It is expected before the experiment that different hardware and software will give different performance results. The hardware specifications are totally different, ranging from different processors to different amount of RAM. For the hardware comparison, four hardware systems were used for the experiment. A part of Chang’An 3D model which is already available in EONTM StudioTM’s EOZ format as shown in Figure 4.31 was used to run on all the systems available.
Figure 4.31. Part of Chang’An Model
It is very clear that the frame rate count is different for all four different hardware specifications as shown Table 4.12. The DELLTM PrecisionTM 650 was tested in both stereoscopic mode and mono mode. Therefore, this proves that different hardware specifications will create different performance of the same 3D model. Hence, the final experiment will be conducted on all hardware systems to see the overall impact they have on the final results.
TABLE 4.12. Frame Rate Performance of Different Hardware Specifications
Hardware Systems
Navigation (Hertz) DELLTM PrecisionTM 650 workstation
(stereo) 10
DELLTM PrecisionTM 650 workstation 5 DELLTM PrecisionTM 670 workstation 45 DELLTM PrecisionTM 380 desktop 15 DELLTM InspironTM 9300 laptop 21.3
For software comparison, EONTM StudioTM ProfessionalTM was compared with
Quest3D®. The same 3D model was constructed in 3ds Max® 8 to be exported to both EONTM StudioTM ProfessionalTM as an EOZ format and Quest3D® as a CGR format. The experiment was countered checked with an external software called Beepa FRAPS Version 2.8.2 Build 6488. It will run behind both software when the simulation starts to give it an external calculation of frame rate as shown in Figure 4.32 and 4.33. In the
simulation, a yellow figure will appear at the bottom left corner of the screen. The self calculation by the software is located at the top left corner of the screen.
Figure 4.32. EONTM StudioTM ProfessionalTM running at 59 - 60 Hz
Figure 4.33. Quest3D® running at 59 - 60Hz
It is proven therefore that the same 3D model produces the same frame rate regardless of
same trend is observed even using different hardware configurations. It is assumed that no matter which software is chosen, at least with the basic four variables used, they will generate the same results. The same cannot be said if more variables were included, especially the secondary ones. To further support the case above, an identical experiment was carried out with the 3d models used for Chapter 4.1.5 (Texture Count). The frame rate count is quite consistent for all the eleven 3d models used. The largest model with the most textures (Tri 2000K with 100 textures) was not used because it crashed for the test in EONTM StudioTM ProfessionalTM. The difference in frame rate is at most 0.2fps difference as shown in Table 4.13.
TABLE 4.13. Frame Rate Performance of Different Software Models \ Details Frame Rate (EONTM
StudioTM ProfessionalTM)
Frame Rate (Quest3D®)
Tri 100k with 10 textures 21.2 21.2
Tri 100k with 50 textures 21.4 21.4
Tri 100k with 100 textures 20.3 20.1
Tri 500k with 10 textures 2.5 2.6
Tri 500k with 50 textures 3.5 3.5
Tri 500k with 100 textures 4.4 4.6
Tri 1000k with 10 textures 3.5 3.5
Tri 1000k with 50 textures 2.4 2.3
Tri1000k with 100 textures 0.7 0.7
Tri 2000k with 10 textures 1.0 1.1
Tri 2000k with 50 textures 0.9 1.0
Therefore, from this point onwards, the experiments were done fully in EONTM StudioTM ProfessionalTM but with different hardware settings to understand the difference. EONTM StudioTM ProfessionalTM was used because the research started with this software and a
lot of projects have already been created in it. Thus, the building of the samples will be a lot easier and faster while there is totally none available yet from Quest3D®.
In conclusion, the independent variable tests have shown how each of the variables influences performance in the VR architectural visualisation. Along the way, time taken, frame rate, texture memory and vertex memory were recorded and analysed. It is clear that triangle size and software do not affect performance at all. Time taken is the same regardless of how complicated the 3D scene gets. Vertex memory and texture memory do increase when simulations get complicated with more complex 3D models but the
amount of memory available today in the graphic cards (up to 1.5GB) and systems (2GB to 8GB RAM) are sufficient to sustain them. Table 4.13 gives a summary of the impact of all these variables against frame rate.
The secondary variables are not explored further from this point because of two reasons.
There is no reason to optimise them first since their use in projects is optional, depending on project features, requirements and demands. Apart from that, they can only exist if the primary variables are in the models. The goal for this research is to optimise the
foundation first, which are the primary variables or the first generation variables. If the first generation variables are not efficient, there is no point in making the second
generation variables efficient since they only exist on top of the first generation variables.
Findings from this chapter will serve as good preparation for the final part of the
research, which is to look into the quantitative and qualitative aspects of the visualisation by conducting an experiment and a survey.
TABLE 4.13. Frame Rate Impact Against Variables
Primary Variable Time Taken Triangle Count x
-√ = mild impact,√ = impact, X = no impact
Primary Variable Frame Rate Triangle Count √
Triangle Size x Geometry Count √ Texture Count √
Texture
Resolution -√
Vertex Count √
-√ = mild impact,√ = impact, X = no impact
Secondary Variable Frame Rate
Light Count -√
Particle Systems √
Programmable / Advanced
Shaders Count √
Scripting √
Collision Detection √ Looped Video / Audio √
Hardware √
Software x
-√ = mild impact,√ = impact, X = no impact