Fortunately all modern CPUs have protections built into them for these sorts of things. If it hits Tj(max), it'll just shut itself down. Which sounds like what was happening.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 Mac Pro 2019 CPU Upgraded to 28 Cores, shut Down unexpectedly. [solved] with More Update
- Thread starter zhpenn
- Start date
- Sort by reaction score
Fortunately all modern CPUs have protections built into them for these sorts of things. If it hits Tj(max), it'll just shut itself down. Which sounds like what was happening.
Quite the opposite . Not all the socket pins had a firm connection with the silicon lands . Had one half or all of that sink been pushing too hard on the array the pins would have squashed down and he would have lost more than the chip and the logic board .
Remember what happened when Anand Shimpi first upgraded the 2009 Dual CPU Mac Pro with off the shelf silicon ? And he was an experienced Builder and tech writer . He basically fried everything ...
Last edited:
Special Thanks to @Snow Tiger on #12 https://forums.macrumors.com/thread...pectedly.2217170/?post=28095222#post-28095222
Reminds me that maybe the screws do not tighten enough.
the heatsink is not tightened enough. (WHICH MAY BE VERY COMMON ISSUE)
Until now it up and running 10+ hours not have any issue.
Reminds me that maybe the screws do not tighten enough.
the heatsink is not tightened enough. (WHICH MAY BE VERY COMMON ISSUE)
Until now it up and running 10+ hours not have any issue.
I did mention this in the more generalised thread about this system. The socket for this CPU is very fragile. Apple has tried to make it a bit easier to work with by using a retention wire but the CPU package is so large it can be difficult to get even mounting pressure to engage all the pins.
Now I'm not saying don't do such upgrades. I'm of the belief we should all tinker more! - But be careful, take photos with your phone to gauge the default angle and distance from the motherboard before removing anything.
If there is an issue you could lose DIMM slots or PCIe channels until the right mounting pressure is attained. At worse, you could fry something by pushing or bending power delivery pins into adjacent data pins.
Now I'm not saying don't do such upgrades. I'm of the belief we should all tinker more! - But be careful, take photos with your phone to gauge the default angle and distance from the motherboard before removing anything.
If there is an issue you could lose DIMM slots or PCIe channels until the right mounting pressure is attained. At worse, you could fry something by pushing or bending power delivery pins into adjacent data pins.
It sounds as delicate as heart surgery. I suspect there will be posts in the future that show a 50/50 chance of getting it correct. But then again, some might be shy about posting their failures....
This is only going to get worse going forward . Next year we'll get LGA 4189 socket personal workstations . 542 extra pins for extra fun . The industry needs to make this simpler , but the PC side simply attaches the silicon to the heatsink and lowers away ... Mac Pro upgraders don't have that luxury .
@zhpenn @crjackson2134 @Snow Tiger what is suggested to stress test a system like this for hours? Link?
I loaded Windows 10 Pro Workstation via bootcamp and ran GIMPS with the heaviest setting . It maxed out processor thread and the entire memory kit concurrently and continuously for hours . There are some utilities used to verify this - one of which is AIDA64 .
Great Internet Mersenne Prime Search - PrimeNet
Great Internet Mersenne Prime Search - Finding world record primes since 1996. GIMPS is an organized search for Mersenne prime numbers using provided free software.
Home | AIDA64
www.aida64.com
I’m curious who does that. I want to see how that works, seems interesting.
For this socket it's the common way to do it.
You'll see here the chips are upside down and attached to their heatsinks already using little plastic clips. The chip on the right is the same one the Mac Pro supports (that CPU would actually fit and boot in a Mac Pro). The chip on the left is also the same series of chip but has an Infiniband fabric extension included on the processor.
You'll notice if you look in the bottom left corner and top right corner of these heatsinks there are silver holes with a very large diameter. These are guide holes which help you insert the heatsink with the CPU pre-attached directly down into the socket.
Here is a picture of the socket with the alignment posts visible.
As you can see here there are metal spikes sticking up to engage with the heatsink and there is no ILM (Independent Loading Mechanism) for this socket. Meaning there's no hinged latching mechanism on the socket that holds the processor down with uniform pressure. This differs to Intels older sockets like LGA 1366, 2011, 2066. For this socket the heatsink alone holds the CPU in the socket and the intention is for the CPU to be pre-attached to the heatsink and the entire assembly to be lowered down into the socket.
I bring all this up not to comment in any way on how Apple has done it, just to provide you with the images you seeked as I too find it quite interesting
- This mechanism is the default way Intel recommends things be done but it's not the only way. Other manufacturers have gone more custom with their mounting solutions for this ginormous socket.
I bring all this up not to comment in any way on how Apple has done it, just to provide you with the images you seeked as I too find it quite interesting
Yeah, thanks for that. I was familiar with the carrier and latch that Threadripper and Epyc use to similar effect to keep everything aligned, but this is new to me.
This post is so great but is still perplexing me:
In pic 1, what makes the CPU stay on the heatsink body (instead of staying on the motherboard)?
In pic 2, how does the heatsink standoff plate get separated from the heatsink?
Pic 1 is an example Pic, really. The CPU stays attached to the heat sink because of the thermal paste, since the bracket is was removed from the logic board instead of removing the heat sink from the bracket.
As for Pic 2: The two mounting screws for the heat sink attach to the standoffs, hidden by the shroud of the heat sink.