rasz_pl wrote on 2024-10-07, 09:07:

So L2 is running at 3-2-2-2? - have you tried lower fsb when playing with L2 timings? Makes me wonder what will happen in PROLIN […]
Show full quote

So L2 is running at 3-2-2-2?
- have you tried lower fsb when playing with L2 timings? Makes me wonder what will happen in PROLINEA MT in 50MHz mode.
- is that with those bodge wires on the back or without? they are bad for signal integrity, ideally should be grounded micro coax 😁
- do you have access to a scope to look at eye diagrams? 😀 maybe those series resistors were doing heavy lifting cleaning signals in original module IF original module can run with 0 waitstates in the first place.
- also scope supply right on each chip, you could try experimenting with bodging more decoupling caps

In theory 2-1-1-1 should be in reach with brand new legit 12ns chips, but even as is its a 2x memory BW bump.

I can only quote the vL82C425 datasheet: "On a read-hit, the VL82C425 performs a 2-1-1-1 (zero wait state) burst read if two banks of SRAM are used. It performs a 2-2-2-2 (three wait states) burst read if one bank of SRAM is used. On a write-hit, the VL82C425 performs a one wait state (three clock cycle) write."
In our case we are talking about writes which are configurable for the VL82C481/3 (and not for the VL82C425) and currently not working when set to 0. I did test it with 256KB, but I didn't test this with the bodge (A18/A19) wires removed, I'll do that too. My oscilloscope supports eye diagrams, but I would appreciate some pointers for reading the measurement, I really have no idea.

>- have you tried lower fsb when playing with L2 timings? Makes me wonder what will happen in PROLINEA MT in 50MHz mode.
No I should and will do that too! Is there a 50MHZ FSB mode? [edit]I checked the MT bios, cache code is shared with the Prolinea 4/xx. By default it also has 1WS for writes.[/edit]

>- also scope supply right on each chip, you could try experimenting with bodging more decoupling caps
I'll do that too!

baronvonkaiser wrote on 2024-10-07, 15:43:

I'm just working on excitment... 🤣

Cheers!, and thank you and everybody for the enthusiasm.

Another update, although I still need to look into the 0WS write issue and analyze the power rails using my oscilloscope, I finally managed to enable the 1MB of cache. Windows NT 4.0/95 are both stable. My modified bios is now able to switch between 128/256 or 1MB, the only problem I'm having is validating if 512KB is inserted or not. As I'm missing the datasheet I can only compare to the HP Vectra VL2 Bios (GA.04.06) image (the only image that supports 512KB/1MB). So it's 256KB or 1MB for now, which seems fine to me, so I'm parking this issue.

As I'm getting to know the Compaq BIOS there is an added bonus. I've made use of the 16KB unused bios space at CC00:0 and added the XTIDE (386 version). My Prolinea is now able to support large drives without any additional ISA cards inserted.

Next I will write a small Compaq VLSI BIOS dump tool, so we can share our BIOS images here.

Here is the biggest part of the patches:

1; nasm patch.asm -fbin -o patch.bin
2
3cpu 486
4
5section .text 
6
7org 4438h
8
9; test the cache size and put the cachesize / 4 into ebx
10get_cache_size:
11
12  mov     ebx, 100000h
13  mov     es:[ebx], word 0
14  mov     dx, 0A5A5h
15  mov     ebx, 140000h
16  mov     es:[ebx], dx
17  mov     ebx, 65536
18  cmp     [dword es:100000h], dx
19  jz      short loc_FAC29
20  mov     ebx, 131072   ; 512KB or the first 512KB from the lower 1MB region
21  mov     al, 0EFh
22  out     84h, al
23  
24loc_FAC29:
25
26  mov     edx, ebx
27  jmp     1391h
28
29  align 2
30
31; test the upper 1MB range
32test_upper_1MB:  
33
34  mov     al, 0E6h
35  out     84h, al
36
37  mov     ebx, 100000h
38  mov     es:[ebx], word 0
39
40  mov     dx, 0A5A5h
41
42  mov     ebx, 140000h
43  mov     es:[ebx], dx
44  mov     cl, 100b
45  mov     ebx, 65536
46  cmp     [dword es:100000h], dx
47  jz      validated_cache
48
49  mov     ebx, 180000h
50  mov     es:[ebx], dx
51  mov     ebx, 131072
52  mov     cl, 101b
53  cmp     [dword es:100000h], dx
54  jz      validated_cache
55
56  mov     ebx, 180000h
57  mov     edx, 131072
58  mov     ecx, edx
59  xor     eax, eax
60

61clear_loc_1MB:
62  mov     es:[ebx], eax
63  add     ebx, 4
64  loop    clear_loc_1MB, ecx
65  mov     al, 0ABh
66  out     84h, al
67  mov     ebx, 180000h
68  mov     ecx, edx
69  shl     ecx, 2
70  mov     al, 0FFh
71
72validate_cache_ffh:
73  mov     es:[ebx], al
74  cmp     es:[ebx], al
75  jnz     no_cache_detected
76  inc     ebx
77  loop    validate_cache_ffh, ecx
78  mov     al, 0A1h
79  out     84h, al
80  mov     ebx, 180000h
81  mov     eax, ebx
82  mov     ecx, edx
83
84fill_range_180000h:
85  mov     es:[ebx], eax
86  add     ebx, 4
87  mov     eax, ebx
88  loop    fill_range_180000h, ecx
89  mov     al, 0A2h
90  out     84h, al
91  mov     ebx, 180000h
92  mov     ecx, edx
93  mov     edi, ebx
94
95validate_cache_180000h:
96  mov     eax, es:[ebx]
97  cmp     eax, edi
98  jnz     no_cache_detected
99  add     ebx, 4
100  add     edi, 4
101  loop    validate_cache_180000h, ecx
102  mov     al, 0A3h
103  out     84h, al
104  mov     ebx, 180000h
105  mov     ecx, edx
106  shl     ecx, 2
107  mov     al, 55h
108
109fill_range_55h:
110  mov     es:[ebx], al
111  inc     ebx
112  loop    fill_range_55h, ecx
113  mov     al, 0A4h
114  out     84h, al
115  mov     ebx, 180000h
116  mov     ecx, edx
117  shl     ecx, 2
118
119validate_cache_55h:
120  mov     al, es:[ebx]
121  cmp     al, 55h
122  jnz     short no_cache_detected
123  inc     ebx
124  loop    validate_cache_55h, ecx
125  mov     al, 0A5h
126  out     84h, al
127  mov     ebx, 180000h
128  mov     ecx, edx
129  shl     ecx, 2
130  mov     al, 0AAh
131
132fill_range_AAh:
133  mov     es:[ebx], al
134  inc     ebx
135  loop    fill_range_AAh, ecx
136  mov     al, 0A6h
137  out     84h, al
138  mov     ebx, 180000h
139  mov     ecx, edx
140  shl     ecx, 2
141
142validate_cache_AAh:
143  mov     al, es:[ebx]
144  cmp     al, 0AAh
145  jnz     short no_cache_detected
146  inc     ebx
147  loop    validate_cache_AAh, ecx
148
149  mov     ebx, 262144
150  mov     cl, 110b
151
152validated_cache:
153
154  jmp     0x14AE
155  
156no_cache_detected:
157  jmp     0x148D
158  
159  align 2
160  
161write_cfg_by_idx:
162  xchg    ah, al
163  out     0ECh, al
164  xchg    ah, al
165  out     0EDh, al
166  retn
167
168read_config_by_idx:
169  out     0ECh, al
170  in      al, 0EDh
171  retn
172
173unprotect_config_registers:
174  out     0FBh, al
175  retn
176
177protect_config_registers:
178  out     0F9h, al
179  retn
180
181xtide_rom_read_slot_write_system:
182  push    dx
183  push    ax
184  mov     al, 01000000b
185  mov     ah, 0Fh
186  call    write_cfg_by_idx
187  wbinvd
188  pop     ax
189  pop     dx
190  retn
191
192xtide_rom_ro:
193  mov     al, 10110000b
194  mov     ah, 0Fh
195  call    write_cfg_by_idx
196  wbinvd
197  retn
198
199xtide_rom_rw:
200  mov     al, 11000000b
201  mov     ah, 0Fh
202  call    write_cfg_by_idx
203  mov     ax, 0CC00h
204  mov     es, ax
205  xor     di, di
206  xor     ax, ax
207  mov     cx, 8192
208  rep stosw               ; clear video ram
209  in      al, 61h         ; PC/XT PPI port B bits:
210                          ; 0: Tmr 2 gate ---? OR 03H=spkr ON
211                          ; 1: Tmr 2 data -+  AND 0fcH=spkr OFF
212                          ; 3: 1=read high switches
213                          ; 4: 0=enable RAM parity checking
214                          ; 5: 0=enable I/O channel check
215                          ; 6: 0=hold keyboard clock low
216                          ; 7: 0=enable kbrd
217  or      al, 1100b
218  out     61h, al         ; PC/XT PPI port B bits:
219                          ; 0: Tmr 2 gate ---? OR 03H=spkr ON
220                          ; 1: Tmr 2 data -+  AND 0fcH=spkr OFF
221                          ; 3: 1=read high switches
222                          ; 4: 0=enable RAM parity checking
223                          ; 5: 0=enable I/O channel check
224                          ; 6: 0=hold keyboard clock low
225                          ; 7: 0=enable kbrd
226  and     al, 11110011b
227  out     61h, al         ; PC/XT PPI port B bits:
228                          ; 0: Tmr 2 gate ---? OR 03H=spkr ON
229                          ; 1: Tmr 2 data -+  AND 0fcH=spkr OFF
230                          ; 3: 1=read high switches
231                          ; 4: 0=enable RAM parity checking
232                          ; 5: 0=enable I/O channel check
233                          ; 6: 0=hold keyboard clock low
234                          ; 7: 0=enable kbrd
235  retn
236
237has_cpuid:
238  pushfd
239  pop     eax
240  mov     ecx, eax
241  xor     eax, 200000h
242  push    eax
243  popfd
244  pushfd
245  pop     eax
246  and     eax, 200000h
247  and     ecx, 200000h
248  cmp     eax, ecx
249  jz      loc_F2CC9
250  xor     al, al
251  jmp     short locret_F2CCB
252
253loc_F2CC9:
254  or      al, 0FFh
255
256locret_F2CCB:
257  retn
258
259xtide_copy_rom_enable_shadow:
260  call    unprotect_config_registers
261  call    xtide_rom_rw
262
263  ; unknown VL82C483 index register (this seems to enable slot read/system writes)
264  mov     al, 53h
265  out     0ECh, al
266  in      al, 0EDh
267  and     al, 7Fh
268  xchg    ah, al
269  mov     al, 53h
270  out     0ECh, al
271  xchg    ah, al
272  out     0EDh, al
273
274  call    xtide_rom_read_slot_write_system
275  mov     ax, 0CC00h
276  mov     ds, ax
277  xor     si, si
278  mov     ax, 0CC00h
279  mov     es, ax
280  xor     di, di
281  mov     cx, 2000h
282  rep movsw
283
284  call    has_cpuid
285  jnz     short loc_F10B8
286  
287  ; unknown VL82C483 index register (this seems to disable slot read/system writes)
288  mov     al, 53h
289  out     0ECh, al
290  in      al, 0EDh
291  or      al, 80h
292  xchg    ah, al
293  mov     al, 53h
294  out     0ECh, al
295  xchg    ah, al
296  out     0EDh, al
297
298loc_F10B8:
299  call    xtide_rom_ro
300
301  call    protect_config_registers
302  retn
303
304xtide_copy_rom:
305  push    bp
306  push    ds
307  push    es
308  push    ax
309  push    bx
310  push    cx
311  push    dx
312  push    si
313  push    di
314  mov     al, 0E0h
315  out     84h, al
316  cld
317  mov     ax, 0CC00h
318  mov     ds, ax
319  mov     ax, 0AA55h
320  cmp     ax, ds:0
321  jnz     short loc_F807B
322  call    xtide_copy_rom_enable_shadow
323
324loc_F807B:
325  pop     di
326  pop     si
327  pop     dx
328  pop     cx
329  pop     bx
330  pop     ax
331  pop     es
332  pop     ds
333  pop     bp
334  retn
335  
336section .data
337  ; program data
338 
339section .bss
340  ; uninitialized data

Not to be a broken record, but this is amazing. This hobby is so much more interesting and fun because of gifted people like you sander!

I just got a Presario CDS 860. I saw the empty cache slot, found the part number, and Googled it. It was a happy surprise to find this active thread! When this kind of thing happened to me in the past the thread was always like 10+ years old.

Like someone else said, I'm shaking my wallet at the screen! I'd love to buy one of these from you with 1MB, if that's possible?

bitslasher wrote on 2024-11-03, 21:41:

Not to be a broken record, but this is amazing. This hobby is so much more interesting and fun because of gifted people like you sander!

I just got a Presario CDS 860. I saw the empty cache slot, found the part number, and Googled it. It was a happy surprise to find this active thread! When this kind of thing happened to me in the past the thread was always like 10+ years old.

Like someone else said, I'm shaking my wallet at the screen! I'd love to buy one of these from you with 1MB, if that's possible?

So your one of my two forever enemies now! Got one of the only 2(aside from my cds 860) that I've ever seen off eBay in August or last week! 2 more days and I would have bought the one last week... Please take care of it.. they seem so rare. And I would gladly purchase it if you ever got tired of it.
And yea we are waiting with diamond forming tension to get one of the 256 or whatever size is available... It still all seems like a miracle after so many decades(1997)..

baronvonkaiser wrote on 2024-11-06, 21:42:

So your one of my two forever enemies now! Got one of the only 2(aside from my cds 860) that I've ever seen off eBay in August or last week! 2 more days and I would have bought the one last week... Please take care of it.. they seem so rare. And I would gladly purchase it if you ever got tired of it.
And yea we are waiting with diamond forming tension to get one of the 256 or whatever size is available... It still all seems like a miracle after so many decades(1997)..

Ha! Hey there. You have one of these systems already? Are you needing any parts? I have a lot of Compaq stuff that I think is interchangeable. Let me know if you need anything! I'll try to be the worst enemy you've ever known 🤣

bitslasher wrote on 2024-11-09, 19:03:

baronvonkaiser wrote on 2024-11-06, 21:42:

So your one of my two forever enemies now! Got one of the only 2(aside from my cds 860) that I've ever seen off eBay in August or last week! 2 more days and I would have bought the one last week... Please take care of it.. they seem so rare. And I would gladly purchase it if you ever got tired of it.
And yea we are waiting with diamond forming tension to get one of the 256 or whatever size is available... It still all seems like a miracle after so many decades(1997)..

Ha! Hey there. You have one of these systems already? Are you needing any parts? I have a lot of Compaq stuff that I think is interchangeable. Let me know if you need anything! I'll try to be the worst enemy you've ever known 🤣

Hello enemy! How are you over in enemy land? Lol
In 97 I got a complete presario cds 860, first real pc(I had an old ibm could only get the basic gorilla throwing a banana in a city game). I still have it, tho it spent 15 years in a Florida garage, and rats decided to make a bed on the sound blaster 16 ct1770... I've cleaned the parts as best I can. But it's just carrion at this point. I was lucky and found a presario 833, 5 or so years ago. And a prolina mt motherboard(same motherboard, BIOS just says prolina instead of presario). But it had the memory left in 3 slots during shipping. And between the summer Florida heat and harsh handling it ripped the retention clips apart. I've glued them and so far they are still holding and registering.
The end of August and 2 weeks ago someone put the only 2 presario cds 860s I've ever seen for sale on eBay. If I had gotten paid 2 days sooner I would have bought the last one for the 150 it sold for. So I'm always on the lookout for one, especially complete. Something about the presario 833\850\860\866 are just magical to me ... Especially with the performance I wrung out of it.. 48mb, a 486 odpr100 and any drive it would recognize. It would chug along with games that called for a Pentium 75 or 90.. it wasn't happy, but my poverty ridden self was happy to play any of those old games.
That's why I say take care of it. I never see them around. I've never seen a presario 866, but my best friend had the 850. And we were always battling for who had the most power. Lol.

I still can't wait for the cache chips... It's the last thing I need. I have scavenged everything to make the 833 run at its best. Just need that magic cache.
Enjoy your Klingon computers!
The mighty Com-PaQ!

Hello and merry Christmas or as the case may be to all here.
Due to impatients and wondering. Mainly to sander, I am sure you are busy like everyone towards the end of the year. But just wondering if any of the test modules are or will be available soonish? Or would trying to navigate pcbway(or some service) be the way to go?
You owe us nothing, especially after all the work you have done and published plans for..
But I had to ask.
Thank you again and have a great new year!

baronvonkaiser wrote on 2024-12-18, 18:08:

But just wondering if any of the test modules are or will be available soonish?

I'm sorry to keep you guys waiting, I've been rebuilding parts of my house. In the meantime I've actually ordered parts for 10 pieces, and added jumpers to the design. So no soldering is required to switch between 1MB/256kb.

Unfortunately it won't be a Christmas present but a 2025 present. I'll keep you posted..

Just an update from me and a thank you to Sander and the community for the work and refinement.
A friend and I built 2 of these over the holiday break and good news, the current design works well. Tested ok on 2 Prolinea 4/50s (upgraded to 100 overdrive).

We found the simplest way to construct these was with solder paste and an controlled PCB oven. My friend is a sound engineer so has access to some cool lab equipment.
In the end we used PCB way for the boards and Mouser Electronics for the parts (as I'm in Australia).
In my use case I left them in the 256kb configuration and it makes a great holiday project.

We are looking to build 2 more (we duplicated the build order in case of any failures) so we will have some spares in the new year.
Otherwise happy new year to all and I look forward to any future refinements.

lordofdaping wrote on 2024-12-31, 02:53:

Just an update from me and a thank you to Sander and the community for the work and refinement. A friend and I built 2 of these […]
Show full quote

Just an update from me and a thank you to Sander and the community for the work and refinement.
A friend and I built 2 of these over the holiday break and good news, the current design works well. Tested ok on 2 Prolinea 4/50s (upgraded to 100 overdrive).

We found the simplest way to construct these was with solder paste and an controlled PCB oven. My friend is a sound engineer so has access to some cool lab equipment.
In the end we used PCB way for the boards and Mouser Electronics for the parts (as I'm in Australia).
In my use case I left them in the 256kb configuration and it makes a great holiday project.

We are looking to build 2 more (we duplicated the build order in case of any failures) so we will have some spares in the new year.
Otherwise happy new year to all and I look forward to any future refinements.

Nice! I’m in Sydney, with two Prolineas (4/66 desktop and 4/50 tower) so I’d buy two off you if you make more! Don’t really have the gear to solder this.

Attached is a utility I wrote to dump the ROM for VLSI (VL82C481, VL82C482, VL82C483 and VL82C486) machines. This tool has the advantage of not requiring you to physically remove the BIOS chip. To use it, run the included utility (source code is also provided in the ZIP file) within a DOS environment on your machine. After extracting the BIOS data, please share the extracted file along with the BIOS version number and the specific machine type here.

To fully utilize the 1MB cache, the BIOS requires specific modifications. However, I can only make these necessary changes to the BIOS if I have access to the original BIOS image.

lordofdaping wrote on 2024-12-31, 02:53:

Just an update from me and a thank you to Sander and the community for the work and refinement. A friend and I built 2 of these […]
Show full quote

Just an update from me and a thank you to Sander and the community for the work and refinement.
A friend and I built 2 of these over the holiday break and good news, the current design works well. Tested ok on 2 Prolinea 4/50s (upgraded to 100 overdrive).

We found the simplest way to construct these was with solder paste and an controlled PCB oven. My friend is a sound engineer so has access to some cool lab equipment.
In the end we used PCB way for the boards and Mouser Electronics for the parts (as I'm in Australia).
In my use case I left them in the 256kb configuration and it makes a great holiday project.

We are looking to build 2 more (we duplicated the build order in case of any failures) so we will have some spares in the new year.
Otherwise happy new year to all and I look forward to any future refinements.

If you go into production, please count me in as a customer.

I've been working on resolving that pesky 0WS write issue, but haven't found a solution. Signal integrity appears fine, and there's no difference with or without the bodge wires. I've also tried using 74F244 series instead of 74AHCT244, but this didn't help. The strange thing is that all cache checks pass (cache is tested with 0WS and works fine!), but the machine hangs after setting the final configuration in the cache register. Access to some extended VL82C483 documentation would be very helpful at this point, but I've decided to leave it for what it is.

I've also ordered the final PCBs (with stencil) today. They will be manufactured with ENIG 2U (gold) and 30-degree edge chamfers. Delivery is expected in approximately 8 days.

The end of last year I already ordered parts for 11 PCBs (including spares) and solder paste. The total cost was +-270 euros for parts (Mouser) and 95 euros for PCBs and stencil (JLCPCB, including shipping and tax), which comes to 36.5 euros per board. As this is a community project, I will not be charging any extra (and you pay for your own shipping) but this will be a onetime production run only as 10 PCBs is already way more than I bargained for 😉

I haven't received any response regarding dumping our BIOS'es. If there is interest in running in 1MB configuration, please respond as I will need to modify the BIOS and order some chips to write the new BIOS content to./i]

here is a prioritized list of previous interest (please let me know it you're no longer interested):

@pentiumspeed 1 2024-10-02, 17:28
@nizmow 1 2024-10-02, 22:37 (buying from @lordofdaping?)
@baronvonkaiser 1 2024-10-03, 17:54
@CharlieFoxtrot 1 2024-10-03, 19:37
@bitslasher 1 2024-11-03, 22:41
@Geoman 1 2025-01-11, 12:16 (buying from @lordofdaping?)

Thanks for the interest Nizmow and Geoman,

Boards are not yet ready though next month in February that should change.
I'm happy to supply once the boards are tested, just 2 at the moment but based on demand I'll see what I can do.
These are the version 1 boards, but I've been using them, and I can't see any issues at all with them so far in circuitry, except sweet, sweet cache.

Sander I'm back from my break now so I'll look to supply you an image of mine to compare, likely next week.
I have a few Prolinea machines so I'm happy to test any bios mods, because, as much as I don't want to tinker and keep it close to Compaq spec, it's going to happen.
I would appreciate your modding of the BIOS when posted because I'm not a hex expert, but I can manufacture cards.
Should hear back more from me next month. Thanks.

sander wrote on 2025-01-22, 20:01:

here is a prioritized list of previous interest (please let me know it you're no longer interested): […]
Show full quote

here is a prioritized list of previous interest (please let me know it you're no longer interested):

@pentiumspeed 1 2024-10-02, 17:28
@nizmow 1 2024-10-02, 22:37 (buying from @lordofdaping?)
@baronvonkaiser 1 2024-10-03, 17:54
@CharlieFoxtrot 1 2024-10-03, 19:37
@bitslasher 1 2024-11-03, 22:41
@Geoman 1 2025-01-11, 12:16 (buying from @lordofdaping?)

I am definitely still interested and waiting.. so awesome to be on the list. Thank you and I will keep checking back for when action is needed on my part.
Thank you for making all this happen, and thank you to everyone else who is interested and making this not just a rarely found dream or prize.
(Watch a horde of 50 oem's be found and thrown on ebay a month after we all get ours, 🤣.)
Still awesome to see something like this so long after the era of the machines.

The PCBs and stencil arrived, and everything looks promising. I'll begin production shortly.

Awesome work - Props!

I spent some time today populating one side of the PCB with components and reflowed them in the oven. It was my first time using lead-free RoHS solder paste, and I'm pleased with how it turned out. I did discover a couple of issues, though: I'm short on capacitors (I ordered too few) and the dip switches I ordered are incorrect. I've placed orders for replacements and additions.

More to come...

Excellent update Sander,

Sorry, I'll look to upload the bios dump of my main system soon.

However, exciting for others (@nizmow and @Geoman). Good news, the boards are now for sale in Australia via Ebay: https://www.ebay.com.au/itm/276870686351

Price reflects parts and labor and I've tested each and worked with another and I'm still amazed at these cards (load tested very well). This is my full inventory at the moment, so if all sold another batch will need to be done. 4 available for purchase today.