MP 1,1-5,1 cMP & nMP {MacPro 4,1 & 5,1} Memory Upgrade Compatibility and FAQ

[flyproductions]

Just a hypothesis that I can't be bothered to test.

So i just gave it a shot...

...just to find the config.plist-"Speed"-entries somehow proven to be plain cosmetic.

Here are the results:

1333 MHz...

...and "667 MHz"

 

[pywm]

So i just gave it a shot...

...just to find the config.plist-"Speed"-entries somehow proven to be plain cosmetic.

Here are the results:

1333 MHz...





...and "667 MHz"

Thanks a lot for testing! Good to know the speed doesn't matter.

 

[VaZ]

Yes correct i noticed the values are swapped but still shows an increase in speeds as i add more pairs of ram and the testing tool i used is called "PC Benchmark - Performance Test" in the Windows Apple Store

 

[jeremywork]

Yesterday I put together an MP5,1 to make my existing 4,1>5,1 redundant since I rely on it heavily.

Installed 6 x 32GB 4Rx4 PC3L-10600R-09-11-AB1-D3 modules (Samsung M393B4G70BM0-YH9) and two x5690 procs.
I've become pretty familiar with OCLP but haven't used Martin Lo's pack before- it was quite easy to install to a USB flash drive and with the CustomMemory flag set to true, macOS had no trouble booting.

I wasn't expecting to see 256GB appear in System Information, however, as I had only installed 192GB. I understand most of the memory flags in the config.plist are cosmetic, but I tried my best to set it up to accurately reflect what's installed. I'm not sure if an incorrect total would actually cause a problem or not.

Procedures:
- Removed OpenCore USB drive, reset PRAM 3 times, booted to EFI Windows 10 SSD, dumped dmidecode to txt
- Ran dmidecode from macOS on 4,1>5,1 which has 6 x 16GB 2Rx4 PC3L-10600R and two x5690 procs; without OC
- Reinserted OpenCore USB, option-booted and selected OpenCore EFI with control held, ran dmidecode from macOS
- Compared three outputs of dmidecode, as well as IORegistryExplorer's IODeviceTree:memory page and System Info
- Made countless edits to config.plist using this info and the links below, rebooting to understand what changed

Resources:

https://forums.macrumors.com/thread...re-acceleration.2180095/page-53#post-28255048
Martin Lo's cMP OpenCore installation kit

https://www.tonymacx86.com/threads/...memory-properties-section.305345/post-2186421
Primer on the new CustomMemory key added to OpenCore 0.6.3 beta & info on using dmidecode to gather source info

https://github.com/acidanthera/OpenCorePkg/blob/master/Docs/Configuration.pdf
OpenCore Reference Manual - See section 10.4, Memory Properties (p.75-78 in rev2024.03.03)

https://www.dmtf.org/sites/default/files/standards/documents/DSP0134_3.7.0.pdf
DMTF SMBIOS Reference Specification - See section 7.17-18, Physical Memory Array, Memory Device (p. 96-109)
(OC CustomMemory key seems only to generate DMI Type 16 and Type 17 objects)

https://www.macuser.de/threads/success-32gb-dimms-in-mac-pro-5-1.923751/page-2#post-11973975
Example of config.plist with two slots not populated

https://dortania.github.io/OpenCore-Post-Install/universal/memory.html#mapping-our-memory
OpenCore Post-Install guide for MacPro7,1 - memory mapping information
(Why are the TypeDetail values different here than in OC Reference and SMBIOS Reference? Both seem to work...)

Results:

- The info in System Information is almost natural. I have not found a way to set Type to be recognized as DDR3 ECC as it does on the 4,1>5,1 so it instead shows only DDR3.

- The info in IORegistryExplorer reveals more discrepancies. 'dimm-info' is all zeros. 'reg' shows "08" where "04" on 4,1>5,1 (this may be normal.) Empty slots' 'dimm-part-number', 'dimm-manufacturer', and 'dimm-serial-number' appear to inherit the specified DeviceLocator key instead of appearing as "" and 'dimm-types' appears as "DDR3" instead of "RAM".
Unpopulated slots probably appear less precisely due to DataWidth, FormFactor, TotalWidth, Type, and TypeDetail keys applying to all devices in the array instead of only populated devices. Specifying these keys within the Devices array does not work.

- The info in dmidecode looks very different. Only DMI types 16 and 17 are populated instead of 16, 17, (18 - Windows 5,1), 19, (33 - macOS 4,1-5,1) & 130. Many differences in the Memory Device parameters for both populated and unpopulated slots.

Screenshots:

5,1 Vim windows (left to right):
Windows non-OC dmidecode beginning with Memory Array through all of DIMM 1 (32GB); all of DIMM 4 (empty)
macOS OC dmidecode beginning with Memory Array through all of DIMM 1 (32GB); all of DIMM 4 (empty)
config.plist beginning with Memory Array through all of DIMM 1 (32GB); all of DIMM 4 (empty)

4,1>5,1 Vim windows (left to right):
macOS non-OC dmidecode beginning with Memory Array through all of DIMM 1 (16GB)
macOS non-OC dmidecode all of DIMM 4 (empty)
(bottom/TextEdit) more readable 'dimm-info' block from IORegistryExplorer

Bonus:

I didn't benchmark my own 4,1>5,1 with 96GB because it's heavily loaded and still on MP51.0089.B00, but I did find a few other examples of 5,1, x5690 x 2, 96GB, 10.13.6 builds to compare memory speeds with. Baseline is this 5,1 with 6 x 32GB 4Rx4 PC3L-10600R (before I fixed all the cosmetics.)

https://browser.geekbench.com/v4/cpu/compare/17280185?baseline=17408125

https://browser.geekbench.com/v4/cpu/compare/16989456?baseline=17408125

https://browser.geekbench.com/v4/cpu/compare/16968235?baseline=17408125

 

[MrScratchHook]

Yesterday I put together an MP5,1 to make my existing 4,1>5,1 redundant since I rely on it heavily.

Installed 6 x 32GB 4Rx4 PC3L-10600R-09-11-AB1-D3 modules (Samsung M393B4G70BM0-YH9) and two x5690 procs.
I've become pretty familiar with OCLP but haven't used Martin Lo's pack before- it was quite easy to install to a USB flash drive and with the CustomMemory flag set to true, macOS had no trouble booting.

I wasn't expecting to see 256GB appear in System Information, however, as I had only installed 192GB. I understand most of the memory flags in the config.plist are cosmetic, but I tried my best to set it up to accurately reflect what's installed. I'm not sure if an incorrect total would actually cause a problem or not.

Procedures:
- Removed OpenCore USB drive, reset PRAM 3 times, booted to EFI Windows 10 SSD, dumped dmidecode to txt
- Ran dmidecode from macOS on 4,1>5,1 which has 6 x 16GB 2Rx4 PC3L-10600R and two x5690 procs; without OC
- Reinserted OpenCore USB, option-booted and selected OpenCore EFI with control held, ran dmidecode from macOS
- Compared three outputs of dmidecode, as well as IORegistryExplorer's IODeviceTree:memory page and System Info
- Made countless edits to config.plist using this info and the links below, rebooting to understand what changed

Resources:

https://forums.macrumors.com/thread...re-acceleration.2180095/page-53#post-28255048
Martin Lo's cMP OpenCore installation kit

https://www.tonymacx86.com/threads/...memory-properties-section.305345/post-2186421
Primer on the new CustomMemory key added to OpenCore 0.6.3 beta & info on using dmidecode to gather source info

https://github.com/acidanthera/OpenCorePkg/blob/master/Docs/Configuration.pdf
OpenCore Reference Manual - See section 10.4, Memory Properties (p.75-78 in rev2024.03.03)

https://www.dmtf.org/sites/default/files/standards/documents/DSP0134_3.7.0.pdf
DMTF SMBIOS Reference Specification - See section 7.17-18, Physical Memory Array, Memory Device (p. 96-109)
(OC CustomMemory key seems only to generate DMI Type 16 and Type 17 objects)

https://www.macuser.de/threads/success-32gb-dimms-in-mac-pro-5-1.923751/page-2#post-11973975
Example of config.plist with two slots not populated

https://dortania.github.io/OpenCore-Post-Install/universal/memory.html#mapping-our-memory
OpenCore Post-Install guide for MacPro7,1 - memory mapping information
(Why are the TypeDetail values different here than in OC Reference and SMBIOS Reference? Both seem to work...)

Results:

- The info in System Information is almost natural. I have not found a way to set Type to be recognized as DDR3 ECC as it does on the 4,1>5,1 so it instead shows only DDR3.

- The info in IORegistryExplorer reveals more discrepancies. 'dimm-info' is all zeros. 'reg' shows "08" where "04" on 4,1>5,1 (this may be normal.) Empty slots' 'dimm-part-number', 'dimm-manufacturer', and 'dimm-serial-number' appear to inherit the specified DeviceLocator key instead of appearing as "" and 'dimm-types' appears as "DDR3" instead of "RAM".
Unpopulated slots probably appear less precisely due to DataWidth, FormFactor, TotalWidth, Type, and TypeDetail keys applying to all devices in the array instead of only populated devices. Specifying these keys within the Devices array does not work.

- The info in dmidecode looks very different. Only DMI types 16 and 17 are populated instead of 16, 17, (18 - Windows 5,1), 19, (33 - macOS 4,1-5,1) & 130. Many differences in the Memory Device parameters for both populated and unpopulated slots.

Screenshots:

5,1 Vim windows (left to right):
Windows non-OC dmidecode beginning with Memory Array through all of DIMM 1 (32GB); all of DIMM 4 (empty)
macOS OC dmidecode beginning with Memory Array through all of DIMM 1 (32GB); all of DIMM 4 (empty)
config.plist beginning with Memory Array through all of DIMM 1 (32GB); all of DIMM 4 (empty)

View attachment 2362361


4,1>5,1 Vim windows (left to right):
macOS non-OC dmidecode beginning with Memory Array through all of DIMM 1 (16GB)
macOS non-OC dmidecode all of DIMM 4 (empty)
(bottom/TextEdit) more readable 'dimm-info' block from IORegistryExplorer

View attachment 2362362

Bonus:

I didn't benchmark my own 4,1>5,1 with 96GB because it's heavily loaded and still on MP51.0089.B00, but I did find a few other examples of 5,1, x5690 x 2, 96GB, 10.13.6 builds to compare memory speeds with. Baseline is this 5,1 with 6 x 32GB 4Rx4 PC3L-10600R (before I fixed all the cosmetics.)

https://browser.geekbench.com/v4/cpu/compare/17280185?baseline=17408125

https://browser.geekbench.com/v4/cpu/compare/16989456?baseline=17408125

https://browser.geekbench.com/v4/cpu/compare/16968235?baseline=17408125

Hey man, you got alot to digest here, but i just found your post and i know what your going thru as far as the memory config, so i just wanna say that the 256 is purely cosmetic, and all the ram should be available but if you want to set them correctly leave the 32gb in the memory section as is as well as the 32768 part for 32gb dimms, and wherever you have no dimms inserted put the gb to 0 as well as the 32768 part to 0. and if you insert 16 gb dimms the other number is 16384. hope this helps

 

[leonardobertaso]

Is anyone using HYNIX 16GB 2Rx4 PC3L-10600R 1333MHZ 1,35V DDR3 ECC REG SERVER on a MP 4,1 > 5,1?

I’m going to buy 6 of this and mount them on my dual X5690.

Technically they should work, but I was reading about someone that was having trouble using DDR3L. The seller says that they’re for server hardwares and not for a PC, but specs seems to be ok with this MP…

 

[VaZ]

Yes that'll work in your Mac Pro server hardware. The low-power mode (L) will be ignored by your Mac Pro and just run on regular power mode.