To tackle this mammoth swath of testing we are using two main test systems, one for all the AMD APUs, thanks to the backwards compatibility, and one for the two Intel Haswell parts. We are using AMD’s FM2+ socket and A88X chipset with Gigabyte’s fantastic G1 Sniper A88X motherboard which we reviewed here. For the Haswell parts on Intel’s LGA 1150 socket we are using the Z87 chipset with MSI’s Z87I-Gaming motherboard, which we reviewed here.
Main Storage Drive – Kingston HyperX 240GB SSD over SATA III interface
Chassis – Lian Li T60 Test Bench
Displays – Dell U2711 Ultra Sharp
Operating System – Windows 7 Ultimate 64 Bit
We would like to thank Gigabyte, MSI, Intel, AMD, Corsair, Kingston, Lian Li, Noctua and all our other partners who supplied us with test system equipment and hardware. Their generosity makes our testing possible and without them we wouldn’t be able to produce the reviews we do, so thank you!
Test Procedures
To conduct this review we have a huge battery of tests, benchmarks and games to battle through. Below we have listed each benchmark and any associated details about how we run the test, how long for (if applicable) and any settings that may be helpful to know.
CPU Tests
Cinebench R11.5 – multi threaded, single threaded and Open GL tests.
Cinebench R15 – multi threaded, single threaded and Open GL tests.
WPrime – a 32 million Pi calculation.
SuperPi 1.9 WP – a 4 million Pi calculation.
AIDA 64 Engineer – CPU and FPU tests: Queen, Zlib, AES, Hash, VP8, Julia, Mandel and SinJulia.
System Benchmarks
PCMark 7 – the PCMark Suite preset of tests.
PCMark 8 – the Home preset with Open CL acceleration enabled.
GPU and CPU Open CL Benchmarks
Basemark Open CL – Basic hardware optimisation running the full suite of tests on the GPU parts only.
Compubench CL 1.1.3 – running only the Physics: SPH Fluid Simulation, Graphics: Raytrace and Vision: Optical Flow tests on both the CPU and GPU.
Luxmark – Open CL computation running on the CPU, GPU and GPU/CPU combined hardware.
Memory Benchmarks
AIDA64 – Memory read, write, copy and latency tests.
Productivity
x264 HD Benchmark 5.0 – one run and two passes using the default settings.
Handbrake – conversion of a short 90 second 1080p video file from MP4 to MKV format.
WinRAR – using the built in benchmark.
JPEG Decoder – AMD’s provided JPEG decoder benchmark that is run using the standard Windows JPEG decoder and AMD’s new HSA decoder.
Photoshop CC – Smart Sharpen test of a sample image to test Open CL acceleration and image processing speed. Timed using digital stopwatch (old school!).
Libre Office – spreadsheet computation and calculation performance.
MuseMage – built in performance benchmark.
Gaming
3DMark11: Performance and Extreme presets.
3DMark (2013): Firestrike and Firestrike Extreme presets.
Battlefield 4 – low preset at 720p (1280 x 720) and 1080p (1920 x 1080). Measured using fraps and a 2 minute sequence of the game in the middle of the first campaign mission. Replayed with as much duplication as possible to ensure accurate results.
Tomb Raider – normal preset at 720p and 1080p, v-sync off, motion blur and screen effects on using the built in benchmark.
Sleeping Dogs – normal preset at 720p and 1080p using built in benchmark.
Bioshock Infinite – medium preset at 720p and 1080p using the built in benchmark.
Batman Arkham Origins – normal geometry, everything else off. 720p and 1080p using the built in benchmark.
Metro Last Light – Medium preset, AF4X and low motion blur at 720p and 1080p using the built in benchmark.
Power and Thermals
Power consumption – total system power draw measured with the system in an idle desktop state, with a CPU load state, GPU load state and CPU+GPU combined load state. Load states were all provided by the AIDA 64 Engineer system stability test.
Thermals – temperatures as recorded by CPUID-HW Monitor (for Richland and Haswell) and AMD OverDrive/AIDA64 (Kaveri). Note that the Kaveri temperature sensors functioned differently so results are not strictly comparable. The load and idle temperatures are for the CPU parts ONLY since the GPU parts did not have accurate temperature sensors on either the AMD or Intel parts. The temperatures were measured after a 5 minute CPU AIDA 64 system stability load test.