The SPECapc working group of
the SPEC/GPC committee wishes to entertain proposals to produce a Maya
benchmark based on the "Maya Complete" product.
The benchmark should tax the
machine under test in order to provide both interesting and relevant data
for the evaluation of the performance of workstations running the application.
It should reflect the breadth of the Maya application, and incorporate
typical use of application features without focusing unduly on any one
area. It should model the workflow of a typical Maya user, and demonstrate
realistic scene complexity.
A sampling of features that
should be included are:
- Modeling using splines
- 3D painting
- Key frame and/or path animation
- Skeletons and Inverse kinematics
- Skinning
- Particle systems
- Marking menus
- Hard and/or soft body dynamics
- Hardware rendering
- Software rendering
Structure
The benchmark should be broken down into multiple scenes, each of which
demonstrates a collection of Maya features that would typically be used
together. An example of scene breakdown could include:
- Character design
- Scene construction
- Animation setup and playback
in preview mode (3D hardware rendering)
- High-quality software rendering
(sample frames from full rendering)
The benchmark will measure
the time to complete each of the main sections, and should measure significant
components within each scene. Time measurements will be written to a file
and compared to those generated by the reference system described below.
Environment
The benchmark will run the Maya application (v4 or later) on all operating
systems supported by Alias|Wavefront at the time of release. It is likely
to assume the existence of a licensed copy of the application on the target
system. It should not use plug-ins or MEL scripts that are not part of
the Maya 4 distribution, unless provided with the benchmark.
Operation
The benchmark should be able to be run multiple times from the operating
system's command line without any user interaction. It will also provide
the option to run each of the three main sections independently. The benchmark
output (timing, sample frames, etc.) will be written to one or more files
suitable for use with the SPEC/GPC reporting tools. The models and software
rendering scripts should be leaveragable to the Maya batch renderer.
Maintenance
The benchmark should be readily upgradeable to future versions of the
Maya application.
Rights & Ownership
Upon acceptance by SPECapc, all materials within the benchmark, including
source code written specifically to develop or maintain the benchmark,
become the property of the SPEC/GPC committees. Models not originally
provided by SPECapc may remain the property of the original owner, but
must be made available, as part of the complete benchmark, without encumbrance
in accordance with the SPEC/GPC benchmark license policies. SPECapc will
distribute the benchmark electronically, free of charge, via its web and
FTP sites. All materials incorporated in the benchmark (including models)
must be available to be included in the free download.
Performance Characteristics
The benchmark should have a duration of approximately 60 minutes on a
reference system defined below. Multiple sections of the benchmark should
be timed, but no individual time measurement within the benchmark should
have a duration of less than 1 minute on the reference system.
Reference System:
-- 700 MHz Intel® Pentium® III processor
-- Intel® 440BX chipset
-- Maya 4
-- Microsoft Windows 2000 operating system
-- Diamond Multimedia FireGL1 graphics card, configured 1280x1024, 24bit
color, 75 Hz refresh rate
Suggestions for Incorporation
into the Animation
SPECapc would like to see some past SPEC/GPC benchmarks referred to in
the animation, either by incorporation of models, logos or screen-shots
of executing benchmarks. These models include a pickup truck, lawnmower,
human figure and photocopy machine, but would need to be converted into
an appropriate form for inclusion in the animation.
|