Workgroup Server 9150 Freezes Without Additional Cooling - Can anyone check their WGS VR temps?

Kind of late replying...

The voltage regulators are resistance based rather than Buck or Switching type. That means that they basically convert all the excess voltage to heat to get the desired voltage. So it's not surprising they get hot.

I've seen similar issues on the PowerMacintosh 7100. In many cases the heat sink grease between the heat sink and CPU has turned to chalk.

Have you removed the CPU heat sink, cleaned it and the CPU and applied new heat sink compound?

Do not overapply. Usually the stuff is conductive and you do not want it running onto the pins of the PPC601. I killed one that way, once....

+1 on applying new thermal paste. Also check if the Peltier is actually good.
Also there's a fan attached to the back of the power supply assembly so it sits right on top of the CPU heatsink once the board is installed in the machine. Make sure that fan is running.

Well, I think my 9150/80 is running into heating issues also (random quits and system errors after a few minutes of uptime). RAM repeatedly tested good. The heat grease had indeed turned into heat not-grease, but a not overly generous application of new compound only made it marginally better. I managed to track down two NOS 50mm Peltier coolers like the /120 has that should fit into the clips on the /80 logic board, and I'll see how well that works when they arrive.

Parenthetically, the cooling in this thing is terrible. The only intake is the case fan in the front. The CPU is trapped behind that huge honking wall of a power supply, so I'm not surprised it starts pooping its pants after awhile.

ClassicHasClass said:

Well, I think my 9150/80 is running into heating issues also (random quits and system errors after a few minutes of uptime). RAM repeatedly tested good. The heat grease had indeed turned into heat not-grease, but a not overly generous application of new compound only made it marginally better. I managed to track down two NOS 50mm Peltier coolers like the /120 has that should fit into the clips on the /80 logic board, and I'll see how well that works when they arrive.

Parenthetically, the cooling in this thing is terrible. The only intake is the case fan in the front. The CPU is trapped behind that huge honking wall of a power supply, so I'm not surprised it starts pooping its pants after awhile.

Click to expand...

All the PPC601 machines that ran at 100MHz or over came with a Peltier equipped heat sink with those four clips. Shreve Systems used to sell them for some absurd price like 10 for $20 or some such. I wish I'd picked up a box at the time, I think.

That said, I did some experiments a couple of jobs ago with Peltier cooling and actually used some of the HSs from Shreve (left them at the job, because they paid for them). The physics of this is that a Peltier will get you a certain temperature gradient per voltage (current?) applied. BUT the Peltier is inefficient and converts some of the power into heat, which is added to the temperature at the heat sink.

So, you have a heat sink at 80C and you think, great, I have this 20C gradient Peltier, I can get the apparent temperature at that heat sink down to 80C - 20C = 60C. BUT, the heat from the Peltier is going to add 15C to the final heat sink temperature, so you're only going to get (80C + 15C) - 20C = 75C...

All numbers made up, but should illustrate the issue.

My conclusion for cooling our chip at my old company was that we weren't really gaining anything other than more power consumption by trying to use a Peltier. I suspect that using a larger heat sink with a Peltier would give much better results. Of course, a larger heat sink by itself has benefits.

I had the thing instrumented with thermocouples on the HS and at the chip/HS interface and I had a temp sensor internal to the chip (on die).

A larger HS should be able to take the higher power from chip + Peltier and give a lower HS temperature, because of it's ability to distribute the heat over more area/more rapid cooling. So the Peltier gradient can hit a lower temperature coming off the larger HS.

It seems like a small HS gets overwhelmed by the increased power and just can't shed the heat fast enough and gives a rise in HS temperature almost as high as the Peltier gradient.

Now if everything is linear, it shouldn't matter. Fluids and convection adn radiation being what they are, I suspect there's something very non-linear going on somewhere, but can't jar my ancient aerospace engineer brain enough to think of it.

Anyway, point being, I'm not sure the Peltier on same size heat sink gains you that much over just having a larger heat sink. Those Apple HSs are relatively tiny. Power Computing sold a waffle iron style HS with Peltier on their early PPC601s which might be better, but I'm not sure it actually has any extra surface area compared to the wings on the Apple Cage-style HS.

Another possibilty for the bold... The LT1087 (LT1086?) voltage regulator for the CPU is resistor controller by two (SM) resistors. One could swap the resistors to try lower or higher voltages on the CPU...

I was going to attach the Linear Tech VR datasheet, but I don't have a copy on a machine from this century. It must be on my old classic OS Mac that doesn't get on the Internet so well. If you want it, I can FTP it into my web space.