Nice work !

Horun wrote on 2023-02-26, 01:58:

Nice work !

Thank you! A lot of work went into it. Hopefully someone can help overcome the cache limitation for the Tualatin and the VIA C3.

Unfortunately I ran into... Video card(s) problems
The GPU is an ATI Radeon R925128DE, that's a 9200 PRO manufactured by Gigabyte.

I can't get the video card to work properly under Win98SE. Under DOS it's fine. Under Windows 98SE it works only with the generic VGA driver.

I tried these drivers on two seperate installs from scratch:

* 6-2_wme_dd_cp_30314.exe
* vga_driver_ati_radeon_agp8x_98me.exe

They install flawlessly but upon the first reboot the screen displays the correct colors (256/65k) then very soon flickers, the background becomes all black and the mouse hangs. After rebooting it's just a blank screen, either with the DVI or VGA connectors. There seems to be a blank signal because the monitor doesn't go to sleep.

Then I tried with a Radeon 9200SE (A9200SE by ASUS), a slower but similar card but the result is exactly the same.

I like these cards because they automatically resize the output to 4:3 in pure DOS, and that's quite comfortable since I'm not using a monitor capable of setting the correct display ratio.

What would you recommend me to try?

I tried reducing the bus speed to 66MHz but it didn't help. I'll try cards from a different brand and see if this is a drivers problem or something more serious. I've read Radeon cards are known to have such problems.

I tried all these cards and they have the same results: R9200SE, R9200 PRO, FX5200, FX5500, MX4000, MX400

ATI
===
I tried the oldest drivers I could find,
ATI CATALYST 3.4 05/14/03 - requires installation of DirectX 8.1 before.

The driver installs fine, the computer restarts and I see the full 256/65k colors and then... the background becomes black and the cursor freezes. That's it. =(

NVIDIA
======
Some drivers I tried with the nVidia cards: 45.23_win9x_international.exe, 81.98_forceware_win9x_english.exe

The behavior is this: After installing the driver the system doesn't start, sometimes you get a black screen with a blinking cursor, sometimes it boots but hangs right after the desktop finishes loading, other times it just goes blank. This depends on if you just installed the drivers or if you're rebooting after it hung the first time.

LIVE CDs
======
I tried booting some CDs:
* DSL (Linux) worked fine even at 1280x1024x24 - I wonder why...
* Hiren's Boot CD (Mini Windows XP) - blank screen. The monitor reports "1280x1024 75Hz"
* XP SP3 setup hangs at "Setup is starting Windows" - I'd love to know why exactly it hangs!

OTHER TESTS and info
==========
I tried changing BIOS options related to AGP: I disabled the onboard sound, serial/parallel ports, everything I could. I also tried unplugging the PS/2 mouse, changed the ram stick, the PSU, unplugged the hard drive. I don't know what else to try.

The cards work fine on DOS and in 16 color mode on W98 though.

What is it? Hardware incompatibility or drivers incompatibility with the system?

WHAT WORKED
===========
* Voodoo3 - Drivers load fine and it works fine.
* TNT2 M64 - Newest one that worked! I used the drivers Detonator 2.08 AGP from here:
old nVidia detonator drivers (TNT and newer)

DIDN'T POST
=========
Other cards didn't even let the CPU start (POST card shows ----).
*MX400-64
*MX4000 PCI

LOTS of QUESTIONS
==============
Is it a hardware limitation? Could it be a BIOS limitation? A chipset limitation?

I expressly wanted that HDMI auto-resizing output GPU. Are there any other cards with that feature that I could get and try? They likely won't work, maybe if we figure out what's the problem.

Can anybody please tell me what's going on? Any ideas?

I think I found a possible answer to this, taken from http://www.playtool.com/pages/agpcompat/agp.html

That's why "If an AGP card fits in an AGP slot then they are compatible" is actually correct if you only consider stuff you can really buy.

One practical matter which must be considered is the fact that some of the original AGP 1.0 motherboards do not provide enough power to operate some newer AGP video cards reliably. For example, some of the original motherboards using the first chipsets which supported AGP (like the Intel 440LX and 440BX) can become unstable if you install video cards which draw lots of power through the AGP slot. The motherboards can't always supply the necessary current for the newer video cards. So if you're adding a video card to an AGP 1.0 motherboard then it's a good idea to install a video card which doesn't consume very much power.

You can also occasionally get memory resource conflicts by installing a new AGP video card into an old AGP 1.0 motherboard. The video card will work properly until you install the display driver. Once you try to install the driver, a memory conflict shows up. The range of conflicting addresses varies from case to case. This problem is very unusual and when it happens it is rarely possible to fix it. I'm not sure exactly what causes the problem but apparently the motherboard and the video card are incompatible in some way which prevents Windows from properly assigning memory addresses to the video card. In the cases I've seen, there doesn't seem to be any way to predict from the video card chipset and motherboard chipset whether there will be a problem. Sometimes a particular video card chipset and motherboard chipset get along well and other times they don't. I'd guess that it's some kind of incompatibility caused by an outdated motherboard BIOS and possibly the video card BIOS. The one thing you can try is to flash your motherboard with the most recent BIOS. But since it's an old motherboard, the manufacturer will most likely not have anything but old BIOSes available. If you're running Windows 95, 98, or ME, it may be possible to manually assign addresses and get it to work but I've seen people try this and the process is about as enjoyable as a root canal and usually fails to fully solve the problem anyway. If you're running Windows 2000 or XP then it's probably impossible to fix because the newer versions of Windows almost always prevent you from manually assigning addresses, IRQs, etc. That's almost always true even if you select the standard PC HAL while installing Windows with the hope that it will allow you to assign resources manually. If you run into one of these memory resource conflicts then you should probably give up and try a different video card. It's rarely

I don't think it's a power issue, I think it's a conflicting memory address issue or something of that sort, something that could be changed with an improved/tweaked BIOS. I'll see if there's a chance to change the memory addresses on Win9x but I highly doubt it. Because as soon as you install the drivers the system doesn't let you do much unless you boot in safe mode.

nah, AGP cards work just fine in LX, and power problem would manifest with Voodoo3.
Are all of your problems while clocked at 83MHz fsb?

rasz_pl wrote on 2023-02-27, 14:40:

nah, AGP cards work just fine in LX, and power problem would manifest with Voodoo3.

I assure you that at least with this board, there are serious problems with most cards.

rasz_pl wrote on 2023-02-27, 14:40:

Are all of your problems while clocked at 83MHz fsb?

I'm doing all tests with a 66MHz bus.

I decided to try again the 9200 PRO and change the memory addresses in safe mode.

These are the settings that can't be changed:

1Memory Range            - 000A0000-000AFFFF
2Memory Range            - 000B0000-000BFFFF
3Input/Output Range      - 03B0-03BB
4Input/Output Range      - 03C0-03DF
5Memory Range            - 000C0000-000CCFFF

Settings that can be changed, let's number them.

11. Memory Range            - 00000000-07FFFFFF
22. Memory Range            - 00000000-0000FFFF
33. Input/Output Range      - 0000-00FF
44. IRQ - 03                         
55. Memory Range            - 00000000-0001FFFF

I'm shooting in the dark because I don't know what to change.
Changing memory ranges can end on a frozen Windows 98 logo (new!) or the usual blank screen upon booting the system (no change).
Changing the input/output range does something stranger. When the system boots instead of the blank signal you get an image of the last safe mode screen you saw, with artifacts.

Can anybody give me possible ranges to try?

Any other suggestions are welcomed!

Can you tell what on the motherboard those ranges are conflicting with? I had a problem sort of like that with a video card running in a 64 bit pci-x slot but it wouldn't tell me what it conflicted with.

I had to snap a pic and manually go though all items on the device manager to see what was using the same resources to be able to figure out what to use. Turned out all the different settings the graphics card wanted to use were used by something else but I was at to change the other devices settings so the graphics card could have what it needed.

tauro wrote on 2023-02-27, 15:05:

I assure you that at least with this board, there are serious problems with most cards.

Do you mean this model in general, or your particular board?
I worked at Zida national distributor in 1998 and while I dont remember us selling any Zida LX boards, all other modes (TX, BX intel and VIA, and my favorite budget option ZX ones) were solid and stable with no quirks or problems. In 98-99 period we maybe got a couple Zida returns, none for Asus, and tons of pcchips 😀.

tauro wrote on 2023-02-27, 15:05:

I'm doing all tests with a 66MHz bus.

with tualatin or coppermine?

tauro wrote on 2023-02-27, 15:05:

Can anybody give me possible ranges to try?

I dont remember ever having to mess with any of that. I would go basic and try installing win95 of win xp. Measuring psu voltages wouldnt hurt either.

tauro wrote on 2023-02-27, 15:05:

These are the settings that can't be changed: […]
Show full quote

These are the settings that can't be changed:

Copy code to clipboard

1Memory Range            - 000A0000-000AFFFF
2Memory Range            - 000B0000-000BFFFF
3Input/Output Range      - 03B0-03BB
4Input/Output Range      - 03C0-03DF
5Memory Range            - 000C0000-000CCFFF

Settings that can be changed, let's number them.

Copy code to clipboard

11. Memory Range            - 00000000-07FFFFFF
22. Memory Range            - 00000000-0000FFFF
33. Input/Output Range      - 0000-00FF
44. IRQ - 03                         
55. Memory Range            - 00000000-0001FFFF

Ok those top ones that cannot be changed are common to many Intel chipsets.
Why is a Video AGP at IRQ3 ? Most video cards PCI and AGP are IRQ'ed above IRQ9, many use IRQ 11 or share with IRQ10 OR use a virtual IRQ like 16 or 17.
In the BIOS do you have PnP OS as enabled ? Even if not something is amiss...

downgrading and resetting the bios are also good ideas just to eliminate potential causes

Is this a known thing, about Coppermine 566 processors being compatible with PPGA socket 370?

Zeerex wrote on 2023-02-28, 04:35:

Is this a known thing, about Coppermine 566 processors being compatible with PPGA socket 370?

not that Im aware of

rasz_pl wrote on 2023-02-27, 23:48:

Do you mean this model in general, or your particular board?
I worked at Zida national distributor in 1998 and while I dont remember us selling any Zida LX boards, all other modes (TX, BX intel and VIA, and my favorite budget option ZX ones) were solid and stable with no quirks or problems. In 98-99 period we maybe got a couple Zida returns, none for Asus, and tons of pcchips 😀.

Cool! The board is great overall, very solid and compact. Those are the reasons I like it too!

rasz_pl wrote on 2023-02-27, 23:48:

with tualatin or coppermine?

Coppermine, since the L1/L2 caches don't work with Tualatin processors on this board with the latest BIOS (1.10)

rasz_pl wrote on 2023-02-27, 23:48:

I dont remember ever having to mess with any of that. I would go basic and try installing win95 of win xp. Measuring psu voltages wouldnt hurt either.

I'm using a very good PSU and I also tested the board with another one. I measured the CPUs and they are getting the right voltage. Another plus for this board, since apparently other boards from this era didn't provide the right voltages for Coppermines and Tualatin.

Horun wrote on 2023-02-28, 01:03:

In the BIOS do you have PnP OS as enabled ? Even if not something is amiss...

Yes, it is enabled.

Grem Five wrote on 2023-02-27, 17:34:

Can you tell what on the motherboard those ranges are conflicting with? I had a problem sort of like that with a video card running in a 64 bit pci-x slot but it wouldn't tell me what it conflicted with.

Great idea @Grem Five!
I didn't think it was possible that Win9x assigned resources to different things and didn't warn you about it.
Were you able to boot the WinXP SP3 installer, Windows XP mini, or any other system that used your graphics card in an otherwise conflicting way? Just to know
I was surprised to see that in that install Windows assigned resources to multiple devices at the same time, the IRQ for example, was the same for the mouse and the AGP video card (R9200).

So I wiped everything and transcribed all the resources that are assigned by Windows 98SE to the different devices upon a clean install. Using a FX5200 instead of the R9200 because the latter creates two video devices and introduces further complexities.

1IRQ
2 0 - System timer
3 1 - Keyboard
4 2 - Programmable interrupt controller
5 4 - Direct memory access controller
6 8 - System CMOS/real time clock
710 - Intel 82371AB/EB PCI to USB Universal Host Controller
812 - Mouse
913 - Numeric data processor
1014 - Intel 82371AB/EB PCI Bus Master IDE Controller
11
12I/O Range
130060-0060 - Keyboard
140064-0064 - Keyboard
1501F0-01F7 - Intel 82371AB/EB PCI Bus Master IDE Controller
1603F6-03F6 - Intel 82371AB/EB PCI Bus Master IDE Controller
17FFA0-FFAF - Intel 82371AB/EB PCI Bus Master IDE Controller
180000-000F - Direct memory access controller
190080-0090 - Direct memory access controller
200094-009F - Direct memory access controller
2100C0-00DE - Direct memory access controller
22D000-DFFF - Intel 82443LX/EX Pentium(r) II Processor to AGP controller
230274-0277 - IO read data port for ISA Plug and Play enumerator
240374-0377 - IO read data port for ISA Plug and Play enumerator
250338-033B - IO read data port for ISA Plug and Play enumerator
260238-023B - IO read data port for ISA Plug and Play enumerator
2704D0-04D1 - Motherboard resources
280CF8-0CFF - Motherboard resources
290010-001F - Motherboard resources
300022-003F - Motherboard resources
310050-0052 - Motherboard resources
320072-0077 - Motherboard resources
330091-0093 - Motherboard resources
3400A2-00BE - Motherboard resources
3500DF-00DF - Motherboard resources
366100-613F - Motherboard resources
375F00-5F0F - Motherboard resources
3800F0-00FF - Numeric data processor
390020-0021 - Programmable interrupt controller
4000A0-00A1 - Programmable interrupt controller
410070-0071 - System CMOS/real time clock
420061-0061 - System speaker
430040-0043 - System timer
44EF80-EF9F - Intel 82371AB/EB PCI to USB Universal Host Controller
45
46
47Memory Range
48FCA00000-FEAFFFFF - Intel 82443LX/EX Pentium(r) II Processor to AGP controller
49DC800000-EC8FFFFF - Intel 82443LX/EX Pentium(r) II Processor to AGP controller
50F0000000-F7FFFFFF - Intel 82443LX/EX Pentium(r) II Processor to PCI bridge
5100000000-0009FBFF - System board extension for PnP BIOS
520009FC00-0009FFFF - System board extension for PnP BIOS
53000E0000-000FFFFF - System board extension for PnP BIOS
5400100000-0FFFFFFF - System board extension for PnP BIOS
55FEC00000-FEC00FFF - System board extension for PnP BIOS
56FEE00000-FEE00FFF - System board extension for PnP BIOS
57FFFE0000-FFFFFFFF - System board extension for PnP BIOS
58
59=========================================================
60PCI bus Properties

61  Settings -> Use Hardware
62
63  IRQ Steering - YES
64Get IRQ table using ACPI BIOS                      - YES
65Get IRQ table using MS Specification table         - YES
66Get IRQ table from Protected Mode PCIBIOS 2.1 call - NO
67Get IRQ table from Real Mode PCIBIOS 2.1 call      - YES
68
69  IRQ Routing Status
70IRQ Steering Enabled.
71IRQ Table read from MS IRQ Routing Specification.
72IRQ Miniport Data processed successfully.
73IRQ Miniport loaded successfully.

This is how the VGA device is configured:

1STANDARD PCI Graphics Adapter (VGA)
2
3000A0000-000AFFFF
4000B0000-000BFFFF
503B0-03BB
603C0-03DF
7IRQ 11
8FD000000-FDFFFFFF
9E0000000-E7FFFFFF
10000C0000-000CF7FF
11FCA00000-FCA1FFFF

Is there anything out of the ordinary?

After installing the drivers this is the configuration assigned by Windows:

1Memory range
2000A0000-000AFFFF   -- No change
3000B0000-000BFFFF   -- No change
403B0-03BB           -- No change
503C0-03DF           -- No change
6IRQ 03 		        -- CHANGED
700000000-00FFFFFF   -- CHANGED
800000000-07FFFFFF   -- CHANGED
9000C0000-000CF7FF   -- No change
1000000000-0001FFFF   -- CHANGED 

So I reverted every setting to the way it was before installing the driver. And... I still get the blank screen.

I tried other memory ranges but I couldn't make any progress.
I'm willing try any suggestions for ranges or IRQs!

In the next post I'm going to make an interesting reveal about the 1.09 BIOS.

rasz_pl wrote on 2023-02-28, 01:44:

downgrading and resetting the bios are also good ideas just to eliminate potential causes

This was key to discover when the AGP incompatibility was introduced.
Luckily I backed up the v1.09 BIOS to a file before flashing v1.10.

I reverted back to that BIOS. I haven’t tested this on Windows 98 yet, but based on live-CDs, these are my findings:

1	v1.09
2Mendocino	AGP cards work fine
3Coppermine	L2 doesn’t work! but AGP cards work fine (486 speed though)
4	
5	v1.10
6Mendocino	AGP cards work fine
7Coppermine	L2 works, but with the aforementioned AGP problems for most cards

So this further ignites my desire to mod the BIOS.
I don't know which way it's better to go, to mod the v1.09 to add Coppermine (and Tualatin) compatibility or to mod the v1.10 to fix the AGP incompatibility. Probably the easiest is to add the correct microcode to the v1.09, but will this break AGP compatibility...?

As I mentioned in my first post, I don't know how to extract the microcode from either of these BIOSes, as they don't seem to be fully compatible with AMIBCP 6.24 / 7.01.01 / 7.60.04 as the P6 Microcode item is not listed and I don't know about other software to mod the BIOS.
Can anybody help me with this? It should be straightforward, shouldn't it?

Using AMIBCP 6.24, both BIOS versions have the same values in "INTEL PIIIX4 BootBlock Table".

AMIBCP 7.01.01 doesn't let you edit the Register Table (it appears as empty).
Inside the option "Edit PCI IRQ Routing Table", both BIOS versions look the same.

There are a few modules that can be erased/modified. I don't think this is the right thing to mod but I tried it anyway:

1st mod
I took the PCI module from the 1.09 bios and inserted it at the same location with the same identifier in the 1.10 bios file. L2 works, but I get the same blank screen upon booting an OS (no change).

2nd mod
Adding the DMI module from 1.10 bios to the 1.09 bios. The L2 cache still doesn't work (no change).

Something to note is that whenever you try to use a Coppermine on the 1.09 you get this message:
"USB Device(s): Keyboard Mouse Hub Floppy Unknown"
It also appears when you try to use a Tualatin on 1.10.
That immidiately gives you a hint, the L2 cache is not going to work...
Also, inside the BIOS menu, the L2 cache is marked as "WriteThrough" instead of "WriteBack" and that also is a hint that the L2 is not going to work.

So I'm completely out of ideas. The good part is that I know it's something doable/fixable but I'm not knowledgeable enough to mod it.

I also tried to use other BIOS files, from different Zida boards. I deduced that the "base" LX model is the 6DLX (since the BIOS for my board has this name in the BIOS string). On this website you can even get the 1.12e BIOS version, but it doesn't fix anything as far as AGP problems are concerned.

EDIT
I'm uploading the original v1.09 BIOS to this post because I haven't found it online, in case someone wants to tinker and add the microcodes 😄 🙏

tauro wrote on 2023-03-02, 20:44:

Were you able to boot the WinXP SP3 installer, Windows XP mini, or any other system that used your graphics card in an otherwise conflicting way

Nobody don`t know why, but LX440 with CPU with SSE not working with Windows 2000/XP.
But works with 95/98.
Unfortunately, LX it`s not "almost BX"
Mendocino/Deschutes is maximum CPU for this chipset

Re: 440LX and Coppermine

shevalier wrote on 2023-03-04, 16:49:

Nobody don`t know why, but LX440 with CPU with SSE not working with Windows 2000/XP. But works with 95/98. Unfortunately, LX it` […]
Show full quote

Nobody don`t know why, but LX440 with CPU with SSE not working with Windows 2000/XP.
But works with 95/98.
Unfortunately, LX it`s not "almost BX"
Mendocino/Deschutes is maximum CPU for this chipset
Re: 440LX and Coppermine

Very interesting. So there's a chipset/hardware limitation instead of a BIOS limitation because the 440LX chipset doesn't fully/correctly support SSE instructions.
So Win9x works because it doesn't use SSE out of box, and WinXP uses it right out of the gate, when you start the Install process.

That got me thinking, maybe it's not a Memory range/IRQ problem but maybe most video drivers use SSE in some way and that's what makes Win9x freeze when it boots with a Coppermine CPU and an AGP video card. I don't know...

I can confirm that the FX5200 and R9200PRO both work fine with a Mendocino on Win9x and the WinXP installer starts too.

For the record, I tried the TNT2 M64 (this one worked with Win9x and a Coppermine with the right drivers) and started the WinXP install program and... it freezes at "Setup is starting Windows". This card works with a Coppermine on Win9x, so it goes beyond an AGP limitation and proves your point of a chipset limitation.

If the 440LX chipset doesn't support Coppermine CPUs that's a black pill. The AGP video cards choice reduces a lot and kind of defeats the purpose of using a faster CPU. Plus, considering the R9200 does automatic 4:3 resizing on DOS... I'll stay with the Mendocino for now.

I'm not sure what will be more stable, the 400@506MHz or the 433@547MHz, I'll do more tests regarding that. They get hot and there's very little room for a medium size cooler to fit.

Take care out there @shevalier!

I was trying bigger fans very gently but somehow I destroyed a capacitor...
I can replace it but I don't know what value it should be.
Can anybody please tell me approximate values and where I could find them? 🙏
I have a few scrape socket 370 boards and others.

rasz_pl wrote on 2023-02-28, 05:25:

Zeerex wrote on 2023-02-28, 04:35:

Is this a known thing, about Coppermine 566 processors being compatible with PPGA socket 370?

not that Im aware of

Yes it's a known thing. They are PPGA, not FCPGA. I have a Coppermine Celeron 566 and it works in PPGA adapters while the rest of Celerons do not. They are the best choice for unmodded PPGA 440LX boards as they will be stable at 710Mhz for sure (8.5 x 83.3). At that speed they should be equivalent to PIII 600 Katmai.

I removed and measured the cap on the left, it's 0.1µF. So I put another 0.1µF in place of the missing one and... sad news.
Mendocinos give me a 00 error on the POST card, that is to say, they don't even post.
Coppermines get further, they initialize, they check the memory and video but then they halt with a 01 error. No POST.
The only one that still works is the 566MHz Coppermine Celeron!

My suspicion is that I didn't break the capacitor right now, it was probably missing/broken long ago, but I only noticed it now because I started having these problems. I suspect that I broke something else, the question is, what? I couldn't find any damaged traces. I only tried a couple of heatsinks that didn't fit and I did it gently.
I removed the mod but it didn't help, Mendocinos still fail to POST with error 00.

Just when I was about to settle with a Mendocino and put everything in a case... this happens... too much. Sad turn of events. Frustration. The fact that the Coppermine 566MHz still works adds more mystery to the problem.

If you have any ideas about what to measure/replace/do, please share them.