MacInTouch Amazon link...

2006-2013 Mac Pro and alternatives

Channels
Apple, Products

Ric Ford

MacInTouch
Here's a memory issue that I didn't readily find a solution to: I bought six used 16GB DIMMs for my two Mac Pro 4,1's and a dual-Xeon server (running Linux). I also have some 8GB DIMMs and many 1GB and 2GB DIMMs. All are ECC. When I use the two 16 GB DIMMs, any other DIMMs are ignored. This is true in all three computers. I tried many permutations of DIMMs in slots on all three computers. Depending on the DIMM arrangement, I can end up with less than the 32 GB of the two DIMMs that are causing the problem....
DIMMs must be placed in exactly the right arrangement in a Mac Pro or you'll lose access to some of the RAM. And, confusingly, the arrangement is different among different years/models!

I saw exactly this problem when a client bought a bunch of memory and installed it himself in a Mac Pro 1,1. I later got the computer back to install Linux and it showed minimal RAM, oddly, so I bought more. When I opened it to install the new RAM, I found it loaded with RAM that wasn't showing up. It took a while, but I eventually got the arrangement correct and all the RAM showed up.

(This, however, doesn't preclude the possibility that you just bought bad RAM.)
 


DIMMs must be placed in exactly the right arrangement in a Mac Pro or you'll lose access to some of the RAM. And, confusingly, the arrangement is different among different years/models.
(This, however, doesn't preclude the possibility that you just bought bad RAM.)
Ric is correct in that you need to install it in the correct order. Follow this guide, based on your model, and pay particular attention to the table at the end that tells you where to install the DIMMs.

Bad DIMMs are also a real problem. I had a bunch of RAM in a Mac Pro 2010 that was giving me all kinds of problems, everything from crashing or not POSTing to various amounts of RAM not being detected. I eventually isolated the problem to a single DIMM that was defective. Removing that from the mix made all the problems disappear.
 


Here's a memory issue that I didn't readily find a solution to: I bought six used 16GB DIMMs for my two Mac Pro 4,1's and a dual-Xeon server (running Linux). I also have some 8GB DIMMs and many 1GB and 2GB DIMMs....
I just did some more testing to refresh my memory. ;)
When any other DIMMs are added beyond the two 16GB ones, the Mac Pro's power light flashes, indicating that no memory is installed. I populated all 8 slots in my Mac Pro – that worked properly until I added the 16GB DIMMs. When testing my dual-Xeon server, it also reported no memory was installed.
DIMMs must be placed in exactly the right arrangement in a Mac Pro or you'll lose access to some of the RAM. And, confusingly, the arrangement is different among different years/models!
I followed the directions in Apple's Technician Guide. I even tried permutations outside of the recommendations.
(This, however, doesn't preclude the possibility that you just bought bad RAM.)
I've tested the RAM and tried swapping and mixing the DIMMs between computers. The only consistent result is that when any of these 16GB DIMMs are used with any other DIMMs, all three computers think they have no memory installed.
 



What are the specs of the 16GB modules and the other DIMMs you’re trying to use them with?
Apple spec' the Mac Pro's memory at PC3-8500 (1066 MHz). The Xeon X5690s I have can use 800, 1066, or 1333 MHz. My 16- and 8-GB DIMMs are 1333, the 2- and 1-GB DIMMs are 1066. I can mix and match the 8-, 2-, and 1-GB DIMMs, I just can't use the 16's with anything else.

The Technician Guide says each DIMM can have up to 18 memory ICs (16 + 2 for ECC); the max size is 4GB/DIMM. My 16- and 8-GB DIMMs each have 36 chips, but the 8GB ones work with additional smaller-sized DIMMs, while the 16 GBs don't. (FWIW: the 2 GB DIMMS have 18 chips, the 1 GB DIMMs have 9 chips.)

One other spec that may contribute: the rank and number of banks, e.g. 2Rx4 means dual-rank with 4 banks. My 16- and 8GB DIMMs are both 2Rx4; my 2GB ones are 2Rx8, my 1GB ones 1Rx8. I know some computers can't mix 2Rx4 and 2Rx8 or 1R and 2R, yet my Mac Pro mixes the 8GB 2Rx4 with 2GB 2Rx8 and 1GB 1Rx8. But it won't mix the 16 and 8 GB DIMMs, which are both 2Rx4.
Ric is correct in that you need to install it in the correct order. Follow this guide, based on your model, and pay particular attention to the table at the end that tells you where to install the DIMMs.
The online guide is more up-to-date than the Technician Guide I've been using. The main difference is that it says each DIMM can have 36 ICs (chips).
Bad DIMMs are also a real problem. I had a bunch of RAM in a Mac Pro 2010 that was giving me all kinds of problems, everything from crashing or not POSTing to various amounts of RAM not being detected. I eventually isolated the problem to a single DIMM that was defective. Removing that from the mix made all the problems disappear.
Although I tested in the past, and swapped the 16GB between different computers, I'll run memtest extensively to be sure. (Since it's DOS based I'll have to put my original GPU back in and hookup a keyboard and mouse. Tomorrow...)
 


Ric Ford

MacInTouch
Here's a memory issue that I didn't readily find a solution to...
I just did some more testing to refresh my memory....
Look what I found, while I was looking for a time- and cost-effective solution to your problem....
OWC said:
32.0GB OWC Memory Upgrade Kit
...
IMPORTANT OWC NOTE:
OWC p/n OWC1333D3MPE8GB (8GB modules) and OWC1333D3MPE16G (16GB modules) require that all installed modules be of the same p/n OWC 8GB and/or 16GB modules. Other existing Apple or 3rd party 1GB, 2GB and 4GB modules are not supported for use with these kits and need to be removed when these modules are added.
 


When I use the two 16 GB DIMMs, any other DIMMs are ignored.
This sounds a lot like you might be trying to mix R(egistered) DIMMS and U(nregistered) DIMMS, which won’t work. Virtually all 16GB DIMMS are RDIMMS, smaller sizes could be either. There’s no way to tell by just looking at them, you have to check the part numbers. (Or hmm, maybe if you boot the machine with just the smaller ones the detailed memory readout from About this Mac/System Report will say which type they are - I can’t quite remember if it does on a Mac Pro or not.)

FWIW I believe this is also the source of the OWC 16GB memory incompatibility Ric mentions above...
 


Look what I found, while I was looking for a time- and cost-effective solution to your problem....
IMPORTANT OWC NOTE:
OWC p/n OWC1333D3MPE8GB (8GB modules) and OWC1333D3MPE16G (16GB modules) require that all installed modules be of the same p/n OWC 8GB and/or 16GB modules. Other existing Apple or 3rd party 1GB, 2GB and 4GB modules are not supported for use with these kits and need to be removed when these modules are added.
This sounds a lot like you might be trying to mix R(egistered) DIMMS and U(nregistered) DIMMS, which won’t work. Virtually all 16GB DIMMS are RDIMMS, smaller sizes could be either. ...
FWIW I believe this is also the source of the OWC 16GB memory incompatibility Ric mentions above...
We have a winner! All six 16GB DIMMs are now in the Mac Pro (it has 96 GB).

Months ago, I tried more than two 16GB DIMMs at once, but apparently I put them in the wrong slots (in spite of having the manual telling me exactly what to do...).

Four of the 16GB DIMMs are an exact part number match. They're made by Micron and labelled by Dell for their servers. The other two are also from Micron, with part numbers very close to the Dells', they were certified by a 3rd-party for Dell servers.

FWIW: I burned a CD with Memtest86 and tested the 32 GB that were in the Mac Pro earlier today. No errors found.

Many thanks to Ric, Todd, and John for sleuthing this out. It's very helpful having the great knowledge of MacInTouch plus new perspectives on a problem.
 


I installed High Sierra on my two 3,1 (2008) Mac Pros over the summer and ran into a perplexing problem.

Both computers had 64 GB of RAM installed, one with a matched set from OWC and the other with a set from Hynix. Both worked fine with Sierra, but when I tried to install High Sierra using the dosdude1 patcher, the install took hours on both. The one with the Hynix RAM was painfully slow to boot into High Sierra and was beachballing every few minutes for long periods once booted. The one with the OWC RAM was also painfully slow booting but never reached the desktop.

I was stumped. A few hours of googling found someone on MacRumors reporting the same thing.

It turns out that removing RAM would make the Mac Pros blazingly fast, but having the full amount would slow them to a crawl. I'm not sure what's causing this but am currently running both with six 8GB DIMMs.
 


I finally found the old GT120 video card that came originally with my Mac Pro 2009; I've been looking for it for months. I installed it today in the one empty PCIe slot (not connected to any monitor), but the Mac wouldn't boot. Rearranging cards to different slots, the Mac simply wouldn't boot with the GTX120 card installed.

My Mac Pro 4,1 (firmware updated to 5,1) is running Mojave and has a Sapphire RX580 8GB video card installed. It has not been flashed from PC to Mac, so I don't get any startup screen or startup options visible. I wanted to install the GT120 because it is a single-slot video card, so if I needed the startup options, I wouldn't have to pull out the dual-slot RX580 and put in the older dual-slot ATI Radeon 5870 that I was running previously to purchasing the RX580.

I know that Mojave needs a Metal-capable video card but can't remember if some aspects of Mojave wouldn't run properly or if the Mac simply won't boot with a non-Metal GT120 installed (but not connected to any monitor).

Anyone with knowledge of Mojave's limitations with non-Metal video cards?

I suppose there might be something wrong with the card, even though it's been in an anti-static bag for years. I'll have to see if I have an older High Sierra boot disk in a drawer to test further.
 


... Anyone with knowledge of Mojave's limitations with non-Metal video cards?
Bottom line: a fully updated Mojave will boot with the GT120 and a Metal GPU, but will not recognize the GT120 after part of the boot process completes.

I tested on my single-CPU 2009 Mac Pro flashed to 5,1 running Mojave 10.14.6 with all updates. I removed one of its two Radeon RX560s and put in the original GT120, hooking up a DVI display to monitor progress. When the startup progress bar got about half-way, the GT120's display froze and never updated after that. Once Mojave booted, the Radeon RX560's display worked. The GT120 did not appear in the Displays System Preference.

This behavior has changed in the last few months, I don't know which update was responsible. Back around May Mojave wouldn't install without a Metal GPU but would boot and run without one. In fact, it booted and ran with only the GT120 installed.

Since I run this Mac Pro headless, I use dummy plugs on the Radeon RX560s' DVI ports so they are recognized and appear in the Displays System Preference. Most dummy plugs fake common standards, mine are set to 1080p (1920x1080).
 


I just tried my 2009 Mac Pro flashed to 5,1 running Mojave with an nVidia GT120 and it works fine. Normally I run it with an r280x card, but I just shut down the machine and removed the r280x and put in its place the GT120 card as the only video card in the machine. It boots up just fine, with video working perfectly.

I'm running macOS 10.14.6 with the latest security updates. The monitor I'm using is a 1280x800 11" monitor, using an HDMI cable plugged into the DVI port of the card via an adapter. Here's an image of the about screen showing the card being used.

I've also just popped in a Catalina boot SSD (built using DosDude's Mac Pro Catalina patch) and selected that drive via the bootscreen. That boots just fine, as well, with video working.
 


I just tried my 2009 Mac Pro flashed to 5,1 running Mojave with an nVidia GT120 and it works fine.
An update on this... I now have the GT120 and the 280x in the flashed 2009 Mac Pro 5,1, with the 1280x800 monitor plugged in to the GT120. The video screen looks like a frozen boot, with the grey screen, the apple logo, and a 2/3-finished progress bar. But I can connect via screen sharing, and the computer is working. I even get the same About this Mac screen, saying I've got an nVidia GT120 with 1280x800 resolution.

When I have monitors plugged into both the 280x and the GT120, then, as Sam Herschbein said, only the monitor plugged in to the 280x worked and the GT120 card does not even show up.

So, in conclusion, it seems that you can only use a GT120 with Mojave if it's the only video card in your computer. It won't work at all if monitors are plugged in to both cards or just the Metal-compatible card. You will get no video output at all if plugged into only the GT120 with a Metal-compatible card unused but still accessible via screen sharing.
 


Since I run this Mac Pro headless, I use dummy plugs on the Radeon RX560s' DVI ports so they are recognized and appear in the Displays System Preference. Most dummy plugs fake common standards, mine are set to 1080p (1920x1080).
If I can ask, which exact plugs are you using?
 



My experience has been that, when using patches to install updates on a Mac, you are limited to the type of video card you can use. While Mojave is still supported on 2009 Mac Pros, Catalina is not, and I believe sometimes it requires that you use a supported Nvidia graphics card. While the GT120 sometimes works as a test card, the Quadro K5000 or GeForce GTX680 may be better choices, since they are Mac-supported graphics cards.

Once the macOS reached Mojave, there were no supported drivers for Nvidia cards other than these two.
 


This is a followup to my post #22399 where I ran into an issue installing High Sierra on a Mac Pro 2008 with 64 GB of RAM installed. With the full set of RAM, the Mac was painfully slow to boot and laggy when running after boot, but would run fine with 48 GB.

I still don't know why this happens, but I found a post on MacRumors about how you can use up to 62 GB of RAM and not have the slowdown. You just need to run a command in Terminal to restrict the amount of RAM the Mac uses
Bash:
sudo nvram boot-args="maxmem=63488"
I then reinstalled the extra 16 GB of RAM and am happily working away with About This Mac showing 62 GB.
 


This is a followup to my post #22399 where I ran into an issue installing High Sierra on a Mac Pro 2008 with 64 GB of RAM installed. With the full set of RAM, the Mac was painfully slow to boot and laggy when running after boot, but would run fine with 48 GB....
I think the same may be true (in a way) with the Mac Pro 2009-2012. If you lower the RAM number down from 64 you'll get better scores from Geekbench and such. I will do this test when I get home today.
 


I think the same may be true (in a way) with the Mac Pro 2009-2012. If you lower the RAM number down from 64 you'll get better scores from Geekbench and such. I will do this test when I get home today.
I tried this on my Mojave 2009 Mac Pro flashed to 5,1 with 96 GB of RAM. Limiting it to 62 GB showed no appreciable difference in Geekbench 5 scores. For the short time I was running at 62 GB, I did not notice any change in responsiveness.

I ran this Mac with 32 GB for a while before upgrading to 96 GB. I never noticed any slowdown after the increase – in fact, performance is better since I can run a lot of memory-hog apps and not run out of memory.
 


I tried this on my Mojave 2009 Mac Pro flashed to 5,1 with 96 GB of RAM. Limiting it to 62 GB showed no appreciable difference in Geekbench 5 scores. For the short time I was running at 62 GB, I did not notice any change in responsiveness. I ran this Mac with 32 GB for a while before upgrading to 96 GB. I never noticed any slowdown after the increase – in fact, performance is better since I can run a lot of memory-hog apps and not run out of memory.
I think the same may be true (in a way) with the Mac Pro 2009-2012. If you lower the RAM number down from 64 you'll get better scores from Geekbench and such. I will do this test when I get home today.
I did try the lower RAM script in Terminal
Bash:
sudo nvram boot-args="maxmem=66488"
and it did not improve anything, and actually made it worse, so I will stay with the 64 GB of RAM. Thanks for this verification.
 


OK, I have a 2009 Mac Pro 4,1 that I successfully flashed to the latest 5,1 firmware and had been usingit fine until the CPU tray had an issue. I acquired a replacement tray and I noticed my memory, which is 1333 bus speed, reverted to 1066 speed.

I zapped the PRAM multiple times to no avail. I had the thought of downgrading the firmware back to a 4,1 to re-flash, thinking something on the CPU tray must be getting updated with the firmware flash. Unfortunately I haven't found any info on anybody successfully downgrading the firmware for Mojave 144.0.0.0 back to any 4,1 version.

I attempted to use both the original and the Github "patched" version of the firmware flash utility with zero success.

I suppose I could live with the lower memory bus speed but would prefer not to, if at all possible. Hoping somebody might be able to offer a reasonable suggestion.
 


OK, I have a 2009 Mac Pro 4,1 that I successfully flashed to the latest 5,1 firmware and had been usingit fine until the CPU tray had an issue. I acquired a replacement tray and I noticed my memory, which is 1333 bus speed, reverted to 1066 speed. ...
The memory speed is determined by which processor variation you have. The W3520, W3540, W3670, and W3680 are limited to 1066. The W3580 and W3590 do 1333.

Your new tray must have a processor that only supports 1066.

You could swap processors. The catch is that removing the heat sink can be tricky because you may damage the CPU socket. Even with all my experience I'm reluctant to do this. I ruined a CPU tray while re-applying heat sink compound, there was no obvious damage to the CPU socket nor to the mobo.
 


I have been using a late 2012 Mac Pro under macOS 10.14 and tried to install the latest Security Update 2020-002 today. The updater ran, the Mac Pro rebooted, and up comes the dreaded circle/slash and no hard disk. Luckily I had a fast backup and ran Carbon Copy Cloner to restore, and all was well after 11 minutes.

Has anyone else run into a problem with this update?

Thanks for your help.
 


The memory speed is determined by which processor variation you have. The W3520, W3540, W3670, and W3680 are limited to 1066. The W3580 and W3590 do 1333.
Your new tray must have a processor that only supports 1066.
I think you gave me a hint. The first tray I had must have had X version processors, as when I flashed the machine initially, the memory speed jumped to 1333 after resetting PRAM. I'll have to fire the machine back up and take a look at what System Report states as the processor type.
You could swap processors. The catch is that removing the heat sink can be tricky because you may damage the CPU socket. Even with all my experience I'm reluctant to do this. I ruined a CPU tray while re-applying heat sink compound, there was no obvious damage to the CPU socket nor to the mobo.
Been there, done that (times two)! I sent out one tray to a so-called "professional" to have the sockets replaced, and they toasted the board, as it worked with the original 2.26GHz processors installed, but it now only max fans the box. The other tray I attempted to hot-air rework myself, and an ill-timed arm movement lead to one or two ripped printed circuit traces that I decided to give up on.

I picked up tray replacements from various vendors at lower than traditional market value. One is equipped with 2.92GHz processors, which I plan on leaving alone (that tray that only reports 1066 memory speed). The second tray I'm picking up on Monday has the 2.26GHz processor, which I plan on replacing with a pair of X5260 Xenons that I had (tested and already delidded).
 


Anyone running the Mac Pro 5,1 (2010) with macOS Catalina 10.15.3 (using Dosdude1's patch) with a GTX 680 Mac Edition?

The install went smoothly, and everything seems to be functional (even with the ancient Nvidia drivers), but if I let the display go to sleep, it will refuse to wake back up, and I have to force-shutdown the computer. The machine is actually awake, but the display is black, and nothing I have tried will bring it back.

I didn't have the sleep issues with a 5700 XT, but the 5700 XT was crashing the whole machine, so I'm back to the GTX 680 for now.
 


The memory speed is determined by which processor variation you have. The W3520, W3540, W3670, and W3680 are limited to 1066. The W3580 and W3590 do 1333.
Your new tray must have a processor that only supports 1066.
I looked up the processor string via Terminal, and it reported back as X5560 - a processor that supports 1333MHz speed, according to the online spec sheet. That's making me believe that the CPU tray must get flashed along with the main backplane board when a firmware update is applied.

My other Mac Pro 4,1, also flashed to 5,1, has a slightly older firmware version (141.0.0.0), so once I get my second CPU tray, I'll install the higher speed processor, test it in machine #1 and then attempt to install Mojave on a clean hard drive in the second machine with hopes that the latest Mojave installer will flash the firmware and replacement processor tray with the 144.0.0.0 firmware.
 


I looked up the processor string via Terminal, and it reported back as X5560 - a processor that supports 1333MHz speed, according to the online spec sheet. That's making me believe that the CPU tray must get flashed along with the main backplane board when a firmware update is applied.
Ah, you jogged my memory. The Mac Pro 4,1 does not support 1333 Mhz; Mac Pro 5,1 does. Have you flashed the new tray?

FWIW: Before doing the usual macOS firmware updates, be sure that Sound Output is not set to Digital. I had many failures installing High Sierra before discovering this.
I looked up the processor string via Terminal, and it reported back as X5560 ...
The X models are "premium" versions. Their TDP (Thermal Design Power) is 95 watts. The W models are hotter, with a TDP of 130 watts. My dual-CPU tray has X5690's.

Since you mentioned de-lidding, you're clearly capable of swapping processors. I also de-lidded some Xeons. I practiced on my original W3520's before buying and de-lidding the more expensive X5690's. It was far easier than I thought, given all the warnings I read.
 


Ah, you jogged my memory. The Mac Pro 4,1 does not support 1333 Mhz; Mac Pro 5,1 does. Have you flashed the new tray?
Yes, as I mentioned in reply #27084 above, I did flash my first Mac Pro 4,1 machine successfully all the way up to 144.0.0.0 (my second 4,1 machine I've flashed up to 141.0.0.0, where I'm hoping to install my second CPU tray to flash both the tray and the machine up to the final firmware version 144.0.0.0).
The X models are "premium" versions. Their TDP (Thermal Design Power) is 95 watts. The W models are hotter, with a TDP of 130 watts. My dual-CPU tray has X5690's.
Good to know - wondered what the difference was between X and W series.
Since you mentioned de-lidding, you're clearly capable of swapping processors. I also de-lidded some Xeons. I practiced on my original W3520's before buying and de-lidding the more expensive X5690's. It was far easier than I thought, given all the warnings I read.
As Murphy's Law goes, I did it for someone else and all went well. I did it for myself and it became a train wreck (twice!), hence the replacement CPU trays. I'm still stumped by the dual X5560's not running at 1333MHz memory speed though. No amount of PRAM resets, main board resets via the push button and battery removal cured it like the previous CPU tray.

I figure I'll wait for my second tray where I'll be installing 3.33 GHz CPUs and hope to flash the second machine up to firmware 144. I'd really like an NVMe boot drive.
 


Amazon disclaimer:
As an Amazon Associate I earn from qualifying purchases.

Latest posts