Closed

Challenging FPS: Testing SLI And CrossFire Using Video Capture - page 2

198 answers Last reply
  1. bystanderPcper.com has finally released their results as well: http://www.pcper.com/reviews/Graph [...] nce-TestinThey compare 7970's and 680's, for those who wanted a more fair comparison.


    How is that a more fair comparison? It'll be nice to read another perspective, especially with different cards and even more especially since they're cards that are more likely to be used in such configurations right now IMO, but fairness doesn't seem impacted.
  2. bwcbwcI don't go quite as far in questioning nVidia's design of the FCAT tool as deliberately biased against AMD. But the fact that almost every test shows a wider disparity in the AMD configuration makes me wonder if nVidia did have issues coding for the AMD architecture. I await a similar tool from AMD or a neutral party for comparison.On the other hand, this could simply be an issue with Crossfire data interchange vs. SLI and the fact that the 7870 is inherently slightly less powerful than the 660 Ti.


    I don't believe a difference in gpu arch makes a difference. Anandtech had a great article yesterday explaining the flaws with fraps and went into specific detail over several pages in stead of a few paragraphs with one graphic:

    You'll see that the gpu is just a link in the chain of events and that nothing in the pipeline within the gpu is getting measured. They also have an article today of using FCAT:


    By reading both it should lay to rest any paranoia people seem to be expressing with using a procedure that nVidia has been using for years to avoid the issues AMD has been admittedly having for years. Really talk about looking a gift horse in the mouth.
  3. ibjeeprSo what effect would running VirutMVP have on this testing method?My understanding is that it eliminates the dropped on runt frames from ever being generated and freeing up the GPU power assiciated with each frame. Would this just even out the lines between hardware and experienced FPS or would this generate an actual improvement in experienced FPS?I've tried VirtuMVP and either didn't do it right or it didn't have any real effect. Based on this testing method it sounds like it would go hand in hand with it.


    That's a great question. I'd like to see this tested with virtual V-Sync to see if anything changes.
  4. cangeliniAnd actually, it'd be nice to see someone like Beepa incorporate the overlay functionality, taking Nvidia out of the equation.


    I am with this sentiment, a independent neutral party needs this kind of data capture technology, while it remains a Nvidia exclusive it lends itself to mala fides particularly the accuracy of measuring competitors parts. Things like software implementation is critical to, Nvidia make it exclusively for its own like so all software and firmware is tailored for Nvidia.

    I don't question the method and principle but will refrain from accepting this as trite on the pure bases that this can be a benchmarketing tool.
  5. blazorthonHow is that a more fair comparison? It'll be nice to read another perspective, especially with different cards and even more especially since they're cards that are more likely to be used in such configurations right now IMO, but fairness doesn't seem impacted.

    At least one poster was complaining that the 660ti was used against the 7870. Pcper uses the two top in cards for each company.

    I personally don't think it mattered for what is being compared, but some were complaining.

    It also goes a lot more in depth with the issue.
  6. FRAPS takes its measurement in the pipeline well before the drivers/GPU portion. That is also where the overlay inserts its bar. It should be intuitively obvious that doing it here is vendor-neutral, UNLESS you're prepared to argue that nVidia's drivers are looking for those bars to come down the pipeline and is handling frames containing them differently.
    ...which ALSO points to a way to handle runts, but would introduce problems of its own, like additional lag between t_game and t_display. If the driver can measure that bar and adjust how long the frame is displayed, runts could be "grown" to normal length. Might something like this be how adaptive vsync works?
  7. Onus said:
    FRAPS takes its measurement in the pipeline well before the drivers/GPU portion. That is also where the overlay inserts its bar. It should be intuitively obvious that doing it here is vendor-neutral, UNLESS you're prepared to argue that nVidia's drivers are looking for those bars to come down the pipeline and is handling frames containing them differently.
    ...which ALSO points to a way to handle runts, but would introduce problems of its own, like additional lag between t_game and t_display. If the driver can measure that bar and adjust how long the frame is displayed, runts could be "grown" to normal length. Might something like this be how adaptive vsync works?


    Adaptive V-Sync just dynamically disables and enables V-Sync at certain FPS points to reduce performance impact at the cost of not not fixing tearing below a certain point, usually 60FPS. What you described is more likely to be a possibility for virtual V-Sync IMO, if anything.
  8. So you got all this data but not one example of what the numbers really mean.

    Lets say you can identify in your data a place in the render where this are not so smooth...what does that look like? Can you play a video example?

    I'd like to see a blind test if ppl can pick the video that has the issue. If they cant to a significant degree then you got a worthless new tool.
  9. Great review. Looks like Practical FPS is a better indicator of performance. I would love to see a Vsync analysis as well, since I play with it on in a lot of older games to eliminate screen tearing.
  10. loopsSo you got all this data but not one example of what the numbers really mean.Lets say you can identify in your data a place in the render where things are not so smooth...what does that look like? Can you play a video example?I'd like to see a blind test if ppl can pick the video that has the issue. If they cant to a significant degree then you got a worthless new tool.


    Here is some video on another site that is doing the same thing today: http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Tes-10
  11. I demand to see this test redone with the HD 7870LE's in xfire.
  12. As I read through the PCPER article, I'm tempted to Dremel-off the Crossfire connector(s) on my AMD cards, lest I am ever tempted to use them...
  13. What I find exceptional is, not that nVidia created this tool, or shows their previous works and success in this direction, but the fact it isnt for end users, and it gives them 1 more opportunity to hit the press.
    I think everyone should be able to see this, dont you?
    Especially at a time where the competition is starting to change the scenery, both in this direction as well as others.
    That being said, cudos to nVidia for their lead here, tho it may not last, they are getting their mileage out of it
  14. JAYDEEJOHNWhat I find exceptional is, not that nVidia created this tool, or shows their previous works and success in this direction, but the fact it isnt for end users, and it gives them 1 more opportunity to hit the press.I think everyone should be able to see this, dont you?Especially at a time where the competition is starting to change the scenery, both in this direction as well as others.That being said, cudos to nVidia for their lead here, tho it may not last, they are getting their mileage out of it

    Given that the primary new tool tests the end result, what is sent to the screen, it still is much better than any other system out there, and one that is hard to make favor one company over the other, unless it really exists.
  15. 7870 vs 660TI?

    Users not knowing with computing will now have an inaccurate result to look to, since you are comparing two $200-$220 cards to like $300 cards... people are going to look at that incorrectly.. its not the first time Toms has done this either.
  16. JJ1217 said:
    7870 vs 660TI?

    Users not knowing with computing will now have an inaccurate result to look to, since you are comparing two $200-$220 cards to like $300 cards... people are going to look at that incorrectly.. its not the first time Toms has done this either.


    According to Tom's tests, the 7870 has held its own against the 660 Ti cards in performance regardless of price, so i's not an issue for this comparison. People can look at it however they want, but Tom's has continually demonstrated this.
  17. bystander said:
    loopsSo you got all this data but not one example of what the numbers really mean.Lets say you can identify in your data a place in the render where things are not so smooth...what does that look like? Can you play a video example?I'd like to see a blind test if ppl can pick the video that has the issue. If they cant to a significant degree then you got a worthless new tool.


    Here is some video on another site that is doing the same thing today: http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Tes-10


    This is even more significant. Interesting... Multiple reputable sources with the same findings.
  18. ubercakeNo tearing on 120Hz monitors until you get over 120fps and even then tearing is no longer perceivable until you hit the mid 400s.Also, that is not the point of the article.This is a great article. It's consistent with others I've read on the subject. It is consistent to what is being published regarding information AMD is also supporting. I look forward to seeing what you do with the tweaks of the FCAT software to further define what equates to a "runt" frame. Seems like that could make an even greater difference. Defining a runt frame seems somewhat subjective. Seems like many more than 21 scan lines or less could define a runt and would seem dependent on the resolution somewhat?

    You can (and will) get a tear at any fps while vsync is not enabled. Once a frame is finished drawing and the buffer is flipped, the monitor scanner will pick up on the new frame. The only exception is VSync.
  19. Thats my point.
    nVidia has gotten their mileage out of this, and everyone ons the same page in the media.
    I dont question the tool, nor its value, I do question some reactions by the press tho, and was my intent in my previous comment, without having to paint it out as such.
  20. So, are AMD engineers crying foul, are they scratching their heads, or did they just fill their pants?
  21. Do they make special screens that refresh quicker? So that it goes top to bottom really fast?
  22. BigMack70Amazing review... really happy to see all the progress being done to get the end user more relevant data about GPU purchases, especially in the world of multi-GPU.My hope is that all of these tools will eventually lead to Nvidia and AMD conquering the microstutter boogeyman that has plagued multi-GPU setups for so long (yes I'm aware that AMD is currently a lot worse than Nvidia on this, no that doesn't mean Nvidia has solved the problem yet).


    Look at all of the misinformed / ignorant AMD fanboys thumbs downing this comment. What he says is fact... Quit being so ignorant. No, I am no an NVIDIA fanboy. Call me a knowledge fanboy if you want... This kind of crap pisses me off. 3 thumbs down at the time of this comment... :(
  23. Onus said:
    So, are AMD engineers crying foul, are they scratching their heads, or did they just fill their pants?


    They've known about this for a long time as was pointed out several months ago when Tom's asked AMD and Nvidia about this stuff. They've been working on it too. None of what you said is accurate.
  24. If they're working on it, then I'd say they're scratching their heads, trying to figure it out. It will be the marketing types that need to change their drawers, at least until the engineers fix it.
    The additional vsync analysis is hopeful though.
    And if anyone wondered, I'm nobody's fanboy, currently running cards from both companies, as I have done back and forth over the years.
  25. OnusSo, are AMD engineers crying foul, are they scratching their heads, or did they just fill their pants?

    Here is an article that was written this week, interviewing the AMD driver guy for single GPU configurations. They are aware of the problem, they weren't aware of it until it was brought to their attention, but are now trying to fix it. They have a target date for July, for fixing Crossfire issues.

    http://www.anandtech.com/show/6857/amd-stuttering-issue
  26. bystanderHere is some video on another site that is doing the same thing today: http://www.pcper.com/reviews/Graph [...] nce-Tes-10

    That PCPer review goes extensively into VSync testing for those who were looking for that.
  27. 17seconds said:
    bystanderHere is some video on another site that is doing the same thing today: http://www.pcper.com/reviews/Graph [...] nce-Tes-10

    That PCPer review goes extensively into VSync testing for those who were looking for that.

    I saw that
    Thanks.
  28. matto17secsThat PCPer review goes extensively into VSync testing for those who were looking for that.

    Well, it looks at one game, but I believe they intend to release more thorough results in the next week or two. They mentioned this was only part of their results. It was also limited to 60hz, but they will be doing 120hz soon, if I recall correctly.
  29. Nvidiots, rejoice!
  30. None of it really matters as long as one can play the latest games at the resolutiion you want. Who cares if one card is faster than another as long as it is able to play the game, and make use of all the latest visual gehaws. Duh.
    Now once you start considering price, and temperature, and watts of electricity used, one has a reason to pick one over another.
  31. roast_porkNone of it really matters as long as one can play the latest games at the resolutiion you want. Who cares if one card is faster than another as long as it is able to play the game, and make use of all the latest visual gehaws. Duh.Now once you start considering price, and temperature, and watts of electricity used, one has a reason to pick one over another.

    I thought the point was getting the best experience you can, and for most people, the cost as well.

    I personally would rather play the game with a better experience, than just play the game with a less enjoyable experience. This is double an issue for me, as these things cause me simulator sickness.
  32. Bystander, exactly if you get the same experience from both cards does it matter if one is a little faster in freame rates, than another, no; what matters is the quality and price, and who has the best drivers, to deliver the best experience, I personally think AMD has a better record as far as drivers go. I use both Nvidia and AMD in my machines. (not in the same machines)
  33. i always use radeon pro for long time and great gaming experience for me with radeon....
    a freeware named radeon pro will mitigate most of latency issues...Tom's has tested that on one of previous benches
  34. roast_porkBystander, exactly if you get the same experience from both cards does it matter if one is a little faster in freame rates, than another, no; what matters is the quality and price, and who has the best drivers, to deliver the best experience, I personally think AMD has a better record as far as drivers go. I use both Nvidia and AMD in my machines. (not in the same machines)


    The point of the article is to help us know what gives the best experience before we buy a card. They are showing new methods to compare, rather than the raw FPS that was used in the past, as there is more to your experience than a number.

    I've used a lot of cards from both sides. I tend to have to fuss around with AMD/ATI drivers a lot more, but I can get them to work, it just takes more effort.
  35. jessterman21 said:
    Great review. Looks like Practical FPS is a better indicator of performance. I would love to see a Vsync analysis as well, since I play with it on in a lot of older games to eliminate screen tearing.


    http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Testin

    That link test with vsync... vsync eliminates a number/most of the radeon xfire issues.

    frankly that article is excellent and answered a number of my lingering questions about this testing methodology.
  36. ingtar33 said:
    jessterman21 said:
    Great review. Looks like Practical FPS is a better indicator of performance. I would love to see a Vsync analysis as well, since I play with it on in a lot of older games to eliminate screen tearing.


    http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Testin

    That link test with vsync... vsync eliminates a number/most of the radeon xfire issues.

    frankly that article is excellent and answered a number of my lingering questions about this testing methodology.


    That was a long one, though much more thorough than Tom's. Silly to use Vsync unless you're running above 60fps 90% of the time. Then it's a must, IMO. Well, except for multi. But that's when you crank the game to low and turn off all antialiasing anyway, so a little frame tearing isn't going to bother anyone.
  37. This is an absurd comparison...not That absurd...but they should still be doing the 660ti vs the 7950...I got my 7950 last week on newegg for 269...that's lower than any of the 660ti's on there.

    The 7870 is a scratch over 200 bucks....
  38. masterofevil22This is an absurd comparison...not That absurd...but they should still be doing the 660ti vs the 7950...I got my 7950 last week on newegg for 269...that's lower than any of the 660ti's on there.The 7870 is a scratch over 200 bucks....

    The good news is that they are not seeing many problems with single GPU configurations anymore. The problems are in Crossfire, so don't even worry.
  39. Very interesting article, thanks.
  40. Wow, very interesting. This just reminds me of why I chose Nvidia for my main rig. Typically you pay for what you get. Nvidia also has a lot more money than AMD has to throw around fixing every little thing and making things perfect.

    I do own an AMD 7870LE Powercolor Myst Edition. So I'm not biased. The 660 Ti is in my main rig. I bought the cheaper AMD card for my secondary. It performs almost on par with my 660 Ti at the settings I play at, but it's hotter, consumes more power, it's louder, and it's not built as good as my EVGA 660Ti. So you pay for what you get. Maybe if they used a 7950 it would have "won" more of the benchmarks but I'm sure it would still have just as many dropped and runt frames.
  41. Before 22 days (at the Crysis 3 comments) I was asking if you will do the test you did today. You said that it was a work in progress. Thank you for keeping your promise :)
  42. masterofevil22This is an absurd comparison...not That absurd...but they should still be doing the 660ti vs the 7950...I got my 7950 last week on newegg for 269...that's lower than any of the 660ti's on there.The 7870 is a scratch over 200 bucks....

    BigMack70It's not an absurd comparison... Toms has always had the 7870 and 660ti as equivalent cards.Rather or not you agree is a different story, but Toms is just being consistent with their previous conclusions and results. I don't know why people are upset about this.http://www.tomshardware.com/review [...] 107-7.html

    Tom's use of 5 gaming benchmarks, one of which is F1, really makes their reviews unreliable. It's often the case that their results stand apart from other review sites that draw conclusions based upon a wider sample of benchmarks.

    Thankfully, these new testing methods may bring about some other performance metric besides FPS to gauge what we all want... the smoothest gameplay possible. Unfortunately, people tend to treat FPS like a basketball game; if one team wins by 1 or 2 points, then they won the contest. There's a lot more to it than that of course.
  43. I'm not really sure it matters. The focus of the article is how Crossfire and SLI are performing, and it does a good job at showing that, though pcper.com has a more in depth picture, but THG does plan to give us more soon. It sounds like they had a system setup problem that caused them to lose a lot of time and data.
  44. 660ti should be up against 7950 non boost editions what a bias article heaping praise on Nvidia while smashing AMD. You know full well Toms that fraps is now a poor reference for frame stuttering. And this is an attempt you to say yer we know its is however we are still right to bash AMD products in the first place. This article reminds me of Garry Kasparov and IBMs Deep Blue.
  45. Wow a 7950 for $269. That is an amazing deal. Plus the crysis 3 free (or did that end?) and the other free game.
    I got my 660 ti from newegg for $230 plus the $150 f2p coupon, but I'd much rather have a 7950 at that price
  46. 777iceman777 said:
    660ti should be up against 7950 non boost editions what a bias article heaping praise on Nvidia while smashing AMD. You know full well Toms that fraps is now a poor reference for frame stuttering. And this is an attempt you to say yer we know its is however we are still right to bash AMD products in the first place. This article reminds me of Garry Kasparov and IBMs Deep Blue.

    You should read some of the other versions of this article. Here are two using the 7970 and 680 to compare SLI and Crossfire:
    http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Tes-12
    http://techreport.com/review/24553/inside-the-second-with-nvidia-frame-capture-tools/11

    Here is another with an interview with an AMD driver representative:
    http://www.anandtech.com/show/6857/amd-stuttering-issue

    There is nothing biased about it and it isn't about comparing a particular card, but about comparing SLI vs Crossfire.
  47. There's a huge conceptual problem with this entire test. Everything smaller than the entire screen is a "runt" frame, and is the price you pay for not using vsync. The "practical" frame rate is always no larger than the screen refresh rate.

    Turn on vsync, and every frame is complete, which makes the rendering rate identical to the practical rate.
  48. Nope sorry fraps has been used for purposes for which it is not designed this has been shown by AMD and backed up by Nvidia, now the industry is in damage control. This was a knee jerk article. This goes far beyond just gpu comparisons as fraps has been used against cpus such as the FX 8350 as well a dual core and quad core system from both sides.
Ask a new question

Read More

Performance Graphics Cards Benchmark