DX12 vs DX11 GPU & CPU Power Consumption
Graphics Card Power Consumption
We chose the four fastest cards that aren't too GPU-limited in Ultra HD, and measured their power consumption for all three render paths.
While the Radeon R9 Fury X adheres to its power limit, the Radeon R9 390X under DirectX 12 blows through the roof compared to DirectX 11. Its 100W-higher power consumption with asynchronous compute activated may be okay when you consider the 20 percent performance increase, but it's just marginally less with this option deactivated, despite significantly lower framerates.
Both Nvidia cards demonstrate increased power consumption with just a moderate performance boost. The GeForce GTX 980 Ti buys itself a mere three frames per second at the cost of 45W+. The GeForce GTX 980, on the other hand, stays rather inconspicuous.
Power Consumption Of The CPU
Of course, we also measured the CPU's behavior. These values were measured over the full benchmark run and contain no VRM losses. Regardless of the graphics card, CPU power consumption drops, particularly when we test with one of Nvidia's boards.
A Look At Efficiency
To draw a more informed conclusion, we add the CPU's and GPU's power consumption. After all, a real efficiency analysis is only possible by looking at the big picture of the individual interactions.
The Radeon R9 390X is really fast under DirectX 12, but consumes a lot of power and lags far behind, resulting in its last-place finish, efficiency-wise.
A Look At The Details
Now we're looking at the corresponding CPU and GPU power consumption values based on their framerate curves over time:
You can see that the Radeon R9 Fury is throttling partially, while the Radeon R9 390X continues to churn along with red-hot glowing wheels.
The CPU power consumption over time, on the other hand, is less spectacular.
Before we draw our final conclusions, we’d like to mention that running the Ashes of the Singularity benchmark wasn’t exactly an easy endeavor. In fact, there was a lot of troubleshooting involved, which cost us a lot of time. The multi-GPU part was particularly problematic, and it seems that this wasn’t just the case for us. Many of our colleagues experienced the same problems. One major drawback of Ashes of the Singularity as it stands is the game’s inability to run in true full-screen mode. Instead, it uses a borderless window as its only option.
This makes the Ashes of the Singularity benchmark vulnerable to any number of external influences, especially at Ultra HD. Running it at that resolution results in bandwidth limitations unless a critical Microsoft Windows 10 update is installed. We’re using the same fresh image for all of our benchmarks, and that cost us (and anyone else in the same boat) hours of troubleshooting.
This is a pretty benchmark that serves up interesting results and compels us to wonder what's coming to PC gaming in the near future? One thing we can say is that AMD wins this round. Its R&D team, which implemented functionality that nobody really paid attention to until now, should be commended.
With asynchronous shading/compute, AMD has a clear winner on its hands. At least for now. But where are the games that take advantage of the technology? It will be a long time until developers use DirectX 12 to its full advantage. By then, GPU architectures like Fiji or Maxwell might already be history, instead of the flagship designs they are now.
Also, nobody knows what Nvidia’s Pascal architecture will do. At least we can be sure that AMD will include these capabilities in Polaris. The jury’s still out on Nvidia’s capability to compensate for hardware disadvantages with software workarounds. The company did show that a lot can be done in this regard for the older beta version of Ashes.