DiRT 3 – HD 7970 versus GTX 680
Dirt 3 is a game that is already running at very high frame rates at 1920×1080 so it comes as no surprise that the scaling in CrossFire and SLI is minimal. In terms of single GPU performance though the HD 7970 stays ahead of the GXT 680 consistently.
This is an interesting plot; while the green and black lines both are pretty thin and tight indicating a solid gaming experience with little stutter or frame variance, both the blue and orange lines of the multi-GPU configs have higher than expected variance. What might be more interesting is that this is the first time thus far that we haven’t found the HD 7970s in CrossFire to be completely abysmal.
Well look at that! Both the observed and FRAPS FPS graphs are nearly the same, telling us that while AMD has some definite CrossFire scaling issues with Battlefield 3 and Crysis 3, DiRT 3 doesn’t appear to have suffering from the same ailment.
The minimum frame rate percentile graph shows us that the HD 7970 is indeed faster than the GTX 680 across the board, though both cards are overpowered for this game at 19×10. CrossFire and SLI both stay in range of each other the entire time as well with SLI falling a bit behind after the 84th percentile.
Our ISU graph here shows an interesting result and different view of performance than the minimum FPS percentile graph above it. While the two cards were close in comparison before the HD 7970s in CrossFire actually look much better than SLI in this case, indicating a “tighter” band of frame times.
The results are starting out the same here at 2560×1440 though there is much more scaling occurring on both AMD and NVIDIA setups by adding a second card.
Hmm, that’s interesting. This was an occurrence we saw nearly every run of Dirt 3 with CrossFire at 25×14 so I wanted to be sure to include it here. At the times between 25 and 30 seconds (it was different times in each run) we saw runts pop up again, almost as if the CrossFire cards were “getting out of phase” or something like that. SLI has a few bigger spikes but nothing ever showed up that consistently.
As a tip to one of my concluding thoughts of this article, I think that this result, and others we’ll show before the end, is indicative of the major problem for CrossFire. It appears that when the GPU is the primary bottleneck, CrossFire just can’t keep up in some way and produces runts. That would explain why would see it at 25×14 here but NOT at 19×12. Just keep that in mind as you keep reading.
As a result of that series of runts, the observed FPS of HD 7970s in CrossFire drops during that time.
The minimum FPS percentiles graph shows the same result in a different way – it is obvious that the runts are causing the frame rates at the back 5% of this chart and the “build up” to and away from that subsection are what causes the decline around the 85% mark.
The frame variance and potential stutter graph here shows the results of that spike of runts as well with a much higher overall variance than SLI or either single card.
Dirt 3 was another game case where 5760×1080 results didn’t pan out for the HD 7970 in CrossFire so we are only seeing single GPUs and SLI results. SLI is scaling very well but the single GTX 680 is getting bested by the HD 7970 the entire time.
All of our results are pretty good here the GTX 680s in SLI produce a fairly tight frame time line and frame rates well above either single card.
Without any runts or drops our observed frame rates are identical to our FRAPS frame rates.
Tight plots of frame times indicate and nearly straight lines here match our real-world game play experiences of smooth frame rates.
While this might look bad for the cards in question – check the scale on the left! The most frame time variance we see on any option here is 1.6ms so we come away from our testing happy with 5760×1080 results for everything but the HD 7970s in CrossFire.
What would a game look like
What would a game look like if
it got a smooth 120+ fps,
was on a 60Hz display,
and in addition to the regular ‘vsync’ spot, it would ‘vsync’ at the ‘1/2’ way spot?
aka update the display at the top and middle, updating at these same two spots every time, and only updating at these two spots.
Would the middle ‘vsync’ spot be annoying? helpful? noticed? informative? etc…? (This sounds like a good way to see how important fps is)
What’s a good name for this?
1/2 x vsync, 2 x vsync, vsync 1/2 x, vsync 2 x, or something else?
What’s the logic behind your pick(s)?
(forward note: i have bad
(forward note: i have bad english.)
1) is there a diff in observer fps between cards with more ram?
i.e. sli of 2x gtx770 2g vs sli of 2x gtx770 4g?
2) can you publish a min/max/var of partial frame per frame?
insted of runt i wanna know how many different “color” are per frame, and if they are evenly spread.
Nice review. I’m interested
Nice review. I’m interested as to how this tech is evolving.
But now I’m curious- I’ve read some of your test methods- but I may have missed something. I’ve seen mostly games that are more single player/first-person. Is that part of your methodology? I’m thinking of more intensive object rendering titles like Rome Total War II that has to render myriads of objects and stress memory more. Have you considered something like that?