Quadra 700 Repair - Screwdriver Heatsink Removal Whoops!

[ymk]

Like I said, so I can read the markings?

I just label the underside of the heatsink. It's not a difficult problem to solve.

 

[Phipli]

I just label the underside of the heatsink. It's not a difficult problem to solve.

Problem solving needs a problem to solve.

There wasn't one.

 

[joshc]

I assumed when ymk said tiny he also meant the height, as you could definitely use a taller heatsink in a Quadra 700/650/900/etc without issue. What would be interesting is some tests of taller heatsinks to see how much difference that really makes with an 040 running at 40MHz. I don't have a thermal imaging camera / probe thermometer to test this but I guess a cheap sensor could be used to take measurements for that sort of test.

Edit: Some further thoughts. It might be useful to use a chunkier heatsink because the main air intake on a Quadra 700 case is the floppy drive opening, and the only fan exhaust is the power supply. Kinda crappy design for good airflow, especially if the case is full of heat generating NuBus cards too (ok, most of them don't generate much heat all, but still). The 650 isn't much better in this regard either, still just one fan for the whole system... Of course, I've never actually encountered a situation where this actually seems to be a problem per-se, but it does seem less than ideal if you're planning to run one of these machines for very long periods, potentially. Apple always said to not run a Quadra 700 with its lid off, so there must have been some legitimate concern at the time in terms of heat dissipation in that case.

 

[Phipli]

I assumed when ymk said tiny he also meant the height, as you could definitely use a taller heatsink in a Quadra 700/650/900/etc without issue. What would be interesting is some tests of taller heatsinks to see how much difference that really makes with an 040 running at 40MHz. I don't have a thermal imaging camera / probe thermometer to test this but I guess a cheap sensor could be used to take measurements for that sort of test.

A fan would be a better bet. I'd use the extra space for even a slow fan. Running a 12v fan at 5v is near silent, but moves enough air to vastly improve heat transfer.

Ultimately, unless you can remove the heat from the heatsink, a larger heatsink primarily acts as a reservoir and delays reaching peak temperature. The higher surface area does improve heat transfer to air, but 040s are not a major heat problem and the thermal resistance at the boundary between the CPU and metal, and the metal and air, is magnitudes more significan in impact than....

Bleh. Nobody ever cares do they. Who am I kidding. The maths is too hard to be bothered to show people. Darn passive cooling stuff is a nightmare.

Edit - a the bigger heatsink will remove more heat because of its higher surface area, but a late model 25MHz 040 is only 3.5w. It's so little, that absolutely any heatsink is a pretty good solution.

 

[eharmon]

Adding my two cents for removing the clips:

I've had good luck abusing a chip lifter (like a wiha 26810) and prying the clips horizontal to the board. The upward curve helps leverage and makes it a little less likely you'll gouge headfirst into the board if you slip.

 

[taddy]

Adding my two cents for removing the clips:

I've had good luck abusing a chip lifter (like a wiha 26810) and prying the clips horizontal to the board. The upward curve helps leverage and makes it a little less likely you'll gouge headfirst into the board if you slip.

How about passing a zip tie through the gap between the clips and pulling it?
I always do that.

 

[ymk]

I assumed when ymk said tiny he also meant the height, as you could definitely use a taller heatsink in a Quadra 700/650/900/etc without issue.

Height too, but mostly footprint. My point was, when the heatsink footprint is larger than the 040, you can pull the CPU out by the heatsink, without prying:

Ultimately, unless you can remove the heat from the heatsink, a larger heatsink primarily acts as a reservoir and delays reaching peak temperature.

Right, and the heat is removed by airflow around the heatsink. I consider fanless and passive different things. Like @joshc mentioned, some Macs shouldn't be run without the cover because the machine depends on incoming air to cool certain components. The 6100 is another example of this. The CPU doesn't have a fan, but depends on the PSU fan drawing air over it; not passive in my opinion. I've set up a Pentium 83 similarly because I don't like listening to tiny fans.

 

[superjer2000]

I assumed when ymk said tiny he also meant the height, as you could definitely use a taller heatsink in a Quadra 700/650/900/etc without issue. What would be interesting is some tests of taller heatsinks to see how much difference that really makes with an 040 running at 40MHz. I don't have a thermal imaging camera / probe thermometer to test this but I guess a cheap sensor could be used to take measurements for that sort of test.

I recently added a Bolle Carrera040 to my IIci, using a 40mhz 68040 I bought from what I understand to be a reputable supplier on eBay. Everything worked fine but I was surprised at how hot the processor was to the touch. I ended up affixing a thermistor to the bottom of the processor, using very fine gauge enameled wire to pass through the processor pins, to see just how hot it was getting.

I was pretty surprised at how quickly the processor temperature ramped up. Admittedly I was trying to stress it out a bit by running some Speedometer tests on loop, but even after stopping the tests there wasn't much of a drop in temps. I then affixed a 50mm x 50mm x 20mm aluminum heatsink from Amazon using some thermal paste. I didn't have any thermal glue so I temporarily just affixed it with Kapton tape. The differences were dramatic - Temp rampup was much slower and hovered around 50C or slightly lower. I hammered it with way more loops of Speedometer tests (I pulled the plug on the tests without the heatsink when I saw temps get up to 105c or so) and the temps held firm. I played around with blowing a fan directly over the heatsink and that got temps down to sub 40C.

I have some thermal glue coming and I am going to attach a fan to the heatsink to keep temps as low as possible.

 

[Phipli]

I recently added a Bolle Carrera040 to my IIci, using a 40mhz 68040 I bought from what I understand to be a reputable supplier on eBay. Everything worked fine but I was surprised at how hot the processor was to the touch. I ended up affixing a thermistor to the bottom of the processor, using very fine gauge enameled wire to pass through the processor pins, to see just how hot it was getting.

I was pretty surprised at how quickly the processor temperature ramped up. Admittedly I was trying to stress it out a bit by running some Speedometer tests on loop, but even after stopping the tests there wasn't much of a drop in temps. I then affixed a 50mm x 50mm x 20mm aluminum heatsink from Amazon using some thermal paste. I didn't have any thermal glue so I temporarily just affixed it with Kapton tape. The differences were dramatic - Temp rampup was much slower and hovered around 50C or slightly lower. I hammered it with way more loops of Speedometer tests (I pulled the plug on the tests without the heatsink when I saw temps get up to 105c or so) and the temps held firm. I played around with blowing a fan directly over the heatsink and that got temps down to sub 40C.

I have some thermal glue coming and I am going to attach a fan to the heatsink to keep temps as low as possible.

View attachment 64498

I have to admit I'm slightly shocked - an 040 shouldn't run at anything near 100°C - it should crash way before that? I'll check a datasheet ...

 

[Phipli]

Yup, max junction temperature 110°C, I'm astounded it was still running with a measured 105 on the exterior of the package, even if it was the metal underside. How accurate is your thermistor and are you using the basic linear calculation to get the temperature, or the more fancy nonlinear formula?

If you want to ballpark check, the phase change in boiling water holds the temperature at close to 100°C so if you boil some water in a pan that should be near 100°, although you might want to waterproof the thermistor to stop water messing with the resistance.

 

[MrFahrenheit]

Can you take a photo of that 68040 clearly and post it ?

 

[zigzagjoe]

It sounds like one of the early masks. They'll do higher speeds just fine, but they definitely need cooling while doing so. What does the chip say on it?

 

[superjer2000]

Yup, max junction temperature 110°C, I'm astounded it was still running with a measured 105 on the exterior of the package, even if it was the metal underside. How accurate is your thermistor and are you using the basic linear calculation to get the temperature, or the more fancy nonlinear formula?

If you want to ballpark check, the phase change in boiling water holds the temperature at close to 100°C so if you boil some water in a pan that should be near 100°, although you might want to waterproof the thermistor to stop water messing with the resistance.

Before I would have said fairly accurate as I had previously calibrated with ice water and boiling water, but I checked again last night and was getting inconsistent results with boiling water - I need to confirm why the calibration hasn't held and then I'll remeasure and update as required but at this point, I would suggest the without heatsink readings may be higher than actual.

 

[David Cook]

View attachment 64498

Amazing chart!

 

[superjer2000]

Well I am going to have to walk back what now looks like an outlandish claim regarding the temperatures my 68040 reached!

I had recently rebuilt my Arduino based temperature sensor and, after the comments here, went back and realized I had connected my 5v into the Vin, instead of the 5v pin. Because there is about a voltage drop when using the Vin pin, the Arduino wasn't getting enough voltage and it was causing it to read a higher voltage than was actually being supplied.

Updated results are quite different (especially for the without heatsink case). With the heatsink, the CPU temp got up to just over 40C and was very stable, despite running Speedometer tests on loop. Without a heatsink, the CPU got to about 60C, with the measurements being the same when taken from the top and bottom of the processor. Again, Speedometer tests were looping.

I'm currently affixing the heatsink more permanently with thermal paste and thermal glue, but with temps around 40C I don't bother with adding a fan.

Sorry for the confusion!

 

[tt]

Thanks for the updated data. I want to see what temp my 040 without a heatsink gets too. It gets too hot to touch so it's at least up to ~50 C.

 

[zigzagjoe]

Here are some notes I had about cooling a 68040.

I'd be curious how much airflow you have on that 040 when testing and what mask you're using. I saw probably +15 degrees C measured on heatsink or the ceramic respectively over what you got.

Mine was an E31F (.65um) @ 45mhz - it gets noticeably unhappy and will crash once temperatures at the heatsink exceed 55 degrees C with mild ambient airflow. Without a heatsink, it becomes unstable very quickly. I found Mpegdec running an MP3 stream was one of the best ways to test stability; it's strenuous and there were noticeable audible cues when the 68040 became too hot. At that point, it would crash if anything else was done. Midway through that video I apply a fan directly and the audio clears up quickly.

 

[superjer2000]

Looks to be an E42K (65 um) if I understand how the coding works. See pic below.

There is pretty good airflow - I'll snap a picture once the thermal glue has cured and it's back in my IIci with the heat sink but the CPU is pretty close to the air intake grills at the back off the PSU. I had replaced this IIci's PSU with a Seasonic and the fan with a reasonably strong Noctua so there is a reasonable amount of air blowing past the CPU and not a lot of heat generated from the PSU. I've done a better job of affixing the heatsink this time around to take advantage of this by turning it 90 degrees so that the PSU fan will pull air through the heatsink.