Big performance gains are possible using DirectX 12
Intel is turning heads at SIGGRAPH 2014 by showcasing its upcoming DirectX 12 API. This year's SIGGRAPH in Vancouver is the 41st intentional conference on computer graphics, and one of the highlights is a demonstration comparing DX11 versus DX12, the latter of which underscores the relationship between performance and power. In the demo, simply switching APIs results in up to a 70 percent boost in graphics performance.
The chip maker is using a Surface Pro 3 with an Intel Core i5 processor to run the demo. Intel's demo tasks the CPU and integrated graphics with rendering a scene filled with 50,000 fully dynamic and unique asteroids in two different modes -- maximum performance and maximum power saving. To compare DX11 and DX12, the app allows the user to switch between the two modes with the tap of a button.
In performance mode, the application runs as fast as possible within the thermal and power constraints of the platform. Using DX11, it pushes 19 frames per second. Switching over to DX12 sees a massive 70 percent jump in performance to 33fps. How can that be?
Intel explains the performance jump by pointing to a power graph showing the distribution of power between the CPU and GPU.
"DirectX 12 is designed for low overhead, multi-threaded rendering. Using the new API we have reduced the CPU power requirement and thus freed up that power for the GPU," Intel explains.
In this demo, DX12 reduces the CPU workload by about half when compared to DX11. It does this by spreading the work across more CPU cores at a lower frequency rather than running a single thread at a high frequency. In addition to benefiting the user with a cooler device and longer battery life, it allows more work to be offloaded to the GPU, which improves graphics performance.
According to Intel, an increase in CPU power efficiency can be used for more performance even if an application is not CPU bound.