Socket3 wrote on 2023-05-05, 23:24:

I've been working on putting together an Athlon XP mobile build for.... a long while now. But, I came across some of my socket A […]
Show full quote

I've been working on putting together an Athlon XP mobile build for.... a long while now. But, I came across some of my socket A boards that allow multiplier selection in BIOS today, and instead of packing and organizing my retro hardware like I was supposed to do, I decided to do some experimenting. One thing led to another, and I ended up building (well, setting up some parts for) said Athlon XP-M build.

Initially I was going to use an Epox KT400A motherboard (EP-8K9A7I), but it turns out that board is (barely) stable at 100MHz FSB, another victim of capacitor rot, like most of my socket A motherboards... no visibly bulging caps, so I put the Epox aside and picked up a couple of MSI KT3 Ultras. Both post, but one has a plethora of bulging leaking Teapo 1000uf capacitors, so I put that one in my "for repair" box and had a look at the other, witch only had 1 bulging Teapo cap. I quickly replaced the bulging cap with an equivalent Rubicon, and to be safe I also replaced the 1000uf near the memory slots (witch turned out to be fine, tested it with my ESR meter). It was not bulging, but I replaced it nonetheless.

I managed to setup a working PC, but the experience left me with quite a few questions:

1. Can you select multipliers higher the 12.5x on older socket A motherboards? My KT3 only goes up to 12.5, while the Epox could do up to 24x (with a Geode - it would hang at post with the 2000+).

2. I've read about people using KT133A based Abit KV7 motherboards with Athlon XP-M CPUs, how does that work exactly? How do you get over the 12.5x multi limit? I tried setmul as well, selecting anything over 12.5x would result in 12.5x

3. Related to the question 1 - it occurred to me that the Athlon XP-M 2000+ I played around with might be locked to 4x-12.5x - could this be the case? I tried the Geodes (1500 and 1750) on the epox board, and that showed much higher multiplier selection in bios. Are the Geodes fully unlocked and the XP-M 2000+ locked to a maximum multiplier of 12.5x, or is it a limitation of the KT333 chipset? (I didn't try the Geodes on the MSI, I only tried them on the Epox).

4. If nr 3 is true, then is there a hardware mod or a bios update that can be done to unlock multipliers over 12.5x on older boards? If so, would this allow higher multipliers to be assigned in software using setmul?

Any info would be appreciated.

I have an EPoX EP-8KTA3PRO and I had an interesting experience with experimenting the CPU frequency settings.

1. Regarding multiplier settings: On these motherboards, 12.5x (13x in BIOS) is the practical max you can set with most of the T-Bred CPUs. The 14x multiplier in BIOS is effectively x5. You can mod the CPU to force a higher multipler if you have the proper tools. However, in that case the actual multiplier may not correspond to the value set in BIOS anymore, due to the way the multiplier setting works. From what I have read, the multiplier part for new CPUs consists of 5 bits but older boards only touch 4, leaving one untouched. The untouched bit will play a major role in setting the actual multiplier depending on the state it's currently on.

(WARNING: You may end up in unbootable values, so be sure you have a backup CPU so you can always go into BIOS to change it until you get a bootable value you want.)

(HINT: Athlon XP 2200+ will run as expected at 133x13.5. Considering its multiplier bit states, the BIOS setting will behave very differently and could be used as a starting point for experimenting, without the need of actual modding.)

(PS: Later Bartons are multiplier-locked. BIOS settings as well as modding will no longer have any effect.)

2. Regarding Athlon XP-M in general: With older motherboards like mine, Athlon XP-M would run at boot-up multiplier (x6) by default and a tool is required to set the multiplier you want it to run later on, such as SETMUL for DOS or CPUMSR for Windows.

I'm not sure whether Athlon XP-M CPUs are capped to the values they were rated, however, as I have an Athlon XP-M 3000+ (Barton) and with CPUMSR I'm not limited to 12.5x. I didn't try SETMUL with these CPUs for higher multipliers at that time: I used it to set it lower for certain speed-sensitive DOS games. Maybe SETMUL has the same limitation as BIOS settings, but I'm not sure.

NOTE: It seems the CPU is not stable with WinXP after setting it with CPUMSR, as I had freezes when I actually set the multiplier that high, so I have to set a lower value. Win98 appeared to be fine from what I tested back then, however. I'm not sure why, maybe it's due to chipset, potentially bad caps, or perhaps the heavy demand on the +5V rail for these CPUs that even modern PSUs may have difficulty to satisfy, since they mostly focus on +12V output.

Finally: This site has everything you need to know. I did my mod and experiments following their guides back then, particularly the painting guide.