PowerTower Pro has decided not to work with XLR8 CPU cards (soldered G3 and ZIF). Thoughts?

[absurd_engineering]

A while back, a fellow hobbyist generously traded me an XLR8 Mach Carrier ZIF card for a Sonnet Crescendo G4 card (I have the rare late revision PowerTower Pro that "cannot and will never be able to" run Sonnet cards, per Sonnet.

And the ZIF card worked pretty well for a while. It was spotty with a G4/500 installed, but pretty reliable with a G3/400.

Over time, though, it was as though something cumulative would happen, until the PTP no longer booted with the card. I'd then have to yank it, put in my totally reliable Newer Tech G3/300, and then reboot a few times, after which time I could reinstall the ZIF card and it would work again.

Then, the ZIF card stopped working altogether. Won't boot with either CPU, nor any clock and multiplier settings. It fails before the system chime, and there's no Open Firmware activity.

I thought perhaps the card had died. It does have some electrolytic caps that are of an age to be funky. My ESR meter gives wonky values when I test the caps in situ, but the one cap I lifted and tested outside the card is actually within spec.

I also recently acquired an XLR8 card with a soldered G3 CPU, thinking it would probably be more stable than the ZIF card. It, too, will not boot the PTP. Same symptoms.

Anybody else experience this with a 9500-based or similar machine, and XLR8 cards? Or any issues with caps or other components on XLR8 cards?

I do not have easy access to another PowerSurge machine to test the cards in, unfortunately.

 

[Phipli]

My first suspicion would be that the computer's power supply is getting weak with age. Perhaps the other card draws less power and so copes better.

Another possibility is that the motherboard caps are starting to lose their capacitance and there is a lot of noise on the power rails.

Do you have another computer to test the card with?

What happens if you unplug everything that uses power other than the bare minimum? I.e. unplug hard disks and any PCI cards.

 

[absurd_engineering]

My first suspicion would be that the computer's power supply is getting weak with age. Perhaps the other card draws less power and so copes better.

Another possibility is that the motherboard caps are starting to lose their capacitance and there is a lot of noise on the power rails.

Do you have another computer to test the card with?

What happens if you unplug everything that uses power other than the bare minimum? I.e. unplug hard disks and any PCI cards.

Yeah, I went there first. I've recapped the logic board, and the power supply is a new 450W ATX PSU.

There's no change with PCI cards unplugged, bare bones RAM, and no drives connected. I've reset everything that can be reset, including in OF and on the logic board, and I've done a deep reset with PRAM battery out. No change.

I will have access to another PowerSurge machine toward the end of the month, but at the moment not.

The machine also boots fine from my 604e CPU cards, which, I believe, use more power than the G3 card.

 

[absurd_engineering]

My first suspicion would be that the computer's power supply is getting weak with age. Perhaps the other card draws less power and so copes better.

Another possibility is that the motherboard caps are starting to lose their capacitance and there is a lot of noise on the power rails.

Do you have another computer to test the card with?

What happens if you unplug everything that uses power other than the bare minimum? I.e. unplug hard disks and any PCI cards.

I should mention that it runs *flawlessly* with my Newer Tech G3 card. It's been burning at 100% CPU for 36 hours, compiling a bunch of Tigerbrew packages, and is perfectly happy and stable.

 

[Phipli]

What if you put a different ZIF in the carrier?

 

[Byrd]

I’m finding the CPU can sag in these Macs (see high end GPUs of today), the little clamp that is meant to hold them down is pretty average. Try cleaning the CPU connector with electronic solvent spray, the CPU connector and see if slight movement of the daughter card (perhaps have board sitting horizontally) helps at all.

 

[trag]

What Byrd wrote.... I've had similar issues and cleaning the CPU socket with something like DeOxit https://www.amazon.com/gp/product/B00006LVEU has relieved the issue at least a couple of times. I had similar issues before I thought to clean the contacts and ultimately replaced the motherboards... Will need to revisit those and clean them one of these days.

Also, perhaps time to use a pink pearl eraser on the CPU card contacts.

Worst case, the CPU socket connectors are available, but pricey and difficult to desolder/solder.

Somewhat related, I had many tubes of 44 pin PSOP Flash chips which had aged a couple of decades. I could not get htem to program reliably until I started cleaning the pins. Pain in the ...

 

[absurd_engineering]

I’m finding the CPU can sag in these Macs (see high end GPUs of today), the little clamp that is meant to hold them down is pretty average. Try cleaning the CPU connector with electronic solvent spray, the CPU connector and see if slight movement of the daughter card (perhaps have board sitting horizontally) helps at all.

I'm running the logic board horizontally, and have thoroughly cleaned everything I can clean, and braced the CPU cards with a clip I fabricated. I've tried ZIF CPUs from 233 MHz G3 through to 500 MHz G4.

I also replaced the four SMD electrolytics on one of the ZIF Carrier boards, mostly because I had to take them off to test values anyway. The old caps tested fine, and there's no improvement with the tantalum replacements.

I've also acquired another ZIF Carrier board, since my initial post, which did the same as the other one: initially fired up, then became progressively flakier, and now does not boot the machine at all, with any installed CPU.

My Newer Tech G3 board, on the other hand, just keeps on ticking.

It's that degeneration that has me the most stymied.

I should reiterate: *every other compatible CPU card works reliably in this machine*.

 

[Phipli]

Check the voltages on your PSU I suggest.

Or actually you said you're using an ATX. Check the right voltages are getting to the right places.

Perhaps the card that keeps working has an onboard regulator?

 

[absurd_engineering]

Check the voltages on your PSU I suggest.

Or actually you said you're using an ATX. Check the right voltages are getting to the right places.

Perhaps the card that keeps working has an onboard regulator?

I just checked the power supply again, all the rails seem good. 12V is reading a bit high (12.6) when under load, but that's within modern spec for ATX.

I was thinking about that voltage regulator thing as well. It seems to be the biggest difference between the cards.

Most of the Newer Tech board is covered by a heatsink that I don't want to remove as long as it's all working, but pretty much the only thing you *can* see is the very prominently mounted LT1580 (VLDO v-reg marketed for CPUs). The XLR8 boards do not have a discrete voltage regulator.

Wondering if I can pull live voltages from somewhere on the XLR8 boards, and see what they're seeing...

 

[absurd_engineering]

Check the voltages on your PSU I suggest.

Or actually you said you're using an ATX. Check the right voltages are getting to the right places.

Perhaps the card that keeps working has an onboard regulator?

Interesting that Interex didn't populate the space they left for a voltage regulator...

I may solder on a header to Vpower here and see what the card is really getting.

 

[absurd_engineering]

Check the voltages on your PSU I suggest.

Or actually you said you're using an ATX. Check the right voltages are getting to the right places.

Perhaps the card that keeps working has an onboard regulator?

So, I installed a pin header at what would be Vpower for the missing regulator (confirmed it has continuity to the power pad on the CPU socket) and one of the ground plane test points, and I'm getting pretty clean 3.3V to the card.

Hmmm.

 

[BetaC]

Have you by chance reset the CUDA yet? I recently worked on a clone that wasn't booting with a G3 upgrade card and after resetting that it suddenly started working.

 

[absurd_engineering]

Have you by chance reset the CUDA yet? I recently worked on a clone that wasn't booting with a G3 upgrade card and after resetting that it suddenly started working.

Yep. Short reset, long reset, reset along with pulling PRAM battery and power, reset then wait overnight.

No difference.

 

[macuserman]

@absurd_engineering Any particular reason you want to use the zif carrier over a standalone card? 500mhz range cards are readily available, I assume you don't have any other compatible machines to test the dying zif carriers in?

 

[absurd_engineering]

@absurd_engineering Any particular reason you want to use the zif carrier over a standalone card? 500mhz range cards are readily available, I assume you don't have any other compatible machines to test the dying zif carriers in?

I haven't found any standalone G3 or G4 cards in that range (other than the Sonnet cards that don't work in the late PT Pro) that are affordable on my hobby budget. I have a couple of 400 MHz ZIF G3s and a 500 MHz ZIF G4. So those would seem to make sense. If they worked.

I do have other machines I can test with eventually, but they're frozen into my storage space until it thaws here.