Frame Rating: GTX 660 vs HD 7870, plus HD 7790, HD 7850, GTX 650 Ti BOOST
Skyrim - GTX 660 vs HD 7870
The Elder Scrolls V: Skyrim (DirectX 9)
The Empire of Tamriel is on the edge. The High King of Skyrim has been murdered.
Alliances form as claims to the throne are made. In the midst of this conflict, a far more dangerous, ancient evil is awakened. Dragons, long lost to the passages of the Elder Scrolls, have returned to Tamriel.
The future of Skyrim, even the Empire itself, hangs in the balance as they wait for the prophesized Dragonborn to come; a hero born with the power of The Voice, and the only one who can stand amongst the dragons.
Our settings for Skyrim
Here is a video our testing run through, for your reference
Unlike the first three games we looked at here, Skyrim at first blush appears to perform as you would expect: both SLI and CrossFire are scaling in our observed, Frame Rating based results.
The Skyrim engine definitely has more stutter and frame variance issues but nothing overly dramatic. The green and black lines that represent the single cards are fairly thin and consistent but both CrossFire and SLI have more variance. All four cards have some pretty bad spikes and "hitches" - as any Skyrim player will report is the case.
Skyrim definitely has an interesting pattern to our FPS percentile graphs with much more gradual declining metrics. Keep an eye on the CrossFire and SLI results and how they do come close to meeting the performance of the single cards towards the end; in fact CrossFire does drop below the HD 7870 single card around the 90th percentile.
Despite those issues, none of the four tested combination really add a lot of variance or potential frame stutter. Our scale on the left hand side doesn't breach the 5 ms mark.
Welp, at 2560x1440, the HD 7870s in CrossFire start to see some issues where the FRAPS FPS and Observed FPS are not matching up! The HD 7870 single card is faster than the GTX 660 but let's see what happened to CrossFire.
There are some obvious issues where CrossFire again falls into the trap of the alternating high and low frame times that we saw in BF3, Crysis 3, etc. This lends more credit to the idea that we theorized in the first article - the more GPU bottlenecked the game session is the more likely we are to see CrossFire problems exhibited.
Based on the minimum FPS data you can see that the HD 7870s in CrossFire quickly become only marginally faster than the single card while SLI stays above the single GTX 660 throughout.
Frame time variance of the CrossFire configuration looks bad once again with much higher variance than the other configurations, but even at the worst case we aren't going much above 10 ms.
At 5760x1080, even without HD 7870 CrossFire results, we some interesting issues with SLI and scaling. While the extra GTX 660 adds performance in some areas, the results are consistent.
Interestingly, even though there are a lot of areas where performance looks to scale well and with tight frame times, there are a lot of spikes and hitches in the game that kind of make SLI a poorer solution that we would expect. That being said, the CrossFire results dropped so many frames from our capture testing that we couldn't even get a result!
The Radeon HD 7870 does perform better than the GTX 660 card from NVIDIA (41 FPS vs 33 FPS) but we do see solid scaling on the SLI configuration (53 FPS vs 33 FPS). The hitching issues seen in the frame time plot show up here as the sudden drop in the minimum FPS right at the 95th percentile.
You can see that not only does the GTX 660 SLI setup have the most frame time variance, it also goes up steeply after the 90th percentile and to some pretty high rates - reaching 40+ ms.