It appears from testing that in some of the titles tested, CF 7970's gives the same fps as a single 7970 and in some cases worse fps. Fraps is showing higher fps but the reality of dropped frames and runts means it isn't giving those frames at all that fraps is showing. This is very bad and if AMD knew about it, very naughty.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Why framerates don't matter: Frame Rating Dissected
- Thread starter bee
- Start date
It appears from testing that in some of the titles tested, CF 7970's gives the same fps as a single 7970 and in some cases worse fps. Fraps is showing higher fps but the reality of dropped frames and runts means it isn't giving those frames at all that fraps is showing. This is very bad and if AMD knew about it, very naughty.
Well according to the article they said AMD mentions that they are simply doing what the game asks while nVidia has internal frame metering:
It also mentions the hopes of nVidia allowing an "off" switch for the frame metering which would be interesting to see the benefits (or downsides) of it being there.
It could be something that AMD could fix on the driver side of things though, and given how well they've been working on them lately it's not out of the question. I may prefer nVidia but I sure as hell don't want them to dominate and it's good to see the single card AMD come out ahead.
It almost seems like AMD wasn't even aware of this problem and just looked at FPS when developing crossfire. They really need to come up with a solution as this type of testing is going to become the norm for graphics cards reviews and it shows them in a terrible light.
Nvidia seem to have been aware of this and have taken steps to address the problems but there is still a long way to go. Hopefully, now that this is all public it will give them more incentive to improve it further.
Nvidia seem to have been aware of this and have taken steps to address the problems but there is still a long way to go. Hopefully, now that this is all public it will give them more incentive to improve it further.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
will be interesting to see the effects of turning off frame metering when benching 
I believe that AMD are already working on a "fix" for frametime issues, which sounds like it will basically be in effect this same frame metering type of solution (with corresponding drops in benchmark results with it turned on)
I believe that AMD are already working on a "fix" for frametime issues, which sounds like it will basically be in effect this same frame metering type of solution (with corresponding drops in benchmark results with it turned on)
Agreed. I also prefer Nvidia but need the competition and AMD have done well of late. I am sure AMD will sort it and they have mentioned that a future driver will allow smoother gameplay or raw performance (user decides in the CCC) but this came out roughly the same time as that article...Coincidence???
Probably not, looks like AMD knew ahead of time that this would see the light of day so makes sense for them to combat the problem as soon as possible. Wonder if this method of benchmarking will take off? I hope so
Associate
- Joined
- 13 Nov 2007
- Posts
- 2,427
Tech Report initiated this aaaaaages ago - since then, most hardware sites have jumped on the band-wagon. Here is another (day-old) excellent article from TR with new investigations based on specific hardware and software provided by Nvidia:
http://techreport.com/review/24553/inside-the-second-with-nvidia-frame-capture-tools
Loving them 680 SLIs
http://techreport.com/review/24553/inside-the-second-with-nvidia-frame-capture-tools
Loving them 680 SLIs
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
Tech Report initiated this aaaaaages ago - since then, most hardware sites have jumped on the band-wagon. Here is another (day-old) excellent article from TR with new investigations based on specific hardware and software provided by Nvidia:
http://techreport.com/review/24553/inside-the-second-with-nvidia-frame-capture-tools
Loving them 680 SLIs
I thought this too, but this new method is claiming to be more reliable, as the frametimes reported from a software only solution doesn't really "see" the end result on the screen
there are big discrepancies between the "old" method and this new one, which a cynic could say has been deliberately invented by nvidia as they know it casts a shadow over AMD... but the fact that AMD have coincided their driver update to include a similar frame metering concept for crossfire indicates that AMD were aware of this prior to it becoming public in the way that it has
interesting times ahead as everyone re-evaluates crossfire vs. SLI with some driver updates coming soon
the July update for amd will adress this with options for users with multi gpu.
they lag behind Nvidia with this due to not really look into the tech they produced and gaming and OS and how they interacted assuming it was outside their driver control.
they are upping their effort and 5 months from now will be good.
they lag behind Nvidia with this due to not really look into the tech they produced and gaming and OS and how they interacted assuming it was outside their driver control.
they are upping their effort and 5 months from now will be good.
AMD have been well aware of it for a long time, their original statement many years ago now was that they considered maximum performance of more value to their customers. There has been a huge attitude adjustment at AMD over the last few months as to how they approach software/hardware support and I think this has been more the driver for the change in functionality in this regard than these results coming to light.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
well, I'm not going to second guess what may or may not have been going on behind closed doors
as I say it will be interesting to see what happens in regards of performance and game benchmarks when both sides get their drivers updated to allow the user to select between frame metering or not when running multi card setups
and whether benchmarking suites / websites etc. require it to be on for results to be considered valid or not
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
as Gregster points out, read the tech report article for a better worded example of the same test method
Soldato
- Joined
- 14 Dec 2010
- Posts
- 3,030
- Location
- Nottingham
It was articles like this which made me pick up a 7970 Ghz ed to replace my 680 but the performance didn't feel to be as good. (Even synthetic benchies scored lower. Odd)
It isn't an issue so much in single card configuration but for CF, it is showing dropped frames all over the place. AMD will fix it and I have read rumours of July (not far off) for the choice drivers (Performance or smoothness).
It seems that the crossfire issues are similar to what was happening when I used VirtueMVP with my old I5. It said I was getting massive fps but play wasn't smotth at all.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
what's 4 months when the issue has been hiding there for over a year without recognition
sorry, I couldn't help myself
sorry, I couldn't help myself
Yes but this could have just been your problem, if you look in the benchmark threads on here 7970's are ahead of 680's. There's even 7950's albeit highly clocked beating some 680's.
As for not feeling as good, I truly think this is placebo as nobody would buy AMD cards if they weren't smooth enough to game on.