Skyscraper wrote:

The Socket-478 P4 EE "Gallatin" 3.4 running in a motherboard with dual channel memory pushes the Geforce 6800 "Ultra" pretty hard, this is somewhat GPU limited.

not really, i got over 33000pts with 6800ultra agp, and i believe the pcie version can be even faster.
just think, even a ti4600 can score over 17000pts, so what makes you think that the 6800ultra is worthy of only 24000pts?

Standard Def Steve wrote:

Skyscraper wrote:

The Socket-478 P4 EE "Gallatin" 3.4 running in a motherboard with dual channel memory pushes the Geforce 6800 "Ultra" pretty hard, this is somewhat GPU limited.

I don't think that's GPU limited. A stock-clocked Radeon 9800 Pro can score 22K with a 2.64GHz A64 behind it, so I'd imagine a 6800 Ultra could easily hit ~35K with a faster CPU.

noshutdown wrote:

Skyscraper wrote:

The Socket-478 P4 EE "Gallatin" 3.4 running in a motherboard with dual channel memory pushes the Geforce 6800 "Ultra" pretty hard, this is somewhat GPU limited.

not really, i got over 33000pts with 6800ultra agp, and i believe the pcie version can be even faster.
just think, even a ti4600 can score over 17000pts, so what makes you think that the 6800ultra is worthy of only 24000pts?

Thats why I wrote "somewhat"

It still scales with higher core speed but not as good as if there wasnt any bottleneck at all.

I will post the score with the Gallatin at 4Ghz later when I get home 😀

Here is the score with the Gallatin at 4 GHz with the Geforce 6800 "Ultra". The score is a bit lower than if the video card diddnt bottleneck the result somewhat.

Here is the Gallatin again at 4 Ghz but with the Geforce 6800 clocked 25MHz higher. The scaling is decent but the video card is also bottlenecked by the CPU.

The dual channel DDR470 2,3,3.5 memory should not bottleneck either the CPU or video card in this case when it comes to bandwidth but lower latency could probably add some points to the score.

The Gallatin 3.4 fought against a Clawhammer @2.4GHz with ECC REG memory back in early 2004 (and lost). Its not near as fast of a CPU as higher clocked, later core revisions of A64 with fast low latency memory. The Gallatin does however manage better 3dmark 2001 results than the ones I have posted (so far 😜) with a faster video card like the Gainward Geforce 7800GS G71 or the Radeon X1950 Pro AGP.

Among a lot of giants I'm here posting my result with my little brother, and I am really happy with that 😀 :

K6-III+ 633MHz (115x5.5)
GA5-AX Rev 5.2 bios F4
512Mbyte RAM PC133 Single stick
Leadtek 5900XT @ 450Mhz/850MHz core/mem GAT2
Promise controller TX2 SATA150
WD Raptor HLFS 74Gbyte
Audigy SB0090

Optimizations via CTU on K6 and Wpcredit on memory

Not bad for a socket 7 😀

Mamba wrote:

Among a lot of giants I'm here posting my result with my little brother, and I am really happy with that :happy: : […]
Show full quote

Among a lot of giants I'm here posting my result with my little brother, and I am really happy with that 😀 :

K6-III+ 633MHz (115x5.5)
GA5-AX Rev 5.2 bios F4
512Mbyte RAM PC133 Single stick
Leadtek 5900XT @ 450Mhz/850MHz core/mem GAT2
Promise controller TX2 SATA150
WD Raptor HLFS 74Gbyte
Audigy SB0090

Optimizations via CTU on K6 and Wpcredit on memory

Not bad for a socket 7 😀

That is really a nice score! 😀

Mamba wrote:

Among a lot of giants I'm here posting my result with my little brother, and I am really happy with that :happy: : […]
Show full quote

Among a lot of giants I'm here posting my result with my little brother, and I am really happy with that 😀 :

K6-III+ 633MHz (115x5.5)
GA5-AX Rev 5.2 bios F4
512Mbyte RAM PC133 Single stick
Leadtek 5900XT @ 450Mhz/850MHz core/mem GAT2
Promise controller TX2 SATA150
WD Raptor HLFS 74Gbyte
Audigy SB0090

Optimizations via CTU on K6 and Wpcredit on memory

Not bad for a socket 7 😀

Well gosh darn, that little K6 is faster than the Northwood Celeron clock for clock!
You should try benching Doom3 on that. If you do, then I'll sit through the D3 slideshow my P2-300/EDO rig will inevitably produce. 😁

I doubled the 3DMark score without even having to double the clock speed. I love it when that happens.
Mobile T5600 @ 2.93/1066, GTX 560, memory at 887MHz 4-4-4-12

Just as with Doom 3, the overclocked T5600 is now very close to the performance of the Q6700 @ 3.33/1333FSB, which scored 42889.

Standard Def Steve wrote:

Well gosh darn, that little K6 is faster than the Northwood Celeron clock for clock!
You should try benching Doom3 on that. If you do, then I'll sit through the D3 slideshow my P2-300/EDO rig will inevitably produce. 😁

come on, i got 0.4fps in quake3 with winchip-100, and now i am wondering if quake3 would run on 486 cpus. 🤣
i also got 1.5fps in doom on cyrix486dx-25.

I'll throw my hat in the ring

2600k @ 4.5ghz
Z77X-UD5H
Sapphire 7970 Ghz Edition
XP SP3 32bit
14.4 catalyst drivers

Had to push it over 30k in Windows 98 SE.
Card: 7800 GTX @ 490MHz/1380MHz
Driver: 81.98 (added 7800 GTX to .inf)
LOD: really bad

Kamerat wrote:

Had to push it over 30k in Windows 98 SE. Card: 7800 GTX @ 490MHz/1380MHz Driver: 81.98 (added 7800 GTX to .inf) LOD: really bad […]
Show full quote

Had to push it over 30k in Windows 98 SE.
Card: 7800 GTX @ 490MHz/1380MHz
Driver: 81.98 (added 7800 GTX to .inf)
LOD: really bad

3dmark_30420.png

So I'm curious how you got this working.. since no one writes drivers for pci-express-enabled-chipsets on motherboards with pci-express for win98se.. and that's kind of needed to get the video part working.... I've been under the assumption all this time that we needed to stick to AGP systems only for Win98se support.

I'm using the standard "PCI standard PCI-to-PCI bridge" driver that Windows installs by itself. The hard disk is also running from an onboard JMB360 SATAII PCIe controller that actually have Windows 98 drivers.

Edit: Modifying the .inf file for the driver isn't necessary, just choose another card that's listed as "NV30" in the driver, like the GeForce 6800 Ultra.

Kamerat wrote:

Had to push it over 30k in Windows 98 SE. Card: 7800 GTX @ 490MHz/1380MHz Driver: 81.98 (added 7800 GTX to .inf) LOD: really bad […]
Show full quote

Had to push it over 30k in Windows 98 SE.
Card: 7800 GTX @ 490MHz/1380MHz
Driver: 81.98 (added 7800 GTX to .inf)
LOD: really bad

3dmark_30420.png

That's awesome man! I keep hearing about adjusting LOD to improve 3DMark01 performance, but I've never actually tried it (dreams of breaking 100K under Win10 🙄 ).
Does adjusting LOD trigger the "benchmark was not run using default settings" message?

Standard Def Steve wrote:

Kamerat wrote:

Had to push it over 30k in Windows 98 SE. Card: 7800 GTX @ 490MHz/1380MHz Driver: 81.98 (added 7800 GTX to .inf) LOD: really bad […]
Show full quote

Had to push it over 30k in Windows 98 SE.
Card: 7800 GTX @ 490MHz/1380MHz
Driver: 81.98 (added 7800 GTX to .inf)
LOD: really bad

3dmark_30420.png

That's awesome man! I keep hearing about adjusting LOD to improve 3DMark01 performance, but I've never actually tried it (dreams of breaking 100K under Win10 🙄 ).
Does adjusting LOD trigger the "benchmark was not run using default settings" message?

Well Kamerat has used a smaller screen resolution and arranged windows (Possibly intentionally?) to cover up every aspect of the program relating to settings and resolution used. So we have no idea if they used the defaults or not based on that screenshot.

EDIT: It's entirely possible to use Windows 98 SE on a 1920x1080 screen before, I've done it many times.

I'm using a Panasonic PanaFlat LC50S LCD monitor from 1999 with 1024x768 as native resolution. LOD tweaking is done at driver level, I used RivaTuner (as in the ol' days) to assist me. The windows doesn't hide anything except the default 3DMark settings.

Standard Def Steve wrote:

That's awesome man! I keep hearing about adjusting LOD to improve 3DMark01 performance, but I've never actually tried it (dreams of breaking 100K under Win10 🙄 ).
Does adjusting LOD trigger the "benchmark was not run using default settings" message?

Thank you. No, it doesn't unfortunately. With a LOD of 15 it looks like crap.

New personal best in Windows 98 SE with 38723 3DMarks, this time with Core 2 Duo E6600 and Asus Radeon X800 XT PCIe. Note that CPU speed detection fails, the real speed is 3776MHz (8x472MHz). 3DMark actually detects the right FSB. No LOD tweaks this time.

Edit: New run, slightly tighter timings.

Ran the same setup with Windows XP SP3. What puzzles me is that the FPS in the "Nature" test is way higher in Windows 98 SE, also "Dragothic" loves 98. I'm using Catalyst 6.2 on both XP and 98.

Here's my main system running 3DMark 2003.

The CPU score on this machine is just nuts. Back in 2003, my 1.8GHz P4 could only muster 6.4 fps on CPU test #2. Today we've got i7s blasting through the same test at 134.3 fps. The CPU test is single-threaded; it's only using 8% of this 12-thread CPU!

The rest of the benchmark, despite being mostly D3D8, is quite GPU limited.

Kamerat wrote:

New personal best in Windows 98 SE with 38723 3DMarks, this time with Core 2 Duo E6600 and Asus Radeon X800 XT PCIe. Note that CPU speed detection fails, the real speed is 3776MHz (8x472MHz). 3DMark actually detects the right FSB. No LOD tweaks this time.

3m01_38723_8x472.png

Edit: New run, slightly tighter timings.

How did you get the P965 chipset to work with Win98? I never heard of anything higher than 925 working with 98.