4.1 Dual CPU Delid? or Leave the lid down?

Discussion in 'Mac Pro' started by Vicromono, Mar 5, 2016.

  1. Vicromono macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #1
    After a great experience upgrading a single CPU 4.1 to 5.1 six core 3.33, I now have the opportunity to upgrade a dual CPU 4.1. Firstly, it's only because of this forum and the helpful information available here. Very little is out there on the subject of "Delidding". So, I am going to document my experiences here. I started by research and observation of others pulling, heating, hammering, cutting and prying the IHS (integrated heat spreader) from the CPU for proper spacing between the mother board and heat sink. The difference in the dual CPU 4.1 is they are held in place solely by the heat sinks screwed to the mother board. The heat sinks are smaller on the dual units and I believe that Apple ordered the Xeons liddles to compensate for the lack of cooling. I have an old Xeon Chip from a previous upgrade that I have used for practice and the results have taught me that cutting the seal from the IHS is the easy part. I used the old style single edge razor blade to cut just deep enough on all four sides. By drawing a line on the blade as a depth gauge I avoid cutting too deep and damaging the small caps inside. Using two thicker box cutting blades, I inserted one on each side between IHS and PCB then rested the blades on small wooden blocks and evenly applied downward pressure to the chip. Way more pressure was needed than I had expected and am certain that I had damaged the CPU at this point but who knows. Maximum operating temperature of the CPU is 172 degrees so I would assume that I can safely heat the IHS to this temp without causing damage while softening the thermal paste for separation. Also, I wonder what solvents can safely be applied that may simply dissolve the paste for the easiest removal of all. The goal is to avoid installing new CPUs with the lids on/avoid the issues that pertain.
     
  2. box185, Mar 5, 2016
    Last edited: Mar 5, 2016

    box185 macrumors member

    box185

    Joined:
    Sep 12, 2013
    #2
    This is an interesting subject - one that I successfully pursued a couple of years ago. In my opinion, installing de-lidded CPU's makes the dual processor 2009 Mac Pro upgrade easy - as easy as upgrading the single processor 2009 Mac Pro.

    You are right to control the depth of cut on the adhesive in order to protect the eleven capacitors under the lid. I have repaired several processors that were damaged by people who did not control the depth of cut - the capacitors are easy enough to replace, but it can be discouraging. Using a double edge razor blade is easier though because they are thinner.

    The IHS is not attached to the processor with thermal paste - it is soldered on, and heat is required to remove it safely. Once the adhesive is cut, the IHS can be removed by applying heat to the lid - controlling the temperature is key to doing this. The melting temperature of the solder can be determined experimentally. The other parameter to control is time, but removing the lid is MUCH easier than installing the lid as Intel did when they manufactured these parts. In removing the lid, you only need to apply heat long enough to melt the solder just under the lid - you do not need to melt the soldier that is attached to the CPU. Again - the key is to control the temperature.

    Once the lid is removed, a double edge razor blade can be used to cut away the remaining adhesive. The solder that remains attached to the CPU can be removed with a single edge razor blade - the solder is malleable enough that no additional heat is necessary.

    Having a tool fabricated to control the depth of cut and using a temperature controlled heat plate makes all of this easier too.
     
  3. Vicromono thread starter macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #3
    Thank you. I am embarrassed that I didn't figure that out. As you stated, the solder is so malleable that is can be scraped off with my thumbnail.
     
  4. beccas macrumors member

    Joined:
    Feb 5, 2016
    #4
    Do either of you have a video tutorial on completing this task?
     
  5. Vicromono thread starter macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #5
    Check youtube.com under Xeon IHS removal. It is youtube so take some methods with a grain of salt.
    --- Post Merged, Mar 5, 2016 ---
    I have a set of X5670 CPUs ordered and I will document my progress. A delidding service is available for the ones who need it done right the first time. I am lucky enough to have some room to play in this area.
     
  6. nigelbb macrumors 6502a

    Joined:
    Dec 22, 2012
    #6
    Plus some spare cash in case you screw up the process too?
     
  7. box185 macrumors member

    box185

    Joined:
    Sep 12, 2013
    #7
    This is why it's necessary to practice on the least expensive compatible processors, but yes - spare cash is necessary to acquire those as well - just not as much.
     
  8. Vicromono thread starter macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #8
    Thanks box, I'm going to take the separated lid and apply heat with a small gun to the outside and see how long it takes to melt the plate
     
  9. rastaman3d macrumors newbie

    Joined:
    Jul 12, 2008
    #9
    I'm actually having the same doubt. I just bought a MP 4,1 and i'm decidinf if to buy lidded or de-lidded cpus. I too scared to do it by myself! My choice at the moment is to buy lidded cpu and install them as they are being very careful not to crash them. But I'll be happy to have your opinions about it.
    Thanks
     
  10. ActionableMango, Mar 8, 2016
    Last edited: Mar 8, 2016

    ActionableMango macrumors 604

    ActionableMango

    Joined:
    Sep 21, 2010
    #10
    People doing 4,1 CPU updates in the past kept the lids on and used washers around the mount points as spacers so the heatsink cannot crush anything. This seems like the simplest and most economical method, so I'm not sure why it's fallen out of favor. I suppose there is risk of crushing the socket if you get the washer height wrong, but people who have done this have posted the correct washer height, and you could always measure your own CPUs with calipers in case you are worried yours are different.

    There are more steps required than just adding washers. Here's the best thread I've seen about it:
    http://forums.macrumors.com/threads/4-1-processor-upgrade-install-help.1701420/#post-18710268
     
  11. box185 macrumors member

    box185

    Joined:
    Sep 12, 2013
    #11
    "simply kept the lids on and used washers around the mount points as spacers so the heatsink cannot crush anything"

    That and cutting the fan connectors free from the plastic housing in the two heatsinks, and installing additional thermal pads for the regulators. Probably covered in the linked thread, but wanted to mention it here for clarity.
     
  12. ActionableMango macrumors 604

    ActionableMango

    Joined:
    Sep 21, 2010
    #12
    Thanks for the clarification, and I've edited my post in order to not mislead anyone.
     
  13. h9826790, Mar 8, 2016
    Last edited: Mar 9, 2016

    h9826790 macrumors 603

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #13
    I think it's because the Xeon is much cheaper and easy to get now. On the other hand, it's still hard to get a cheap CPU tray. So, it's now worth to pay more (and take the risk to damage the CPU) to get a pair de-lidded CPU, which make the upgrade much much easier (but not take the risk to damage the CPU tray).

    Also, there was no reliable de-lidding service some time ago, but now Box185 can provide that. Which make this option much more attractive as well.
     
  14. rastaman3d macrumors newbie

    Joined:
    Jul 12, 2008
    #14
    Thanks guys for the infos.
    The only thing that is not still clear to me is about the fan connectors, what should i exactly have to do to them? Is there any videos/pics explaining this step?
     
  15. ActionableMango macrumors 604

    ActionableMango

    Joined:
    Sep 21, 2010
    #15
  16. prvt.donut macrumors 6502a

    prvt.donut

    Joined:
    Jan 1, 2008
    #16
    Oh that sounds like me except I damaged mine too much to make repairs worthwhile!:oops:
     
  17. Bubalight macrumors newbie

    Joined:
    Apr 22, 2015
    #17
    You also need to test the processors before you attempt decapitating them. If you received a dead or buggy cpu when you bought used from the cheapest seller you could find, you will have no recourse once you have chopped it up and then find out that it doesn't work, they will rightfully blame you.

    Also a potentially serious problem is uneven tightening. There are several pictures of cpus that have burned a pad off due to the 200 amps that is supposed to be spread out evenly over 200 pins getting funneled through not enough of them.
    Frequently lidded installs have great trouble getting all the memory channels working and they continue to tighten and sometimes it finally works. This is due to not all of the pins contacting. And several people that did a lidded install years ago with CPUs slower than X5690s have later decide to upgrade their mac to the max and after removing they find they have bent the pins on the board, although it had been working, it's a huge risk trying again.
     
  18. box185 macrumors member

    box185

    Joined:
    Sep 12, 2013
    #18
    Testing the processors or purchasing tested processors is excellent advice - excellent.

    The risk of damage to the socket's pin field is significantly lower once the lids have been removed. The most significant problem I have witnessed is that people doing this upgrade will mix the installation process for the two types of CPU's.

    Once the lid has been removed, there is no longer a need to carefully count the turns and "feel" the engagement of the processor - the processor board has standoffs that mate with the heat sinks and those standoffs limit how far the heat sink can compress the socket's pin field. This is the same approach people try to achieve when using washers, except that the washer method also requires the additional thermal pads and cutting the fan connectors loose.
     
  19. Vicromono thread starter macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #19
    A very good point is made here that most of us have no ability to practice. How can you test a cpu without a test fixture? I know the risk involved and know it's a crap shoot as to weather or not the processor is a dud (before or after). I feel a little better knowing they are Xeon.
     
  20. Bubalight macrumors newbie

    Joined:
    Apr 22, 2015
    #20
    I know it seems like they are really expensive, but the guy that sells most of the delidded CPUs has talked about it sometimes, a large part of the extra cost he looses in eBay fees. He also sells them on his website for a discount because of the lower fees. He also has to guaranty they will still work after being chopped apart.

    I have tried the vice method on some junk processors once, it works...usually (3 out of 4 tries) but on one it one cracked off a corner of the die, so that one would have been ruined. With that practice, I'd happily offer to delid CPUs for people for $10, but with no warranty. But for pair of X5690s (which run around $700 right now) I still wouldn't promise success for $400 extra.
     
  21. box185 macrumors member

    box185

    Joined:
    Sep 12, 2013
    #21
    The problem with selling de-lidded CPU's on eBay is that the seller essentially pays the eBay and PayPal fee when they purchase the lidded processors, and then has to add another instance of these fees when they are resold as de-lidded CPU's - taxes are also be a factor. The eBay and PayPal fees together are close to 13% - the real winners here are eBay and PayPal - especially for the now higher cost of the X5690.

    Success can be guaranteed at a reasonable cost if the processors are known good before de-lidding and the de-lidding process uses conservative engineering principles to remove the heat spreader.
     
  22. rastaman3d macrumors newbie

    Joined:
    Jul 12, 2008
    #22
    What do you mean for "conservative engineering principles"?
    Thanks
    --- Post Merged, Mar 10, 2016 ---
    Anyone has tried this method?

    Seems to be the most safe, my only worry is about the temperature that the cpu reach when the iron is sticked to it. could the high temperature fry the cpu?
     
  23. Vicromono thread starter macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #23
    I watched that also and had some questions with that scenario. If the razors are not in deep enough, will the chip just keep cooking? If the iron is to hot, will it cook the chip? Did that one work when he was done? You can't see if it pulls loose until it drops out. I would try it in a glass jar with a pad at the bottom. And find a way to check the temp of the iron. I would call that "seat of your pants" engineering. The thermal paste seems like a good idea especially when using an iron made for clothing where even temperatures are not a big concern.
     
  24. Daze513, Mar 10, 2016
    Last edited: Mar 10, 2016

    Daze513 macrumors newbie

    Joined:
    Sep 22, 2015
    #24
    I have dual X5675s with the lids on. I took my time, laid everything out and did it very methodically, and I didn't have to cut fan connectors or anything. No washers, and temps are better than they were with the OEM CPUS, although that can be due to more efficient paste.

    I can't speak for everyone, but if you take your time, test one processor at a time after install, and use the right tools, and a bit of common sense, you can get it installed, with no modifications to the connector.

    Only thing different on mine are 5MM thermal pads that provide better contact and thus more security for me. As matter of fact my brother ion law and I did it at the same time and his also is the same way. YMMV

    EDIT: Actually now I remember I did mine twice because the first time I used thermal pads that weren't as good as the ones I have now. They were not as dense as the silicone one I replaced it with. So yea, 2 times with 2 processors each time.
     
  25. Vicromono thread starter macrumors member

    Vicromono

    Joined:
    Nov 11, 2015
    Location:
    Oregon
    #25
    I received a pair of 5670s yesterday and started the process of cutting and separation. The one thing I did not prepare for was when the solder reaches it's melting point and the blades between the CPU and IHS release the "stored energy" in the form of spring tension. Here is what happens...
    --- Post Merged, Mar 12, 2016 ---
    It worked... kind of. I need one more CPU now so it will be a bigger gamble now to find one that may or may not match up with the single 5670 I currently have. One CPU functions perfectly and one does not. As I have asked from before, How do I test a cpu without a test fixture? Will a 5670 run in a 5.1 single CPU tray? That is the only way for me personally to test. So, if anyone wants a DOA Xeon x5670 to play with, it's all yours. Does anyone know how vulnerable these CPUs are to pairing. I don't know what tolerance levels need to be for them to function as a pair. Learning what happens has been really fun. It only took one wrong way to find the right way (In your face Edison!). I will use painters masking tape to support the blades and hold things in place next time so the CPU can be "flicked" downward from the IHS as soon as the solder releases it's bond. The single edge razors are nearly the perfect width to cut under. The longer blades are from a wall paper removal tool and supplied a perfect amount of tension and support. With all the blades in place, they also serve as a heat shield from the gun. With all of this in place, The proper temperature does not need to be known because the IHS is released at the exact time of solder melt. It will always bug me though as to what and when the problem occurred with the first CPU. Thanks to Box185 for the help.
     

Share This Page