Danger Manfred wrote on 2023-07-25, 22:24:

I was wondering if anyone still had a source for Epox 8KTA3+ (plus, non-pro) BIOS images . […]
Show full quote

I was wondering if anyone still had a source for Epox 8KTA3+ (plus, non-pro) BIOS images .

Unfortunately, with my current BIOS, my AMD Athlon XP-M 2600+ (Barton) is identified as an XP 2000+ with 133x12.5 clock rates.
Normally, these mobile Athlons shouldn't be locked, but all multipliers I can select are 8,9,10,11,12,13,14 - with 13 being interpreted as 12.5x and 14 as 5x.
I would expect this Athlon to have multipliers enabled from 5 to 24, that would make a hell of a difference on this board! Even just 16 would already be enough to make the Barton run really fast and make the board perfect for dual boot Win98 and WinXP gaming.
Also, the Vcore range is very small, I can't even run the CPU at default voltage (1.45V) because the BIOS only lets me choose voltages higher than that. Not much of a problem though, these Mobile Bartons are supposed to run anywhere from 1.45 to 1.7 within spec.

But the missing multipliers really bother me, the CPU could run at least 2.5 GHz if only I had some way of increasing the damn multiplier

A proper BIOS should do the trick, but Epox' page is down, and all Google spits out are BIOSes for the pro or non-plus versions of this board.

This any use? - https://theretroweb.com/motherboards/s/epox-p … 8kta3#downloads

Athlon XP-M (mobile ones) boots with a lower multi than rated default (usually 6x) -- something else is needed to actually configure the CPU to operate at rated speed.

The 13x being 12.5x and 14x being 5x are a common issue for KT133(A) AFAIK because the BIOS multi setting only touches four of the five multiplier bits on those CPUs. SetMul/CPUMSR can correctly set the multi to any possible value, but it depends on the other components on the motherboard to be able to let the CPU operate reliably at the desired multiplier.

Taken from a EP-8KTA3+ REV 1.0. Hope it helps! Please report back.

Danger Manfred wrote on 2023-07-25, 22:24:

I was wondering if anyone still had a source for Epox 8KTA3+ (plus, non-pro) BIOS images . […]
Show full quote

I was wondering if anyone still had a source for Epox 8KTA3+ (plus, non-pro) BIOS images .

Unfortunately, with my current BIOS, my AMD Athlon XP-M 2600+ (Barton) is identified as an XP 2000+ with 133x12.5 clock rates.
Normally, these mobile Athlons shouldn't be locked, but all multipliers I can select are 8,9,10,11,12,13,14 - with 13 being interpreted as 12.5x and 14 as 5x.
I would expect this Athlon to have multipliers enabled from 5 to 24, that would make a hell of a difference on this board! Even just 16 would already be enough to make the Barton run really fast and make the board perfect for dual boot Win98 and WinXP gaming.
Also, the Vcore range is very small, I can't even run the CPU at default voltage (1.45V) because the BIOS only lets me choose voltages higher than that. Not much of a problem though, these Mobile Bartons are supposed to run anywhere from 1.45 to 1.7 within spec.

But the missing multipliers really bother me, the CPU could run at least 2.5 GHz if only I had some way of increasing the damn multiplier

A proper BIOS should do the trick, but Epox' page is down, and all Google spits out are BIOSes for the pro or non-plus versions of this board.

Your motherboard is realy wrong for this cpu .No matter if you update bios or not you wont get what you want .

Okay, first of all thank you all for the replies, I'll reply in order.

PC Hoarder Patrol wrote on 2023-07-25, 23:52:

This any use? - https://theretroweb.com/motherboards/s/epox-p … 8kta3#downloads

Thanks, this one at least suggests I can use an Epox 8KTA3 (non-plus) BIOS, since it says "also known as".
However, the 04/09/2003 BIOS is exactly the one I am already using.

LSS10999 wrote on 2023-07-26, 01:36:

Athlon XP-M (mobile ones) boots with a lower multi than rated default (usually 6x) -- something else is needed to actually configure the CPU to operate at rated speed.

Actually it does boot with the default multiplier, I just want a higher one to make up for the fact that the board doesn't support the default FSB of the CPU (166 vs 133).

LSS10999 wrote on 2023-07-26, 01:36:

The 13x being 12.5x and 14x being 5x are a common issue for KT133(A) AFAIK because the BIOS multi setting only touches four of the five multiplier bits on those CPUs. SetMul/CPUMSR can correctly set the multi to any possible value, but it depends on the other components on the motherboard to be able to let the CPU operate reliably at the desired multiplier.

It's great to know that SETMUL can still set higher multipliers! My problem might be solved then! The motherboard is supplying enough voltage and its FSB isn't raised, so that should be fine. At least that's how I've seen it countless times on another such Board/CPU Combo: the Asrock 4CoreDual-VSTA (FSB 1066 max) with the Intel Core 2 Quad QX9650 (FSB default 1333) - you just run the CPU at a lower FSB than default and crank up the multi instead, as long as the resulting frequency is the same, it shouldn't be a problem for the board, given it can deliver enough power. And the Mobile Bartons use significantly less power (45W instead of 65W), so that should work™.

tauro wrote on 2023-07-26, 06:20:

Taken from a EP-8KTA3+ REV 1.0. Hope it helps! Please report back.

I need a few more details before I flash that, what is the date of this BIOS? The one I am already using is 04/09/2003. Have you used it and been able to access multipliers lower than 8 or higher than 12.5?

ciornyi wrote on 2023-07-26, 06:48:

Your motherboard is realy wrong for this cpu .No matter if you update bios or not you wont get what you want .

Don't worry, as I have written above, I have already seen and done similar things (run a CPU that's not officially supported at a lower FSB yet higher multiplier to match its original speed) on several occasions, from early socket 7 boards with K6-II processors to the aforementioned Asrock/QX9650 combo. It usually works fine as long as you don't stress the boards power delivery too much and cool the CPU well.

Again, thank you all very much!

Danger Manfred wrote on 2023-07-26, 07:27:

tauro wrote on 2023-07-26, 06:20:

Taken from a EP-8KTA3+ REV 1.0. Hope it helps! Please report back.

I need a few more details before I flash that, what is the date of this BIOS? The one I am already using is 04/09/2003. Have you used it and been able to access multipliers lower than 8 or higher than 12.5?

I can't tell what version it is or if it's compatible with that CPU because the board is not working.
I think you already have the latest version. Both 8KTA3 and 8KTA3+ share the same bios. Here's the archived link anyway, just to be sure:
http://web.archive.org/web/20040615003534/htt … ios/socketa.htm
http://web.archive.org/web/20051231124349/htt … OS/8KT33409.zip

Oh well, for now I think I'll install 3DMark01 and confirm that SETMUL does the jerb - if so, I don't even need another BIOS. I'll just underclock the CPU for Win98SE and crank it back up to 2600+ for XP. I'll report back in a few hours!

So... the CPU works great (see picture 2) until I try to set *anything* with setmul. Instant Freeze. (see picture 1).

I tried 18 or 16 first, then, thinking it might just be too much, I tried 14 (the first multiplier the board turns into 5x instead) and 8 (which is a severe underclock from the default 12.5).

No matter what I do, result is an instant freeze at picture 1.

I feel a little stupid for having made pictures with my phone but you know... a frozen machine won't do any screenshots, plus it was much more convenient since the PC isn't in my network yet and I would have had to save the file, copy it onto some physical medium and carry that to my living room PC, so... it is what it is.

Anyway, 9245 3DMarks is a respectable result for a system that still uses 133 MHz SDRAM!

Did this mainboard that was released at the peak of the capacitor plague ever undergo a recap in it´s VRM section?
This is what comes to my mind when there are any stability issues at higher frequencies - it doesn´t explain why setmul fails to set lower multipliers though.

Yes, it has been completely recapped, and the new caps are in pristine condition.
However, this is not an instability issue to begin with.
I can loop benchmarks for as long as I want, no problem.
The only situation where it freezes is whenever I run setmul.exe, instantly. And even with a LOWER than default clock!

Well done, so at least this problem can be ruled out. Strange things can happen when the VRMs are not 100% stable.

Yeah, it was a good call, I should have said at the beginning that I already ruled that out.

Apparently, I would have to "change a bit", according to this (see "Chipset requirements":

https://web.archive.org/web/20170822082459/ht … _Multiplier.htm

Unfortunately, either it doesn't explain how to do this on KT133A or I'm just too tired to understand it.

But I think this page is really important, since it explains a lot of AMD Athlon Mods, including how to unlock way more multipliers (3x-24x).

It would be so amazing if I could get this to work, because it would enable me to literally set the CPU speed between 400 and at least 2000 MHz. Possibly more, depending on stability.

If using SetMul you probably need to do it before actually entering Windows. I think raw I/O while in Windows is not really a good idea.

There's a tool for Windows called CPUMSR that can be used to change multi on these AMD K7 CPUs. However in my case setting the CPU to rated multi in XP caused instability (freeze) that did not immediately occur. The CPU was more stable when using Win98, though.

The board I'm having is a 8KTA3PRO. I'm not sure about the differences between the PRO and non-PRO boards.

I haven't recapped my board but the board hardly had issues with T-Breds as well as Bartons (locked or not), just some minor ones, namely the board can only reliably use 1GB of RAM. If trying to install 1.25GB or 1.5GB the system becomes very unstable in any version of Windows (freezes and sometimes BSODs). With DOS most things appear fine, though. The VIA chipset drivers are not perfect, and if you're using PCI sound cards like SBLive! you need to patch the system to fix instabilities and in some cases data corruptions.

It's just that I might need a PSU that's more 5V focused. Usually older, lower-rated PSUs tend to be rated for much higher amperes on 5V rail than newer ones that are more focused on 12V ones.

Yeah no, at the provided link, the author of CPUMSR literally states that to use mobile CPUs on desktop boards will always result in this freeze with SETMUL *until you change that certain bit* that enables the changing of P-States.
He says it right there under "Chipset Requirements".
So I really have to find out how that is done.

I'm experimenting with a similar build to yours - Athlon XP-M 2000+ ( Thoroughbred) + Abit KT7A-RAID and 256MB of PC133 SDRAM. My board post and detects the CPU as an Athlon XP 2000+ (non-M). I can't select any multiplier over 14x from bios, but I'm OK with that, considering I want to run Windows 98se and DOS games on it. It can drive the CPU at 1866MHz just fine, stable at stock voltage, so I'm happy with that, but SetMUL does not work for me either. Instead of freezing like in your case, setmul reports it has changed the multiplier, but nothing happens, the multiplier remains at the same value set in BIOS. I've also tried a couple of AMD Geode CPUs as well (geode NX 1500 and NX 1750) - they are detected as "AMD Athlon 4" - 1000 and 1400MHz respectively. Selecting any multiplier from 3x to 14x works correctly in BIOS, but setmul fails to change the multiplier to anything over the default ones. I can change the multiplier to 3x, 4x and so on, but nothing over 10.5x will work (in the case of the NX1750). Like with the 2000+ mobile, setmul will say "multiplier now at 14x" for example - but the multiplier remains at 10.5x.

The solution is definitely in that link I posted, it just goes over my head.
In the end, I tried an old Desktop XP 2400+ from my drawer, but it seems that one was dead.
But I found out I made a mistake with my CPU: it wasn't a 2600+ Mobile, but a 2500+ Mobile (AXMH2500QQ4C).
I don't know what that thing's normal FSB is like, but will try the real 2600+ Mobile (AXMA2600FKT4C) next.

Unfortunately, I don't seem to have any FSB 266 Desktop T-Breds. My two Desktop 2600+ are FSB333 (although an FSB266 model does exist), my 3000+ is an FSB333 Barton and my 3100+ is an FSB400 Thorton.

Anyway it definitely IS possible to make the multiplier changeable in SETMUL, it's right there on this archived website, and we WILL be able to solve this once someone smarter reads that damn article and can explain how to actually do what the OP suggests.