Test setup

• Windows 2000, SP2 + SRP, VIA 4-in-1 drivers v. 4.38(2)v(a)
• Abit KT7 Motherboard
• VIA KT133 Chipset

Tested modes

• No patches
• George Breese's PCI Latency patch v.0.19
• George Breese's PCI Latency patch v.0.19d
• VIA's "RAID performance patch" v.1.04
• VIA's "RAID performance patch" v.1.05

For the PCI Latency patch, see http://www.networking.tzo.com/net/software/
For VIA's "RAID performance patch", see http://www.viaarena.com/?PageID=66 (it's at the bottom).

What was tested?

The PCI Config Space was dumped for:

• bus0,dev0,func0 (the north bridge)
  PCI ID: 00:00.0 Class 0600: 1106:0305 (rev 02)
• bus0,dev1,func0 (the AGP port)
  PCI ID: 00:01.0 Class 0604: 1106:8305

The config space was dumped in Linux and all modes listed above in Windows 2000.
The config spaces were then compared.

Results

• North Bridge
• Linux -> Windows unpatched
  

  Change in byte at 07, a2 -> 22
  Change in byte at 55, 09 -> 89
  Change in byte at 84, e0 -> c0
  Change in byte at 88, 02 -> 03
  Change in byte at 89, 80 -> f0
  Change in byte at 8a, 4d -> 00
  Change in byte at 8b, 1d -> 00
  Change in byte at b7, 0c -> 0f
  

• Linux -> Windows + PCI Latency 0.19
  

  Change in byte at 07, a2 -> 22
  Change in byte at 0d, 08 -> 00
  Change in byte at 75, 81 -> 83
  Change in byte at 76, 62 -> c2
  Change in byte at 84, e0 -> c0
  Change in byte at 88, 02 -> 03
  Change in byte at 89, 80 -> f0
  Change in byte at 8a, 4d -> 00
  Change in byte at 8b, 1d -> 00
  

• Linux -> Windows + PCI Latency 0.19d
  

  Change in byte at 07, a2 -> 22
  Change in byte at 0d, 08 -> 00
  Change in byte at 70, c8 -> cc
  Change in byte at 75, 81 -> 85
  Change in byte at 76, 62 -> 52
  Change in byte at 84, e0 -> c0
  Change in byte at 88, 02 -> 03
  Change in byte at 89, 80 -> f0
  Change in byte at 8a, 4d -> 00
  Change in byte at 8b, 1d -> 00
  

• Linux -> Windows + VIA Patch 1.04
  

  Change in byte at 07, a2 -> 22
  Change in byte at 75, 81 -> 07
  Change in byte at 84, e0 -> c0
  Change in byte at 88, 02 -> 03
  Change in byte at 89, 80 -> f0
  Change in byte at 8a, 4d -> 00
  Change in byte at 8b, 1d -> 00
  

• Linux -> Windows + VIA Patch 1.05
  

  Change in byte at 07, a2 -> 22
  Change in byte at 55, 09 -> 89
  Change in byte at 84, e0 -> c0
  Change in byte at 88, 02 -> 03
  Change in byte at 89, 80 -> f0
  Change in byte at 8a, 4d -> 00
  Change in byte at 8b, 1d -> 00
  Change in byte at b7, 0c -> 0d
  

• AGP Port
• Linux -> Windows + any patch
  

  Change in byte at 2c, 00 -> ff
  Change in byte at 2d, 00 -> ff
  Change in byte at 2e, 00 -> ff
  Change in byte at 2f, 00 -> ff
  

• One addition when comparing Linux -> Windows + PCI Latency 0.19
  

  Change in byte at 42, 08 -> 18
  

Note: bytes 84, 88, 89, 8a and 8b in the northbridge deal with the
AGP aperture AFAIK so their differing values may not be very interesting.

Raw data

Can be found at http://www.student.nada.kth.se/~i99_hnd/via/raw_data/