Gaming Benchmarks on Kaveri
Many enthusiast gamers won’t see a visual benefit from system memory. The discrete GPU (and associated GDDR5 memory) will be doing most of the graphical heavy lifting. That equation changes when using an integrated GPU though – without the dedicated graphics memory, on-board graphics cores share the DDR3 RAM with the rest of the system; quickly shifting the graphics bottleneck to memory bandwidth. Here, memory speed and latency can make a very real difference.
I tested the HyperX Fury RAM kit at the stock 1866 MHz and overclocked 2133 MHz speeds, then threw in a few comparison modules for reference. While it isn’t a comprehensive list of benchmarks, hopefully it will help illustrate how memory speeds can make a difference if you’re gaming using the integrated Radeon graphics on an APU.
Motherboard: Gigabyte G2-F2A88XN-Wifi
System Memory: (See graphs for memory tested)
Processor: A10-7850K 3.7 GHz / 4.0 GHz Turbo
Video: Integrated Radeon R7 @ 960MHz
Disk Drive 1: Intel 335 Series 240GB SSD
Operating System: Windows 8.1 Pro 64-bit
Running on a modified Unreal Engine 3, Irrational’s Bioshock Infinite won its fair share of awards in 2013. Utilizing DX11 specific features like Contact Hardening Shadows, Diffusion Depth of Field and HDAO Ambient Occlusion, Bioshock Infinite’s floating city of Columbia was visually breathtaking yet ran surprisingly well on a range of hardware.
Repeating much of the same story from the synthetic benchmarks (this time with more relevant values), the HyperX Fury kit still shows an improvement over a typical DDR3 1600 CL9 set. With a very easy overclock, a decent 16% increase in frame rates was very attainable over the common DDR3 1600 CL9 kits. Overall, the increase in speed from the higher-clocked HyperX Fury kit made exploring Columbia an even better experience if you were using an APU.
Tomb Raider (2013)
Square Enix’s Tomb Raider reboot is built with some AMD-specific technologies – that being said, the A10-7850K still isn’t quite powerful enough to run special features like TressFX in full 1080p resolutions. Still, on medium presets and lower resolutions Tomb Raider is very playable on modern integrated graphics chipsets.
Using the Normal preset for the built-in benchmark and running at a resolution of 1280×720, the HyperX Fury RAM hovered around the 60fps mark. Overclocking gained a bit of headroom and smoothed out the minimum FPS drops, but the overall trend continued; even down into the more memory-constrained High preset at a resolution of 1920×1080.
Unigine Valley / 3dMark11
The UNIGINE Engine powers a popular series of GPU benchmarks. Starting with the Sanctuary and Tropics benchmarks, Unigine Corp. is probably most known for its popular Heaven benchmarking utility. The new version, “Valley,” contains procedurally generated rocks and vegetation and advanced visual technologies like dynamic sky, volumetric clouds, sun shafts, depth-of-field and ambient occlusion. I used the most recent Valley benchmark for the Kaveri integrated graphics testing, and ran the “Basic” preset.
Like Unigine Heaven/Valley, most enthusiasts are also familiar with Futuremark’s 3DMark 11. A DirectX 11 powered video card benchmark suite, it runs a series of graphics and physics tests to arrive at a composite score. For the Kaveri APU, I’ll be using the Performance preset.
The scores are pictured above, and follow similar trends set by the game engines. Frames-per-second during the Valley benchmark averaged around 23fps, with all memory kits dipping into the low teens and showing bursts into the mid 40s range.