4,1 CPU upgrade: pre-delidding test

[cdf]

I'm in the process of upgrading the CPUs in my 2009 Mac Pro. I'm thinking of removing the IHS on them to keep everything as original as possible on the computer, but I'd like to make sure that they function properly before. What would be the best way to test the CPUs using only my computer? It seems that to test them, I'd have to go through the process of installing them, albeit one at a time as CPU A, using something like the washer method (requiring that I modify the fan cable), which would defeat the purpose.

If everything goes well, I'll document the delidding here. I just need to test the CPUs first. Any advice would be great!

 

[RedTomato]

You're more likely to get a good answer at a cpu focussed or modding forum.

Maybe you could test the chips on a spare computer to avoid wear and tear on your mac pro internals?

 

[AkuskaUK]

Good luck!

 

[ActionableMango]

I've never done the 2009 CPU swap, so I don't know how much clearance there is. However, if there's enough room, you could theoretically remove the fan connector from the heatsink entirely and use a pair of alligator clip leads to electrically connect the fan connector to its matching socket on the logic board.

If you haven't bought the CPUs yet, I've noticed that on Ebay occasionally there are servers that go for around the same price as just the bare CPUs (or in rare cases, some times the server is actually cheaper than the bare CPUs). So instead of buying bare CPUs, you could buy a whole server and test the CPUs before you move them over to your Mac Pro.

You could call around to some PC recycle shops in your area to see if they'd be willing to test the CPUs in an old Westmere PC for a reasonable charge.

You could buy an old Westmere PC on Craigslist cheap, test your CPUs in it, and then resell the PC. Assuming you got it at market price, you should be able to resell it at market price.

That's all I can think of. I hope one of these ideas works for you. For me though, I'd just assume they were okay and do without testing them first. CPUs are pretty reliable and I've never bought a bad one, used or new. You'd be playing with the odds a little bit, but I think they are good odds.

 

[raymanster]

I took the risk and de-lidded without testing. I got the x5677 which is significantly cheaper, so could afford to risk it. If I was going for the x5690 at 4-5 times the price I'd be a lot more cautious.

Perhaps you should just go for a vendor selling delidded processors? Or one that provides a delidding service?

 

[DanSilov]

If everything goes well, I'll document the delidding here. I just need to test the CPUs first. Any advice would be great!

I've recently delidded a bunch of CPUs for me and my friend, and it's in fact not that scary or difficult. But it takes some practice.

1. We've bought 2x X5675 and 2x X5680 for our Mac Pros, but those were hard to chase on eBay for good prices, so I've also purchased about 5 different CPUs for testing about $5 each. For example, X5550 is compatible with 4,1 tray and can be tested all the way.

2. Before delidding I've tested every CPU to make sure they are actually ok. For this you need to do the following: On a two-CPU tray remove CPU B cooler and CPU itself. Install lidded CPU into slot A. Then carefully pull fan connector from the assembly, place it over, and after that you can carefully place the cooler over the lidded CPU, connect the fan, make a few light turns for every screw and then try to boot.

You can boot to OSX to check the CPU and then shut down. Fans will kick in full speed without CPU B, but that's ok. If it doesn't boot, CPU might be bad, or most probably you need one or two additional turns on cooler screws.

Testing took a lot of time of careful connectiion and disconnection of the cooler. If you do not rush, it will be ok. Fan connector is tricky to keep in place, once it's disconnected from the assembly, but it allows to install lidded CPU without cutting the wire, and then you can install the connector back the way it was.

3. After testing we started delidding. Actually, turns out the most problematic part of the process is not heating the CPU (more on that later), but careful cutting of the resin around the lid. It super tight, razor must be very thin, and you need to be careful not to cut any resistors, which are very close to the resin. If you don't cut all of it, heating will not separate the lid from the CPU, you will think you need to more heat and you might overheat it. Cheap test CPUs were very helpful to get the hang of it.

4. After you cut the resin, you need to heat the CPU enough to desolder the lid. The very first try was with the lighter-torch. I've placed the CPU over the cup, holding with some razors stuck between the lid and the CPU. The problem was that after it got delidded, everything feel into the cup, and the lid quckly stuck back to the CPU. I've then had to reheat again, and the CPU was dead (tested).

5. After that I've tried using hot iron method, and it worked magically. Spread some cooling paste over the lid, turn the hot iron on. Once it's hot enough, press into the CPU, it will glue to it because of the paste. Then hold over a cup or glass. As solder melts enough, it will pop, CPU will fall into the cup, and the lid will remain glued to the hot iron.

You can see this method here:

Seriously, this is the easiest way, I've delidded 7 CPUs one by one using it, all survived.

6. After CPU is cool enough, take the razor and remove solder from the die. Don't worry, you can damage the die with the razor, only solder will be removed. Just be careful not to cut any resistors.

7. After that I've also used some very fine sandpaper (3000, 5000 and then 7000 grit), but that's optional.

All CPUs are now functioning perfectly, even though I've accidentally cut half on resistors on one of them.

 

[SolidCake]

I delidded 2 Xeon X5690's myself with nothing but a vice, that's a "vise" for you Yanks

The upgrade turned out great and I highly recommend this method as you don't risk damaging the caps with razors and such. Not to mention you don't need to apply heat to the chips.

 

[DanSilov]

I delidded 2 Xeon X5690's myself with nothing but a vice, that's a "vise" for you Yanks
The upgrade turned out great and I highly recommend this method as you don't risk damaging the caps with razors and such. Not to mention you don't need to apply heat to the chips.

From what I've seen, this is actually quite risky, and works well only for non-soldered heatsinks. If the heatsink is soldered to the die, pressure from the vice might tear the die from the board.

 

[SolidCake]

From what I've seen, this is actually quite risky, and works well only for non-soldered heatsinks. If the heatsink is soldered to the die, pressure from the vice might tear the die from the board.

Yes, it is risky, so are the other methods. The reason I find it the 'safest' is that there are a lot less variables for errors. Applying pressure, slowly and gradually is the way to go. The solder is softer than the diode and should come loose easily when doen right. (Check the video). The picture you posted looks more like heat damage than anything else. Also, the vice method would not work with your CPU because it has caps al around, leaving no safe path for the heat shield to travel.

 

[DanSilov]

Yes, it is risky, so are the other methods. The reason I find it the 'safest' is that there are a lot less variables for errors. Applying pressure, slowly and gradually is the way to go. The solder is softer than the diode and should come loose easily when doen right. (Check the video). The picture you posted looks more like heat damage than anything else. Also, the vice method would not work with your CPU because it has caps al around, leaving no safe path for the heat shield to travel.

I'm actually surprised it worked for you, but I have no reason to doubt it, I guess it also works. I'm not worried that the die is softer than solder. But I think the die connection to the PCB board is the weakest point. The image was definitely from from the vice method, just something I've found quickly. Before, when I researched this topic, I've only seen vice-method applied to non-soldered heatsinks.

In any case, I think buying 2-4 cheap test CPUs would be very helpful for anyone, who wants to try delidding himself.

 

[cdf]

Thank you all for the great info! I went ahead and delidded the processors without testing.

With the lids on, the heatsinks must be tightened just right for the processors to be detected; too tight, and the CPU socket can be damaged. I rather take a loss on the processors than on my machine's CPU tray.

To delid the processors, I used the razor-blades-and-heat method, similar to DanSilov's excellent guide in this thread. See also youtube.com/watch?v=aPhDfUkll-o. I must admit, though, that SolidCake's vice method is intriguing.

For me, cutting the adhesive was not an issue. I used a flat no. 17 x-acto blade with tape on it to indicate the depth of the cut (3mm, just shy of 1/8'', is enough). I inserted the blade straight in, then removed it. I moved the blade over and repeated. I did this twice around the perimeter of the lid. (See the picture below.)

To release the lid, I inserted 4 razor blades to create tension (3mm deep and overlapping each other at the edges). I placed the processor lid down in a rectangular cutout of a piece of cardboard and taped the ends of the razors to the cardboard. I suspended the cardboard and used a pencil torch to heat the IHS. I kept the outermost flame in contact with the lid. After about a minute and a half, the lid fell off.

The last step was removing the solder from the die. I used an old credit card to gently peel off the solder. Then, I used 600 grit sand paper and rubbing alcohol to remove the last layer, revealing the smooth glass-like surface of the die.

For me, heat was an issue. The first CPU worked, but not the second. For the second processor, I kept the brighter, innermost flame of the torch in contact with the lid. That killed it.

Before I delid a replacement, I will record the temperature using an infrared temperature sensor pointed at the underside of one of the lids that I removed. I will document the temperature using the inner and outer flames of the torch and using an iron.

 

[DanSilov]

Before I delid a replacement, I will record the temperature using an infrared temperature sensor pointed at the underside of one of the lids that I removed. I will document the temperature using the inner and outer flames of the torch and using an iron.

Sorry to hear about the killed CPU. I strongly suggest giving a hot iron a go, it's much more reliable heat source, because it spreads it equally and very fast, but it doesn't get too hot to kill the CPU fast.

 

[cdf]

Sorry to hear about the killed CPU. I strongly suggest giving a hot iron a go, it's much more reliable heat source, because it spreads it equally and very fast, but it doesn't get too hot to kill the CPU fast.

What setting did you set the hot iron to? I did the tests that I mentioned in my last post, and surprisingly the pencil torch that I used (both with the inner and outer flame) creates a more gradual heat than the iron (set on max). The iron is hotter and faster. Is that better?

 

[DanSilov]

What setting did you set the hot iron to? I did the tests that I mentioned in my last post, and surprisingly the pencil torch that I used (both with the inner and outer flame) creates a more gradual heat than the iron (set on max). The iron is hotter and faster. Is that better?

Pencil torch doesn't heat the lid equally, and temperature varies depending on the distance of the torch. It's very hard to control (at least in my experience). With hot iron you set it to the max or close to it. Solder heats and melts everywhere simultaneously, and then lid pops very quickly and it sticks to the iron (use thermo paste). I've killed the first test CPU with the torch, but then 5 CPUs have been delidded without any issues with hot iron.

 

[cdf]

I ended up using the iron method on the replacement and it worked perfectly. I have now successfully upgraded my Mac Pro. The iron method is indeed a very controlled way of delidding. The temperature is just right for the task.

The indium solder that is used between the die and the IHS melts at 160°C (320°F). The maximum temperature that my iron reaches is 175°C. (I measured this temperature using an infrared sensor pointed at a spot on the iron painted in thermal grease. Without the grease, the surface is too reflective for the sensor to work properly.) Once the iron reached 175°C, I placed it onto the processor's grease-covered IHS. After about 45 seconds, I measure the temperature of the underside of the CPU: 80°C. After one minute, the lid popped off. There was very little solder on the lid; most of it was still on the die. I measured the temperature of the CPU again: still 80°C. So it's really just a top layer of solder that ever reaches 160°C.

To make sure that the lid popped off as quickly as possible, I suspended a 100g (about 3.5 oz) weight to the board of the CPU. (In the picture below, you can see the strings of the hanging weight.)

Many thanks to DanSilov who pointed me in the right direction!

 

[DanSilov]

Many thanks to DanSilov who pointed me in the right direction!

Glad it worked out well for you, congratulations your old-new speedy beast!

 

[jggorman]

I delidded 2 Xeon X5690's myself with nothing but a vice, that's a "vise" for you Yanks

The upgrade turned out great and I highly recommend this method as you don't risk damaging the caps with razors and such. Not to mention you don't need to apply heat to the chips.

I second that. I tried the razor method and ended up slicing through my resistors.

 

[MIKX]

As you cut. . . you ALIGN the angle of the razors UPWARDs = AWAY from the top of the resistors.

Does this sound illogical ?

It worked 3 times for me.
======================

After cleaning the top of the CPU and of course having previously covered your now exposed CPU tray CPU sockets. . . . of course you covered them .. . .

==============Easy bits are now done.

Most unsuccessful 4.1 CPU upgrades are from either TOO MANY . . . . or . . . .. NOT ENOUGH turns on the heatsink bolts and countless other reasons that do not occur to those who are in the "Must DO this TONIGHT ! " group.

if you are contemplating upgrading a 4,1's CPUs but are not a methodical, thinking, patient person . . . don't blame any MarRumors members.


HINT : Once either of the Dual Tray CPU heatsinks are removed ... . COVER the now exposed CPU SOCKET pins with a non-conductive material - never, ever, ever touch the CPU socket pins.

 

[jggorman]

You totally got me pegged with:
"those who are in the "Must DO this TONIGHT ! " group. "

 

[MIKX]

No offense intended. 4,1's are notorious for CPU upgrading.

I'm kind of lucky as my 4,1 is a Dual Quad 2.93 Ghz which is enough for my needs.

5,1's are dead easy by comparison.

 

[Amsterdamned]

The easiest method I’ve found so far, for delidding, is using a vice. But not a bench vice, a drill press vice. They have a much smoother, straighter action so the lid simply slips off. No fuss at all. I’ve done a fair few and all worked perfectly. The edge of the lid is also a great tool for removing the solder from the top of the die.

If you’re in Adelaide and need some help, let me know

 

[ArchStudent]

Sorry to hear about the killed CPU. I strongly suggest giving a hot iron a go, it's much more reliable heat source, because it spreads it equally and very fast, but it doesn't get too hot to kill the CPU fast.

Are you adding paste like arctic silver 5 to the iron / heat spreader and then “ironing“ it?

Can the paste be easily cleaned from the iron, or would it be advisable to get a second hand iron just for this purpose?

 

[DanSilov]

Are you adding paste like arctic silver 5 to the iron / heat spreader and then “ironing“ it?

Any paste will do, just as long as it is sticky to keep the lid "glued" to the iron when the CPU falls in the cup. Otherwise the lids falls as well, and then get immediately soldered back to the CPU.

Can the paste be easily cleaned from the iron, or would it be advisable to get a second hand iron just for this purpose?

Just place some tinfoil between the CPU and the iron. Paste can be cleaned, but most irons have holes for steam, the paste gets there and can be a pita to clean.