Are you sure you get 6.5 x 133 and not 12.5 x 66? I recently had a similar problem and it turned out to be related to a broken DIMM: The SPD EEPROMs and the clock generator are on the same I2C bus. The broken SIMM shorted out the I2C bus, so the BIOS couldn't initialize the clock generator, and it stayed at its default frequency of 66MHz instead of switching to 100 or 133 as set up in the CMOS setup.

Tried a whole batch of cpus and this is the only one.
I guess it's a mobile cpu that needs manual adjustement.

I tried with all configurations of DDR - 2100/2700/3200; 1/2/3 sticks.

Two different mobos. 😀

According to the L bridge configuration, it should be desktop Athlon Tbred with Vcore 1.6 and a multiplier of 12.5x. Are you sure this isn't just about misidentifying the CPU? Have you done some performance tests and comparisons with other XPs in the database like Sandra?

@mkarcher: I don't know of any board with s.462 (A) that can FSB 66MHz - the lowest frequency starts at 100MHz, so this won't be the cause of the problem.

Ydee wrote on 2022-11-12, 11:12:

According to the L bridge configuration, it should be desktop Athlon Tbred with Vcore 1.6 and a multiplier of 12.5x. Are you sure this isn't just about misidentifying the CPU? Have you done some performance tests and comparisons with other XPs in the database like Sandra?

@mkarcher: I don't know of any board with s.462 (A) that can FSB 66MHz - the lowest frequency starts at 100MHz, so this won't be the cause of the problem.

Ran PcMark 02 + 05, results are in line with XPs (not T-bird); if I make a mean value it falls around what it gives.
'02: cpu 2506, ram 2024; 2nd run cpu 2528 ram 2117;
'05: cpu 987, ram 894; 2nd run cpu 998, ram 901;

On the 2nd motherboard they were close, a little higher as it was on a Via 600 chipset.

1.4Ghz T-bird gives cpu 3607 '02, 1571 '05.
Didn't test ram.

Voltage was 1.65V.

Ydee wrote on 2022-11-12, 11:12:

@mkarcher: I don't know of any board with s.462 (A) that can FSB 66MHz - the lowest frequency starts at 100MHz, so this won't be the cause of the problem.

My board, the dreaded Elitegroup ECS K7S5A (my copy works perfectly, knocking in wood), also doesn't (officially) support FSB66 and it doesn't allow the user to select that clock. Nevertheless, the clock synthesizer chip powers up with a default frequency of 66MHz until the BIOS reprograms it. If the I2C bus is inoperative, the clock chip keeps running in the power-up default configuration.

Even if I upped the FSB, it had a 6.5x fixed.
Maybe it's the huge scratch on the surface that affects the multiplier?

@Ydee, where did you find info on L bridges? Interested 😀

I think the resistor next to L5 is broken, maybe that has something to do with it.

I could try putting a new one, but it's a scratch and reads 478 ohms.
Same as other working cpus.

Edit: you were right, it was broken. I'll definitely solder a new one. After removing it, reads 578 ohms from opposite pin row.

https://www.ocinside.de/workshop_en/tbred_painting/

Maybe it went wrong?

mkarcher wrote on 2022-11-12, 15:10:

My board, the dreaded Elitegroup ECS K7S5A (my copy works perfectly, knocking in wood), also doesn't (officially) support FSB66 and it doesn't allow the user to select that clock. Nevertheless, the clock synthesizer chip powers up with a default frequency of 66MHz until the BIOS reprograms it. If the I2C bus is inoperative, the clock chip keeps running in the power-up default configuration.

Oh, so you thought the clock generator was stuck on a basic frequency (66MHz here)?. I didn't even know it could be. Thank you for the new information.

Nexxen wrote on 2022-11-12, 21:40:

@Ydee, where did you find info on L bridges? Interested 😀

Ever since ancient times, everything about tweaking AMD CPU has been drawn from this page: http://fab51.com/index-e.html

The bridge configuration at K7 describes this page:http://fab51.com/cpu/barton/athlon-e23.html

As you can see, something is wrong with your Athlon - according to the bridges, the Vcore should be 1.6V and the multiplier should be 12.5x. The multiplier of 12.5x is the last in the bottom set, the top set from 13x is switched by opening the FID4 at bridge L3. I don't have any XP2000+, I only have XP1700+ and then Barton 2600+, but I can test the XP1700+ in the KT600 board, in the PCMark - I have an open multiplier on it, so I'll drop it on the 6.5x.
But I'm afraid there's something wrong with your CPU and I'll get similar results.

So as I suspected, the results with the 6.5x multiplier are similar to yours, so the CPU is really running on this multiplier. For comparison, I give the results with a multiplier of 12.5x.
I don't know what's wrong with yours CPU and unfortunately you can't help yourself by interrupting fid4 with L3 and switching to a higher set of multipliers only, because in your current state you'd get a multiplier 20x. This could work with the FSB 100MHz, but I'm not sure even 133MHz. For a more usable lower multiplier, you'd have to intervene in other FIDs (interrupt or join).

Ydee wrote on 2022-11-13, 10:29:

So as I suspected, the results with the 6.5x multiplier are similar to yours, so the CPU is really running on this multiplier. For comparison, I give the results with a multiplier of 12.5x.
I don't know what's wrong with yours CPU and unfortunately you can't help yourself by interrupting fid4 with L3 and switching to a higher set of multipliers only, because in your current state you'd get a multiplier 20x. This could work with the FSB 100MHz, but I'm not sure even 133MHz. For a more usable lower multiplier, you'd have to intervene in other FIDs (interrupt or join).

Goldmine of information, fab51 bookmarked. Thanks!

Yes, something is off. The 102 resistor hasn't much to do with the multiplier, but I'll report back when resoldered.
It's nice when you have a lot of dead athlons to source things from.

I don't think it matters, but the orientation of those 102 isn't important?

I had a whole batch of these cpus, and cleaned them thoroughly. A few had signs of silver paste.
If some ended in the L cuts that could explain shorting to ground? I'm not stressing about this, but it is odd - 🤣

There are other components on those cpus, brown/-ish, around the core.
Any idea what are those for?

Ydee wrote on 2022-11-13, 09:20:

mkarcher wrote on 2022-11-12, 15:10:

My board, the dreaded Elitegroup ECS K7S5A (my copy works perfectly, knocking in wood), also doesn't (officially) support FSB66 and it doesn't allow the user to select that clock. Nevertheless, the clock synthesizer chip powers up with a default frequency of 66MHz until the BIOS reprograms it. If the I2C bus is inoperative, the clock chip keeps running in the power-up default configuration.

Oh, so you thought the clock generator was stuck on a basic frequency (66MHz here)?. I didn't even know it could be. Thank you for the new information.

Yes, exactly that. As this happened to me recently, I wanted to point out that an unexpected low core clock might not necessarily be caused by a wrong multiplier, but also by a wrong FSB clock. As the OP already did enough tests to disprove the idea of "wrong FSB clock", this is just a side story in this thread. Having a memory module influence the FSB isn't something I would have thought about in advance. So even with the OP stating that two boards were used, it could still have been that it was the same DIMM.

mkarcher wrote on 2022-11-13, 17:14:

Ydee wrote on 2022-11-13, 09:20:

mkarcher wrote on 2022-11-12, 15:10:

My board, the dreaded Elitegroup ECS K7S5A (my copy works perfectly, knocking in wood), also doesn't (officially) support FSB66 and it doesn't allow the user to select that clock. Nevertheless, the clock synthesizer chip powers up with a default frequency of 66MHz until the BIOS reprograms it. If the I2C bus is inoperative, the clock chip keeps running in the power-up default configuration.

Oh, so you thought the clock generator was stuck on a basic frequency (66MHz here)?. I didn't even know it could be. Thank you for the new information.

Yes, exactly that. As this happened to me recently, I wanted to point out that an unexpected low core clock might not necessarily be caused by a wrong multiplier, but also by a wrong FSB clock. As the OP already did enough tests to disprove the idea of "wrong FSB clock", this is just a side story in this thread. Having a memory module influence the FSB isn't something I would have thought about in advance. So even with the OP stating that two boards were used, it could still have been that it was the same DIMM.

If it makes you happy I actually took note of your experience in my K7 notes bundle.
I've read too much here to take anything as "impossible".

Let alone strange behaviours before repairs. Life wouldn't be retro spicy otherwise 😀
---

I'm going to test Crystalcpu and try to modify the multiplier by software.
Date code is 0233, week 33 of 2002, not locked. Unless it is.
Complete is:
AXDA2000DUT3C 9390505281461
AIRGA 0233DPMW mc 1999 AMD

It's weird, another cpu story. If I started a topic on this it could yield big in oddities 😀

So far I'd like to thank everybody.

Nexxen wrote on 2022-11-13, 13:54:

I don't think it matters, but the orientation of those 102 isn't important?

Afaik, it´s rezistor array 1kOhms (1000 Ohms) and orientation isn´t important, is passable in both directions.

Nexxen wrote on 2022-11-13, 13:54:

There are other components on those cpus, brown/-ish, around the core.
Any idea what are those for?

I think it's SMD capacitors. Palominos have them at the bottom between pins, Tbreds and Bartons at the top next to the core.

Ok, replaced the broken resistor and it now works @12.5 x 133, 1667mhz 😀 and Vcore from 1.6 to 1.65 as expected.

Fab51 has been an interesting read for days. I'm going to try to repair some more cpus.

Benchmarks;
2002: cpu 4280, mem 2656
2005: cpu 1988, mem 1437

Bottom line:

- resistors must be checked if everything else isn't working, could be a dead SMD
- date codes are important to understand if the multiplier can be modified via L bridges or is hard coded

Still have to figure out what the big brown capacitors do.

Nexxen wrote on 2022-11-19, 21:29:

Still have to figure out what the big brown capacitors do.

Maybe some decoupling capacitors?

Ydee wrote on 2022-11-20, 12:13:

Nexxen wrote on 2022-11-19, 21:29:

Still have to figure out what the big brown capacitors do.

Maybe some decoupling capacitors?

Unfortunately I can't use these cpus without a h/s 😀 to measure.
I'll look around, for sure there are infos. Probably on teh XP papers AMD released.