I'm just curious if anyone else has noticed a possible change in the demands of the benchmark, in terms of AGP usage?
The system: Epox 8kta+ (KT133), 256 MB PC133, Athlon TBird @ 1.41, WinME, AIW Radeon (32MB if you don't know)@ 185, 9016 drivers. Rage3D tweak beta 4 is installed, but with only the per application gamma enabled in game util. I didn't investigate this program further due to how the issue was corrected.
The example: I was trying the VIA 410 AGP driver in standard (as opposed to turbo) to see what the difference was. Well, besides some PHENOMENAL slowdowns in Deus Ex, the other result was I noted that 3dmark 2001 SE was giving me low results (3000) and 3dmark 2001 was giving me similar results to before I tried the switch (3400) (this is with 32 bit zbuffer + 8 bit stencil, what I actually use for gaming).
The experiments: I tried picking 24 bit zbuffer from 3dmark, I tried disabling all versions of fog in the registry (due to something Uriel was trying on Rage3d), changing the overclock, and I still couldn't get a better score. However, 3dmark 2001 Original still got around 3400/3300 in the same boot.
The successful variation: reinstalling the drivers in Turbo mode brought the SE score back up (what does it change, exactly? I couldn't find a BIOS setting that changed, and I thought I would...probably just missed it...I checked PCI delay, all my AGP settings in Advanced Chipset menu)
Conclusions: Well, my environment was way too uncontrolled for any definite conclusions, but some possibilites that struck me are:
1) some driver thing with ATi that didn't recognize the 3dmark 2001SE (there are any number of possibilities here), but this is hardly Quake/Quack since scores went back up when I changed back to Turbo. I didn't notice a difference in IQ in any of the tests.
2) some sort of probably straightforward but non-announced change in texturing management in 3dmark 2001 SE that is impacted adversely by a 32MB video card with whatever the "standard" mode AGP installation of 410 did for my motherboard.
3) some unintended side effect of whatever the fix was for Kyro cards (which may in fact end up being slightly modified versions of the previous two).
4) since I'm not up to speed on all the various changes from Original to SE, it could be Something Completely Different.
I'm just curious if anyone has/will observe anything similar on their systems? I'm sure there are some 32MB Geforce 2 owners that could try something similar and we can find out if this was just a localized fluke or something interesting. Or maybe something not at all surprising for someone but that will be educational for me when they respond.
EDIT: Come on, no one is interested (or bored) enough to check this out on their rig? Or does no one else have both 3dmark 2001 versions, a VIA chipset, and a 32MB card? BTW, "First post in this group to break 100 views without a reply."
<font size=-1>[ This Message was edited by: demalion on 2002-02-17 04:36 ]</font>
The system: Epox 8kta+ (KT133), 256 MB PC133, Athlon TBird @ 1.41, WinME, AIW Radeon (32MB if you don't know)@ 185, 9016 drivers. Rage3D tweak beta 4 is installed, but with only the per application gamma enabled in game util. I didn't investigate this program further due to how the issue was corrected.
The example: I was trying the VIA 410 AGP driver in standard (as opposed to turbo) to see what the difference was. Well, besides some PHENOMENAL slowdowns in Deus Ex, the other result was I noted that 3dmark 2001 SE was giving me low results (3000) and 3dmark 2001 was giving me similar results to before I tried the switch (3400) (this is with 32 bit zbuffer + 8 bit stencil, what I actually use for gaming).
The experiments: I tried picking 24 bit zbuffer from 3dmark, I tried disabling all versions of fog in the registry (due to something Uriel was trying on Rage3d), changing the overclock, and I still couldn't get a better score. However, 3dmark 2001 Original still got around 3400/3300 in the same boot.
The successful variation: reinstalling the drivers in Turbo mode brought the SE score back up (what does it change, exactly? I couldn't find a BIOS setting that changed, and I thought I would...probably just missed it...I checked PCI delay, all my AGP settings in Advanced Chipset menu)
Conclusions: Well, my environment was way too uncontrolled for any definite conclusions, but some possibilites that struck me are:
1) some driver thing with ATi that didn't recognize the 3dmark 2001SE (there are any number of possibilities here), but this is hardly Quake/Quack since scores went back up when I changed back to Turbo. I didn't notice a difference in IQ in any of the tests.
2) some sort of probably straightforward but non-announced change in texturing management in 3dmark 2001 SE that is impacted adversely by a 32MB video card with whatever the "standard" mode AGP installation of 410 did for my motherboard.
3) some unintended side effect of whatever the fix was for Kyro cards (which may in fact end up being slightly modified versions of the previous two).
4) since I'm not up to speed on all the various changes from Original to SE, it could be Something Completely Different.
I'm just curious if anyone has/will observe anything similar on their systems? I'm sure there are some 32MB Geforce 2 owners that could try something similar and we can find out if this was just a localized fluke or something interesting. Or maybe something not at all surprising for someone but that will be educational for me when they respond.
EDIT: Come on, no one is interested (or bored) enough to check this out on their rig? Or does no one else have both 3dmark 2001 versions, a VIA chipset, and a 32MB card? BTW, "First post in this group to break 100 views without a reply."
<font size=-1>[ This Message was edited by: demalion on 2002-02-17 04:36 ]</font>
