Civilization Beyond Earth Benchmark – Mantle vs DirectX Nikolas Nikolaou October 31, 2014 Benchmarks Turn-based strategy games aren’t usually a good fit for PC performance benchmarks but Sid Meier’s Civilization: Beyond Earth has a slight twist to it. The developers over at Firaxis have added Mantle support for both single and Crossfire AMD graphic cards. They mention using Split Frame Rendering (SFR) for Mantle in CrossfireX instead of the usual Alternate Frame Rendering (AFR). Due to Mantle’s flexibility developers have the freedom to use the technology and customise it to their game engine’s specific needs. This is something that Firaxis has put extra emphasis on , to maintain a smooth game experience out of the box for multi GPU setups, something scarce in this day and age for the PC Master Race. Civilization: Beyond Earth is a turn-based strategy game where some of the most talented and lucky people from Earth are sent to a new world. You build a small city, explore and fight against alien enemies until you grow your empire. I’m not an RTS fan nor have I played any of the older Sid Meier games (even though it’s a very big franchise) so I won’t continue on what the game is about as it’s not the focus of the benchmark. We will be looking at the performance of DirectX vs Mantle in HD and 4K on a single AMD R9-290 and two R9-290s in Crossfire. Testing Components and Methodology CIV Beyond Earth Benchmark System CPU i7 3930K 4.3Ghz Motherboard Asus P9X79 Deluxe RAM Kingston Hyper-X 16GB 1600mhz HDD/SSD Crucial M550 256GB GPU 2XMSI R9-290 Gaming 4S Monitor Asus PB287Q 28″ 4K 60Hz 1MS PSU Corsair AX1200W OS Windows 8.1 Drivers Catalyst 14.9.2 Beta Benchmarking CIV Beyond Earth isn’t done by using Fraps or FCAT, like we usually do. The nice thing is that you can use specific commands to run an internal benchmark that saves a CSV file into your CIV game folder. By right-clicking properties in Steam and running -benchmark “pick save file name here” CIV Beyond Earth runs a prebuild game map filled with units. It lasts 50 seconds with the mouse hovering over the map and demonstrates, in my opinion, a very good simulation of what would be going on in-game while playing. The files can be interpreted through Excel with some custom formulas. It’s a bit of a hassle if you’re not an Excel pro so a few members over at the Overclockers.co.uk forum created an easy tool that converts the data to meaningful numbers and charts that you can find here. (Thanks Havana_UK, GoogalyMoogaly and Gregster). After running the benchmark open the saved file with Notepad, select all the data and insert it into the text holder on the CIV Beyond Earth Benchmark Interpreter and press the Calculate FPS button . You should get a minimum, average and maximum FPS count as well as a graph similar to the one below: CIV Beyond Earth 1920×1080 Benchmark The CrossfireX setup on DirectX doesn’t scale well at Ultra with MSAA at 8X in HD, the purpose of setting everything to the maximum was to avoid any bottlenecks and stress the GPUs as hard as possible. We see big variance between maximum and minimum frame rate on our Crossfire setup, our assumption is that Firaxis left AFR (Alternate Frame Rendering) as the default setting for DX. See anything unusual between the R9-290 and the CFX R9-290s using Mantle ? As mentioned above Firaxis has optimized SFR for CFX which has tighter frame rates with less fluctuation. The difference between Mantle and DX are visible as the wild fluctuations in DirectX causes the game to have a slowdown feel while in Mantle its frame rate range is a lot closer giving a better feel and overall smoothness. We see scaling for the R9-290s in DX getting better (60%+) while we move up to 3840×2160. The horsepower needed to run everything maxed out on 4K is not an easy task even for an RTS. We see big differences again between DirectX and Mantle, big fluctuations on DX and smoother gameplay on Mantle. The single R9-290 performs better than the multi gpu setup in Mantle. Conclusion The capabilities for developers using Mantle seem limitless, but theres more to it for the actual implementation. There is no significant bump in framerate with Mantle (in a margin of error) on both a single and multi GPU setup. The mention of a smoother experience with Mantle is solely due to lower frametimes. Those that have more than 2 GPUs are going to have to disable them to run Mantle. Turning off your second GPU and using VSync is a better option rather than dealing with the hiccups of any multi GPU setup in Civilisation Beyond Earth. We haven’t added our Nvidia GTX970 single and SLI benchmarks but something worth a mention is that Nvidia’s frametimes are lower (smoother gameplay) and with higher frame rates without the need for any custom optimizations. I like the effort put in from Firaxis on the Mantle API but this is something AMD should have dealt with a long time ago on DirectX. Not all developers will put the time and effort into ironing out problems that have been there for ages. See Pricing If you have any questions about the benchmark please ask in the comment section below and I will be glad to answer.