Doom 3 v. Quake 4 Performance

Written by Michael Larabel in Linux Gaming on 22 October 2005 at 01:00 PM EDT. Page 1 of 3. Add A Comment.

With Quake 4 being available for retail sales here in the United States for a couple of days now, and even the Linux binaries, we've already posted over 120 in-game single player screenshots here and on release date posted a slight preview into the world of Quake 4, but now is our first time delivering frame-rate performance results from inside Quake 4. Making this more than the usual performance testing, we've compared our Quake 4 results against that of Doom 3. As Quake 4 is powered by a tweaked version of id's Doom 3 engine, it will be interesting to see the difference in frame-rate, if any, between these two popular games.

For testing today, we relied on our AW8, 7800GTX, Pentium D system we have used extensively in recent weeks for our NVIDIA Q4-2005 Driver Performance comparison among other articles. Although the Abit AW8 may lack the bells and whistles of other i955X-based motherboards that include additional onboard controllers, we have found the AW8 to be a terrific performer under Linux. Even though Quake 4 isn't able to largely utilize SMP (Symmetric Multi-Processing) we went ahead and used an Intel Pentium D processor as the times are coming to multi-core computing and we also went ahead to overclock the processor marginally to ensure our Doom 3 and Quake 4 tests wouldn't be as much CPU bound. The memory used on the system was 1GB of Crucial Ballistix that was running at DDR2-800 speeds. Finally, the graphics card used in this testing system was a Leadtek PX7800GTX 256MB, with the G70 presently being the fastest NVIDIA GPU. For the software side of things, we performed all of our testing with the Quake 4 v1.0.2147 GNU/Linux binaries and Doom 3 using version 1.3.1302; both of which are the latest releases at the time of testing and available from id Software's FTP server. Onto the drivers, we are using NVIDIA's latest 1.0-7676 proprietary drivers, which are the last in the 7XXX series. id Software's minimum system requirements for Linux Quake 4, which are similar to that of the Windows counterpart, is a Intel Pentium 4 2.0GHz or AMD Athlon XP 2000+, 512MB DDR RAM, 2.4/2.6 kernel, glibc 2.2.4, ALSA 1.0.6, and SDL 1.2. Of course for optimal Quake 4 performance under Linux you'll probably want to stick with a NVIDIA graphics cards, due to rather poor ATI drivers, and for the GeForce series the 3/Ti, 4/Ti, FX, 6, and 7 are supported by the game engine but for best-results you'll need to stick to the 6/7 series.

Hardware Components
Processor: Intel Pentium D 820 (2.80GHz) @ 3.35GHz
Motherboard: Abit AW8
Memory: 2 x 512MB Crucial Ballistix DDR2-800
Graphics Card: Leadtek PX7800GTX 256MB
Hard Drives: Western Digital 160GB SATA2
Add-On Devices: Creative Labys Audigy 2
Power Supply: Enermax Whisper II 535W SLI
Software Components
Operating System: FedoraCore4 (Stentz)
Linux Kernel: 2.6.12-1.1398
GCC (GNU Compiler): 4.0.0
Graphics Driver: NVIDIA 1.0-7676
Xorg: 6.8.2

Due to the actual in-game differences between Quake 4 and Doom 3, we put fourth our best effort into creating a similar benchmarking demo environment between the two games. For all of the Doom 3 testing, we used the standard demo1 that ships with the actual game and is used in hundreds of benchmarks here at Phoronix. Attempting to minimize the visual differences between Doom 3 and Quake 4, while sticking to their stock maps, we spent a great deal of time determining a single-player area to record our in-house demo. Taking the lighting, enemy combat, size of map, object distances, and overall environment into consideration, among other things, we found a particular portion of the single-player Purification Center level to be most like the Doom 3 demo1. Below are some screenshots from the Purification Center time demo we recorded. Although our Quake 4 demo isn't 100% comparable to that of Doom 3's demo1, it should offer a good representation of the frame-rate difference between the two games in general.


Of course, during testing all system settings in both games were kept identical to ensure accuracy and in each situation, we ran each benchmark twice and recorded the second number to allow the first trial run to cache all of the various details, etc... When changing the AntiAliasing and Anisotropic Filtering, these settings were specified over-rode through the nvidia-settings panel. For the various benchmarking scenarios today, we used 1024 x 768 - Low Quality, 1024 x 768 - Medium Quality. 1024 x 768 - High Quality, 1024 x 768 - Ultra Quality, 1280 x 1024 - High Quality - 2x Quincunx AA/2x AF, 1280 x 1024 - High Quality - 4x AA, 9-tap Gaussian/4x AF, 1280 x 1024 - High Quality - 4x Bilinear AA/8x AF, 1280 x 1024 - High Quality - 8xS AA/8x AF, and 1280 x 1024 - High Quality - 16x AA/8x AF.


Related Articles