VOGONS


Reply 20 of 40, by cydvog

User metadata
Rank Newbie
Rank
Newbie
Disruptor wrote on 2026-02-27, 08:12:
Please don't compare Covington with Katmai Pentium III. The first one does not have L2 cache at all. The latter one at has L2 ca […]
Show full quote
st31276a wrote on 2026-02-27, 07:54:

A geforce 4 mx440 is solidly a pentium4 era graphics card.

"Terrible" performance on a Covington / Katmai is expected.

Please don't compare Covington with Katmai Pentium III.
The first one does not have L2 cache at all.
The latter one at has L2 cache and introduced SSE instructions and a faster FSB.
There should be a performance difference of about 4 times and with SSE optimized code even more.
However, Katmai power still may be too low for the GeForce 4 MX.

thanks for all very informative answsers. so should i switch to another gpu ? maybe like ( fx5200 or riva tnt2 m4? thank.

Reply 21 of 40, by Garrett W

User metadata
Rank Oldbie
Rank
Oldbie

That depends. What games are you trying to play? You might find that even after swapping a faster GPU in your system, the rest of the components (CPU and RAM mainly) may be still bottlenecking the GPU.

Reply 22 of 40, by cydvog

User metadata
Rank Newbie
Rank
Newbie
Garrett W wrote on 2026-02-27, 12:57:

That depends. What games are you trying to play? You might find that even after swapping a faster GPU in your system, the rest of the components (CPU and RAM mainly) may be still bottlenecking the GPU.

mainly Games like motocross madnes 1 , midtown madness 1 . so im now using Pentium 3 500 mhz , and geforce 4 mx 440 ( 64 mb , asus v9180se rev 1.00 ) 64 mb sdram pc100.

and performance of system not good. slow even games period 97-99.

Reply 23 of 40, by Garrett W

User metadata
Rank Oldbie
Rank
Oldbie

Motocross Madness 1 & Midtown Madness 1 should be running okay on your system and they are definitely lighter games than NOLF that you quoted in a previous post. Perhaps you are facing a GPU driver issue?
Unfortunately your video card is not an MX 440, but rather an MX440-8X, which is slightly newer, so the oldest driver you should be able to use under Win9x is 31.40, can you give that a try?

Reply 24 of 40, by Jasin Natael

User metadata
Rank Oldbie
Rank
Oldbie

Someone might have already mentioned it but are we sure that the MX440 is a 128bit card?
I have one of the Micron Dell OEM 440-8X cards and it's a 64bit DDR card. Ram can overclock to the moon but it's still slower than any of the 128bit cards.
I'm sure that later driver is also contributing as well.

Reply 25 of 40, by Garrett W

User metadata
Rank Oldbie
Rank
Oldbie

Yes, this has been discussed and it's a 64bit card, so performance is likely slightly slower than a GeForce 2 MX due to the lower bandwidth. Still should be faster than any TNT2, so it's a bit weird that performance in the two games OP mentioned just above is not at least decent. Unless OP is expecting 60fps at all times, which is unrealistic. A description or video of the issue could help us troubleshoot better, for example OP may just be experiencing the effects of double buffered Vsync tanking his framerate.

Reply 26 of 40, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t
Garrett W wrote on 2026-02-27, 19:07:

Yes, this has been discussed and it's a 64bit card, so performance is likely slightly slower than a GeForce 2 MX due to the lower bandwidth.

Unless the 64bit Geforce4MX is running SDR or very low clocked DDR, it should not have lower bandwidth than a Geforce 2 MX (which basically always have SDR). Geforce4 cards normally use DDR that is clocked the same or higher than the SDR that a Geforce2 MX uses, so at worst the bandwidth should be the same unless the GF4MX has been saddled with exceptionally slow DDR. A quick search online shows some ASUS v9180SE Rev.1.00 cards with 5ns chips rated at 200Mhz (400Mhz DDR) and others with 7.5ns chips rated for 133Mhz, so it could be anything in between.

A Geforce 4MX should also have a much higher core clock compared to the Geforce2 MX, in addition to architectural improvements that improve efficiency.

I know these types of videos aren't great tests, but this seems like a decent example:
https://www.youtube.com/watch?v=O9GvI7QUL9M
Notice that the EVGA card is sometimes slower because it has very slow 133Mhz (266 DDR) 64bit memory, compared to the GF2 MX 400's 166Mhz 128bit SDR. The others are basically always faster though, and sometimes the difference is huge.

Now for some blitting from the back buffer.

Reply 27 of 40, by st31276a

User metadata
Rank Member
Rank
Member
Disruptor wrote on 2026-02-27, 08:12:
Please don't compare Covington with Katmai Pentium III. The first one does not have L2 cache at all. The latter one at has L2 ca […]
Show full quote
st31276a wrote on 2026-02-27, 07:54:

A geforce 4 mx440 is solidly a pentium4 era graphics card.

"Terrible" performance on a Covington / Katmai is expected.

Please don't compare Covington with Katmai Pentium III.
The first one does not have L2 cache at all.
The latter one at has L2 cache and introduced SSE instructions and a faster FSB.
There should be a performance difference of about 4 times and with SSE optimized code even more.
However, Katmai power still may be too low for the GeForce 4 MX.

Did I compare them?

Both were tried and found to be too light as far as I can tell.

The reason being, both are the bottlenecks and not the card.

A fast coppermine or tualatin might turn the table around a bit in some cases.

However, I think max geforce 2 in a p3 system, given what was available in the era.

Unless the performance bottleneck is not offending you, ofc. In that case, just appreciate the cpu performance as you play, as that would be the limiting factor. Installing a graphics card that is the limiting factor would be objectively slower overall.

Reply 28 of 40, by Garrett W

User metadata
Rank Oldbie
Rank
Oldbie
Ozzuneoj wrote on 2026-02-27, 19:31:
Unless the 64bit Geforce4MX is running SDR or very low clocked DDR, it should not have lower bandwidth than a Geforce 2 MX (whic […]
Show full quote
Garrett W wrote on 2026-02-27, 19:07:

Yes, this has been discussed and it's a 64bit card, so performance is likely slightly slower than a GeForce 2 MX due to the lower bandwidth.

Unless the 64bit Geforce4MX is running SDR or very low clocked DDR, it should not have lower bandwidth than a Geforce 2 MX (which basically always have SDR). Geforce4 cards normally use DDR that is clocked the same or higher than the SDR that a Geforce2 MX uses, so at worst the bandwidth should be the same unless the GF4MX has been saddled with exceptionally slow DDR. A quick search online shows some ASUS v9180SE Rev.1.00 cards with 5ns chips rated at 200Mhz (400Mhz DDR) and others with 7.5ns chips rated for 133Mhz, so it could be anything in between.

A Geforce 4MX should also have a much higher core clock compared to the Geforce2 MX, in addition to architectural improvements that improve efficiency.

I know these types of videos aren't great tests, but this seems like a decent example:
https://www.youtube.com/watch?v=O9GvI7QUL9M
Notice that the EVGA card is sometimes slower because it has very slow 133Mhz (266 DDR) 64bit memory, compared to the GF2 MX 400's 166Mhz 128bit SDR. The others are basically always faster though, and sometimes the difference is huge.

Hey nice catch!
The information I saw online placed the OP's card in line with that EVGA card, so 133MHz memory. But really cool to see that later titles actually pull ahead even with such a low memory bandwidth, likely due to the much higher core clock. Even older titles are not too far off.

Reply 29 of 40, by cydvog

User metadata
Rank Newbie
Rank
Newbie
Garrett W wrote on 2026-02-28, 09:15:
Ozzuneoj wrote on 2026-02-27, 19:31:
Unless the 64bit Geforce4MX is running SDR or very low clocked DDR, it should not have lower bandwidth than a Geforce 2 MX (whic […]
Show full quote
Garrett W wrote on 2026-02-27, 19:07:

Yes, this has been discussed and it's a 64bit card, so performance is likely slightly slower than a GeForce 2 MX due to the lower bandwidth.

Unless the 64bit Geforce4MX is running SDR or very low clocked DDR, it should not have lower bandwidth than a Geforce 2 MX (which basically always have SDR). Geforce4 cards normally use DDR that is clocked the same or higher than the SDR that a Geforce2 MX uses, so at worst the bandwidth should be the same unless the GF4MX has been saddled with exceptionally slow DDR. A quick search online shows some ASUS v9180SE Rev.1.00 cards with 5ns chips rated at 200Mhz (400Mhz DDR) and others with 7.5ns chips rated for 133Mhz, so it could be anything in between.

A Geforce 4MX should also have a much higher core clock compared to the Geforce2 MX, in addition to architectural improvements that improve efficiency.

I know these types of videos aren't great tests, but this seems like a decent example:
https://www.youtube.com/watch?v=O9GvI7QUL9M
Notice that the EVGA card is sometimes slower because it has very slow 133Mhz (266 DDR) 64bit memory, compared to the GF2 MX 400's 166Mhz 128bit SDR. The others are basically always faster though, and sometimes the difference is huge.

Hey nice catch!
The information I saw online placed the OP's card in line with that EVGA card, so 133MHz memory. But really cool to see that later titles actually pull ahead even with such a low memory bandwidth, likely due to the much higher core clock. Even older titles are not too far off.

so what is you suggest to me ? 😀 upgrading ram ? cpu or gpu ? thanks.

Reply 30 of 40, by cydvog

User metadata
Rank Newbie
Rank
Newbie
Thandor wrote on 2026-02-19, 17:58:
The GeForce 4 MX is available in all shapes and forms and some of them are really, really slow. According to the specs you’ve li […]
Show full quote

The GeForce 4 MX is available in all shapes and forms and some of them are really, really slow. According to the specs you’ve linked to your card is one of those slow versions. Based on that I wouldn’t expect performance, especially with a Celeron 300 (without L2 cache!)

Perhaps you can compare with benchmarks and results online to see if your performance matches up.

Of course make sure you have installed chipset drivers, preferably use vanilla Windows 98SE and don’t install too much programs that run in the background (in other words; don’t install crap 😜).

Using CPUMark 99 I scored 14,5 points with a Celeron 300 ‘Covington’ SL2YP.

my cpu score is 33.7 with a pentium 3 500 mhz . is it good ?

Reply 31 of 40, by feda

User metadata
Rank Member
Rank
Member

Your whole system is weak for 3d games. Upgrade everything.

Reply 33 of 40, by Repo Man11

User metadata
Rank l33t
Rank
l33t

I would suggest running some 3d benchmarking programs such as 3D Mark 99 and 2000; these would give both you and people reading this thread an idea of where your system is at, and if your system is performing as well as it ought to or if it is being held back by drivers or misconfiguration of jumpers etc.

For comparison, here is my Tyan S1854 with a Quadro 2 Pro (close to a Geforce 2 Ultra), and the same system with a coppermine P3 1000.

After watching many YouTube videos about older computer hardware, YouTube began recommending videos about trains - are they trying to tell me something?

Reply 34 of 40, by cydvog

User metadata
Rank Newbie
Rank
Newbie

in manual from retroweb it says it supports pentium 2 processors. so is it possible not support pentium 3's or even celerons ?

Reply 35 of 40, by devius

User metadata
Rank Oldbie
Rank
Oldbie

No. If it boots, it supports those CPUs.

We need numbers like RepoMan mentioned otherwise no one will be able to help you properly.

Reply 36 of 40, by cydvog

User metadata
Rank Newbie
Rank
Newbie
devius wrote on 2026-03-01, 17:19:

No. If it boots, it supports those CPUs.

We need numbers like RepoMan mentioned otherwise no one will be able to help you properly.

here some benchmark results. driver version 31.40. with pentium 3 500 mhz , 64 mb pc 100 sdram , and geforce 4 mx 440 - 8x , windows 98 se.

Reply 37 of 40, by Repo Man11

User metadata
Rank l33t
Rank
l33t

CPU/z is a good utility for making sure everything is configured correctly. For instance, boards such as this could be set to run the memory speed asynchronously in order to run slower memory such as PC66 (memory was expensive!) at its rated speed - if you have PC100 memory, you want to make sure the memory is set to run at 100 MHz.

After watching many YouTube videos about older computer hardware, YouTube began recommending videos about trains - are they trying to tell me something?

Reply 38 of 40, by sfryers

User metadata
Rank Member
Rank
Member
cydvog wrote on 2026-03-01, 19:09:

here some benchmark results. driver version 31.40. with pentium 3 500 mhz , 64 mb pc 100 sdram , and geforce 4 mx 440 - 8x , windows 98 se.

Your 3DMark 2000 score does seem quite low for that CPU/GPU combo. Looking at my records, I got a score of 3339 with a P3-500 and a 64-bit GF4 MX440 on an Intel 440BX-based motherboard.

VIA Apollo chipsets are known to be slower than intel ones, but I don't think that's the entire bottleneck. 64Mb RAM is really the minimum you want on a Pentium 3 gaming system- my suggestion would be to try again with 128Mb or more installed.

MT-32 Editor- a timbre editor and patch librarian for Roland MT-32 compatible devices: https://github.com/sfryers/MT32Editor

Reply 39 of 40, by Repo Man11

User metadata
Rank l33t
Rank
l33t

Looking through the motherboard's manual, it appears there is no way to allow memory interleave in the CMOS settings. This setting can cause instability with some memory, but usually it works fine and it does provide a small boost to memory bandwidth. Luckily, a fellow by the name of George Breese wrote a software patch that enables it. It's free and worth a try.

After watching many YouTube videos about older computer hardware, YouTube began recommending videos about trains - are they trying to tell me something?