Jump to content

Research on Gigabyte GA-970A-DS3P: Fail


ragnargd

Recommended Posts

In a Nutshell: I was not able to rund Windows 98 SE on a Gigabyte GA-970A-DS3P.

Reason (probably): Despite the chipset, being well-known for its compatibility, some other onboard components map memory into the space W98SE cannot control, resulting in access violation or (even worse) memory corruption.

Details:

As my ASRock 890FX Deluxe5 is slowly dying (i already lost two of them), which i attribute to the enormous heat the voltage regulaters produce for my 125W TDP FX-8350, i have to find an alternative.

Ready for use lies my trusted ASRock 970 Pro3 R2.0 beneath me. I don't know any board exept for the 890FX Deluxe5 that can provide PCIe 8x on the second 16x slot to make XP or W10 usable for performant full-HD gaming, but x4 on the second 16x slot on the 970 Pro3 would still deliver reasonable framerates (about 15% loss against 8x or 16x) with, i.e., a GTX 970.

But then i stubled across the GA MoBo. I did not consider it at first, as the revisions known to me (1.0, 3.0) did not support neither my FX-8350 nor my FX-4320, but, lately, i found out that the Revision 2.1, which is only sold in Europe, does support both. As the GA MoBo resembles the ASRock 970 MoBo in most details (chipset, PS/2 ports, USB, you name it), even providing a better layout (having one more PCIe 1x slot free, and providing a better thermal position for my Audigy2), i bought one fresh from the shop, hoping for the best.

How did i test?

I did my usual routine, with components successfully used several times on other MoBos:

- I switched everything off (LAN, Sound, USB, UEFI, and everything else i could)

- put in a trusted bar of 1GB DDR3 (working fine in both my ASRock Boards with W98SE)

- a 256MB NVidia 7600 GT PCIe, passively cooled, sporting a VGA interface

- and a SATA II SSD (formatted, and filled with all files needed for installation)

Upon installation, i was immediately greeted with "out of memory", which already hinted to components using ressources that will conflict with W98SE (normally, a compatible system works without any modification with 1GB installed, although rloew reported cases where even this was too much. I did not find 512 MB bars for DDR3 in shops).

At first, although i did not have more that 1GB of memory installed, i used himemx.exe with the usual routine, which already got me a reboot further in the setup. But after the second reboot, memory violation stopped me from reaching the desktop.

I then used rloews memory-patch, at first on its own, then with the /M option, and finally with SPLIT8MB.EXE in autoexec.bat. All to no avail, sometimes (with drivers injected) i got to the desktop, but experienced severe memory corrution (display-artefacts, corrupted fileallocation tables), or simply got access violation before getting to the desktop.

What i also did: I successfully tested the MoBo with the HalfLife 2 "Lost Coast" benchmark on XP and W10-x64, to make sure the board is intact.

I also repeated every step with W98SE on my ASRock 970 Pro3 R2.0, and got a rock-solid system

If you can think of any step i still might try, please tell me, but otherwise, i think that pretty much is it.

Verdict: I still think that the AM3+ boards working with W98SE are more or less a lucky, but freakish incident.

Edited by ragnargd
content, grammar
Link to comment
Share on other sites


You, Sir, are a scholar. I did not even know these parameters exist... ;-)

Without PCI-bus, that would mean no soundcard, as the Audigy2 is PCI.

But, well, it is at least worth a try for scientific reasons... ;-)

btw: I just read "Exploring "SETUP /P I;S=detectbus" and Grub4Dos: 8½ times Win98se on USB-drive". Well done!

Edited by ragnargd
Link to comment
Share on other sites

I should have written "install W9x without PCI-bus firstif succesfull try your luck with wizard New Hardware." After installing the PCI bus and reboot once you will have full control which Device to install and which not. Do not reboot after installing PCI devices, first check Device Manager (view Connection) for double entrances. Especially the non-PCI VGA device should be deleted manually before reboot.

But you have already red my "Exploring SETUP /P I;S=detectbus......"-Members project. :cool:

The /P S= switch is very badly documented. All entrances in MSDET.INF can be used to skip during installation (seperated with ).

Link to comment
Share on other sites

The problem here is that of choice: Do i have to force this board into subservience, or do i work with something that does the job happily, as does the ASRock 970 Pro3 R2.0?

The ASRock board is newer, cheaper, and easier to get in shops (ASRock alsways offers their mainstream boards longer than any other company).

AND it is much easier to handle. We have a clear winner here, and it's not clear that even with some effort the GA MoBo will pass the finishing line at all.

Reason:

Even with the parameters given, success on the GA ist not guaranteed, as memory access violation is not the same as components hanging during detection:

The offending memory used by the (unspecified) onboard components is already occupied at boot time, and will stay so, while detection and ressource-/driver-assignment happens later.

But that's theory, we'll see.

 

Link to comment
Share on other sites

I agree that memory problem can be unsolvable.

Once I tried (from USB drive) MS-DOS7.1 on a Gigabyte board, if I remember well a Z170 species. Didn't work!

I am a happy user of my 2010 Asrock 960GM-GS3 AM3+ board (with FX 6100/DDR3 1600), compatible with everything except Win98se ACPI. 

Edited by deomsh
FX 6100, not FX 8300
Link to comment
Share on other sites

Hm, the Bios Setup of your GA board doesn't look very promising. I assume you run the SATA ssd in IDE mode. And what about the extreme low vcache settings I mentioned in my USB project?

If *something* is mapped in memory you can try Jack Ellis XMGR.SYS instead of HIMEMX.EXE (who gave you already an extra reboot).

Edited by deomsh
Typo
Link to comment
Share on other sites

The BISO-Setup actually is not worse that that of the ASRock, and regarding the PS/S emulation via USB, actually this could help on other boards without true PS/2. Not that it helps with the problem, of course. And yes, i run the SSD used for W98SE on the controller set to IDE Mode (on channels 5 + 6) - a feature only GigaGyte and ASRock offer (MSI does not, and Asus not any more). This will burn the SSD faster, but i don't use W98SE for everyday use, so no big deal here. If i had one wish for free, i'd ask the vendors to put in the bios-option to boot from the PCIe 4x slot (like was possible on the compatible GA-870A-DS3)...

I did not try XMGR.SYS yet - i like the fact it is in active development. Worth a try.

The low vcache... have to think about this... i don't understand yet, what that would help with memory violation.

After these tests, which i will do to learn, and of course to honor your efforts ... ;-) ... , i will sink rather more time in looking for other working boards. With Ryzen taking over the AMD-fanboybase, used AM3+ boards flood the market for cheap... :-D

 

Link to comment
Share on other sites

9 hours ago, ragnargd said:

The low vcache... have to think about this... i don't understand yet, what that would help with memory violation.

Has to do with the so called "System Arena" in win9x memory architecture. Since vcache is using it too, sometimes smalller is better. See https://www.pc-experience.de/wbb2/thread.php?threadid=1903

In *some* win95 KB (cannot find it anymore) MaxFileCache=8192 is mentioned as value to solve stability problems.

If MinFileCache is not set, MaxFileCache should > 1/24 of memory (RLoew's Rule of Thumb - he didn't mention MinFileCache needed).

According to my experimental findings lower values are possible, but always MinFileCache <= MaxFileCache.

In my "Smartdrive Revisited"-investigations I found a vcache of value 1024 (=1MB) as "best" (not related to win98 Setup, but to some problems using a win3.1 HD audio driver). In my USB-project the extreme low vcache value has especially the purpose to minimize System Arena footprint if Win98se is running on a Memdrive. 

My standard Win98se Setup values, set in SYSTEM.INI after copying files:

[386Enh]

MaxPhysPage=20000

ConservativeSwapfileUsage=1

DMABufferSize=64

[vcache]

MinFileCache=1024

MaxFileCache=1024

 

Link to comment
Share on other sites

Umm...try using the AHCI mode, and not IDE/Legacy/Compatibility for the connected SSD.

With the AHCI driver being publicly available, using compatibility mode is not required anymore:
https://archive.org/details/ahci_win9x

XMGR.SYS might fix things, I've never really used it, HIMEMX.EXE always solved everything for me, but I've seen a few times where XMGR worked better than HIMEMX.

Can you try a different GPU (even with the generic vga.sys/VBEMP driver?)

Link to comment
Share on other sites

Out of lack of time I will focus on more promising things first, before sinking more time in the GA-MoBo (which i will still do, but later - the memory-problems are something that may be solved).

I will get an ASRock 990FX Extreme4 soon, which, being very similar to the 890FX Deluxe5, has a good prospect of working quite well. Let's see.

The AHCI-driver is something i will definitely try out on that occasion, now that the drivers are available. Although, honestly, the IDE-Mode was working just fine. The SSD will burn faster, but as i don't work on that SSD every day, just a few times per year, i will most probably never notice.

Yeah, now i'll have some sleep. Thank you for your support, this is what got us to where we are.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...