Testbed and Methods

To estimate the testing participants' performance I used the same benchmarks as in the previous reviews. As I have already pointed out multiple times, these benchmarks are recommended for testing in 3D Studio max by the program developer company.

However, I have added another three scenes to the traditional complex one, as they will help us to consider the performance of the today's graphics accelerators in greater detail. The descriptions of the benchmarks will follow.

I decided not to run any tests with the enabled Anti-Aliasing, as all the contemporary graphics accelerators can do 3ds max anti-aliasing without performance losses.

We tested our hardware in the following conditions:

• Background Texture Size: 512x512
• Texture Size: 256õ256
• Texel Lookup: Nearest
• Mipmap Lookup: None
• Display wireframe objects using triangle strips: No
• Use triangle strips: Yes
• AntiAliased Lines: No

In our test session I will compare the results obtained on ATI FireGL 8800 and NVIDIA Quadro 750XGL and 900XGL with the performance of their gaming "forefathers": ATI RADEON 8500 (275MHz/550MHz) and NVIDIA GeForce4 Ti 4600 (300MHz/600MHz). I didn't level out all the working frequencies on purpose. First of all these frequencies are really close to one another, and secondly, there is a certain chip-to-memory frequency ratio recommended by each developer, which will be automatically ruined if all testing participants will be brought to the same level. I also added the results for Quadro DCC (born from GeForce3 200MHz/460MHz) to compare them with the performance of new Quadro XGL solutions to find out if Quadro DCC is worth replacing with the newcomers.

We ran all the tests on the following test platform:

• Pentium 4 2.2GHz CPU;
• ABIT TH7II mainboard;
• 1024MB PC800 RDRAM;
• Graphics cards: NVIDIA GeForce4 Ti4600, Quadro4 750XGL and 900XGL, Quadro DCC, ATI Fire GL 8800, RADEON 8500;
• 20GB IBM DTLA 7200rpm HDD.

We used the following software:

• Windows 2000 SP2;
• 3ds max 4.26 (OpenGL rendering), 1280x1024 32bit.

We used the following drivers for the graphics cards tested:

• Quadro4 750XGL and 900XGL, Quadro DCC: Maxtreme v4.00.18 for 3DMAX and Detonator XP version 27.70 for OS.
• GeForce4: Detonator XP version 27.70.
• Fire GL 8800: driver version 3024.
• RADEON 8500: driver version 6025.

Performance: 3ds max Recommended Benchmarks

Benchmark 1

The first benchmark is a certain "stress-test". It plays an animated scene in four viewports at the same time. However, the rendering methods set in these viewports are different. The scene in two upper viewports is in "Wireframe" mode, in the lower left viewport - in "Smooth + HighLights" + "Edged Faces", and in lower right viewport - in "Smooth + HighLights".

This scene contains very few polygons, only 28 thousand. However, since the animation was played in all four viewports simultaneously, the resulting fps rate appeared not very high.

• Polygons: 28868
• Light sources: 1
• Mode: Wireframe, Smooth + Highlights

The first thing that catches your eye is the almost 50% advantage of FireGL 8800 over Quadro4. Secondly, we see that the increase in the chip working frequency of Quadro 4 compared with the predecessor led to a tiny performance growth in this benchmark.

Moreover, it is very illustrative that all NVIDIA graphics cards show almost similar results. Firstly, it means that Maxtreme drivers do not have any advantages over the Detonator drivers when playing animation in 4 viewports, and secondly, it means that the chip frequency of NVIDIA solutions doesn't matter at all: even Quadro DCC clocked 1.5 times lower performed almost as fast as Quadro4 900XGL.

On the other hand, the performance difference between FireGL 8800 and RADEON 8500 is too high. This fact also means a few things to us. Firstly, it implies that ATI software developers did a great job on the professional drivers. But at the same time it means that unlike NVIDIA, ATI's gaming drivers are much less optimized for such applications as 3ds max, which has already become traditional for ATI solutions: the very first drivers for RADEON 256, for instance, couldn't make it work in viewports at all.

Benchmark 2

This benchmark shows how graphics cards can cope with very complex geometry. It shows the graphics cards performance in Smooth + Highlights mode in scenes with complex geometry.

Here I decided to expand our tests a little bit and added a few tests in Wireframe mode to the Smooth + Highlights mode of both geometric benchmarks.

• Polygons: 200270
• Light sources: 1
• Mode: Smooth + Highlights, Wireframe

So what do we see? In Smooth + Highlights mode Quadro4 cards manage to get a bit ahead, however, in Wireframe mode FireGL 8800 rushes forward showing almost twice as fast performance. We also see very clearly that the difference in working frequencies between Quadro4 graphics cards doesn't have any influence on the results.

Benchmark 3

The third benchmark is devoted to testing the graphics cards abilities in terms of processing complex geometry only. This time the amount of polygons nearly doubled and made 376 thousand. The same surface, as in the previous benchmark scene, is now covered with buildings.

• Polygons: 376875
• Light sources: 1
• Mode: Wireframe, Smooth + Highlights

The results obtained in this benchmark are similar to what we have just seen in the previous benchmark, however, we should point out that the relative performance difference between the graphics cards is smaller here than in the previous test, because the geometry here is much more complex.

Judging by the results of the last two geometric benchmarks, we can say that the difference in the working frequencies doesn't matter at all for Quadro4 graphics cards, and for complex geometry we would recommend FireGL 8800, which falls a bit behind in Smooth + Highlights mode but defeats all the competitors in Wireframe mode.

Quadro DCC, the previous generation graphics solution managed to run neck and neck with Quadro4 in Smooth + Highlights mode, however, in Wireframe mode its performance worsened by 20-30%.

The gaming solutions from NVIDIA managed to prove several times faster in geometric benchmarks than the gaming card from ATI. Of course, NVIDIA owes its victory to the "universally" optimized drivers, which RADEON 8500 doesn't have.

Benchmark 4

This benchmark tests how fast the multiple light sources can be processed. Since most graphics cards do not support more than 8 light sources, this test and the ongoing two tests contain 8 light sources of different types. This test offers 8 SpotLights, which are constantly moving and lighting some kind of asteroid.

We should point out that imitating the effect made by SpotLights is a much more resource-hungry process than the imitation of Omni or Directional lighting.

• Polygons: 39600
• Light sources: 8
• Mode: Smooth + Highlights

This is the first benchmark where the clock frequency difference between the Quadro4 graphics solution matters for the performance. The performance of FireGL 8800 is not very impressive here. It is comparable with that of the gaming RADEON 8500 and is lower than that of GeForce4 Ti4600.

Benchmark 5

Here we see the same asteroid, but this time it is lit by 8 Directional lights. Directional lights are slower than Omni lights, but faster than SpotLights.

• Polygons: 39600
• Light sources: 8
• Mode: Smooth + Highlights

When processing Directional lights, which are less resource hungry than Spot lights, Quadro4 performance has grown much higher, unlike the performance of GeForce4 Ti 4600. The performance of ATI graphics cards, the gaming one and the professional one, grew even more, which means that the driver developers paid much more attention to optimization for Directional lights.

Benchmark 6

One more time we see the same asteroid with the lights coming from 8 different sources. But now these are all Omni lights, the fastest lights in 3D MAX.

• Polygons: 39600
• Light sources: 8
• Mode: Smooth + Highlights

Here the results shown by both: "professional" Quadro4 and gaming GeForce4 Ti 4600 increased compared with the previous cases, and those of both ATI solutions dropped down. Since Omni lights are the least resource demanding, then we can explain the performance drop of ATI solutions only by poor driver optimization for these particular light sources.

Benchmark 7

This is a scene with 1 light source and very simple geometry, including only 4500 polygons, which occupies the entire viewport. It is aimed at testing the rasterizing speed in Smooth+ Highlights.

When the camera is moving, the graphics card should rasterize big and small polygons (relative to the screen size).

• Polygons: 4684
• Light sources: 1
• Mode: Smooth + Highlights

In the Rasterization benchmark professional cards from ATI and NVIDIA showed close results. Also this benchmarks shows us that the GPU working frequency is of great importance here (by the example of Quadro4 cards). We also see that the performances of the gaming GeForce4 Ti 4600 and professional Quadro4 750XGL are nearly equal, which is actually not surprising for us, since the testing conditions of the raterizing test are very close to games. That is why the card shows good results here. The last one in the race is RADEON 8500, as usual.

Benchmark 8

The next benchmark is devoted to work with textures. The file contains a lot of textures and very little. As for the benchmark, it is just a rotating spherical polygon with the facets covered with 48 different textures.

Very little geometry and many textures involved give us a perfect idea of how fast the graphics cards can process these textures.

• Polygons: 224
• Light sources: 1
• Mode: Smooth + Highlights

The results shown by FireGL 8800 in this texture benchmark are dramatically low: it is almost twice as slow as Qoadro4 based cards. The gaming card from NVIDIA also beats ATI's professional solution almost twice.

I decided to go a little bit farther with this benchmark, which you should know very well already from our previous reviews. I suggest taking a closer look at how well the graphics cards cope with miniatures of different size. I usually test graphics cards in texture benchmarks with the enabled setting: "display texture miniatures in 256x256 pixels viewports", which is a default setting for OpenGL drivers of our today's testing participants. However, in order to test the dependence of the graphics cards performance on the miniature size more thoroughly, I ran the same Benchmark 8 with the miniature textures settings from 128x128 pixels to 1024x1024 pixels.

The results are given in the table below:

Since OpenGL driver doesn't allow setting the texture size over 512x512 pixels, we have no data for the gaming graphics cards with the settings exceeding 512 pixels.

The results of a more detailed testing session provided in this table prove that NVIDIA cards retain the twofold priority with all the settings. The miniatures of 1024x1024 and up (Maxtreme drivers allow setting miniature size up to 4096x4096 pixels) make the texture rich scenes impossible t process, as these miniatures do not fit even into 128MB of the local memory and the graphics card performance is simply killed by the texture transfer rate via AGP.

Benchmark 9

This is a fully textured room with a camera moving inside. This benchmark is very close to real applications because it has a lot of textures, not very simple geometry and several light sources. It shows what graphics cards are capable of when processing pretty complex scenes in Smooth + Highlights mode.

• Polygons: 12413
• Light sources: 8
• Mode: Smooth + Highlights

The situation here is nearly the same as in the previous benchmark, however, both gaming cards, RADEON 8500 and GeForce4, showed considerably lower performance compared with "professional" cards than in the previous case: RADEON 8500 performed twice as slow, and GeForce4 Ti 4600 - 1.5 times slower.

Benchmark 10

Animated "waves" with a texture laid over them show how fast the card can process and modify the textures.

• Polygons: 880
• Light sources: 1
• Mode: Smooth + Highlights

Summing up the results of all texturing benchmarks we can conclude that ATI software developers really left out the driver optimization for texture-rich scenes. Even the gaming GeForce4 Ti 4600 is about 1.5-2 times faster than "professional" FireGL 8800. Therefore, if you are planning to work with highly textured scenes, ATI products are hardly the right choice to make: the relatively cheap (compared with the professional cards) GeForce4 Ti 4600 will suit much better for these needs.

Benchmark 11

This benchmark runs with different speeds in the Wireframe mode. 111 thousand polygons in Wireframe mode will be a really tough test for any modern graphics card.

• Polygons: 11270
• Light sources: 1
• Mode: Wireframe

In Wireframe mode the situation is totally different from what we saw in texture-rich benchmarks: FireGL 8800 is almost twice as fast as Quadro4 family. It looks as if ATI software team put all their knowledge and effort into optimization for Wireframe modes. And I am absolutely sure that these are solely drivers that matter, because RADEON 8500 gaming graphics card showed very poor results here, being even slower than GeForce4 Ti 4600.