AMD Frame Pacing Explored: Catalyst 13.8 Brings Consistency to Crossfire
by Ryan Smith on August 1, 2013 2:00 PM ESTTotal War: Shogun 2
Our first detailed benchmark is Shogun 2, which is a continuing favorite to our benchmark suite. Total War: Shogun 2 is the latest installment of the long-running Total War series of turn based strategy games, and alongside Civilization V is notable for just how many units it can put on a screen at once. Even 2 years after its release it’s still a very punishing game at its highest settings due to the amount of shading and memory those units require.
For the sake of completeness we’re posting our frame rate charts for each of our individual games, but in general there’s nothing here we haven’t seen before in the 7990 review, in other reviews, or in Bench. The 7990 and GTX 690 still swap places fairly regularly.
Looking at our expanded delta percentages for Shogun, we can see how the 7990 and other Crossfire solutions stack up to the GTX 690 and other SLI solutions. For all AFR configurations the results match what we saw in our summary, with NVIDIA’s solutions offering lower deltas than AMD’s even with the new drivers.
This is actually AMD’s weakest game, with both the 7970GECF and 7990 exceeding 20% variability on this game. However it’s also the only non-action game in this collection, so it’s the game least affected by higher levels of variation and consequently the game AMD can afford to do the worst at. Nevertheless the improvement over Catalyst 13.6 without frame pacing is nothing short of amazing.
Meanwhile we’ll hit upon this a few times, but as a reminder AMD’s frame pacing improvements apply to older cards too, so the 6990 has its frame pacing problems resolved like the rest of AMD’s multi-GPU cards. It actually does better than the rest, we believe due to the fact that the lower framerate and higher frame times give AMD’s drivers more time to analyze and schedule frames.
Looking at the FCAT graphs, we can see that the higher variability of the 7990’s frame times is represented well. Though NVIDIA’s frame time spikes are more extreme than AMD’s.
Finally we have our 95th percentile frame times. Despite the fact that AMD’s framerates are down slightly versus Catalyst 13.6, their 95% percentile times are way up. Simply by instituting frame pacing they’ve dropped from 36.2ms to 21.5ms per frame.
Facebook
Twitter
RSS
102 Comments
View All Comments
boot318 - Thursday, August 1, 2013 - link
I've read a couple people got "black screened" when they did this update on one GPU. I'm not saying that will happen, but you better prepare for it if you do.Bob Todd - Thursday, August 1, 2013 - link
I may have missed this when I skimmed through the results, but have you heard anything about rough estimates from AMD about a frame pacing release supporting Eyefinity (e.g. Q4, H1 2014, etc.)? I know it's still a tiny percentage of users, but there are relatively cheap 1080p IPS panels now so building a nice looking 5760x1080 setup is pretty affordable these days. After playing games this way, it's something I wish I had done earlier, and I'm eager to see a frame pacing driver supporting this setup.Ryan Smith - Thursday, August 1, 2013 - link
Sorry, AMD didn't give us an ETA on that one. Let me see if I can still get one out of them.DanNeely - Thursday, August 1, 2013 - link
HardOCP says DX9 and Eyefinity support should be available in a driver update later this month.http://www.hardocp.com/article/2013/08/01/amd_cata...
DeviousOrange - Thursday, August 1, 2013 - link
I am hoping this will also improve Dual Graphics, will give it a test over the weekend.Homeles - Thursday, August 1, 2013 - link
Well I'll be damned. They did it. Not quite as good as Nvidia, but at this point, the difference isn't really one worth mentioning.xdrol - Thursday, August 1, 2013 - link
The link is bad for the driver, please remove "-auth" from the URL.chizow - Thursday, August 1, 2013 - link
Like watching a baby crawl. Good first step for AMD, but still a long way to go.AMD and their fans can thank the press (mainly TechReport and HardOCP, sorry Derek, you guys were way late to the party and still not fully onboard with FCAT measurements) and Nvidia fans for making such a big stink of this. Lord knows AMD and their fans were too busy looking the other way to address it, anyways.
Hopefully AMD and their fans take something away from this: if you want to improve your product, don't try to sweep it under the rug, address it, own it, and demand a fix for it.
chizow - Thursday, August 1, 2013 - link
Sorry my above post should reference the author Ryan, not Derek (was thinking of your predecessor), when referring to AT not being at the forefront of this runtframe/microstutter issue.Also, I feel the accolades given to TechReport, while not completely undeserving, should also be given to PCPer's Ryan Shrout and some of the German publications like PCGamesHardware. While TechReport did start the ball rolling with some new ways to measure frame latency/microstutter, Ryan Shrout really harped on the runtframe issue until Nvidia worked with him in unveiling FCAT. Also, the German sites have been hammering AMD for years about their much worst microstuttering in CF, largely ignored by the NA press/blogs. And finally Kyle at HardOCP has said for years SLI felt smoother than CF with some Pepsi challenge type user testing, but not so much hard evidence as presented here as well as other sites.
Finally Ryan, are these new metrics you've done an excellent job of formulating going to make it into future benchmarks? Or are you going to just assume the issue has been fixed going forward? I would love to take AMD's word on it but as we've seen from both vendors in the past, driver regression is commonplace unless constantly revisited by users, reviewers, and the vendors alike.
Ryan Smith - Friday, August 2, 2013 - link
"Finally Ryan, are these new metrics you've done an excellent job of formulating going to make it into future benchmarks?"They'll be in future articles in a limited form, similar to how we handled the GTX 780 launch. It takes a lot of additional work to put this data together, which isn't always time we have available. Especially if it becomes doing hours of extra work to collect data just to say "yep, still no stuttering."