I had a fair share of 3DMark2000 random crashes even in Win2k, with 3DMark 2001 working stable. This has something to do with SSE optimized rendering implementation.

The Serpent Rider wrote on 2025-11-28, 09:57:

I had a fair share of 3DMark2000 random crashes even in Win2k, with 3DMark 2001 working stable. This has something to do with SSE optimized rendering implementation.

Really? With a Thunderbird, it will also crash after 24 hours... thunderbird is 3DNow and MMX only.

I don't know that I would hinge the stability of a build on one program, regardless of whatever program that is.

Jasin Natael wrote on 2025-11-29, 14:35:

I don't know that I would hinge the stability of a build on one program, regardless of whatever program that is.

Perhaps, but the crashing has got me thinking:

1) Is 3DMark2000 crashing, or is Windows 9x crashing?
2) If the latter, is Win9x suitable for testing stability at all?

Only way to find out is to test with Windows 2000.

But I'm not going to because @The Serpent Rider is probably correct. Mind you, later VIA 4-in-1 4.56 did vastly improve things in this regard when I was using nVidia 45.23 drivers. But I'll just stick with Thunderbird and VIA 4.35 drivers, I think eliminating SSE altogether from the equation is the safer bet.

Maybe. 98 isn't always the most stable OS out there. 2K should be more stable. Can't comment too much on the Athlon build, I've got one 462 build but it is an SiS chipset.
I had my Athlon 1.4 in that board before and it seemed quite stable and very fast. But it ran crazy hot. I've got some later XP chips I should probably test them sometime.

No one wants to try and beat my rather middling score?
I'm kind of curious what a C3 with a faster than 133 bus speed might do.

Jasin Natael wrote on 2025-12-17, 16:47:

No one wants to try and beat my rather middling score?
I'm kind of curious what a C3 with a faster than 133 bus speed might do.

I've just done some testing with a Nehemiah on an i815 board (Gigabyte GA-6OXE) and have concluded that you've found more or less the ideal configuration for running 3DMark 2000 with this CPU. I don't think memory bandwidth is a hugely limiting factor for this CPU at FSB133, and as has often been demonstrated, it's difficult to beat an overclocked 440BX setup within this generation. My best results were as follows:

• FSB 133 x 11.5 (1533MHz) with GF3 Ti200, driver version 7.76 : 6601 3D marks
• FSB 150 x 10.0 (1500MHz) with GF3 Ti200, driver version 7.76 : 6783 3D marks
• FSB 147 x 10.5 (1542MHz) with GF3 Ti200, driver version 7.76: 6805 3D marks - this was the best result I could get without overclocking the GPU, and was the furthest I could push this CPU. I couldn't complete a 3DMark run at 148x10.5.

The theoretical performance gain from later GPU generations is more than cancelled out by the additional CPU overhead from the later drivers:

• FSB 147 x 10.5 (1542MHz) with GF4 Ti4200 8x 128Mb, driver version 41.09: 6634 3D marks
• FSB 147 x 10.5 (1542MHz) with GF FX5900XT, driver version 56.64: 6379 3D marks

Your score is likely beatable with a faster GF3 or a first-gen GF4 Ti on earlier drivers, but I don't think the Nehemiah CPU itself has much more to give with stock cooling and voltages.

Screenshots of my results below.

@sfryers - Awesome! Thanks for taking the time to run these tests.
I haven't tried my chip on a 815 board as the only one I have(Intel D815eea )doesn't support it.
Good to know that I am up against the performance wall, I do have a disputably "faster" Radeon 8500(actually a FireGL 8800), that I could throw in and see if it is any faster.
It scores higher than the Geforce3 in my Tualatin machine.
I could also overclock the Ti200 I suppose, I know that it will sustain vanilla Geforce 3 clocks.
The only other faster AGP cards I have are Geforce 6 stuff with 98 driver issues.

I only have old 3DMark2001 results, at 1266 MHz max.

Tualatin Celeron vs Williamette Celeron

3DMark2001SE build 330 - Full run at default settings 1024x768x32 NoAA:

VIA C3 Nehemiah+ ( 6.9.8 )
_7.5 x 100 = _750 MHz: 3272 3DMarks
_9.5 x 100 = _950 MHz: 3828 3DMarks
_9.0 x 133 = 1200 MHz: 4814 3DMarks (default setting)
_9.5 x 133 = 1266 MHz: 4857 3DMarks (had to raise the voltage to 1.6V)

Test System
Motherboard: GA-6BXC rev 2; i440BX chipset; AGP divider 2/3; Upgradeware BIOS. Either Slot-T or MS6905 Slotket.
Memory: 128MB SDRAM; a single DIMM at CL3.
Disk Drive: SD-Card 8GB on IDE adapter; southbridge IDE controller.
Operating System: Windows 98 SE, clean install; DirectX 9.0c.
Graphics Card: Asus V8420/TD Geforce4 Ti4200 AGP; 128MB DDR; Clocks 250/550; Driver v30.82 at default settings.

gerwin wrote on 2026-01-05, 21:18:

I only have old 3DMark2001 results, at 1266 MHz max. […]
Show full quote

I only have old 3DMark2001 results, at 1266 MHz max.

Tualatin Celeron vs Williamette Celeron

3DMark2001SE build 330 - Full run at default settings 1024x768x32 NoAA:

VIA C3 Nehemiah+ ( 6.9.8 )
_7.5 x 100 = _750 MHz: 3272 3DMarks
_9.5 x 100 = _950 MHz: 3828 3DMarks
_9.0 x 133 = 1200 MHz: 4814 3DMarks (default setting)
_9.5 x 133 = 1266 MHz: 4857 3DMarks (had to raise the voltage to 1.6V)

Test System
Motherboard: GA-6BXC rev 2; i440BX chipset; AGP divider 2/3; Upgradeware BIOS. Either Slot-T or MS6905 Slotket.
Memory: 128MB SDRAM; a single DIMM at CL3.
Disk Drive: SD-Card 8GB on IDE adapter; southbridge IDE controller.
Operating System: Windows 98 SE, clean install; DirectX 9.0c.
Graphics Card: Asus V8420/TD Geforce4 Ti4200 AGP; 128MB DDR; Clocks 250/550; Driver v30.82 at default settings.

Awesome, sounds like a pretty tightly dial rig.
I will try and find the time to run some 3dMark 2k1 tests, I know that I did some testing but I can't remember any of the results.

Edit: Ran some tests....
Ok this is wild...I installed 3DMark 2K1 (and Dx 8.1)

All defaults 1024x768X32, No AA
At 1500mhz 100x15=2397
At 1466 133x11=2059
At 1533 133x11.5=2103

Everything thing else is the same.
I know that the GF4 ti4200 is faster than a GF3 ti200....but is it THAT much faster?!

Edit Revisited:
I'm down a couple thousand points in 3DMark 2K as well, so something is going on. Not sure what it could be I haven't changed anything on the build that I can recall. Other than installing Dx 8.1.

Ok, well I'm a dummy.
I figured out why the 3DMark 2K results were off. I had stupidly forgot that I couldn't just run my script and OC my CPU without also using CPUFSB to overclock the bus as well.
Once I did that my performance was of course back to normal.

That said I ran the 3DMark 2K1 tests and still barely cracked 2100 points....any ideas?

Jasin Natael wrote on 2026-01-08, 14:03:

Ok, well I'm a dummy. I figured out why the 3DMark 2K results were off. I had stupidly forgot that I couldn't just run my scrip […]
Show full quote

Ok, well I'm a dummy.
I figured out why the 3DMark 2K results were off. I had stupidly forgot that I couldn't just run my script and OC my CPU without also using CPUFSB to overclock the bus as well.
Once I did that my performance was of course back to normal.

That said I ran the 3DMark 2K1 tests and still barely cracked 2100 points....any ideas?

Thanks for trying. But no idea.what is wrong there, or if that is just the norm for Geforce 3.. I don't own a Geforce 3 to test. Just this Geforce 4 and below that a variety of MX 440's + Voodoo 3's.
I should check my 3DMark2001 benchmark list at home, or look up some results on the internet. Even if these are Pentium III or Athlon benchmarks, they should give some indication of what to expect.

Edit, for example: Visiontek Geforce 3 on Athlon "T-Bird" 1 GHz https://www.pcstats.com/articles/726/7.html -> 5600 3DMarks2001
Geforce 3 Ti-200 on Athlon "T-Bird" 1,4 GHz https://www.computerbase.de/artikel/grafikkar … test.49/seite-6 -> 6022 3DMarks2001