eliot_new wrote on 2026-05-03, 07:23:

Hello everybody […]
Show full quote

Hello everybody

Does anybody have an explanation about a phenomenon?

I have ECS P5HX-A rev2.1 with a BIOS 1.2 from 08/26/96
This is the very first BIOS and it accepts a 2nd TAG-SRAM Chip.

As soon as I flash the BIOS 09/12/97 the board refuses to boot untill I remove the 2nd TAG-SRAM:
https://theretroweb.com/motherboards/s/ecs-p5 … er.-2.0-pcb-2.1

What is amazing and looks like a trade-off:
1. 96 BIOS accepts a 2nd TAG but shows only 8GB HDD and no USB enabled option visible
2. 97 BIOS doesn't accept 2nd TAG but shows fully 32GB HDD (i.e. the SanDisc sdcard) and I see USB enabled within the integrated peripherals

Yesterday I even run Speedsys and Cachechk and it seems that the 2nd TAG is really working with 128MB EDO
Now I have to decide, before I setup a dual-boot DOS/WIN98SE what is more important.

1. using 128MB RAM, limiting to only 8GB (I have to repartition the 32GB sdcard, currently it has 2GB FAT16, 8GB FAT32, 22GB FAT32) and using an extra USB 2.0 PCI card
2. using only 64MB RAM, having full 32GB recognized and using the integrated USB 1.1 Controller

Are there any "WIN9x" games in the era 1996...1999 which needs more than 64MB RAM?
I bought this board for cheap because I always wanted an i430HX based board.

What I really appreciate is, that:
1. ATX, i.e. I can plugin PS/2 mouse and PS/2 keyboard without an adapter
2. ATX power plug (better than Gigabyte 586HX or Asus P55T2P4 with their BAT plugs)
3. RTC Dallas Chip sits on a socket and is easily removable with a screw driver
4. It accepts a 2nd TAG-SRAM Chip and according to speedsys it caches > 64MB RAM
5. It has white PCI and ISA slots, which looks really awesome
6. It has an USB connector (the Gigabyte 586HX rev1.54 hasn't)
7. The minimum VCore is 2.5V, this means that maybe at least a K6-3/400 nonplus with 2.4V will work

The respective BIOS release notes don't point to anything much between 1.2 & 1.2q - https://web.archive.org/web/20010621174750id_ … e/p5hxaa10.html. I do see one old forum post elsewhere saying 1.2 worked on the board but 1.2q didn't, though it didn't mention board revision or the nature of the failure.

Can't imagine it'd help, but the original 1.2q update file can be foud at https://www.elhvb.com/mobokive/archive/Ecs%20 … bios/hxa12q.zip

Did you use 16Kx8 or 32Kx8 for the second TAG RAM?
Do you have 512K L2 cache onboard or did you upgrade from 256K with a COAST stick? What cache (COAST) module did you use to upgrade L2 cache? With just 256K (onboard) L2 cache there´s no cacheable area larger than 64MB. Note that to achieve this it´s not sufficient BIOS "accepts" the 2nd TAG chip but there must also be 512K L2 cache present and detected correctly.
I remember that all the common COAST sticks didn´t work.
I did some testing with this mainboard a while ago:
ECS / Elitegroup P5HX-A L2 cache upgrade

USB BIOS option only means whether BIOS sees USB keyboard not USB functioning in OS.

I would patch eariler BIOS with Bios Patcher 4.23 which would remove HDD size limitations.

majestyk wrote on 2026-05-04, 04:46:

I did some testing with this mainboard a while ago:
ECS / Elitegroup P5HX-A L2 cache upgrade

Hi majestyk,

Thank you for your write-up about the L2 cache mods on this P5HX-A board. This is great information!
I’m helping the OP via PM to find a better BIOS, or otherwise patch the v1.2q BIOS, for better CPU support and fix the 32GB and 64GB IDE HDD limit bugs.

I’ve found an Award upgrade BIOS by Unicore Software for the ECS P5HX-A Rev 2.x board. This 03/2000 BIOS has all the usual goodies like K6-2+/III+ and 128GiB drive support.
If you still have the P5HX-A, could you help test this Unicore upgrade BIOS on your modded board? Thanks in advance.
Here is a copy:

Paadam wrote on 2026-05-04, 05:30:

I would patch eariler BIOS with Bios Patcher 4.23 which would remove HDD size limitations.

Hi Paadam,

Using Apple_ROM’s BIOS Patcher tool on the 09/1997 Ver. 1.2q BIOS would indeed remove the 32GB and 64GB drive limit bugs and add better CPU support.
But this won’t work on the earlier 1996 BIOS. This 1996 BIOS has no Int 13h extensions and is therefore limited to 8GB drives.
Afaik, the BIOS Patcher tool can’t add these extensions and is unable to break the 8GB limit. It can only fix the limit bugs in BIOSes that already have the Int 13h extensions.

Other solutions are needed to break the 8GB barrier in these earlier BIOSes, like EZ-Drive, XTIDE, or a BIOS upgrade.

Cheers, Jan

Chkcpu wrote on 2026-05-04, 18:50:

Hi majestyk, […]
Show full quote

majestyk wrote on 2026-05-04, 04:46:

I did some testing with this mainboard a while ago:
ECS / Elitegroup P5HX-A L2 cache upgrade

Hi majestyk,

Thank you for your write-up about the L2 cache mods on this P5HX-A board. This is great information!
I’m helping the OP via PM to find a better BIOS, or otherwise patch the v1.2q BIOS, for better CPU support and fix the 32GB and 64GB IDE HDD limit bugs.

I’ve found an Award upgrade BIOS by Unicore Software for the ECS P5HX-A Rev 2.x board. This 03/2000 BIOS has all the usual goodies like K6-2+/III+ and 128GiB drive support.
If you still have the P5HX-A, could you help test this Unicore upgrade BIOS on your modded board? Thanks in advance.
Here is a copy:

The attachment 2A59FE1C.zip is no longer available

Paadam wrote on 2026-05-04, 05:30:

I would patch eariler BIOS with Bios Patcher 4.23 which would remove HDD size limitations.

Hi Paadam,

Using Apple_ROM’s BIOS Patcher tool on the 09/1997 Ver. 1.2q BIOS would indeed remove the 32GB and 64GB drive limit bugs and add better CPU support.
But this won’t work on the earlier 1996 BIOS. This 1996 BIOS has no Int 13h extensions and is therefore limited to 8GB drives.
Afaik, the BIOS Patcher tool can’t add these extensions and is unable to break the 8GB limit. It can only fix the limit bugs in BIOSes that already have the Int 13h extensions.

Other solutions are needed to break the 8GB barrier in these earlier BIOSes, like EZ-Drive, XTIDE, or a BIOS upgrade.

Cheers, Jan

Dear Jan, thx so much! I will test the Bios asap and will report the result here. Kind regards

Chkcpu wrote on 2026-05-04, 18:50:

Cheers, Jan

Dear Jan, as I m back home now, I have tested your bios.
Unfortunately I get this error message after starting AWDFLASH and tipping: 2A59FE1C.bin as source.
Maybe you can modify anyway an older BIOS for the ECS P5HX-A, please?

Have you tried Uniflash?

Or just /F parameter to force flashing
(Warning! This can brick your board) 😀))

eliot_new wrote on 2026-06-13, 11:09:

Dear Jan, as I m back home now, I have tested your bios. Unfortunately I get this error message after starting AWDFLASH and tipp […]
Show full quote

Chkcpu wrote on 2026-05-04, 18:50:

Cheers, Jan

Dear Jan, as I m back home now, I have tested your bios.
Unfortunately I get this error message after starting AWDFLASH and tipping: 2A59FE1C.bin as source.
Maybe you can modify anyway an older BIOS for the ECS P5HX-A, please?

You get this error message from AWDFLASH because the BIOS IDs don’t match.
ECS changed the BIOS ID in your 09/12/97 BIOS to (P5HX-A)C, while the Unicore BIOS has the original 2A59FE1C BIOS ID.

You can force the flash with the /Py switch. So try this command to flash:
AWDFLASH 2A59FE1C.BIN /Py /CC /CP /CD

The three /Cx switches clear the CMOS, PnP, and DMI data of the old BIOS. Adding these three switches often helps for a successful BIOS flash.

Cheers, Jan

Chkcpu wrote on 2026-06-13, 20:06:

You get this error message from AWDFLASH because the BIOS IDs don’t match. ECS changed the BIOS ID in your 09/12/97 BIOS to (P5H […]
Show full quote

eliot_new wrote on 2026-06-13, 11:09:

Dear Jan, as I m back home now, I have tested your bios. Unfortunately I get this error message after starting AWDFLASH and tipp […]
Show full quote

Chkcpu wrote on 2026-05-04, 18:50:

Cheers, Jan

Dear Jan, as I m back home now, I have tested your bios.
Unfortunately I get this error message after starting AWDFLASH and tipping: 2A59FE1C.bin as source.
Maybe you can modify anyway an older BIOS for the ECS P5HX-A, please?

You get this error message from AWDFLASH because the BIOS IDs don’t match.
ECS changed the BIOS ID in your 09/12/97 BIOS to (P5HX-A)C, while the Unicore BIOS has the original 2A59FE1C BIOS ID.

You can force the flash with the /Py switch. So try this command to flash:
AWDFLASH 2A59FE1C.BIN /Py /CC /CP /CD

The three /Cx switches clear the CMOS, PnP, and DMI data of the old BIOS. Adding these three switches often helps for a successful BIOS flash.

Cheers, Jan

Dear Jan, thank you very much!

I have succeeded in flashing bios with:

AWDFLASH 2A59FE1C.BIN /Py /CC /CP /CD

But now a small issue appears:

Neither DOS nor Win98SE recognizes my LG DVD ROM along with the IDE2SATA Adapter.
With the older Bios 1.2q from 1997 it works fine.

Could be the reason, that the Unicorn Bios works only with K6, K6-2 or K6-3 CPUs?
I have remained with the P166MMX after flashing to test stability and behaviour.

Tomorrow I will reflash again the newest bios, install my 3x CPUs K6, K6-2, K6-3 and write the results here.

Chkcpu wrote on 2026-06-13, 20:06:

ECS changed the BIOS ID in your 09/12/97 BIOS to (P5HX-A)C, while the Unicore BIOS has the original 2A59FE1C BIOS ID. […]
Show full quote

ECS changed the BIOS ID in your 09/12/97 BIOS to (P5HX-A)C, while the Unicore BIOS has the original 2A59FE1C BIOS ID.

You can force the flash with the /Py switch. So try this command to flash:
AWDFLASH 2A59FE1C.BIN /Py /CC /CP /CD

The three /Cx switches clear the CMOS, PnP, and DMI data of the old BIOS. Adding these three switches often helps for a successful BIOS flash.

Cheers, Jan

Dear Jan, now I have installed a K6-2 400 as well as a K6-3 400 (non plus).
Both work and are recognized, I can even play Quake 1 in pure DOS 6.22.

But unfortunately neither the CDROM nor the PS/2 Mouse work, I had to reflash the old 1997 bios.
Nor do Win98SE start properly, it freezes few seconds after the desktop appears.

In case you had some free time this summer, I would highly appreaciate if you could please make a modified bios based upon the 1997 edition?
Below I add some pictures for documentation.

I’m sorry to hear the the Unicore BIOS gives you problems. This sometimes happens with these upgrade BIOSes, but most of them work fine.
Anyway, it was worth a try.

I have now put the 1997 1.2q BIOS on my “to be patched“ list but it will take some time before it is ready. I have several projects to finish first. 😉

I will let you know when the patched ECS P5HX-A BIOS is ready.
Greetings, Jan

Update: Jan sent me the link to the patched BIOS.
Thank you very much dear Jan, you are a genious!
I confirm that it works:

http://www.steunebrink.info/bios/P5HXA_J2.zip

My results:
1. DVDROM drive on sata2ide works 100% I can install games,Win98SE, Creative driver CD etc
2. 128GB,64GB,32GB SanDisk are recognized
3. K6-233 ANR works as K6-200 because the max multi is 3.0x
4. K6-2 400 2.2V works although min VCore jumper is 2.5V: setting 2x Multi, 66Mhz, I can activate write allocation and write combining with K6 Tool in autoexec.bat
5. K6-3 400 2.2V boots but PC freezes on message ...ESS Update success...
6. K6-3 400 2.4V even doesnt boot, but it makes a beep sequence
7. Games tested:
Doom1
Doom2
Duke3D
Quake1 Software and GLQuake with V1 4MB
Quake2 with V1 4MB: only 3.20 works, 3.21 crashes immediately
Tomb Raider 1

Win98SE: it shows that I have only 62MB EDO instead of 64, but this seems to be only a display bug.
Furthermore DirectX 7.0 works with dxdiag: I see the rotating cube, everything ok.

SETMUL can only deactivate L1.
I cannot slow down K6-2 400 like the P166MMX with Phils slow,middle,fast.bat
Nor can I throttle it with cpudspd like it works on my ALi5 boards

Hardware:
ECS P5HX-A 2.1 ATX i430HX
K6-2 400
64MB EDO
Spea Mirage V7 2MB
Diamond Monster 3dfx
AWE64 value
PicoGUS 2.0
MSDOS6.22 dualboot Win98SE

Q1 timedemo 30 fps
Q2 timedemo 29.5 fps
DeusEx1 timedemo 15 fps with lowest settings

Why is the ECS a charm?
1. Cheaper than Asus P55T2P4 ATX Version
2. All slots are looking cool: White
3. It supports a 2nd sram 256kb to cache >64MB or you can use the Brown slot

Thanks for testing the patched P5HX-A 1.2q BIOS. It’s great to hear it works well. 😀

About the issues with the K6-III nonplus, I can think of 2 reasons that may cause this.

1) I know that on older socket 7 chipsets like the Intel 430TX, 430VX, ALi Aladdin IV(+), or SiS5582, you may have to set "System BIOS Cacheable" in the CMOS Setup to "Disable" to get it working with a K6-III, K6-2+, or K6-III+.
It is possible that this is also needed on your 430HX board. Please try if this helps.
2) The P5HX-A still has linear voltage regulators. For the Vcore an LT1584 or equivalent low dropout regulator is usually used, and this regulator can supply a maximum current of 8A. This is enough for a Pentium MMX 233 or K6/233, but the K6-III/400 can draw up to 12 to 14A, overloading the regulator. This will cause the Vcore to become unstable.

Note that the K6-2/400 can draw up to 10A under full load, so this is also a mild overload. Although the K6-2 works well, it causes a lot of heat to be dissipated by the regulator and may burn it out on the long run.
So I advice an interposer with its own regulator when using a K6-2/K6-III on this board.

The K6-2+/III+ uses a lot less power and could be used with a linear regulator. However the lowest Vcore setting on this board is 2.5V. This is much to high for these 2.0V CPUs. Maybe there are undocumented lower Vcore settings on the P5HX-A?

Cheers, Jan

Chkcpu wrote on Yesterday, 16:30:

1) I know that on older socket 7 chipsets like the Intel 430TX, 430VX, ALi Aladdin IV(+), or SiS5582, you may have to set "System BIOS Cacheable" in the CMOS Setup to "Disable" to get it working with a K6-III, K6-2+, or K6-III+.
It is possible that this is also needed on your 430HX board. Please try if this helps.

What is the story behind that? Does the larger L2 internal cache interfere with the BIOS sizing algorithm for the external cache? Is this a software or hardware issue?

jakethompson1 wrote on Yesterday, 17:48:

Chkcpu wrote on Yesterday, 16:30:

1) I know that on older socket 7 chipsets like the Intel 430TX, 430VX, ALi Aladdin IV(+), or SiS5582, you may have to set "System BIOS Cacheable" in the CMOS Setup to "Disable" to get it working with a K6-III, K6-2+, or K6-III+.
It is possible that this is also needed on your 430HX board. Please try if this helps.

What is the story behind that? Does the larger L2 internal cache interfere with the BIOS sizing algorithm for the external cache? Is this a software or hardware issue?

Lacking a logic analyzer, I was unable to find the root cause of this cacheability problem, but I believe it is a hardware timing issue between these classic socket 7 chipsets and CPUs with on-die L2 cache.
Here is the story:

I became first aware of this issue back in 2002. After successfully patching several 1999 super socket 7 BIOSes for K6-2+/III+ support, I tried to do the same on 1997/1998 BIOSes for i430TX boards.
Here I experienced a BIOS hang at the end of POST, just before the BIOS hands-over control to the bootloader of the active drive. Disabling the External cache didn’t help, but when disabling the Internal cache, the boot completed normally.
After a lot of trail and error, I found that disabling the ‘System BIOS Cacheable’ option in the BIOS fixed the problem and allowed normal boot and stable operation with all caches enabled on a K6-III, K6-2+, or K6-III+ CPU.

When I found this issue also on other TX boards, I thought this was an issue with the i430TX chipset alone, but later I found the same issue with the 430VX, ALi Aladdin IV(+), and SiS5582.
Note that the issue is not present on super socket 7 chipsets like VIA MVP3 and Ali Aladdin V. These boards function normally on K6-2+/III(+) CPU with System BIOS Cacheable enabled.

Hypothesizing about the cause, I keep coming back to a hardware issue.
The System BIOS Cacheable BIOS option directly toggles the bits in a chipset register that controls the ability to cache the F000h-FFFFh region in the Internal CPU cache(s). Although no BIOS routines are involved here that could explain a timing issue, the larger Internal L2 cache of the K6-2+/III(+) allows more of the BIOS runtime code to be cached. Potentially this could mean that a piece of time sensitive BIOS code could now be executed much faster from the CPU cache, instead of from shadow RAM. However, I consider this unlikely because the super socket 7 BIOSes are not bothered by this issue, while they practically use the same interrupt service routines as their one year older socket 7 counterparts.

Jan

That is interesting.
That disabling external cache doesn't help rules out my sizing idea (that 256KB L2 cache could confuse the BIOS trying to size the 256KB external cache).
I wonder if a BIOS might ever flip chipset registers causing F000-FFFF shadow RAM to point to different data, and forget to do a wbinvd immediately afterward.
That idea could be tested by writing a DOS program to wbinvd and then flip the chipset register to turn F000-FFFF caching back on.