Overclocking And Power Consumption
Overclocking: CPU, GPU Or Both?
We might as well say it now: overclocking the CPU, which you should be able to get stable all the way up to 4.5GHz, doesn't really yield any performance gains if the processor is truly used as an APU (that is to say, you utilize its on-die graphics engine).
It's a lot more important that the integrated GPU can be overclocked from its stock 866MHz all the way to 1040MHz and more. The additional horsepower isn't just measurable; it can also be felt subjectively, especially backed by our fast DDR3-2400 RAM.
Power Consumption For Different Scenarios
First, we measure the power consumption of various tasks. It's easy to see that the APU can exceed 125W if the GPU sits idle. Of course, it took Prime95 to get the A10-7890K to 128W. But real-world applications that use all four threads (we did this with a photovoltaics simulation involving solar radiation and shading) can peak at 123W, with averages in the 117W range. AMD's official 95W TDP is clearly exceeded in a big way.
You might assume that the power measurements would be even worse in games, since the CPU and GPU are active. In reality, though, gaming has the opposite effect. Our results fall somewhere around 90 to 100W.
To understand these seemingly contradictory results, we need to take a look at how power consumption is controlled in AMD's APUs. Once the GPU starts drawing a lot of power, which can easily end up in the vicinity of 50W, the so-called Power Control cuts the CPU's consumption dramatically. It achieves this by significantly cutting into the host processor's clock rate.
Allow us to demonstrate by logging a sample session. The 4GHz base clock AMD promised us drops to 3GHz. Never mind the hope for some Turbo Core speed-up. And it doesn’t matter if the CPU is manually overclocked in your BIOS, either. Once the GPU kicks in, the CPU's frequency takes a back seat.
This single graph is enough to suggest you'll see barely any performance difference between AMD’s A10-7870K and A10-7890K in games. The latter’s higher clock rate vanishes as soon as the GPU starts its work.
Interestingly, overclocking the integrated GPU doesn’t affect the CPU’s frequency. This is why we put so much emphasis on overclocking the Radeon engine compared to the CPU.
Here are the power consumption results for AMD’s A10-7890K in a few different scenarios:
There is really only one significant difference between AMD’s A10-7870K and A10-7890K. The latter yields a higher stable GPU overclock. With only two test samples, it’s impossible to say if this is due to an improved manufacturing process or not, though.
And here are the specifications of our online gaming PC build:
Power Consumption | |
---|---|
Test Method | Contact-free DC measurement at PCIe slot (using a riser card) 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 |
Gaming Performance | |
Test System | AMD A10-7890K, AMD A10-7870K, AMD A10-7850KWraith CoolerAsus A88X-Pro2x 8GB Radeon Memory DDR3-24001x Crucial BX200, 240GB SSDKolink KL 400 80 PLUS BronzeAerocool GT-RS ATX Cube Windows 10 Pro (All Updates)Intel Comparison SystemCore i7-6700K, MSI Z170A Gaming M7Core i3-4160, MSI Z97A Gaming 6Intel Stock Cooler2x 8GB Corsair Dominator DDR3-2133 |
Drivers | AMD: Radeon Software 15.301 B35 (Press Beta Driver, February 2016)Intel: Beta 15.40.18.4380, 02/09/2016 |
This is what our gaming cube looks like completely assembled. We installed the Intel system’s components instead for the comparison results.