MSI MS-5169 VER:2.1, AMD K6-2 350, Riva TNT2 M64, 128/384 MiB PC100 CL3, Windows 95 OSR2

128 MiB of RAM: 2499 3DMarks, 4851 CPU 3DMarks
384 MiB of RAM: 2084 3DMarks, 4346 CPU 3DMarks

Just for fun:
128 MiB of RAM, L1 cache disabled: 96 3DMarks, 148 CPU 3DMarks
128 MiB of RAM, 380 MHz (4× 95): 2662 3DMarks, 5057 CPU 3DMarks

GemCookie wrote on 2024-08-05, 20:31:

MSI MS-5169 VER:2.1, AMD K6-2 350, Riva TNT2 M64, 128/384 MiB PC100 CL3, Windows 95 OSR2

128 MiB of RAM: 2499 3DMarks, 4851 CPU 3DMarks
384 MiB of RAM: 2084 3DMarks, 4346 CPU 3DMarks

This looks like your amount of RAM exceeds the cacheable area of your current mainboard configuration.

Disruptor wrote on 2024-08-05, 21:05:

This looks like your amount of RAM exceeds the cacheable area of your current mainboard configuration.

Indeed; that's why I initially ran the benchmark with 128 MiB of RAM installed.

bloodem wrote on 2023-03-28, 05:19:

Did you apply all of the known patches?
I've attached my 3dmark99 patch pack which I think combines multiple patches that I've found over the years (works on a Ryzen 7 5800X3D with RTX 4070Ti).
Depending on which CPU you have (i.e. if you get 0 - 0.1 FPS on all tests), you might also want to apply mkarcher's NOTC fix

Hi. Where does one find this 3DMark 99 NOTC fix? Google points to your post.

Bruno128 wrote on 2024-09-08, 10:08:

bloodem wrote on 2023-03-28, 05:19:

Did you apply all of the known patches?
I've attached my 3dmark99 patch pack which I think combines multiple patches that I've found over the years (works on a Ryzen 7 5800X3D with RTX 4070Ti).
Depending on which CPU you have (i.e. if you get 0 - 0.1 FPS on all tests), you might also want to apply mkarcher's NOTC fix

Hi. Where does one find this 3DMark 99 NOTC fix? Google points to your post.

https://drive.google.com/file/d/1RBV2-D3E5vUk … ?usp=drive_link

In this package - all patch 3DM99/2000/2001, NOTC too.

Bruno128 wrote on 2024-09-08, 10:08:

Hi. Where does one find this 3DMark 99 NOTC fix? Google points to your post.

The fix is actually called "NOTSC", not "NOTC". This is likely the reason why you don't find it. The original post about that patch is Re: 3dmark99 MegaThread . This patch is only relevant for 486-class CPUs and the Cyrix 6x86. All newer processors include a "time stamp counter" (TSC) that counts the cycles (what exactly that is, especially with dynamic frequency switches varies from generation to generation) since powerup, and 3DMark99 works without problems on this kind of processors.

AntonPC wrote on 2024-05-21, 23:00:

Here are the test results for the ATi Radeon 9250 with 256MB VRAM and a 128bit bus. (without the stock fan). One of my favorite cards for a mid end Windows 98 setup. (although these results are in Windows XP) 😀

ive got exactly the same card, only difference i can see is mines got samsung memory at 200mhz, and a later bios.

AntonPC wrote on 2024-05-01, 23:44:

Following the tests, here are the results for the Geforce FX 5200 64 bit with 128MB DDR VRAM bus from EVGA. I have to say that I found the 2D quality of this card very crisp 😀

🤣 just noticed you've posted a couple of other cards i've just benchmarked

this one is an ASUS V9520 Magic/T - FX5200 128MB 64BIT, this one has the lower 166mhz memory as apposed to the standard 200mhz, i think these types are why the FX5200 gets a bad name, it does take a hit in performance, so those are probably best avoided.

this one is an MSI-8917 FX5200 128MB 64BIT, this is what i use in my dos-98/Xp system. the 200mhz memory makes a difference of 800 points over the 166mhz of the ASUS.

DudeFace wrote on 2024-09-08, 17:29:

ive got exactly the same card, only difference i can see is mines got samsung memory at 200mhz, and a later bios. […]
Show full quote

AntonPC wrote on 2024-05-21, 23:00:

Here are the test results for the ATi Radeon 9250 with 256MB VRAM and a 128bit bus. (without the stock fan). One of my favorite cards for a mid end Windows 98 setup. (although these results are in Windows XP) 😀

ive got exactly the same card, only difference i can see is mines got samsung memory at 200mhz, and a later bios.

The attachment XP Celeron D 360 ATI 9250.jpg is no longer available

The attachment Radeon 9250 256mb 128bit.jpg is no longer available

The attachment Radeon 9250 256mb 128bit 2.jpg is no longer available

Wow! That's a huge jump in performance and that classic ATI headsink looks stylish. I guess my Pentium 4 at 1.4Ghz is holding the GPU back. 😀

This is a LEADTEK GEFORCE 4 MX440 64MB 128BIT, strangely this scored higher than the MSI FX5200, despite having half the pixel shaders and ROP's and zero vertex shaders, it has half the vram of the FX5200 but twice the bus speed, also its agpx4 whereas the FX5200 is agpx8, even stranger still it somehow scored higher than the radeon 9250 which also has twice the pixel shaders and ROP's as well as the addition of 1 vertex shader, it has 4 times the amount of vram and also has a 128bit bus and agpx8 over the MX440's agpx4, it does have an extra 30mhz on the gpu clock over the radeon, but still i expected the radeon to come out on top. someone explain!

AntonPC wrote on 2024-09-08, 18:57:

Wow! That's a huge jump in performance and that classic ATI headsink looks stylish. I guess my Pentium 4 at 1.4Ghz is holding the GPU back. 😀

i usually stick a small fan on top of the heatsink like yours, makes a hell of a difference to the temps and even the score, but yeah seems the cpu does effect the score as well, my motherboard is using AGPx8 as well, also i noticed your memory is 166mhz and mine is 200mhz , if you look at the 2 FX5200's i just posteed one has 166mhz memory and the other has 200mhz memory theres an 800 point difference in the score, also are you using the patched version of 3dmark99 that has the frame limit unlocked? before i was using it i was getting half the score i am now.

DudeFace wrote on 2024-09-08, 19:11:

AntonPC wrote on 2024-09-08, 18:57:

Wow! That's a huge jump in performance and that classic ATI headsink looks stylish. I guess my Pentium 4 at 1.4Ghz is holding the GPU back. 😀

i usually stick a small fan on top of the heatsink like yours, makes a hell of a difference to the temps and even the score, but yeah seems the cpu does effect the score as well, my motherboard is using AGPx8 as well, also i noticed your memory is 166mhz and mine is 200mhz , if you look at the 2 FX5200's i just posteed one has 166mhz memory and the other has 200mhz memory theres an 800 point difference in the score, also are you using the patched version of 3dmark99 that has the frame limit unlocked? before i was using it i was getting half the score i am now.

It's just incredible how they sold such different performing memory chips under the same GPU name and model. Well, those sketchy practices are even used today hehehe.

Regarding the AGPx8 interface, and the chipset itself could explain part of the difference indeed. I will try to run the same tests again on an intel 845 or 865 motherboard next time.

mkarcher wrote on 2024-09-08, 11:57:

This patch is only relevant for 486-class CPUs and the Cyrix 6x86.

And VIA C3 Nehemiah. Thanks it is confirmed to fix 0 fps there!

Bruno128 wrote on 2024-09-09, 07:10:

mkarcher wrote on 2024-09-08, 11:57:

This patch is only relevant for 486-class CPUs and the Cyrix 6x86.

And VIA C3 Nehemiah. Thanks it is confirmed to fix 0 fps there!

What scores do you get now on your C3?

AntonPC wrote on 2024-09-09, 04:03:

It's just incredible how they sold such different performing memory chips under the same GPU name and model. Well, those sketchy practices are even used today hehehe.

i dont think it is down to the memory chips, the asus fx5200 with the 166mhz memory has exactly the same samsung chips (same part number) as the chips on the 200mhz 9250,
i know some manufacturers clock their cards at different speeds, but as ours are the same sapphire versions, the only thing i could put it down to is that your is a slightly earlier revision with an earlier bios, if you flashed yours with the bios from mine it might give you the extra 34mhz, if you're interested i could dump my bios.

I don't really sets remember it, but it starts to adjust, refinement is still needed, but it seems that after a long rest, the config is in the element of the idea.
MSI_AB does not know the video card, the old gpuz is well indicated by the sensors, I set it up to display MAX values.
Prima is the MSI am2+ sheet and the little 565 goes well.

Win98 isntall and little tuning :

DarthSun wrote on 2024-09-23, 17:53:

Win98 isntall and little tuning :

The attachment PHIIW98FI_3DM99.jpg is no longer available

Love that setup! Never had the chance to run a Phenom during it's production time, since my last AMD CPU was a Sempron for socket 754, before going Intel Pentium D 805 and then back to AMD FX 6300. It's nice to see how it's running under W98SE and WXP, which makes me consider a dual boot system now. Thanks! 😀

DudeFace wrote on 2024-09-09, 09:55:

AntonPC wrote on 2024-09-09, 04:03:

It's just incredible how they sold such different performing memory chips under the same GPU name and model. Well, those sketchy practices are even used today hehehe.

i dont think it is down to the memory chips, the asus fx5200 with the 166mhz memory has exactly the same samsung chips (same part number) as the chips on the 200mhz 9250,
i know some manufacturers clock their cards at different speeds, but as ours are the same sapphire versions, the only thing i could put it down to is that your is a slightly earlier revision with an earlier bios, if you flashed yours with the bios from mine it might give you the extra 34mhz, if you're interested i could dump my bios.

Thanks for the offer! I will give it another run as soon as I clear some space in the only corner of the house I have left for projects 😀