Test Setup & Overclocking
Test System and Processors
| Core i7-6950X | Core i7-6900K | Core i7-6850K | Core i7-6800K |
Motherboard & Memory
We’re using MSI's X99A Gaming Pro Carbon as the basis for our benchmarks and overclocking attempts. As mentioned previously, this is a refreshed X99-based platform that modernizes Intel's aging chipset a bit with USB 3.1 Gen 2 support, Type-C port connectivity and a U.2 connector for high-end storage.
If you're new to building PCs, MSI's Game Boost is an interesting one-button overclocking feature. Of course, experienced power users are going to tweak their settings manually, but it's still nice to have options.
One of the most important components of a stable overclock is the right memory. We’re using four 8GB modules of G.Skill Trident Z DDR4-3200 for our X99, X79 and Z170 platforms. This should assure our Broadwell-E overclocks aren't limited by memory.
The only difference between our performance benchmark system and the one for overclocking/power consumption testing is cooling. We don't want to risk hitting a thermal ceiling, so the platform we're tuning employs an open-loop water cooling setup that can get rid of as much as 500W of waste heat. These CPUs shouldn't come close to that number, but you never know...
Benchmarks
| Ashes of the Singularity | 2560x1440, Point Lights: High, Glare, High, Terrain Detail Object Level: High, Shading Samples: 16, Terrain Detail Shading Samples: 32, Shadow Quality: High |
|---|---|
| Bioshock Infinite | 2560x1440, Ultra quality preset, Light Shafts: On, Ambient Occlusion: Ultra, Object Level of Detail: Ultra |
| F1 2015 | 2560x1440, 4x SSAO, 16x anisotropic filtering, Post-processing options enabled, Shadows enabled, Smoke shadows enabled, Vehicle reflections enabled, Weather effects enabled, Ground cover enabled. |
| 3DMark | Fire Strike version 1.1 |
| 7-Zip | Version 16.02, LZMA2 method, Ultra compression, 4.52GB workload, mixed file types |
| Blender | Version 2.75a, Cycles engine, BMW benchmark |
| Cinebench | Version R15, Single- and multi-threaded CPU benchmark |
| HandBrake | Version 0.10.2, x264, Constant quality: 20, variable bit-rate |
| LAME | Version 3.99, .wav to .mp3 conversion, 160 Kb/s fixed bitrate |
| PCMark 8 | Version 2.7.613, Hardware acceleration enabled, Adobe Creative Cloud and Microsoft Office suites |
Overclocking & Stability
Our overclocking methodology includes increasing each CPU's clock rate until we discover the maximum for a fully utilized processor and even faster Turbo Boost settings for each core. We also adjust voltages as-necessary to achieve a balance between stability and waste heat, since this directly affects cooling.
With a configuration that boots successfully, we run different load scenarios to verify stability. After completing Aida64’s integrated stress test (CPU, FPU, cache), the system is faced with workloads created by a Creo 3.0 SPECapc run, 4K video compression using the Adobe CC Media Encoder, a Witcher 3 sequence, Prime95 and AVX.
A clock rate configuration is only considered successful if those workloads finish without any errors and within reasonable voltages. Before we get to the results, here’s a table with our test system’s technical specifications:
| Power Consumption Test Setup | |
|---|---|
| Test Method | Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply Real-Time Infrared Monitoring and Recording |
| Test Equipment | 2 x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function 4 x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) 4 x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) 1 x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function 1 x Optris PI640 80Hz Infrared Camera + PI Connect |
| Water Cooling | Pump: Alphacool VPP655 (Undervolted, 50 percent)CPU Cooler: Lian LiReservoir: Phobya Radiator: Alphacool, 24cm (6cm thickness)Fan: 2x 12cm Noiseblocker eLoop, PWM controlled (up to 1500 RPM) |
MORE: Best CPUs
MORE: All CPU Content
MORE: Intel & AMD Processor Hierarchy
