Repasting a Sonnet Duet G4

[ziggy29]

Just want to see if anyone has done this successfully, and make sure there are no hidden "gotchas" involved. I have acquired a 1.6 GHz Sonnet Duet G4 for my Digital Audio, and it works... for a few minutes. Eventually I get crashes and freezes, I think almost certainly because the CPU is overheating even when I keep the side panel open. And the fan on the card is working fine.

So my suspicion is that maybe the card needs a new application of thermal paste, but it looks like the unit is sealed in a way that discourages "users" from opening it up (for good reason in the general case). It looks like there are four security Torx screws to remove in order to access the innards. I have not tried it yet, but I plan to go out soon and pick up the screwdriver for it. But before I make that investment I'm hoping to see if there might be something I am missing, or some issue with doing this. So if someone has been there and done that, I'd appreciate some insights!

 

[Fami7455]

Dried up thermal paste could perfectly be the cause of thermal issues. However, I've only seen one case so bad a machine became unstable because of it. And that was because some doofus used sub-par quality paste when redoing an MDD.

Are you using the original power supply? Those tend to degrade with time and get unstable when asked to deliver higher currents. The Duet is certainly a lot more power hungry than the stock processor, so a flaky power supply could also be the cause of instability here.

If it turns out to be the paste, I'd say the repasting process shouldn't be that hard. You'll need to make sure of a couple of things:
1) Spread the paste evenly across the die surfaces, or you could get hotspots.
2) Don't overtighten the heatsink screws when putting the thing back together, or you could crack a die.

 

[ziggy29]

Thanks. I'm no stranger to repasting -- I've successfully opened up a Quad G5 to do it -- just wanted to make sure there were no surprises when I open this thing up to get to the CPU. I also have a pretty hefty GPU in there, so maybe I'll test with a stock GPU that pulls a lot less power, and if that is a lot more stable maybe the PSU is an issue, but this thing does feel *really* hot when it quits so I think there are thermal issues regardless.

Would also add that OS 9 runs a lot longer than OS X before freezing or crashing. I suspect that's because OS 9 isn't taxing the CPU as much.

 

[bunnspecial]

I did my dual 1.8.

You'll have to ignore the warnings on it and cut the stickers to get to it.

Aside from that, IIRC I just took my time and tightened the screws slowly and carefully when I put it back together.

 

[ziggy29]

I did my dual 1.8.

You'll have to ignore the warnings on it and cut the stickers to get to it.

Cool, thanks. I'm not worried about voiding a warranty that expired 10 years ago!

 

[barbu]

1) Spread the paste evenly across the die surfaces, or you could get hotspots.

Not trying to over-rule you here - and correct me if I'm wrong - but is it not the correct technique to apply the paste as a small pea-sized ball and NOT spread it around? I followed this when I upgraded the Xeons in my old MP1,1. I put a small ball down on each CPU, and installed the heatsinks. When I had to disassemble shortly thereafter, the paste had perfectly spread and conformed itself to the chip very evenly and neatly. I think spreading can cause the paste to squish out the side which is obviously undesirable.
Just my 0.02$

 

[ziggy29]

If you put a small "ball" of paste in the middle of it and push straight down (not at an angle), the paste should evenly spread itself. Otherwise it could be uneven, inadequately cover some parts of the die and could ooze out the side; whether or not it causes problems depends on what it oozes into and the conductivity of the paste, though in any event it would be a sloppy job.

 

[Fami7455]

Not trying to over-rule you here - and correct me if I'm wrong - but is it not the correct technique to apply the paste as a small pea-sized ball and NOT spread it around? I followed this when I upgraded the Xeons in my old MP1,1. I put a small ball down on each CPU, and installed the heatsinks. When I had to disassemble shortly thereafter, the paste had perfectly spread and conformed itself to the chip very evenly and neatly. I think spreading can cause the paste to squish out the side which is obviously undesirable.
Just my 0.02$

If you put a small "ball" of paste in the middle of it and push straight down (not at an angle), the paste should evenly spread itself. Otherwise it could be uneven, inadequately cover some parts of the die and could ooze out the side; whether or not it causes problems depends on what it oozes into and the conductivity of the paste, though in any event it would be a sloppy job.

I recommend the ball technique for devices with an integrated heatspreader (most modern socketed CPUs and those pesky Sanyo STK amplifier modules, for example), since the surface area tends to be bigger and more irregular. The pressure of the heatsink helps the paste get to all the nooks and crannies.
However, I've found the spread method better for bare dies, because the area is usually small and very flat. These two things combined usually eliminate the possibility of air pockets. And the small excess of paste squishing out the sides shouldn't be a problem when using a non-conductive, non-capacitive paste. So far I haven't had any issues with any of the chips I've done this way.

In the end it all comes down to personal preference, I guess. I just don't like the risk of having a lack of coverage on a die caused by a miscalculation of the amount of paste I should have applied.

 

[bunnspecial]

I generally spread on CPUs without an IHS, which includes every G3/G4/G5 I've ever seen. I don't want to take a chance of not having paste on part of the die.

The MP 1,1 is a bit different since all of the CPUs for that DO have IHSs. In fact, the only MP I know that doesn't have them is the dual 4,1.

 

[barbu]

I appreciate the info, good thread.

 

[ziggy29]

Success! I performed the 'surgery' this morning (needed a T15 Security Torx), and the paste was all but obliterated and just a little powdery crud. No wonder it was crapping out within five minutes! I'm actually surprised it ran long enough to boot up to the login screen.

Cleaned it up, applied new paste (used Noctua NT-H1), closed it up tightly without forcing the screws, let it rest for a few hours.

I installed it and fired it up half an hour ago, have been stress testing it and it is holding up nicely! Thanks to all who chimed in. And it's always nice to "save" one of these, since they are not making more of them.

 

[ziggy29]

A little Geekbench "before and after", just for grins. Still rock solid.

 

[Fami7455]

A little Geekbench "before and after", just for grins. Still rock solid.

View attachment 857459 View attachment 857460

 

[ziggy29]

The sad thing is, how many of these (and other stuff that is coveted now) went into the trash because they needed a fairly simple fix? (This took less than 15 minutes.)

 

[z970]

The sad thing is, how many of these (and other stuff that is coveted now) went into the trash because they needed a fairly simple fix? (This took less than 15 minutes.)

The thought sickens me, and usually comes down to the conclusion that people, in fact, suck.

EDIT: I just remembered that this directly ties into why that fixit shop recycling all those machines was so horribly painful, that they died an easily preventable death, if only someone was there in the right place, at the right time to save them. Of course, it only rubs salt in the wound when said fixit shop heavily focuses on iPhone and MacBook repairs too, when said products will themselves only get replaced again in another two years...

Thanks, Apple. Instead of going truly green and advocating for recycling of usage and owners, you advocate for recycling of perfectly functioning, user-serviceable machines for the production of specially screwed, glued together pieces of shiny junk. All for your ever-loving pursuit of fatter pockets.

The reality of it all makes one grit their teeth...

 

[eyoungren]

The sad thing is, how many of these (and other stuff that is coveted now) went into the trash because they needed a fairly simple fix? (This took less than 15 minutes.)

The rest of them are on eBay through a couple of sellers who know the value and consequently give them a serious markup. There is one particular vendor who is known to snap up the private sales and then turn them around for ridiculous prices.

 

[ziggy29]

Thanks, Apple. Instead of going truly green and advocating for recycling of usage and owners, you advocate for recycling of perfectly functioning, user-serviceable machines for the production of specially screwed, glued together pieces of shiny junk. All for your ever-loving pursuit of fatter pockets.

Yeah, they talk a mean game about being "green" but the truth is that more than any other vendor, in the Cook era they engineer their systems to be all but impossible to service or upgrade any more. So when one component breaks or the whole thing becomes too slow because you can't upgrade the RAM or the CPU or the GPU... into the trash it goes for a lot of folks. Really green, isn't it?

 

[redheeler]

Yeah, they talk a mean game about being "green" but the truth is that more than any other vendor, in the Cook era they engineer their systems to be all but impossible to service or upgrade any more. So when one component breaks or the whole thing becomes too slow because you can't upgrade the RAM or the CPU or the GPU... into the trash it goes for a lot of folks. Really green, isn't it?

And now the last Mac ever made with all these things upgradable (unless you happen to have $6,000 lying around somewhere, that is) officially can't run the latest MacOS Catalina. How convenient. Apple is more than happy to greenwash while simultaneously exercising the planned obsolescence of perfectly good older Mac hardware, and I am always surprised by the amount of people on this forum willing to defend this behavior - well, it is 7 years old. To me it is simply inexcusable. For a company the size of Apple and the prices these Macs are being sold at, 11 years of security updates should be the bare minimum.

 

[ziggy29]

Apple is more than happy to greenwash while simultaneously exercising the planned obsolescence of perfectly good older Mac hardware, and I am always surprised by the amount of people on this forum willing to defend this behavior - well, it is 7 years old. To me it is simply inexcusable. For a company the size of Apple and the prices these Macs are being sold at, 11 years of security updates should be the bare minimum.

Yep. This is a response to Moore's Law slowing down and older machines not going technically obsolete as fast as they used to. Fact is, even a 10-year-old machine with maxed out RAM and an SSD is still perfectly capable for most things. Apple knows it, which is why they have intentionally time-bombed systems perfectly capable of running the next OS by preventing the installer from working.

 

[AphoticD]

Success! I performed the 'surgery' this morning (needed a T15 Security Torx), and the paste was all but obliterated and just a little powdery crud. No wonder it was crapping out within five minutes! I'm actually surprised it ran long enough to boot up to the login screen.

Cleaned it up, applied new paste (used Noctua NT-H1), closed it up tightly without forcing the screws, let it rest for a few hours.

I installed it and fired it up half an hour ago, have been stress testing it and it is holding up nicely! Thanks to all who chimed in. And it's always nice to "save" one of these, since they are not making more of them.

Well done @ziggy29, excellent outcome!

I’ve always lightly spread the thermal paste on my PowerPC chips and MacBooks/MBPs. The intel desktops and later model MBPs with bigger IHS take a dollop and spread fine with pressure. I wouldn’t be too concerned with a little spill over, just go light-on to begin with.

Fun fact; The now superseded fanless 12-inch MacBooks used no thermal paste or pad at all. The CPU IHS presses against a carbon coating on the inside of the bottom case (yes, bottom case and not the typical top case). I’m glad Apple dropped this design!

 

[jmilan0302]

I recommend the ball technique for devices with an integrated heatspreader (most modern socketed CPUs and those pesky Sanyo STK amplifier modules, for example), since the surface area tends to be bigger and more irregular. The pressure of the heatsink helps the paste get to all the nooks and crannies.
However, I've found the spread method better for bare dies, because the area is usually small and very flat. These two things combined usually eliminate the possibility of air pockets. And the small excess of paste squishing out the sides shouldn't be a problem when using a non-conductive, non-capacitive paste. So far I haven't had any issues with any of the chips I've done this way.

In the end it all comes down to personal preference, I guess. I just don't like the risk of having a lack of coverage on a die caused by a miscalculation of the amount of paste I should have applied.

Maybe I'll try repasting my MDD like you suggested, I could never get good coverage on the middle CPU (not sure which is exactly considered second/first) so there very well might be better to spread it. But I probably also have PSU issues so I can't really test temps or anything.