Benchmarks for Radiance
Results for Various Machines

Written by Paul Bourke
Original: October 1997, Last updated: April 2002

In an attempt to quantify the relative merits of various platforms for rendering with Radiance, two scenes were designed each with quite different attributes. The first (bench1.tar.gz) has a very small memory footprint, the second (bench2.tar.gz) is much more demanding both in terms of memory and runtime.

Results

The statistics that follow are the time it took the respective machines to render the scenes in seconds (lower the better). Radiance reports user and system times, these were checked against the results from the UNIX "time" command. In general, special compiles were not created.

Machine benchmark 1 benchmark 2
Intel processor P4, 2GHz 166 485
Intel P4, 1.5GHz 190 696
Compaq ES40, 667MHz, Alpha 21264 ev6 202 1002
Dell workstation, 933MHz, Intel Pentium III (Coppermine) 260 950
Compaq XP1000, 600MHz, Alpha 21264 ev6 280 1356
Digital DPW, 600MHz, Alpha 21164 ev56 443 2187
SGI Origin 2000, 195MHz, IP27 670 2600
SGI O2, 180MHz, R5000 1560 4820
SGI Indigo, 250MHz, R4000 1660 5500
SGI Power Challenge, 90MHz, R8000 2690 6615
SUN Ultra SPARC, 167MHz 1550 11700
SUN SPARCStation 10 5616 35157

Benchmark 1

Reflective mirror scene
6 polygons, 3 cylinders, 1 sphere

oconv mirrorbox.rad > mirrorbox.oct
rpict -t 30 -vf v1.vf -lr 100 -ab 1 \
        -x 1000 -y 1000 \
        mirrorbox.oct > mirrorbox.pic
pfilt -e 3 -x /2 -y /2 mirrorbox.pic > mirrorbox_e.pic

Benchmark 2

Mixed gasket and sponge
87000 4 vertex facets

oconv scene.rad > scene.oct
rpict -t 30 -vf v1.vf -ab 1 \
        -x 1000 -y 1000 \
        scene.oct > scene.pic
pfilt -x /2 -y /2 scene.pic > scene_e.pic