mockingbird wrote on 2025-02-24, 05:36:

Interesting suggestion, but if that's the case, then the slotket should also have these types of low-esr caps, and I don't know of a single one that does. I'd love to see someone design a new one.

The search returns pictures of adapters that are more decent than those from MSI.



Considering that there was no reference design, manufacturers created minimally working products.

The P3 slot1 Coppermine usually contains several capacitors on the cartridge.

Not sure why we're talking about capacitors. 😀
The discussion was about my AA370TS being unstable in certain niche conditions. I only mentioned the capacitors because I replaced the original ones, just to eliminate them as a potential cause. But, no, the instability persists.

bloodem wrote on 2025-02-24, 07:33:

Not sure why we're talking about capacitors. 😀
The discussion was about my AA370TS being unstable in certain niche conditions. I only mentioned the capacitors because I replaced the original ones, just to eliminate them as a potential cause. But, no, the instability persists.

Sure)
The slotket is a non-standardized side-product that was developed by engineers with much less experience than the developers of the SECC2 cartridge.
With some boards and processors, some slotkets will be completely stable, with others they will not.
Like "Asus Photoshop bug" - not enough stable Vtt voltage, occurs under very specific conditions

Therefore, change the capacitors to adequate ones, or look for another slotket.
It was not designed engineers from Santa Clara (Intel's headquarters).

My point is that the original electrolytic caps on this slotket were not even low-ESR, they were very cheap general purpose 33uF caps.
Now, whether capacitors are the actual cause (or play a role) when it comes to this slotket's instability, or if it's just a bad design overall, I can't really say. It's certainly way above my paygrade to even attempt to discuss it. 😀

shevalier wrote on 2025-02-24, 10:14:

Like erlier specified 100-560uF *6.3V polymer capacitors These definitely make sense with an ESR of 10-30 mOhm. And this type of […]
Show full quote

Like erlier specified 100-560uF *6.3V polymer capacitors
These definitely make sense with an ESR of 10-30 mOhm.
And this type of capacitance can to some extent offset the shortcomings of the board routing, for example, the small cross-section of the power polygons.[/details]
Therefore, change the capacitors to adequate ones, or look for another slotket.
It was not designed engineers from Santa Clara (Intel's headquarters).

I swapped the 5 capacitors on my Abit Sloket iii (Had Nichicon PM on it I installed years ago - 400 or so mOhm) with Chinese counterfeit (but good quality - opened one up and examined the polymer quality) UCC PSG series 16V 100uF parts (40mOhm - part doesn't exist in datasheet so I tested with my LCR meter). I also obtained a second Nehemiah 1.2Ghz CPU.

So far, testing on my 3.3V modded Asus P2B it seems that it has made a difference in stability, as I have 3DMark running for almost 20 hours at 1024x768x32bpp on a GeForce4 MX440 (45.23 driver) and it's still going.

I will test this on my GA-6BXC to confirm.

Thanks for the capacitor suggestion @shevalier. (if it is indeed the 10x reduction in ESR of the slotket caps and not the new CPU).

mockingbird wrote on 2025-04-04, 18:03:

I swapped the 5 capacitors on my Abit Sloket iii (Had Nichicon PM on it I installed years ago - 400 or so mOhm) with Chinese cou […]
Show full quote

shevalier wrote on 2025-02-24, 10:14:

Like erlier specified 100-560uF *6.3V polymer capacitors These definitely make sense with an ESR of 10-30 mOhm. And this type of […]
Show full quote

Like erlier specified 100-560uF *6.3V polymer capacitors
These definitely make sense with an ESR of 10-30 mOhm.
And this type of capacitance can to some extent offset the shortcomings of the board routing, for example, the small cross-section of the power polygons.[/details]
Therefore, change the capacitors to adequate ones, or look for another slotket.
It was not designed engineers from Santa Clara (Intel's headquarters).

I swapped the 5 capacitors on my Abit Sloket iii (Had Nichicon PM on it I installed years ago - 400 or so mOhm) with Chinese counterfeit (but good quality - opened one up and examined the polymer quality) UCC PSG series 16V 100uF parts (40mOhm - part doesn't exist in datasheet so I tested with my LCR meter). I also obtained a second Nehemiah 1.2Ghz CPU.

So far, testing on my 3.3V modded Asus P2B it seems that it has made a difference in stability, as I have 3DMark running for almost 20 hours at 1024x768x32bpp on a GeForce4 MX440 (45.23 driver) and it's still going.

I will test this on my GA-6BXC to confirm.

Thanks for the capacitor suggestion @shevalier. (if it is indeed the 10x reduction in ESR of the slotket caps and not the new CPU).

Well, I just replaced the capacitor in my PS/2 keyboard.
And I thought my CUBX-L burned out - it didn't respond to keys and took a long time to enter BIOS, beeping with the speaker.
Because its keyboard is powered through a diode and comes in at 4.75 V.
With more modern motherboards, the keyboard worked without problems.

This is what an electrolytic capacitor is... In the wrong place at the wrong time. 🙁

shevalier wrote on 2025-04-05, 17:44:

Well, I just replaced the capacitor in my PS/2 keyboard. And I thought my CUBX-L burned out - it didn't respond to keys and took […]
Show full quote

Well, I just replaced the capacitor in my PS/2 keyboard.
And I thought my CUBX-L burned out - it didn't respond to keys and took a long time to enter BIOS, beeping with the speaker.
Because its keyboard is powered through a diode and comes in at 4.75 V.
With more modern motherboards, the keyboard worked without problems.

This is what an electrolytic capacitor is... In the wrong place at the wrong time. 🙁

You are full of good ideas -- I had a PS/2 keyboard that was not working. I took it apart and swapped out the 1uF and 10uF caps on the logic board and cleaned the connector ribbons which were very oxidized... I'm not sure if it was the caps or the ribbon cable cleaning but after doing both, the keyboard now works perfectly (and it is a nice quality keyboard too!).

Unfortunately, the GA-6BXC was not fully stable in 3DMark2000 for 24 hours (crashed to desktop). So I think I'll be swapping to the P2B with the modified 3.3V circuit. The only disadvantage is the bugginess of going from 50Mhz FSB to 66Mhz FSB with RayeR's SMB, but I can live with that. I will also now test with my GeForce FX on the P2B which I also had stability issues with on the GA-6BXC. I'd rather use it than the MX440.

I will reveal at this point that the Slotket I was experimenting with on the Gigabyte board was a Gigabyte GA-6R7+ Rev 1.1. It's a nice slotket and like the Abit Slotket iii, it has a voltage clamp. There is 1 33uF small electrolytic cap on the PCB which shares a footprint for an SMD tantalum -- so I might try installing a very low ESR tantalum there some day in place of the electrolytic to see if that improves things. At 900Mhz, the gigabyte board/slotket combo was stable. At 1.2Ghz, it was hit and miss.

Well, I'm abandoning the VIA C3 Nehemiah build. I have tried everything, different slotkets, different motherboards, different videocards, nothing I do can guarantee 100% stability. On my Abit Slotket III, I swapped the electrolytic capacitors with 30mOhm polymers. On my Gigabyte slotket, I swapped the 33uF capacitor with an aluminum polymer SMD cap (Kemet T520) (80mOhm at 1kHz).

For comparison (and a sanity check), I can run 3DMark2000 perfectly fine with my Asus P2B 1.04 on the lin-lin/Abit slotket iii with a Celeron 1.4 for an eternity.

I can kind of get the Nehemiah stable with only one videocard I tried with this benchmark, but even that's uncertain.

I accept the reports that this setup is stable for others, but my own testing does not confirm this.

In my opinion this is a unicorn build that should be avoided and substituted for real hardware (i.e. seperate 386 and 486) which I already possess.

Did you also try another CPU and also different GPU drivers? I know that 3DMark can be buggy with certain hardware & driver combos (which can cause a crash to desktop).
What about other games / 3D apps? Did you experience any instability during actual usage?

bloodem wrote on 2025-05-06, 04:46:

Did you also try another CPU and also different GPU drivers? I know that 3DMark can be buggy with certain hardware & driver combos (which can cause a crash to desktop).
What about other games / 3D apps? Did you experience any instability during actual usage?

Yes, it worked perfectly with a Tualatin Celeron -> lin-lin -> Abit Slotket iii combo on a P2B 1.04 and a GeForce 4 MX440 8X.

No, the stability issue existed only in 3DMark2000 and 3DMark2001 AFAIK.

If 3DMark is crashing to desktop, then the system is not stable. The hardware should be transparent to the programs.

Ezra-T was stable, but I can't use Ezra-T because of the broken write combining.

The build was cool, in theory -- go all the way down to 386 performance and then back up as fast as a 600Mhz Katmai... The combo just didn't work for me, was expensive and time consuming, and there are much better and easier alternatives.

Sorry to hear that. It's a shame, I'm sure there has to be an explanation, since it's always worked very well for me.

All in all, I agree that there are better alternatives (for me, as I said many times before, it's the Athlon XP on VIA chipsets).

Recently, I was planning to do a final build with a Pentium 60 on a socket 4 motherboard, but I eventually gave up on that idea and went (again) with an Athlon XP Barton 2800+ (I converted it to mobile and unlocked all multipliers up to 24x) on a Biostar M7VKD / KT133A board. What I like about this board, in particular, is the fact that you can change the CPU multipliers from within Windows / DOS without the need to modify chipset register 55 (something you have to do on boards such as the Abit KT7A) - on this board's BIOS, the second bit is already set to "1". With a GeForce 4 Ti 4400 + a Gainward Voodoo 2 & Audigy 2 ZS for Windows + an ESS ES1688 non-PNP for DOS, this thing is a beast - realistically, I don't really need anything else for my retro needs... I could just sell all of my other builds and keep this one. Too bad I'm a crazy bastard, so I must. have. them. all. 🤣

mockingbird wrote on 2025-05-06, 13:49:

Yes, it worked perfectly with a Tualatin Celeron -> lin-lin -> Abit Slotket iii combo on a P2B 1.04 and a GeForce 4 MX440 8X. […]
Show full quote

bloodem wrote on 2025-05-06, 04:46:

Did you also try another CPU and also different GPU drivers? I know that 3DMark can be buggy with certain hardware & driver combos (which can cause a crash to desktop).
What about other games / 3D apps? Did you experience any instability during actual usage?

Yes, it worked perfectly with a Tualatin Celeron -> lin-lin -> Abit Slotket iii combo on a P2B 1.04 and a GeForce 4 MX440 8X.

No, the stability issue existed only in 3DMark2000 and 3DMark2001 AFAIK.

If 3DMark is crashing to desktop, then the system is not stable. The hardware should be transparent to the programs.

Ezra-T was stable, but I can't use Ezra-T because of the broken write combining.

The build was cool, in theory -- go all the way down to 386 performance and then back up as fast as a 600Mhz Katmai... The combo just didn't work for me, was expensive and time consuming, and there are much better and easier alternatives.

I would recommend using a Socket 370 motherboard on VIA chipset. This combination usually supports C3 out of the box. Nehemia support should be added in BIOS updates. This is the easiest way to evaluate the capabilities of the VIA C3.
But as noted earlier, the idea of ​​using this platform in theory sounds better than it turns out to be in practice. Stability is harder to achieve than on a purely Intel 440/8xx+P3/Celeron build.
Personally, I have mixed feelings. On the one hand, I like to tinker with VIA hardware, it gives interesting possibilities. On the other hand, it is often buggy hardware and buggy drivers. So I understand if someone gives up on the path to building with C3. My Socket 478 i865 setup has worked perfectly for years, both as a main computer and as a retro, and after using it I don't want to lose that trouble-free nature and stability in pursuit of flexibility. That's why I didn't limit myself to one build.