3ds max Viewset (3dsmax-04)

This test is based on graphics created in 3ds max 3.1. We used an OpenGL plugin (driver) from Discreet for tracing. Models were taken from SPECapc 3ds max 3.1 test. Each model was measured with two different lighting models to reflect a range of potential 3ds max users. The high-complexity model uses five to seven positional lights as defined by the SPECapc benchmark and reflects how a high-end user would work with 3ds max. The medium-complexity lighting models uses two positional lights, a more common lighting environment. For a detailed description of test scenarios and weight coefficients please visit the 3dsmax-04 page.

We've noticed two patterns here. Firstly, a higher result always corresponds to a more powerful graphics card. Secondly, the high-clocked dual-core processor is the best choice here, while Q6600 performs even a tad worse than E4700.

CATIA Viewset (catia-02)

This test imitates operation in CATIA V5R12 from Dassault Systemes. We used models from CATBench2003 developers and CATIA community. The car model contains more than two million points. SPECviewperf replicates the geometry represented by the smaller engine block and submarine models to increase complexity and decrease frame rates. After replication, these models contain 1.2 million vertices (engine block) and 1.8 million vertices (submarine). For a detailed description of test scenarios and weight coefficients please visit the catia-02 page.

The test description makes it clear that the entire load will fall on a graphics card here. What concerns CPU ratios, they are practically the same as in the previous test. Like the previous test, it illustrates how weak gaming graphics cards are in professional tasks. Our GeForce card is 2-3 times as slow as the equally-priced Quadro card. The same situation can be seen in most other tests, except two.

EnSight Viewset (ensight-03)

CEI EnSight is a software product for engineering and scientific visualization. CEI contributed the models and suggested workloads. The model data is replicated by SPECviewperf 9 to generate 3.2 million vertices per frame. But the rendering load is significantly increased. For a detailed description of test scenarios and weight coefficients please visit the ensight-03 page.

Note how FX 570 breaks away from FX 370. But already FX 1700 does not bring special advantages. Roughly speaking, nothing has changed for CPUs, any one will do from the point of view of this test.

Maya Viewset (maya-02)

This test imitates user operations in Alias Maya 6.5. The models used in the tests were contributed by artists at NVIDIA. Wireframe complexity differs greatly (from 2.2-2.4M to 23M vertices). For a detailed description of test scenarios and weight coefficients please visit the maya-02 page.

FX 570 breaks away from FX 370 again. However, there are still no reasons to consider a more expensive FX 1700. GeForce 9600 GT also performs well in Maya and EnSight, only slightly slower than FX 370 in the last test. We might have skipped our comments on differences between CPUs -- previous tests somehow revealed slight differences between them, but results of this test easily fall within a measurement error.

Pro/ENGINEER Viewset (proe-04)

These tests are based on the old Pro/ENGINEER 2001. However, this is a conservative engineering suite, so we cannot say that its load has changed much in its modern versions. So it can still be used for a combined estimate. If you work with new versions of this program, it makes sense to analyze results of alternative benchmarks, not from SPEC. For a detailed description of test scenarios and weight coefficients please visit the proe-04 page.

Graphics cards line up in a similar way. What concerns CPUs, we'd prefer a dual-core processor, subject to a proviso that only one core is used in this case, now even two.

SolidWorks Viewset (sw-02)

The same situation: even though there is a SPECapc test for modern SolidWorks 2007, SPECviewperf still uses SolidWorks 2004. Two models are used for rendering: GTX car and Suzuki engine. We have no exact information about their complexity (about 3M vertices in the engine). For a detailed description of test scenarios and weight coefficients please visit the sw-02 page.

It's one of those rare tests, where graphics cards line up in a uniform way, that is we can see the advantage of FX 1700. But there are still no differences between CPUs.

UGS Teamcenter Visualization Mockup Viewset (tcvis-01)

This test imitates user activities in UGS Teamcenter Visualization Mockup. Its various scenarios use models with 5M-11M vertices. For a detailed description of test scenarios and weight coefficients please visit the tcvis-01 page.

Results depend solely on a given graphics card. To all appearances, a more powerful graphics card would improve test results. No signs of CPU dependence.

UGS NX 3 (ugnx-01)

This test imitates user operations in UGS NX 3. It's a popular engineering modeling suite (not the latest version again). Model complexity is very high -- the volume of data used to render a single frame reaches 800 MB. Test scenarios use models with 3.9M-30M vertices. For a detailed description of test scenarios and weight coefficients please visit the ugnx-01 page.

Results depend solely on a given graphics card. A more powerful graphics card is welcome again. No CPU dependence.

Conclusions

Quadro FX 370 and FX 570 cards are officially positioned for the Entry-Level segment and FX 1700 is a Mid-End card. So it's only natural that performance of real-time modeling and visualization applications does not depend on CPU. In other words, it makes sense to buy a more powerful graphics card. On the other hand, FX 1700 does not always break away from FX 570. The latter looks like an optimal choice for an inexpensive workstation. If you plan to work with professional software, it apparently makes no sense to buy a gaming graphics card. GeForce 9600 GT is greatly outperformed by the equally-priced Quadro FX 370 in most tests.

Besides, SPECviewperf tests imitate user activities, which is a valuable feature allowing to find out how good a given configuration is for a user (performance and comfort levels). However, some applications use final rendering, when a model is rendered by CPU using algorithms different from algorithms that utilize GPU at previous stages. As this process is automated and needs no user input, one can schedule it for free time. Or, if your work is very intensitve, rendering can be offloaded to a dedicated server. Still, if you want to build a universal workstation for applications like 3dsmax or Maya, you should consider your typical model rendering times.

Write a comment below. No registration needed!