VOGONS


First post, by nemail

User metadata
Rank Member
Rank
Member

hi

currently i'm playing around with my socket 7 fic pt-2007 board which has 512kb onboard pipeline burst level II direct-mapped write-back cache according to the manual and according to the chips i can see on the mainboard.

now the strange thing ist, i get only barely different results in 3dmark 99 when i disable that cache in bios (option is named "external cache: disabled/enabled").

3dmark result with cache enabled:
1022 3DMarks
1564 CPU 3D Marks (Synthetic CPU 3D Speed)
1867 3D Rasterizer score

3dmark result with cache disabled:
1003 3DMarks
1561 CPU 3D Marks (Synthetic CPU 3D Speed)
1862 3D Rasterizer score

i also didn't really notice a difference in general system performance (mainly boot process) with cache enabled or disabled...

the rest of the system currently consists of:

pentium 233mmx
256MB SD RAM (i know 430TX can only cache 64MB, could that be the issue?)
Voodoo 3 2000 PCI 16MB

any advices? should i run another benchmark tool?

thank you all for your help!

Reply 1 of 3, by PeterLI

User metadata
Rank l33t
Rank
l33t

Is 3DMARK not simply about GPU performance? MOBO cache probably has limited input on that.

Reply 2 of 3, by RacoonRider

User metadata
Rank Oldbie
Rank
Oldbie

Windows 98 maps memory starting from the last cell, so yes, most of it turnes out to run uncached if you use 256 MB. The difference appears only once you exceed 256MB barrier.

Reply 3 of 3, by alexanrs

User metadata
Rank l33t
Rank
l33t

I'd try using the MAXPHYSPAGE setting in system.ini or loading an 192MB ramdisk (using a switch to force it to allocate its memory from the top) so Windows can only use the first 64mb. Once you do that, see if the scores improve.