For years, gamers using multi-GPU setups have noticed that despite higher frame rates, games often didn’t seem to perform to expectations. Even with a confirmed high frame rate, animations would be jittery and laggy. Nvidia addressed this in their last iteration of GeForce, but AMD haven’t, until recently. So who do we have to thank for AMD’s frame pacing driver, which by most accounts seem to be effective in remedying the issue?
It wasn’t until Tech Report started testing for frame time delivery that the laggy frame effect was quantified. Using FRAPS and showing the actual delivery time of each individual frame, they were able to determine that despite high “FPS” some games would still appear to not play very smoothly at all, based on what the user sees on his monitor.
We switched to that method of testing last year, but the revelation still fell upon deaf ears at AMD and their fans. FRAPS based testing had a flaw that allowed AMD and their fans to dismiss results in that they were recorded very early in the rendering pipeline. Although the FRAPS method could be proven with video evidence, it wasn’t something that could be counted on for reliable results in every single game.
Everything changed though, when Nvidia, with the help of PC Perspective, began working on a testing method dubbed “FCAT” that could record frame time delivery at the monitor level – the actual frames we see. Once ready for release, Nvidia distributed FCAT testing kits to three more of the busiest hardware sites – Anandtech, Tom’s Hardware, and Tech Report. Along with PC Perspective, the indisputable truth was finally proven – Nvidia’s hardware based frame time delivery method worked in most cases, while AMD has major issues when using a multi GPU setup. AMD finally took notice, and began working on a driver that would resolve issues on a game-to-game basis.
Yesterday, AMD released the first public beta of a driver that resolves the issue on a limited range of games and resolutions so far. For now, the driver only works on DX11 and DX10 games, and in resolutions 2560×1600 and below. This is sure to be refined later, but it’s good to see that whatever solution they are coming up with is working.
Below are the results from Tech Report’s testing of the beta driver:
As you can see, in Crysis 3 at least, the new driver makes a huge difference. While the average FPS went from 30 to 29, the actual frame delivery time was significantly improved.
I also appreciate that Tech Report continue to use FRAPS to test for frame latency. Because only four sites have access to FCAT testing hardware, a lot of us have to rely on FRAPS to test for this. In most games, it appears to be a viable tool for this type of testing, despite being very early in the rendering pipeline. We will keep an eye on how games are affected when reviewing video cards in the future.
While it’s unfortunate that it took so long for AMD to address the issue (the first article showing it was published in 2011), it’s good that they finally did once testing methods improved and distributed to influential media. Although I don’t often recommend multi GPU setups for other issues, it’s good that it is finally viable to consider a multi GPU Radeon setup if you think it’s the right choice for you performance-wise.