You're running XP SP3 (with only 256MB of memory at that) and comparing to benchmarks that were likely conducted on machines running 98SE or, at worst, Windows 2000. You're also using a fairly inconsistent metric for comparison - Quake 3 doesn't have an explicit built-in benchmark like say, Hitman Absolution, so you're comparing timedemo performance, potentially between different timedemos. You also have to remember that review benchmark systems are often fairly stripped-down compared to normal usage machines - they tend not to have things like anti-virus suites, lots of background services, etc enabled as reviewers are going after the most consistent numbers between test-runs. You're also comparing PCI cards to tests that are using AGP cards, which will have an impact on performance as well.

Basically, you're wrestling with multiple factors, some of them beyond your control. The most significant things you could do would be to dump XP, and remove background "junk" (you don't really need anti-virus, firewall, etc since a machine like this really has no business on the web). You will never match the performance of an AGP system with this HP, but you should be able to improve from where you are.

Finally, you never mentioned your hard-drive situation - that could also be a bottleneck.

also its seems you are using an i810 wich performs worse than an intel bx440 (chipset in the test)

on top of that all the pcs that arent clones usually performs below a same spec clone...

Yup, good points.

Luckily in reviews they usually mention the driver version and patch version of the game. That helps quite a bit. Best is to use drivers and software from that era. Later drivers focus on newer games.

For Geforce 2/4 MX I've heard there is quite a performance drop with the Forcewares, and that you should stick to 4x.x Detonators.

alexanrs wrote:

For Geforce 2/4 MX I've heard there is quite a performance drop with the Forcewares, and that you should stick to 4x.x Detonators.

Some GeForce 4 MX series cards can't use such old drivers (e.g. MX 4000, which is newer than many GeForce FX boards), but the 2 MX series (like any other GF2) can actually use 3x.xx drivers, which are usually suggested as being best for compatibility with older games in other threads.

Geforce2MX really sucks with forceware 8x, even if with 440BX. And being PCI you are probably looking at tnt2 performance.

As an addition to Quake III, the game was published almost the same date that the Celeron 533MHz was launched. I don't remember ever top performance for a game to be achievable using a budget CPU from the same era.

A few clarrifications:

• The XP OS has been fully optimized. There is no anti-virus running and system services have been stripped down. Its booting up at 70mb RAM usage on SP3, so its pretty lean.
• The Q3 version is the latest patched version of the game
• I have tested out the Forceware and Detonator Drivers. Currently installed are Detonator 28.32, which I have tested against the GeForce 2MX (have yet to test the GeForce 4 against the Detonators to see what happens).
  http://www.nvidia.com/object/winxp-2k_archive.htm
  http://download.nvidia.com/Windows/28.32/28.32_winxp.exe
• I am averaging 5 timedemo tests to get a more balanced framerate result for the timedemo on each config.
• The Harddrive is a 7200 RPM 320GB Blue IDE Drive (Model #: WD3200AAJB)

Interesting point about the i810 chipset!

What impact on the fps would windows 98 have vs winxp? Remember I was also testing out Quake 2 and it was stuck at 50fps max with vsync on, with a refresh rate of 75hz, 640x480, 16bit, OpenGL.

I also came across this link, which spefically tested a GeForce 4 MX 420 PCI card in a shuttle system. The CPU is an AMD Athlon XP 2000+ 1.67GHz, but at least its a benchmark for the PCI version of this card which is hard data to find these days:
http://techreport.com/review/3813/shuttle-ss4 … ebones-system/7

I'd still drop to either an earlier iteration of XP, like SP1, or to 98SE. It will likely improve the system's overall performance (and 70MB is not much less than my completely untouched XP Pro SP2 SMP machine uses; by contrast that machine running Win2k Pro with no optimizations is using like 50-80MB, and 98SE can be even lighter than that). Performance wise you're likely also hitting a wall with the PCI cards and very slow CPU - you're comparing either to significantly faster CPUs (like AthlonXP 2000+, which is faster than basically any Pentium 3, some early Pentium 4s, etc) or to machines with similar CPUs, but AGP graphics and other improvement features. You're also probably getting a more comprehensive FPS average than what some reviews are showing you - I know Demo001 was a popular choice for Quake 3 benchmarks years ago, and if you're instead averaging all of them together that will produce different results.

In the plus column, the hard-drive shouldn't be a problem. 😀

The latest patch for Quake III replaces the two demos that the originally came with. Instead, it replaces them with a single demo, DEMO004. This is a far more demanding benchmark than the previous two demos.

Reviewers probably didn't use Demo004 at the time. Try reinstalling Quake and using an earlier patch, I typically use 1.16 IIRC!

Windows XP SP1 or earlier, installed without ACPI will yield a noticable improvement in framerate as well. I used to do comparisons back before I switched to SP3 because everything was so fast it didn't matter. I wish I had an SP1 disc, still so I could use it for this purpose... I'll have to look around.

r_swapInterval 0

Thanks all for the tips and info.

I did some more testing, and even threw some more graphics cards at it. The CPU is definitely bottlenecking the cards. I am going to stick with the GeForce cards I have, specifically the GF4 MX 420, as I can run older drivers. Any of the ATI driver bundles I have run across are too modern and use up a lot more system RAM.

It turns out, that I had GL Extensions: On. Setting this to off gains 7-8 frames, bringing the average framerate for demo004 to 47fps. This is much better, and overall the experience is pretty pleasant.

Later on down the road I will try out XP SP1 or 98SE to observe more improvements on this rig. At the moment this machine will be SP3, as its supposed to be a little companion XP PC to a Pentium 4 for some LAN gaming.

Turning gl extensions off is bad and should lead to worse performance. The option is there incase some cards' drivers implement some extensions improperly (most 1996-1998 hardware). You should have that on

You should also try changing r_primitives to 1 or 2. 2 is usually better on ATI. Setting it t0 0 autodetects, and sometimes it assumes wrong.

For whatever reason turning GL Extensions off on this setup lead to improved performance. I ran 5 benchmarks all coming out to 46.5-47.5 fps versus the 39fps average I was getting with GL Extensions turned on. I tested this with Forceware and Detonator Drivers. The result was also observed with ATI cards as well. On any other machine, such as my Pentium 4 for instance, I will leave GL Extensions set to On. However on this 533mhz turning it off yeilded a huge performance gain.

I will experiment with r_primitives along with other custom configuration options as well.

Just FYI, the GL extensions Q3A uses are all in the interest of raising performance than increasing graphical quality. By disabling GL extensions, you'll disable compiled vertex arrays, texture compression, and multitexturing.