iXBT Labs - Computer Hardware in Detail

Platform

Video

Multimedia

Mobile

Other

Testing 3D Accelerators in Linux

August 13, 2003




The word Linux can be often seen in news on the Internet sites and in printed periodicals. Certainly, most our readers are well aware that it is an operating system for modern PCs alternative to Microsoft's products. It's often mentioned as an office or even home OS rather than a network one. And taking into account that home PCs are inseparable from games, it becomes very interesting to look into this aspect. The next link in this logical chain will be 3D video accelerators. Their tests under the Windows operating systems take a solid part of our site. But their operation in the Linux remains a mystery.

This article shouldn't be considered a complete and unarguable research, and I hope that it will initiate a constructive discussion of testing video cards (and not only) in the OSes different from the Microsoft Windows.

PC configuration

Testbed:

  • Processor: Intel Pentium 4 2.53 GHz (FSB 533 MHz, L2 512 KB)
  • Mainboard: Gigabyte 8PE667 Ultra (i845PE chipset)
  • Memory: 256 MB PC2700 DDR SDRAM
  • Video cards: NVIDIA GeForce4 Ti 4200 64 MB and ATI RADEON 9500 128 MB
  • Hard drive: Quantum Fireball Plus AS 30 GB

Such configuration is not top, but it certainly can be referred to the middle-end class.

Although the video cards belong to different classes, it's the first time we test accelerators under an OS other than the Windows and we won't directly compare accelerators in the style of ATI vs NVIDIA, but focus on comparison of their performance in different OSes.

Software

For the Windows we used the following suite:

  • Windows XP Pro SP1
  • Intel Inf 5.00.1012
  • IAA 2.3

As to the Linux, the range of its distributives is really great. They actually differ in software versions, installers, programs supplied and localization.

Besides, the outcome may depend on a distributive (and its version). But if the Windows version can be easily described as the "Windows XP Pro Eng SP1", for such a dynamically developing OS as Linux any distributive recorded on CDs gets outdated by the time it appears on the market :)

A while ago I worked with the RedHat 7.3 and found that OS quite handy. Let's try to use it once again. The latest available version is RedHat 9. No additional packets and OS upgrades were used. The installation variant was Full.

Video cards' drivers

Like in case of the Windows, the drivers in the Linux were installed successfully. They were NVIDIA Detonator 40.72 and ATI 7-84-030228a1-008040c.

The only thing we changed was making the vertical sync off in the drivers. For that purpose we had to use an additional control panel for the ATI's driver.

It's a bit more difficult to install the drivers for the Linux than for the Windows, though it's a matter of experience. For the NVIDIA's video adapter we used the driver v1.0-4363 downloaded from the company's site. Its installation was very transparent - one simply has to start the file. The only thing that bothers is that the README.txt is over  120 KB (almost 3000 lines)! But it's highly recommended to read it before installing the driver.

In case of ATI searching for the drivers took a bit more time. On their site they offer drivers for the XFree86 4.1 and 4.2 (this is the graphics shell's core in the Linux), while the RH9 comes with the v4.3. However, I could find on the Internet the driver v2.9.12 written exactly for the XFree86 4.3. For correct installation of this driver we deleted the libGL packet included in the RH9 so that the driver could install its own version of its library.

There is one more problem with installation of the drivers for the Linux. The NVIDIA's drivers use variable environment for their settings. You can choose the FSAA mode, anisotropy and enable vertical sync (it's disabled by default).

For the ATI's cards you can only adjust gamma and enable the dual-head support. The vertical sync parameter is determined during the driver installation. Besides, since the driver was developed for professional cards, you can choose a profile for different applications (but what settings are used there is not clear).

Probably, graphics programs can control image display modes themselves, but it's difficult to ascertain that, and we will leave this issue for the next time.

Tests

Now, when we have the PC, two OSes and drivers, the only thing to do is to find applications which would work under these OSes and with these drivers.

Since we are comparing the video accelerators in different OSes (not one with the other), applications should be identical. Fortunately, we have found such:

  • Quake III Arena, bepcč˙ 1.32
  • Unreal Tournament 2003 DEMO 2206
  • SPECViewperf 7.1

These programs are available both for the Windows and for the Linux. There are more 3D games which can be used for the OS comparison (for example, RtCW, Serious Sam) but today we will use only the above listed.

The Quake III Arena and Unreal Tournament 2003 games were ported to the Linux by the developers themselves, that is why their quality is not questionable. The SPECViewperf comes with the source code which can be compiled for an OS required. The outcome may depend on a compiler, but we hope (and we will check it some day) that this program which simply transfers requests to the OpenGL driver doesn't depend much on the compiler optimization.

The Linux doesn't use the Direct3D. The only API supported here is the OpenGL. There is also the WineX project which can run games in the Linux from the Windows working with the DirectX. We will try to study it next time.

Tests

Although applications often use special optimizations and settings to boost performance, we will use the standard settings where possible.

Let's start with the Quake III. Unfortunately, the format of demo files changes from version to version, that is why we to decided on a relatively simple version four.dm_68 included into the program's build. The game is pretty old, and I used two video modes: standard High Quality and heavy configuration files (with the index H on the diagrams) widely used in our Video Section.

The Linux outpaces the Windows XP in all resolutions and with both variants of the settings! The advantage reaches 25% with the standard settings, and 80%(!) in the heavy mode (at 800x600).

It's interesting that the angle of inclination of the curves in the Windows XP and the Linux is different. Besides, the speed noticeably rises with the transition from 1280x1024 to 1600x1200 in the Linux with the heavy configs. Since all the measurements were taken 5 times, these figures are not doubtful. 

The game looks identical in the Windows XP and in the Linux. That is why if your aim is to get as more fps as possible and you use the NVIDIA's card, the OS upgrade can bring good dividends.

Now the ATI's situation.

Surprisingly, in the Linux the card falls behind by 10-30% depending on a resolution.  This time the angle of inclination of the curves is almost identical.

But remember that new drivers, both for the Linux and Windows, can easily change the results.

Now the Unreal Tournament 2003 DEMO. This game can use both the Direct3D and the OpenGL under the Windows, but only the latter one under the Linux. The factory settings weren't touched. We changed only the resolution. The game comes with its own benchmarks, and we used one of them - flyby-antalus. Note that the video cards' results published on our site were obtained with the same benchmark, but another version, that is why you should compare them very carefully.

Let's start with NVIDIA again.

In this application the Linux looks inferior (as compared to the Windows version of the OpenGL) - it's almost twice as slow in 800x600. Although the gap shrinks to 21% in 1600x1200, the Linux comes out beaten.

In case of ATI, the gap is 38% in 800x600 and shrinks to 2 fps in 1280x1024 and 1600x1200.

The fact that fps is limited by 70-75 for two lower resolutions in the Linux looks strange. But I can't explain it even after running the additional tests. Later we will come back to this problem, but now take it 'as is'.

The last application is SPECViewperf. Although this is a professional test, and the cards used are of the game class, let's try it anyway.

In case of the NVIDIA's card the test ran into a system error in the light-06 subtest. The problem is probably in the video card's drivers because the card worked well with the previous version. Since that was the only problem I left the latest drivers and calculated the scores in the light-06 by averaging two tests out of four.

In the most tests the video card demonstrates equal performance under different OSes, which indicates that the drivers used are based on the common source code.

ATI has the opposite results once again :) - the leaders in the dx-08 and drv-09 have exchanged heir positions. Besides, the double advantage of the Linux in the ugs-03 looks alarming. In the other subtests the Windows XP outscores the other OS by 17-32%.

Conclusion

So, we have looked at the performance of two popular video cards from two leading manufacturers under the Windows XP and Linux OSes. In general, we didn't have a lot of problems in the tests. And we admit that 3D games can be played not only in the Windows, - the Linux (with the respective software support) doesn't fall much behind. The orientation towards the Direct3D API can make it difficult to port games to the Linux platform where the OpenGL is used, but I hope that the example of id Software and Digital Extreme/Epic Games will be catching.

I can't say that one or the other OS shows better performance - the Windows has it better in most cases, but sometimes the Linux takes the lead.

Kirill Kochetkov (kochet@ixbt.com)


Write a comment below. No registration needed!


Article navigation:



blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook


12

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.