Yes they are weak, but we have said a million times that 3dmark2001SE does NOT show realworld grachips performance 😜

You can gain a LOT of points just by altering system's FSB, timings, cache or even software things like drivers or OS tweaking...

Pentium 4 2.4GHz Northwood, Soyo P4I-845PE, 512mb DDR-333 ram, NVIDIA GeForce 6600 256mb AGP (82.69 Win9x beta drivers), DirectX 8.1b, Windows 98SE

GeorgeMan wrote:

Yes they are weak, but we have said a million times that 3dmark2001SE does NOT show realworld grachips performance 😜

You can gain a LOT of points just by altering system's FSB, timings, cache or even software things like drivers or OS tweaking...

Altering things like FSB and timings however do offer real-world results, that's changing the physical function of the processor/ram.

Lets party like it is year 2004! 😀

AMD Athlon 64 FX55, A8N32-SLI, 2x1GB DDR480, Geforce 6800Ultra 92.91, onboard audio, XP-SP3.

Random question.. isn't PCI-E 1.1 16x almost the same bandwidth as AGP-8x ? I have a 6800 ultra myself but AGP version.. I'm wondering if it runs as fast as the PCI-E version.

kithylin wrote:

Random question.. isn't PCI-E 1.1 16x almost the same bandwidth as AGP-8x ? I have a 6800 ultra myself but AGP version.. I'm wondering if it runs as fast as the PCI-E version.

In the Doom 3 timedemo the AGP version of GF6800GT/Ultra perform just as well as the PCI-E version.
Perhaps even better because I think GF6800 is native AGP and uses a PCI-E bridge chip for PCI-E versions?

Skyscraper wrote:

In the Doom 3 timedemo the AGP version of GF6800GT/Ultra perform just as well as the PCI-E version.
Perhaps even better because I think GF6800 is native AGP and uses a PCI-E bridge chip for PCI-E versions?

http://www.gpureview.com/show_cards.php?card1=179&card2=201

I found this, which essentially means the PCI-E version is the exact same "paper specs" as the AGP version.. just the PCI-E version supports SLI and can go up to 512MB of ram on-board. EDIT: Ah, I looked up my old screenshot from a few months ago when I had it in the 3ghz socket 754 chip, apparently my AGP version came in @ 29k in 2001se, so it is as fast as the PCI-E version.. yay 😄 I just need to get that system repaired. I have it in an AthlonXP system now and I've been meaning to run it through 2001se, I might do that today.

kithylin wrote:

Skyscraper wrote:

In the Doom 3 timedemo the AGP version of GF6800GT/Ultra perform just as well as the PCI-E version.
Perhaps even better because I think GF6800 is native AGP and uses a PCI-E bridge chip for PCI-E versions?

http://www.gpureview.com/show_cards.php?card1=179&card2=201

I found this, which essentially means the PCI-E version is the exact same "paper specs" as the AGP version.. just the PCI-E version supports SLI and can go up to 512MB of ram on-board. EDIT: Ah, I looked up my old screenshot from a few months ago when I had it in the 3ghz socket 754 chip, apparently my AGP version came in @ 29k in 2001se, so it is as fast as the PCI-E version.. yay 😄 I just need to get that system repaired. I have it in an AthlonXP system now and I've been meaning to run it through 2001se, I might do that today.

It seems there are 3 different full spec 6800 cores, NV40, NV45 and NV48. I guess NV40 is native AGP and the other two native PCI-E.
The specs are the same so they should perform the same with the same CPU. AGP should not be a bottleneck for these cards.

[Edit]My PCI-E card gets identified as NV40 so it seems NV40 is for PCI-E or the core naming scheme has nothing to do with with AGP/PCI-E[/edit]

[Edit] I clocked the FX55 to 2.9 and got 28500 points so your 29K points @3.0 is just about right 😀 [/edit]

Skyscraper wrote:

It seems there are 3 different full spec 6800 cores, NV40, NV45 and NV48. I guess NV40 is native AGP and the other two native PC […]
Show full quote

It seems there are 3 different full spec 6800 cores, NV40, NV45 and NV48. I guess NV40 is native AGP and the other two native PCI-E.
The specs are the same so they should perform the same with the same CPU. AGP should not be a bottleneck for these cards.

[Edit]My PCI-E card gets identified as NV40 so it seems NV40 is for PCI-E or the core naming scheme has nothing to do with with AGP/PCI-E[/edit]

[Edit] I clocked the FX55 to 2.9 and got 28500 points so your 29K points @3.0 is just about right 😀 [/edit]

If you're wondering this was my score before: http://www.outfoxed.net/athlon-754/Athlon64-4 … 001se_maxed.jpg

EDIT: Fixed the correct URL to show 2001se .. not 3dmark 2000, and it's just 28,544 so right about in-line with what you had, almost exactly the same actually.

Using a AMD Socket 754 mobile 4000+ chip in a desktop motherboard, very minor overclock to 3055 Mhz CPU Speed (stock is 2.6 ghz) and single-channel ram at I dunno what speed, and a very minor overclock on the 6800 ultra. Sadly that motherboard won't run anymore until I re-cap it some day, so.... all I have for now is an AthlonXP Throughbred chip for the 6800 ultra, at the moment. If I can get the energy up to go boot that thing and run it through 2001se I'll put it in and show the results. I think it's just 12k though.

I think it's really interesting seeing your result there. Apparently the high-end chips from 754 aren't much different in performance than the high-end 939 single-core chips. Even though your system has dual channel ram and my 754 system did not, and you had PCI-E 6800 ultra and I had AGP.

kithylin wrote:

gh.

I think it's really interesting seeing your result there. Apparently the high-end chips from 754 aren't much different in performance than the high-end 939 single-core chips. Even though your system has dual channel ram and my 754 system did not, and you had PCI-E 6800 ultra and I had AGP.

150 Mhz core speed and a little overclock on the GPU makes up for the lack of dual channel 😀
The performance difference clock for clock between different K8 platforms is really slim.

Some integraded "action":
SiS 630 integrated graphics - Asus TUSC motherboard - Pentium III EB 866 MHz - 512MB PC133 - Windows XP SP3

Kamerat wrote:

Some integraded "action": SiS 630 integrated graphics - Asus TUSC motherboard - Pentium III EB 866 MHz - 512MB PC133 - Windows X […]
Show full quote

Some integraded "action":
SiS 630 integrated graphics - Asus TUSC motherboard - Pentium III EB 866 MHz - 512MB PC133 - Windows XP SP3

377 3D Mark points! I'm guessing this was just a test and that you normally use a graphics card as opposed to the onboard graphics?

borgie83 wrote:

377 3D Mark points! I'm guessing this was just a test and that you normally use a graphics card as opposed to the onboard graphics?

It's an Asus Terminator Tualatin barebone and I'm not using this system much. The motherboard lacks AGP slot and the fastest PCI graphics I got is a 3dfx Banshee (I will try this soon). The built in PSU is defective so I hooked to an ATX PSU. But I got two other Pentium III (BX) rigs if I want to do some serious 3D gaming 😀

Just increased the score to 407 3Dmark (8%) by changing memory timings from 3-3-3-5 to 2-2-2-5 (and SDRAM Refresh Mode form "Staggered 1T" to "Simultaneous") 😀

Kamerat wrote:

Just increased the score to 407 3Dmark (8%) by changing memory timings from 3-3-3-5 to 2-2-2-5 (and SDRAM Refresh Mode form "Staggered 1T" to "Simultaneous") 😀

Increased score to 524 3Dmarks, 39% faster than the first run at 377 3DMarks. 😎
Removed one stick of RAM, now only running 256MB. (This is the reason for most of the increase.)
Made a few other BIOS tweaks.
Changed the driver to the one supplied by Windows.
Had to disable the "High Polygon Count" tests because the Windows driver makes the system crash running it.
Increased the core frequency of the IGP from 101MHz to 175MHz. 😈 (This only gained a few points, looks like the IGP is heavily starving for memory bandwith.)

Voodoo3 in an upgraded Dimension T550

Celeron-1400 on a Lin-Lin adapter on a Slotket.
512MB of PC100 2-2-2
Voodoo3 3000 AGP
Dell version of the Intel SE440BX-3
SB Live!
Win98SE

Overclocked 6800GT + Overclocked PIII-S + Ancient driver = 13K in 3DMark01 😁

PIII-S @ 1585MHz, 151MHz FSB, 512K L2
1.5GB SDRAM @ 151MHz, 2-2-2
EVGA 6800GT @ 385/540, Forceware 81.98
X-Fi Extreme Music
Asus TUV4X (VIA 694T)
XP Pro SP3

By the way, a stock-clocked 6800GT scores 12646. The synthetic tests and Game Test 4 - Nature seem to like the GPU overclock. The other game tests are too CPU limited.

Here's the overclocked PIII-S pushing a PCI Radeon 9250 to the limit. System specs are listed in the post above.

This one was built by ATI themselves and has a 128-bit memory bus, according to GPU-Z. Default core speed is 240MHz, but I gave it a slight bump to 250MHz, effectively turning it into a Radeon 9200. Yes, the 9200 was slightly faster than the 9250. This was ATI's marketing team at their best. 🙄

Interestingly, the 9250 managed to edge out the 6800GT in Car Chase - High Detail. This test appears to prefer ATI's drivers over nVidia's; the Radeon 9800 Pro was also able to beat the 6800GT by quite a bit, scoring 46.5 fps. The 6800GT was able to greatly outperform the 9250 in all of the other tests, as expected.

The overclocked PCI bus in this machine may also be helping out. The driver used was Catalyst 5.6.

The 9800 Pro is completely stretched to the limit with an Athlon 64.

The S939 A64 3700+ was overclocked to 2.64GHz, the highest frequency it could handle at default vcore (1.35v). But even when pushed to 2.8GHz (1.425v), the score only increased by 14 points, indicating a completely maxed out video card.

Still, I'm very impressed that a stock-clocked 9800 Pro can hit nearly 22K!

Very impressive!

Im pretty sure that beats my best score with a Radeon 9800 with a large margin.