Comparison of Frame-rates in
Quake 2 Using Voodoo 3D Cards

Quake II using DEMOMAP DEMO1.DM2

NOTE: Frame-rates for Quake 2 are completely different than for GLQuake. Quake 2 uses a different animation and rendering engine than Quake, so comparing their frame-rates is like comparing apples and oranges. The minimum acceptable frame-rate for solo play in Quake 2 is about 12 fps, and maybe around 15 for netplay. More is certainly better. Quake 2 puts even more strain on a 3D graphics subsystem than GLQuake does.

THE TEST:

The test used is the same one Brian Hook described using in his .plan for his frame-rate tests on a souped up P-II 300 system.

If you would like to perform the exact same test I've performed here, using the same settings so you can compare your system to the following chart, you can download the test by clicking here... Unzip the files into an empty directory and follow the instructions in the included README.TXT file.

USE THESE TEST RESULTS ONLY AS A GUIDE! Please be aware that frame-rate tests are not very precise. It's not unusual for two systems with identical processors and 3D cards to have scores that differ by as much as 8 fps due to differences in motherboard design, BIOS settings, design of the secondary cache algorithm, etc. There are many variables involved! Even the same system can test differently on two different days depending on the temperature inside the case (the 3D card's processors, and even your CPU, are sensitive to temperature - one good reason to invest in a cooling fan for your 3D card). Don't e-mail me pissing and moaning that your P-200 gets 5 fps less than the one tested here, and you want to know how to fix it - DON'T WORRY, BE HAPPY! Again, use these test results only as a guide. Your mileage may (will!) vary.

"BUT WHY SHOULD I TEST MY FRAMERATE WITH THE SOUND DISABLED?"

These test configs set all display variables to defaults and disable sound as recommended by Brian Hook). Why disable sound? In a nutshell, having sound enabled will cap your framerate in the timedemo. A timedemo can run much faster than a normal game (you can't actually play Q2 at the speed that a timedemo runs at, can you now?). It can run so fast that the sound card buffer fills up. Once this happens, the 3D card and CPU must wait until the sound card clears it's buffer before it can render a frame, thus the system is bottlenecked at the sound card.. At that point you're not testing the fps of the 3D video subsystem anymore - instead you're testing the fps of the sound card since the rest of the system is waiting on it.

BEFORE YOU WRITE IN COMPLAINING ABOUT THE WAY I DO MY TESTS...

1. The purpose of the tests I've performed is to make a COMPARISON. The game doesn't matter and Quake2 is as valid a testbed as Q3 or UT. I use Q2 because I already have a VAST database of systems benchmarked running Q2 with EXACTLY the same settings using the Brian Hook standard timedemo test. I don't have time to retest all these systems on every other game out there. I have started to test some systems running Quake3 and I will post the results here as they come in. Don't nag me to test other games as I have more important things to do - like watch porn and play computer games.

2. It's logical to assume that if you see a 25% improvement in framerate going from CPU A to CPU B in Quake2, then you'll likely see the same 25% improvement in other first person shooters. Yes, I know that's simplistic, but we're talking RELATIVE comparisons of different CPUs running 3Dfx hardware. If you want ABSOLUTE numbers get them from some other sites, but unless the tests are done EXACTLY the same way with EXACTLY the same settings, they're meaningless for purposes of comparison.

3. Along the same lines as #2 above, even if you have a different brand of video card, you can use this chart to see POTENTIALLY what kind of performance increase you'll get by upgrading your CPU. Again, if I got a 25% improvement in FPS going from CPU A to CPU B on a V3-3000, you'll probably see ROUGHLY the same percentage of improvement on a G400, a TNT2 or a GeForce using the same processors (yes I know some of these other cards are less CPU dependent than others, but it still applies as a valid comparison guide). I only own 3Dfx hardware (I'm not a 3Dfx advocate, that's just the hardware I've chosen to use - something better comes along that I can afford, I'll buy it).

4. The purpose of my page is to help people decide on a CPU upgrade, NOT to sell you on a 3Dfx card or a particular CPU. I'm trying to answer this question which you constantly see posted in various forms on 3D hardware forums: "Is it worth it for me to upgrade my <blah blah> CPU to a <blah blah> CPU, and how much of difference will I see?"

5. I don't give a rat's behind what framerate you're getting on YOUR system with YOUR video card with YOUR settings. Please don't e-mail me to brag. I fully agree that there may be better hardware out there, this is just the hardware I'm using and have access to. You want to mail me your video card for a few weeks so I can do tests? No problem. I'm more than happy with the framerate I'm getting on my personal system thank you, and bragging is NOT the point of my framerate page (see #4 above). Don't treat me like a deprived child, and don't start any flame wars. I'm trying to provide information here.

TEST 1: System Performance Comparison in Software 2D Mode

This section is for the benefit of you poor deprived folks out there who STILL don't own a 3D card. Quake2 doesn't perform very well in software 2D mode, except on very high-end systems with very fast 2D graphics cards. Even so, there's no way you can get the performance in software mode that you can get in 3D mode. 

So you can better appreciate what a 3D card will do for you, here's some systems tested running in 2D software mode.

All following tests were done with the viewsize set for 100 and the sound handler turned off. All computers are running version 3.14 (or later) of Quake 2.

COMMENTS: I would consider the 30 fps barely acceptable for serious net-play. All the other systems would only get acceptable performance at lower resolutions (400x300 or 320x240).

TEST 2: System Performance Comparison in 3Dfx OpenGL 3D Mode

This is what we're all really here for. This is a comparison of a variety of systems using either Voodoo, Voodoo2 or Voodoo3 video cards in a variety of resolutions.

All following tests were done with the viewsize set for 100 and the sound handler turned off. All computers are running version 3.14 (or later) of Quake 2. Vsync was disabled for best performance and to avoid an artificial framerate cap.

NOTE: The latest drivers for 3Dfx video cards don't allow you to disable VSync. There is no need to disable VSync in normal play, and doing so will give you horrible image tearing. However, with VSync disabled your framerate will never exceed your monitor's vertical refresh frequency in a timedemo test. To perform benchmark testing properly you should temporarily disable VSync for the test. If your driver supports it, disable VSync in the the advanced video properties tab. Otherwise, you can download the V2 or V3 Overclocker utility form my Files & Drivers Page which adds a new display properties tab with VSync and clock adjustments.

Systems running with a Voodoo 1 card are displayed with blue text.

Systems running with a Voodoo 2 card are displayed with red text.

Systems running with a Voodoo 3 card are displayed with green text.

Any of these systems will give you okay performance in solo play (although it's going to bog down sometimes on anything less than a P-133), but realistically you need at least 30 fps to enjoy netplay. Competative netplay requires about 40 fps. I would consider a minimal system for competative netplay to be a P-233 with a Voodoo2 or Voodoo3, and I'd disable dynamic lighting (set gl_dynamic 0) to keep the framerate near 40 fps.

You should also look carefully at the scores for the P2-400 with the single Voodoo2 card as opposed to the scores for the P2-400 with dual cards running in SLI mode. The single card system slightly outperformed the dual card system at 512x384 because it's CPU is over-clocked to 448 MHz. However, it lags far behind the dual card system at 800x600 resolution. This clearly demonstrates the improved performance you get when using dual-card SLI on a high performance system - the framerates on the dual card system remain practically the the same at all resolutions (although there's nothing shabby about the 64 fps the single card system got at 800x600, I might point out).

Many of you have expressed surprise (or dismay, as the case may be) at the showing of the Cyrix PR-200+ system. However, this performance is not surprising considering the fact that the Cyrix CPU has a very weak floating-point unit compared to Intel processors and even compared to the K6. However, at 15.3 the Cyrix is at least (barely) playable in solo mode, although the framerate is not acceptable for serious deathmatch play. And before you ask, I wouldn't expect the score to be any higher if you put a Voodoo2 in the Cyrix machine - you can't make a cow run like a gazelle...

VOODOO 2: You might be surprised that Voodoo2 cards performed so poorly on low-end systems, but this is to be expected. The Voodoo 2 only really shines when it's running on a high-end Pentium II or better system. However, it is worth noting that even on a low-end system, the Voodoo 2 produced excellent framerates at all resolutions on up to 800x600. Also, while it's not reflected in these scores, the Voodoo 2 plays much smoother on low-end systems than the Voodoo cards do. This is mostly thanks to the Voodoo 2's dual texture units and larger texture memory, which results in a reduction in frames dropped due to texture thrashing. That's worth the price of a Voodoo2 alone - it sucks getting fragged all the time because your system slows to a crawl when there's a lot of Hyberblaster fire going on. It's worth noting here that on a Voodoo2 you probably won't be playing Quake 2 at 512x384, as  the Voodoo 2 is capable of maintaining a high frame rate while running at 640x480 and 800x600 (see Test 2 below for the results of Voodoo2 tests on low-end systems at various resolutions) - and oh boy does it look pretty at those high resolutions, childrens....

VOODOO 3: Notice that the Voodoo3 essentially performs on par with Voodoo2 SLI. However, image quality is much better on the V3. Overclocking the V3 generally didn't make much of a difference until the Pentium 3 came along - especially the Coppermine CPUs. Check out the overclocked V3 results for the P3-550E. The Coppermine CPU itself also overclocks easily (too easily!), and most people are doing just that. The P3-550E results posted above are from my own personal system. This system easily overclocked to a fully stable 666 MHz using PC100 SDRAM. It ran stable enough at 733 MHz to run benchmarks, but it would occasionally lockup. It's likely that with PC133 SDRAM it would run fine at 733. I've seen this processor get as high as 800 MHz on some systems.

Compiled by Robert Osorio, "Flying Penguin (Mercenary)"
Last updated on 3/22/00