Multithreaded Paraboloid Mapping - Practical Rendering and Computation with Direct3D 11 - page 594

133.3 Empirical Results

With this concept in mind, we will turn our attention to some empirical tests taken on two
different development computers. One of the configurations uses a dual-core processor and
the other uses a quad core processor, so this should provide some indication of the relative
performance difference when the same application is executed over more processor cores.
The relevant specifications of the two test machines are listed below.

System
Memory

Processor „ Video Card Operating System

Processor Video Card Operating System

Memory

._ „ AMD Phenom II X4 „ „„ „„„ „„„ Windows 7

AMD Phenom II X4
@ 3.2 GHz

Windows 7
Ultimate 64 bit

PC 1

4 G B AM D 578 0 4 GB AMD 5700 PC1

9 3.2 GHz

Ultimate 64 bit

Intel Core 2 Duo
§ 2.4 GHz

Windows 7
Ultimate 64 bit

PC 2 4 GB

AMD 5830

§ 2.4 GHz Ultimate 64 bit

The tests were performed on each configuration with three reflective objects in the
scene, and a varying number of non-reflective objects. By changing the number of non-
reflective objects, we can control the number of draw submission API calls (which includes
the pipeline configuration, as well as draw calls) that must be performed in each frame.
This will let us characterize how multiple threads operating on multiple cores process
these API calls in parallel. The results of these tests are shown in Figure 13.15. The graph

Figure 13.15. A graph showing the time required to generate a frame under varying configurations and
scene conditions.