OpenGL Performance Evaluation on Multiple Computer Platforms
Sean Ho · 2001
With the recent advent of 3D graphics hardware for personal computer (PC), it is worthwhile to exploit the cost effectiveness and OpenGL performance issues among currently available commercial off-the-self (COTS) computers.Graphics hardware vendors typically list several gross measurements of system performance when releasing new graphics hardware.Often these coarse or subjective figures do not represent how a software application performs.On the other hand, one seldom sees the same benchmark performed on machines across multiple platforms and operating systems, i.e., Intel-based PCs and RISC-based UNIX workstations.This document reports the results obtained from running two OpenGL benchmark programs, SPECviewperf 6.1.2and SPECglpcrf 3.1.2,on existing computer workstations at ARL. References 28 Distribution 29Documentation Page 311.This graph shows the performance of the four target systems from running the Awadvs-04 benchmark in SPECviewperf 6.1.23 2.This graph shows the performance of the four target systems from running the DRV-07 benchmark in SPECviewperf 6.1.24 3.This graph shows the performance of the four target systems from running the DX-06 benchmark in SPECviewperf 6.1.24 4.This graph shows the performance of the four target systems from running the Light-04 benchmark in SPECviewperf 6.1.25 5.This graph shows the performance of the four target systems from running the MedMCAD-01 benchmark in SPECviewperf 6.1.25 6.This graph shows the performance of the four target systems from running the ProCDRS-03 benchmark in SPECviewperf 6.1.26 7.This graph shows the plotted WGM results for the six viewsets in SPECviewperf 6.1.27 8.This graph shows the performance of the four target systems on the rendering of disjoint lines in immediate, RGB, and flat-shaded mode 8 9.This graph shows the performance of the four target systems on the rendering of disjoint lines in display-list, RGB, and flat-shaded mode 8 10.This graph shows the performance of the four target systems on the rendering of lines strips in immediate, RGB, and flat shaded mode 9 11.This graph shows the performance of the four target systems on the rendering of lines strips in display-list, RGB, and flat shaded mode 912.This graph shows the performance of the four target systems on the rendering of triangle strips in immediate, Z buffer, and smooth shaded mode with 1 infinite light source 913.This graph shows the performance of the four target systems on the rendering of triangle strips in display-list, Z buffer, and smooth shaded mode with 1 infinite light source 9 IV 14.This graph shows the performance of the four target systems on the rendering of quads in immediate, Z buffer, and smooth shaded mode with 1 infinite light source 10 15.This graph shows the performance of the four target systems on the rendering of quads in display-list, Z buffer, and smooth shaded mode with 1 infinite light source 10 16.This graph shows the performance of the four target systems on copying pixels of varying image sizes within the frame buffer 11 17.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in immediate and RGB mode 11 18.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in display-list and RGB mode 11 19.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in immediate and RGBA mode 12 20.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in displaylist, and RGBA mode 12 21.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in immediate, 2x zoom, and RGBA mode 12 22.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in displaylist, 2x zoom, and RGBA mode 13 23.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in immediate, 0.5x zoom, and RGBA mode 13 24.This graph shows the performance of the four target systems on writing pixels of varying image sizes to the framebuffer in displaylist, 0.5x zoom, and RGBA mode 13 25.This graph shows the performance of the four target systems on reading pixels of varying image sizes in RGB mode from the framebuffer 14 26.This graph shows the performance of the four target systems on reading pixels of varying image sizes in RGBA mode from the framebuffer 14 27.This graph shows the fill rate of each target system in various modes 14 28.This graph shows the performance of the four target systems on rendering varying sizes of line strips in immediate, and flat shaded mode 15 29.This graph shows the performance of the four target systems on rendering varying sizes of line strips in display-list, and flat shaded mode 15 30.This graph shows the performance of the four target systems on rendering varying sizes of line strips in immediate, and flat shaded mode with the Z buffer turned on 16 31.This graph shows the performance of the four target systems on rendering varying sizes of line strips in display-list, and flat shaded mode with the Z buffer turned on 16 32.This graph shows the performance of the four target systems on rendering varying sizes of line strips in immediate, and flat shaded mode with both the Z buffer and antializsing turned on 16 33.This graph shows the performance of the four target systems on rendering varying sizes of line strips in display-list, and flat shaded mode with the Z buffer turned on 16 34.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in immediate and flat shaded mode 17 35.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in display-list and flat shaded mode 17 36.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in immediate and flat shaded mode with Z buffer turned on 17 37.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in display-list and flat shaded mode with Z buffer turned on 18 38.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in immediate and smooth shaded mode 18 39.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in display-list and smooth shaded mode 18 40.This graphs shows the performance of target systems on the rendering triangles strips of varying sizes in immediate and smooth shaded mode with Z buffer turned on 19 41.This graphs shows the performance of target systems on rendering triangles strips of varying sizes in display-list and smooth shaded mode with Z buffer turned on 19 42.This graph shows the performance of the four target systems on rendering triangle strips with varying number of light sources in immediate and smooth-shaded mode with Z buffer turned on 20 43.This graph shows the performance of the four target systems on rendering triangle strips with varying number of light sources in display-list and smooth-shaded mode with Z buffer turned on 20 VI 44.This graph shows the performance of the four target systems on rendering quads with varying number of light sources in immediate and smooth-shaded mode with Z buffer turned on 20 45.This graph shows the performance of the four target systems on rendering quads with varying number of light sources in display-list and smooth-shaded mode with Z buffer turned on 20 46.This graph shows the performance of the four target systems on clearing the color buffer in various modes 21 47.This graph shows the performance of the four target systems on the rendering of points with various modes turned on 21 48.This graph shows the performance of the four target systems on rendering disjoint lines in various modes 21 49.This graph shows the performance of the four target systems on rendering disjoint triangles in various modes 22 50.This is one of the five graphs that show the performance of the four target systems on rendering disjoint quads in various modes 22 51.This is second of the five graphs that show the performance of the four target systems on rendering disjoint quads in various modes 22 52.This is the third of the five graphs that show the performance of the four target systems on rendering disjoint quads in various modes 22 53.This is forth of the five graphs that show the performance of the four target systems on rendering disjoint quads in various modes 23 54.This is last of the five graphs that show the performance of the four target systems on rendering disjoint quads in various modes 23 55.This graph shows the performance of the four target systems on the rendering of 10-sided disjoint polygons in various modes 23 56.This graph shows the performance of the four target systems on rendering text strings in various modes 23 57.This graph shows the performance of the four target systems on rendering images of various sizes in RGB format 24 58.This graph shows the performance of the four target systems on rendering images of various sizes in RGBA format 24 59.This graph shows the performance of the four target systems on rendering mipmapped textures of various sizes in RGB format 24 60.This graph shows the performance of the four target systems on rendering mipmapped textures of various sizes in RGBA format 25 61.This graph shows the performance of the four target systems on bounding non-mipmapped textures 25 62.This graph shows the performance of the four target systems on bounding mipmapped textures 25 VllTables 1.This shows the hardware specifications on the three target systems 1 2. Results in WGM for the six viewsets in SPECviewperf 6.1.27Vlll