Benchmark Results: SPEC Tests
The Standard Performance Evaluation Corporation was established in 1988 as a cooperative effort between workstation vendors to create a set of standardized, realistic performance tests. As it grew, it became an effort to both allow vendors to demonstrate their performance advantages and also have the tests based on existing applications, instead of older, less-relevant pieces of software.
Developed by the SPECgpc (Graphics Performance Characterization) working group, viewperf tests the performance of a system running under OpenGL using the graphics engines of various 3D applications from the content creation and CAD markets.
|Viewset||Composite Score||Multi-Sample Performance|
Viewperf is designed to only stress the graphics subsystem of a computer under OpenGL. It doesn’t test CPU loading and only needs 3 GB of installed memory to run.
Most of the tests in this revision of viewperf are based on CAD-related software, with the animation software tests being LightWave and Maya. The graphics tests for LightWave and Maya are also repeated in their respective SPECapc tests. The tests marked with a multi-sample performance of 4x ran the tests with both no AA and 4x AA, while those without it ran only with no AA.
The SPECapc tests are application-specific benchmarks that run an extended script, which performs various operations to evaluate performance during part of a normal workflow. It tests interactivity, real-time 3D performance, and rendered 3D performance.
SPECapc LightWave 9.6
The LightWave test was co-developed with Newtek and tests character animation interactivity, OpenGL performance, and software rendering. The multitasking segment of the benchmark, in which single-threaded tasks are run concurrently with rendering, helps gauge total system performance in a multitasked environment.
SPECapc LightWave performance is normalized against a 2.0 GHz Dell Precision 690 Workstation, a dual-core processor based on Intel's Core microarchitecture, with a Quadro FX 570 graphics card. So, even without other systems in the charts, you can compare the performance of this system against an older Core 2 machine at a glance.
The render time improvement of 7.48x over the baseline is particularly encouraging, but the less-impressive improvement reflected in the interactivity tests is likely indicative of some other variable.
SPECapc Maya 2009
SPECapc Maya 2009 features updates to the older Maya 6.5 workload, plus an additional scene created by AMD. Its output is also normalized against the same Dell machine as the LightWave test above. The Maya test shows a little bit less performance improvement over the baseline than the LightWave test does, likely because the Maya test is run full-screen, while the LightWave test regulates its own window size.
SPECapc 3ds Max 9
The 3ds Max 9 test is the oldest of the three SPECapc tests that we use, and is normalized against a much older machine (a NetBurst-based Xeon at 2.4 GHz using RDRAM and a Quadro 1000). Thus, while the results are easily comparable to other runs of this test, they are not easily related to the results from the other two SPECapc tests.
Consider its inclusion here to be a data marker, as SPEC is promising a new SPECapc for 3ds Max 2011 soon.