by FastSite
12/17/2002 | 12:00 AM
Since we have already discussed it in our article called "Intel Pentium 4 3.06GHz CPU with Hyper-Threading Technology: Killing Two Birds with a Stone…", you should remember that Hyper-Threading technology emulates a dual-processor platform with only one physical CPU in the system. Therefore, I suggest that you also try to recall what the advantages of a dual-processor system are in 3ds max 5 application. Having hard time? Let me help you then :) <%BANNER[article]%>
As you know, the final renderer in 3ds max is optimized exclusively for dual-processor platforms. It results into an 80% performance growth, which is quite a usual thing for 3ds max. Also dual-processor platforms have a few advantages for processing massive geometry in viewports due exactly to the fact that there are two CPUs instead of one: we all know that the geometry calculations in 3ds max are performed solely by the system CPU, and the graphics card chip has nothing to do about it (unlike the situation in games).
However, dual-processor platforms also have a big drawback: when the scene to be displayed in the viewports is built with not very much geometry, dual-processor platforms demonstrate somewhat lower fps rate than uni-processor ones. It seems to me that this phenomenon has a lot to do with the fact that two streams created by the CPUs cause certain delays required to shift between them. This way, when there is little geometry to be processed, these delays eat up the performance gain due to the use of two processors.
Now that we have singled out all highs and lows of the dual-processor platforms, let's turn to some tests, which will help us to find out how well this theory corresponds to real life.
Since the work in 3ds max can be split in two major parts: creating the scene in viewports and performing the final rendering of this scene, the article will also have two parts. Namely we will dwell on tests ran in viewports and on testing the final rendering speed.
For our test session I used the benchmarks you are already familiar with, which are recommended by the 3ds max 5 developer. The tests in viewports were run at 1280x1024x32bit and the results were taken in fps. During the final rendering tests, the resolution was set to 800x600 and we measured the time it took each of the testing participants to complete the tests (so, the lower is the value, the better).
Testbed 1:
Testbed 2:
We used the following software:
For GeFroce4 Ti 4600 we used driver version 30.82 with V-Sync off.
This benchmark checks if the graphics card is capable of refreshing more than one viewport in 3ds max simultaneously. There is some animation played in all viewports, and to load the graphics accelerator even more each viewport is displayed in a different mode: from Wireframe to Smooth + Highlights.
Since this 3ds max 5 benchmark is quite precise, we can state a slight performance increase (by about 2.5%) with the enabled Hyper-Threading technology, when there is animation played in viewports.
Here the graphics cards have to work with animation in a single viewport. During the test the camera is flying above the rocks and hills of the moon surface landscape, which is built by 400 thousand polygons displayed in Smooth + Highlights mode.
The situation here is similar to that in the previous benchmark: the performance is growing a little bit.
This test is none other but the same moon surface picture from the previous benchmark. However, besides the landscape itself, we now have some flying objects, like pace crafts or planes.
Here the systems proved equally fast, which is most likely to be because of the GPU, which had to visualize a great lot of polygons on its own. I would like to draw your particular attention to the fact that this benchmark checks how well the scenes with massive geometry can be displayed in viewports in Smooth + Highlights mode, which involves the graphics card chip to the full extent. As you may notice, the situation in Benchmark 13 is completely different, as it is run in Wireframe mode.
This benchmark deal with the processing of multiple light sources. Since most graphics cards do not support more than 8 light sources, this test as well as the next two work with 8 lights of different types. Here we will have 8 SpotLight light sources, which move and light some geometrical object.
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.
Well, this is finally the first benchmark where our system with Hyper-Threading appears slower. The reason for that is the drawback of the AGP OpenGL drivers, which we have described in the very beginning of the article.
Here we have the same object, but this time it is lit by 8 Directional lights. Directional lights in 3ds max 5 are the fastest unlike the previous package version.
The situation repeats again with different light sources.
Again we've got the same object and 8 light sources. But this time these are all Omni lights, which involve average resources between SpotLight and Directional lights described above.
In the last lighting benchmark the system with Hyper-Threading again appears the slowest. Summing up the performance in lighting tests we can conclude that Hyper-Threading causes about 5% performance drop here.
This scene boasts "easy" geometry and a couple of light sources. It is displayed in a single viewport on the entire screen thus being an excellent test of the rasterizing speed in Smooth + Highlights mode.
In rasterization benchmark both systems perform equally fast, which is most likely to be connected with the fact that this benchmark requires both: geometrical calculations alongside with the GPU visualization ones.
This test is intended to show how fast the graphics cards are when it comes to multiple textures processing. The file contains a lot of textures and very little geometry.
This benchmark emulates the work on the game level, as it contains both: sufficient geometry and numerous textures. The animation is arranged in such a way that the entire scene could be displayed completely.
This test reveals the ability of the graphics accelerators to display textures on the deforming geometry.
The results in texturing benchmarks show that Hyper-Threading hardly has any positive effect on the performance. On the contrary, its influence is sometimes quite negative.
This test is aimed at showing what the graphics cards are capable of in terms of transparent textures processing.
The new 3ds max 5 features not only the transparency remaining from the previous version, which is imitated by dithering:
But also the 'real" transparency implemented by blending the pixel color of the overlapping objects:
You can shift between the transparency modes in the viewport control panel:
Of course, you can guess that the more correctly implemented transparency will be slower.
Well, as we see, the picture here is quite predictable. The CPU and GPU workloads here are nearly equal, that is why in the more or less balanced scene with reasonable geometry (up to 100,000 polygons) and not too many light sources, the advantages and drawbacks of the Hyper-Threading technology level out.
Here the camera flies through the rocks and hills of the moon surface landscape built of 400 thousand polygons, i.e. the scene is the same as in Benchmark 2, actually. However, the picture is displayed in the Wireframe mode.
However, in benchmarks with huge geometry, Hyper-Threading finally gets the chance to show its best, especially in Wireframe mode, where the graphics card is loaded considerably less than in Smooth modes. So, the processor's ability to cope really fast with geometry calculations moves to the forefront.
This is the Benchmark 3 scene in wireframe mode:
Well, more geometry, more advantages provides Hyper-Threading technology.
So, summing up the results obtained during tests in viewports we can conclude that in case massive geometry needs to be processed, especially in Wireframe modes, Hyper-Threading appears really demanded, though not that much: The performance grows up within only 5%. Sometimes, if the geometry is really gigantic (Benchmark 13) - up to 10% at the most.
At the same time, when we have to process scenes with little geometry, but quite many light sources and textures, the situation appears just the opposite: the system with Hyper-Threading gets about 5%-10% slower. In general we can state that both platforms showed quite similar results.
This benchmark is based on a well-known ktx_rays scene borrowed from the very first 3ds max version, when the whole package used to be called 3D Studio MAX developed by Kinetix Company, which wasn't yet a part of Autodesk. Now the file is called 3smax5_rays, which anyway doesn't change its essence. Unlike the first version of this scene, the current benchmark includes not only the demonstration of the volumetric light but also a reflective surface at the bottom:
We have come across this file in the previous 3ds max versions, though it wasn't included into the benchmarks package at that time. This benchmark features two lights (one Raytraced and one Shadow Mapped), which light geometry built of 50 thousand polygons:
This test represents a scene from the underwater world and includes numerous reflective maps. It can be rendered in 5 passes:
As we see in the final rendering benchmarks, the platform with enabled Hyper-Threading shows results, which are stably 15%-20% higher.
To run these tests, I opened two 3ds max 5 applications and started rendering VolumeLight file from Benchmark 5 in one of them. In the other application I started Benchmark 1 from the First part of our Viewports performance tests. This way, we could see how the testing participants would cope with the animation played against the background of the final rendering. Here are the results obtained:
As you can see, the system with enabled Hyper-Threading proved extremely fast: the performce got 5 times better!
Here I started two tasks at a time: archiving with the maximum compression level with the WinRAR utility and rendering the Benchmark 2 file from the previous part of the performance analysis.
Again the results are fantastic: over 25% performance increase!
As is known, when we work with a single processor system (without the Hyper-Threading technology) with two or more tasks in Windows environment, the priority lies with the active task. However, it is not always the case: sometimes the priority goes to a more resource hungry task. In case of a dual-processor system with two tasks running at a time, these tasks are processed in parallel by both processors. The same thing happened when we tested the system with enabled Hyper-Threading technology: independent of the active window both tasks were processed at a time, which results into a tremendous performance growth. Especially is the background task is more resource-hungry than the "active" task (such as the Benchmark 1). Of course, we can hardly hope to catch up with the real dual-processor system in terms of performance, however, compared with the uni-processor one running without Hyper-Threading, the improvement is simply gigantic.
Well, Intel managed to move from the simple increase in the core clock frequencies to introduction of optimized algorithms, which will definitely make the further frequency increases much more efficient. In 3ds max 5 application Hyper-Threading technology proved highly powerful, as it ensured at least 10%-20% performance growth during single-task final rendering and a multiple performance increase during multi-task viewports processing.