I haven't experienced microstutter running NV hardware in SLI since the 6xxx days....

Smooth as a baby's arse here.
 
i don't know if you guys looked the videos there:

http://www.pcper.com/reviews/Graphi...ils-Capture-based-Graphics-Performance-Tes-10

I've seen reviews on 7950 in crossfire and usually tehy push better frame rate but man it looks choppy as fuck compare to the 660's. People are calling this thread a flame bait but to me it's an eye opening. Specially if you are looking at enthusiastic that might opt for a dual card setup. From what I've read amd kinda dropped on ball hard this time. Just look at all the tearing it's insane.
 
i don't know if you guys looked the videos there:

http://www.pcper.com/reviews/Graphi...ils-Capture-based-Graphics-Performance-Tes-10

I've seen reviews on 7950 in crossfire and usually tehy push better frame rate but man it looks choppy as fuck compare to the 660's. People are calling this thread a flame bait but to me it's an eye opening. Specially if you are looking at enthusiastic that might opt for a dual card setup. From what I've read amd kinda dropped on ball hard this time. Just look at all the tearing it's insane.

Wow thats crazy. I didn't see much difference on Skyrim but on Sleeping dogs it's ridiculous.
 
My 7970s are my first dual card setup after coming from TNT2, GeForce MX440, 7950GT, and 9800GTX+. This is my first real build last year and I wanted the best of the best for BF3, so I was the few that forked out $600 for the 7970s when they first came out. I didn't know what microstutter was at that time, but I knew something was off. I was getting 80-100 fps, but it didn't feel smooth and enjoyable at all. Disabled second card, and even at 50 fps, it was smoother than 2 cards at 60 fps. Turning on v-sync and limiting framerates to 60 in MSI AB seemed to have fixed it, but that doesn't fix other crossfire issues like actual stuttering, which is more annoying.

I have learned my lesson. My next upgrade will probably be just the best single card. I would lose too much money now if I try to sell my 7970s, and go for 680s. Sometimes I am even hesitant to purchasing a game because I'm not sure if it'll play nicely with my 7970s. It's funny how I have a power gaming machine, but am afraid to buy games because I'm afraid it won't be a smooth experience.
 
I'm suprised at how the data stacks up. Not sure I could say a second AMD card is useless. At 2560x1600, I started with one unlocked 6950 with 6950 clocks. That wasn't enough so I went Xfire with 2 unlocked 6950's with stock 6950 clocks. That seemed to me to be enough. I felt like I was driving my monitor. Now, I didn't push every gadget doohicky to Ultra and BBQ high settings, but I played a variety of games just fine.

The 680's dropped, and I picked up one. Again, it was just shy of not enough so did SLI where I am now. And, sure, the 680's are much smoother and much faster than the 6950's, but I really cannot see any issues in what I'm playing. Maybe I'm just not that picky. The 6950's seemed fine to me.
 
:confused:I though Kyle did a review on multi GPU setups and said that second GPU comes in handy when you have several screens working.
 
Status
Not open for further replies.
Back
Top