Having access to an Asus PCI/I-486SP3G board (Intel 420ZX chipset, Saturn II), I tried to obtain the datasheet for the north bridge (the Intel 82424ZX cache/DRAM controller, supported by the 82423TX data path unit), I failed to turn up anything but the 2 page product brief. So I bit the bullet to buzz out the connections on that board. I had the revision 1.6 at hand, which doesn't come with a dirty RAM for L2 (and the Saturn can't combine tag/dirty). It has a well-known layout mistake that would prevent the dirty tag from working anyways, as pointed out by the German PC magazine c't in the 10/94 issue: DIRTYQ (at the chipset) is connected to the VALID tag bit instead of the dirty output. Compensating for that mistake, I came up with the following pinout:

1 - ADS#
2 - D/C
3 - M/IO
4 - W/R
5 - DIRTYD
6 - DIRTYQ (+HITM)
7 - DWE# (+INV)
8 - A4
9 - A3
10 - VCC
11 - A2
12 - A6
13 - A7
14 - A8
16 - A10
17 - A11
18 - A5
19 - A9
20 - A16
21 - A13
22 - A20
23 - A22
24 - A15
25 - A12
26 - GND
27 - VCC
28 - A24
29 - A21
30 - A18
31 - A14
32 - A19
33 - A30
34 - A17
35 - A23
36 - A29
37 - A25
38 - GND
39 - A26
40 - A31
41 - A28
42 - A27
43 - VCC
44 - LCAa
45 - LCAb
46 - HCAa
47 - HCAb
48 - CWE3#
49 - CWE2#
50 - CWE1#
51 - CWE0#
52 - COE#
53 - GND
54 - LCCS#
55 - HCCS#
56 - CALE
57 - RAS0#
58 - CAS1a#
59 - CAS1b#
60 - GND
61 - VCC
62 - RAS1#
63 - CAS2a#
64 - CAS2b#
65 - RAS2#
66 - CAS3a#
67 - CAS3b#
68 - RAS3#
69 - CAS0a#
70 - CAS0b#
71 - GND
72 - RA0a
73 - RA0b
74 - RA1
75 - RA2
76 - RA3
77 - RA4
78 - GND
79 - VCC
80 - RA5
81 - RA6
82 - RA7
83 - RA8
84 - RA9
85 - RA10
86 - RWE#

102 - PCIRST#
103 - AD31
104 - GND
105 - AD30
106 - AD29
107 - AD28
108 - AD27
109 - AD26
110 - GND
111 - AD25
112 - AD24
113 - C/BE3#
114 - VCC
115 - AD23
116 - AD22
117 - AD21
118 - AD20
119 - GND
120 - AD19
121 - AD18
122 - AD17
123 - AD16
124 - C/BE2#
125 - FRAME#
126 - GND
127 - IRDY#
128 - TRDY#
129 - DEVSEL#
130 - STOP#
131 - VCC
132 - LOCK#
133 - SERR#
134 - SMIACT#
135 - GND
136 - PAR
137 - C/BE1#
138 - AD15
139 - AD14
140 - AD13
141 - AD12
142 - GND
143 - AD11
144 - AD10
145 - AD9
146 - AD8
147 - VCC
148 - C/BE0#
149 - AD7
150 - AD6
153 - GND
152 - AD5
153 - AD4
154 - AD3
155 - AD2
156 - AD1
158 - AD0
159 - REQ#
160 - GNT#
161 - FLSHREQ#
162 - MEMREQ#
163 - MEMACK#
164 - GND
165 - VCC
166 - PCICLK1 (PCI1/PCI2)
167 - PCICLK2 (PCI3/SCSI)
168 - GND
169 - TOE#
170 - TWE#
171 - T0 / cfg
172 - T1 / cfg
173 - T2
173 - T3
174 - T4
175 - T5
176 - T6
177 - (T7)
178 - (T8)
179 - VALID
180 - GND via 0R
181 - CLKIN2 (from CPUCLK)
182 - CLKIN (25/33)
183 - VCC
184 - SL strap via 1K (some output)
185 - (ALT_RST# AND KBDRST#) via U22a, SRESET (input)
186 - PWROK
187 - INIT (to CPU, aka SRESET)
188 - SMI/CPURST
189 - BE3#
190 - BE2#
191 - BE1#
192 - BE0#
193 - AHOLD (and-ed with ADS# by U26b)
194 - EADS#
195 - BOFF#
196 - KEN#
197 - GND
198 - PCICLK3 (SB)
199 - CPUCLK
200 - GND
201 - VCC
202 - RDY#
203 - BRDY#
204 - PCD/CACHE
205 -
206 - HLOCK#
207 - CLK? via 33R to 95
208 - GND

I intend to make a more detailed post about that board and its peculiarities after bodging the mistake the next days.

Buzzing out all the pins made me scratch the solder pads of the north bridge too much, and some eroded solder caused a couple of shorts between neighbouring pins. I feel sorry for both the board and it's owner, who experienced how I damaged his board by probing. But don't be afraid, I managed to fix it with flux, wick, and a lot of retries...