Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Microsoft Corporation

macrumors newbie
Original poster
May 16, 2018
18
0
Hi guys,

Ive got (what once was) a working mac pro 4.1 flashed to 5.1 and bought a pair of used x5690s on ebay. I wanted to upgrade the stock CPUs to the x5690s on my dual cpu mac pro. Im aware that these must be delidded to avoid thermal problems, however, prior to delidding i wanted to test them first.

I read forums that using one cpu will boot up with the fans on full blast. None of this happened for my case, ive installed one of the x5690s on socket “a” with socket “b” empty. I get no boot chime, nothing being displayed, no onboard led errors, other than the power led suddenly blinking quickly and continuously on the first attempt with the fans running like idle.

At that point ive closely checked the sockets for debris, bent or broken pins (also inspected the cpus), made sure the video connections were secured, tried doing the pram reset multiple times, also did the smc reset by leaving the computer unplugged for 20 secs, gently tightened more the cpus which none of this worked.

Ive then use the other x5690 and hook it up on cpu a with b still being empty, the issue still remained, the only difference noted is that the power led was solid white.

Afterwards, i hookup both x5690s on both sockets, making sure they’re seated properly and the issue still remains.

Finally, I hookup both the original cpus and strange enough the issue still remains.

Ive found a forum with someone facing similar issues but doesn’t really discuss the issues im facing. At this point im out of ideas and currently stuck with a non working mac pro.

A theory i came up with is that one or both 5690s might have been previously used in a fried or damaged motherboard that could’ve ruined my cpu tray that not even the original cpus that came with my mac would suddenly not work anymore, but i knock on wood and hope that this isn’t the case.
 
I would imagine that the CPU sockets got damaged when using the 5690 cpu's when not delidded. its easy to apply to much pressure to the socket when tightening down the heat sinks. also there should be a red CPU light that comes on to say the CPU is faulty or not seated properly.
 
I would imagine that the CPU sockets got damaged when using the 5690 cpu's when not delidded. its easy to apply to much pressure to the socket when tightening down the heat sinks. also there should be a red CPU light that comes on to say the CPU is faulty or not seated properly.
None of the sockets appear damaged, the pins are in shape. I occasionally got a red led errors due to the cpu not being secured. For that, i would finger tighten the heatsinks with the screwdriver till feeling a mild resistance. Ive also count the turns when screwing or unscrewing.

Even so, the system still detects loose cpus which makes me think that its not really a faulty contact. I can also feel the heat being dissipated in the heatsinks which I wouldn’t leave it running for long.
 
Last edited:
When lidded CPUs are installed the heatsink fan power connector will not reach, so did you mod the power connector to fit? If not that may be your problem. I suspect that it won’t boot because it’s not getting a good fit.
 
When lidded CPUs are installed the heatsink fan power connector will not reach, so did you mod the power connector to fit? If not that may be your problem. I suspect that it won’t boot because it’s not getting a good fit.
Good call, i did not. Would i need to clip off the clips that hold onto the heatsink and connect it into the board directly? Also will both need to be modified and connected even if there isn’t a cpu in socket “b”?

One more thing, wouldn’t it already run with the original delidded cpus?
 
A X5690 can fry a Rev.A early-2009 CPU tray. CPU trays made before early May 2009 don't work with 130W Westmere Xeons, just with 95W ones.

First thing is to find the week of the build date from the CPU tray SSN to confirm that it was made past at least week 19, 20 to be sure (something like J5920________, third character is the year, then the next two numbers are the week).

Then you should borrow an early-2009 single CPU tray, single CPU tray requires a normal lided Xeon, and test each of the X5690 and de-lid the working ones.

You can buy a dual CPU tray bare PCB for an early-2009 for around $130 - it's just the bare PCB and you will reuse the aluminium tray base + heatsinks + screws:

661-4998 Apple Processor Board 8-Core (without processors) for Mac Pro Early 2009 820-2336-A
 
Last edited:
When lidded CPUs are installed the heatsink fan power connector will not reach, so did you mod the power connector to fit? If not that may be your problem. I suspect that it won’t boot because it’s not getting a good fit.

You have to remove the power connector clips from the heatsink and then push it down so that it fits in the connector if you haven’t delidded. I’d do it with just the A heatsink for now, test the CPUs one at a time. The power connector is made to reach with original delidded CPUs installed and won’t reach with lidded CPUs installed. Once you are confident that the processors are working then I would delid. The vice grip method is easiest.
 
I would suggest making high res macro photos of the cpu sockets and post them in this thread.

I would never use lidded cpus in a dual 4.1 board. Too easy to damage the pins.

They also change color when shortened or overloaded, so asking for pictures.

Of course what Alex mentioned can be (also) the cause, early cpu boards what are not capable of running high power Xeons.
 
update:

last night I’ve hooked up the original CPUs and all i did is to tighten them up a little more and finally things are back to normal again. Strange that no error leds were on.

Ive noticed that the flash was unsuccessful this whole time apparently after looking up in system profiler for the serial number on my cpu tray. So after numerous attempts it wouldn’t flash. Ive googled around and didn’t find much so my first thought is due to the os im running on which is os x 10.13.

For the little i know, Mac os 10.11 is known to be the last supported system for the 4.1 models. For those questioning on how am i running 10.13, it has already been preinstalled since ive got the mac. (Im not sure how)

Im now facing another issue where im trying to get a copy of any supported versions like 10.10 or 10.11 to see if the flash would work. Ive originally tried booting to network recovery with “shift - option - cmd - r” and “option - cmd - r” which none of them worked. Im not sure if something needs to be configured because not even the startup disk menu doesn’t give me the option to connect with wifi (unless in the 4.1 firmware was never a thing).

Moreover, finding a older copy of mac os has been another source of wicked pain the the butt. Im aware that the official Apple website has older versions of mac os available to download, the issue with getting a older copy from apple wouldn’t let me install the pkg file on my newer macbook (which is the only alternative mac device i have). So i google around to see if i can get a copy, unfortunately many are just shady sites and fake adware downloads. Then lastly i end up looking in torrent sites like thepiratebay or kickasstorrents to see if i could find something but many are for hackintosh systems.

To wrap things up for now, could using a newer unsupported os be a factor to flash issues? Also, is network recovery really only accessible to a newer firmware or could something be wrong? And finally, when installing a single cpu, will i still need to connect the heatsink to the empty socket? If so, would i risk ruining the pins if i left the heatsink sit over it?
 
Last edited:
Hacked installs don't update the Mac Pro firmware, you need to install 10.11 natively. Internet Recovery is only supported with Mac Pros after late-2013 one - your's don't have it.

A BootROM reconstruction service can bypass all this hassle and get you to 144.0.0.0.0 firmware fully upgraded.
 
You can get Mavericks from the Internet Archive:


Read this article to know how to create a bootable installer:

 
you’re right, el capitan was just pure problems. Thanks to tsialex for the installer, ive managed to flash the mac with no problems.

Ive finally tested the processors and both work great. I managed to Delid both with the iron method, it was nerve racking but successful. Now im at a point where i need to remove the solder from the die, blading it would be risky.
 
D7E48F07-E6FE-4BAC-8B2B-3C11F6E3B3D0.jpeg
Update:

It appears that ram slot 3 and 4 are not being detected and im now getting 2 led errors. Initially I double checked the CPUs and did discover debris on the CPUs themselves but no bent or dusty pins in the sockets which both appear in good shape. All ram slots worked before and remained untouched until i Inspected slot 3 and 4. Ive tested all 8 ram sticks which all make up 64gb and work fine. Slot 3 did happened to have some debris.

Here’s what ive done so far:

- Tighten both cpus to max (inspected the sockets after)
- cleaned out the cpus
- air blew the sockets
- inspected each ram stick
- cleaned ram slot 3 and 4
- swapped around all ram sticks
- pram reset
- rtc reset

None of these methods worked. Ive followed another forum of someone facing similar problems but their methods didn’t worked for me. Any ideas?

Forum:

 
Did you swapped the Xeons to confirm if it's the CPU tray or the Xeons? If you swap the Xeons between A and B and the defective slots continue to be the same, it's the CPU tray.

This is one of the adverse effect of using lided Xeons with an early-2009 CPU tray, CPU sockets pins are easily bent or damaged. Inspect the pins with a magnification glass.
 
RAM errors usually are caused by a bad socket connection. Check for damage, swap CPUs, reseat heat sinks and retest like tsialex recommends.
 
Did you swapped the Xeons to confirm if it's the CPU tray or the Xeons? If you swap the Xeons between A and B and the defective slots continue to be the same, it's the CPU tray.

This is one of the adverse effect of using lided Xeons with an early-2009 CPU tray, CPU sockets pins are easily bent or damaged. Inspect the pins with a magnification glass.
I’ll give it a shot, ive delidded them both so i hope it’s not a problem.

One thine to note is that im using the same black plastic gaskets that go over the cpu
 
Did you swapped the Xeons to confirm if it's the CPU tray or the Xeons? If you swap the Xeons between A and B and the defective slots continue to be the same, it's the CPU tray.

This is one of the adverse effect of using lided Xeons with an early-2009 CPU tray, CPU sockets pins are easily bent or damaged. Inspect the pins with a magnification glass.
You’re the man! The swap actually detected the full ram set, thanks alot!
 
A X5690 can fry a Rev.A early-2009 CPU tray. CPU trays made before early May 2009 don't work with 130W Westmere Xeons, just with 95W ones.

First thing is to find the week of the build date from the CPU tray SSN to confirm that it was made past at least week 19, 20 to be sure (something like J5920________, third character is the year, then the next two numbers are the week).

Then you should borrow an early-2009 single CPU tray, single CPU tray requires a normal lided Xeon, and test each of the X5690 and de-lid the working ones.

You can buy a dual CPU tray bare PCB for an early-2009 for around $130 - it's just the bare PCB and you will reuse the aluminium tray base + heatsinks + screws:

661-4998 Apple Processor Board 8-Core (without processors) for Mac Pro Early 2009 820-2336-A
where actually can you find out which revision your CPU tray is, and what week it is built?
 
where actually can you find out which revision your CPU tray is, and what week it is built?

Just look at the CPU tray serial number label, like the early-2009 CPU tray (Rev.A) serial number label below, I've explained it already in the post you quoted:

First thing is to find the week of the build date from the CPU tray SSN to confirm that it was made past at least week 19, 20 to be sure (something like J5920________, third character is the year, then the next two numbers are the week).

For example, see this example CPU tray serial number label:

screen-shot-2022-02-04-at-18-36-15-png.1954559


J591300Z01LUC - CPU tray is from 13th week of 2009, the week from 23 to 29 of March, so it's a Rev.A.

Whenever I need an early-2009 CPU tray, I ask for the serial number and I only buy if it's from end of July onwards to be absolutely conservative and sure that I'm not getting a Rev.A - something like this example below:

J5929xxxxxxxx - 29th week of 2009 (3rd digit is the year, 4th and 5th are the week number), this is not a Rev.A.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.