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

[h9826790]

Thanks! I would assume chosing either PC3L 1333Mhz or PC3 1600Mhz memory will not be different from getting the normal PC3 1333 Mhz: the first one will run at 1.5V consuming the same power as a PC3, the other will run at 1333 Mhz (in CL9), so practically all three routes yield the same results in the cMP.

Can the low voltage be better silicon to begin with?

We may assume both 1600MHz or PC3L has better silicon. Because one of the chip can run stably at 1600MHz, the other one can stably run at 1333MHz with just 1.35V. So, on the cMP, the chip will be either downclocked or has more voltage to further increase stability. But I don't think practically make any difference.
[doublepost=1551182722][/doublepost]

Thank you. Very helpful.


What about ECC vs non-ECC? That should make a difference in speed shouldn't it?
Or should one expect crashes and nothing but trouble if going for non-ECC?
(My office is about 30 meters above sea level at roughly 59° latitude - if that matters to anyone familiar with bit flipping caused by cosmic rays.)

ECC is always the way to go (if available). Speed difference is very negligible in real world.

I ran non ECC DIMM in my Mac for quite a few year (at that time, high capacity ECC RAM are more expensive, and my computer is just for normal home use, doesn't really matter). It's completely stable. And 99% of modern computer are non ECC anyway, the chance of have single bit error cause by cosmic rays which eventually cause whole system crash should be very very low.

Just 30m AMSL shouldn't be a matter at all. I usually work at above 10000m, and I never see a single sudden crash that can point to cosmic radiation (the aircraft's computer of course are protected, but all other computer equipments are just normal PC, not even using ECC RAM). Even on the "high solar activity year (which occur every 11 years)", still nothing strange happen. So, I personally won't worry about this too much.

But if a single undetected bit error can cause catastrophic failure in your workflow. Then I will say you have no choice, but only use ECC RAM.

 

[HunPro]

Thanks!

I can actually get a combo:

MICRON MT36KSF2G72PZ-1G6E1FE 16GB 2RX4 PC3L-12800R DDR3

Low voltage and 1600 MHz. I also see this type sold in a German mac store, so it shouldn't be risky.

 

[HunPro]

Update: I installed the sticks, I went with these: DDR3 16GB 1600 MHz REG ECC HP 672612-081 HMT42GR7MFR4C-PB

It was pretty wild.

1. Normal boot, OSX comes on, only 32 GB recognized in slot 1 & 2
2. PRAM reset
3. Normal boot, OSX comes on, only 32 GB recognized in slot 1 & 2
4. PRAM reset
5. Normal boot, folder with a question mark

6. Removed stick from slot 4
7. Normal boot, OSX comes on, only 48 GB recognized in slot 1, 2 & 3
8. Added stick from previous slot 4 to slot 3 (to test if the stick was bad)
9. Normal boot, OSX comes on, only 48 GB recognized in slot 1, 2 & 3

10. removed all sticks, installed all four
11. PRAM reset
12. Boot fails, slot 1 led red
13. Switched the sticks in slot 1 and 2 (to see if the stick in slot 1 was bad)
14. Normal boot, 64 GB of ram recognized. No red leds.

Running Rember now.

 

[bsbeamer]

Did you clean out the RAM slots at all?
How many sticks were being used previously?

 

[HunPro]

Did you clean out the RAM slots at all?
How many sticks were being used previously?

2x8GB sticks. I did not, I wasn't sure how to clean them.

It's been running since with 64 GBs, Rember is yet to finish. I don't have any writeable CDs ready to burn memtest on.

 

[MRisberg]

ECC is always the way to go (if available). Speed difference is very negligible in real world.

I ran non ECC DIMM in my Mac for quite a few year (at that time, high capacity ECC RAM are more expensive, and my computer is just for normal home use, doesn't really matter). It's completely stable. And 99% of modern computer are non ECC anyway, the chance of have single bit error cause by cosmic rays which eventually cause whole system crash should be very very low.

Just 30m AMSL shouldn't be a matter at all. I usually work at above 10000m, and I never see a single sudden crash that can point to cosmic radiation (the aircraft's computer of course are protected, but all other computer equipments are just normal PC, not even using ECC RAM). Even on the "high solar activity year (which occur every 11 years)", still nothing strange happen. So, I personally won't worry about this too much.

But if a single undetected bit error can cause catastrophic failure in your workflow. Then I will say you have no choice, but only use ECC RAM.

Love your detailed response. Thank you.

Curious: Is there a way of checking if there has ever been an ECC error during the lifetime of the cMP?
Checking Memory under System Information always show up as "OK" to me. Is this because the status is reset every time I reboot my machine or something like that? Or did they actually never fail me during the years of service?

 

[MVMNT]

Does anyone know who vends the same memory as DATARAM?

My cMP came with a good 32GB installed (4x8) but if I want to go triple channel on a dual CPU setup, and buy another 16GB, I could do with a good match. Not having much luck.

 

[h9826790]

Love your detailed response. Thank you.

Curious: Is there a way of checking if there has ever been an ECC error during the lifetime of the cMP?
Checking Memory under System Information always show up as "OK" to me. Is this because the status is reset every time I reboot my machine or something like that? Or did they actually never fail me during the years of service?

I am quite sure memory error that catches by ECC will be logged. However, I never really try to find out the answer. I expect I can see something in Console if ECC pickup something.

One of the main reason of having ECC is allow the user replace that “start to have problem” DIMM ASAP. If the error is not logged or report. Then how can a user know the DIMM is failing?

 

[LightBulbFun]

I would not be surprised if ECC errors would get logged to console

I can say for sure, that if you get ECC Errors, system profiler will report them under Memory

status OK, will change to "ECC errors"

I remember catching a Random ECC Error probably caused by cosmic rays on my Mac Pro that way once when randomly checking system profiler

 

[HunPro]

The first Rember test passed, on 58 gigs of ram. Not ideal, but then, who knows how that was allocated.

Since all these mishaps occured during boot, and after hours of memtesting all rams report OK at the system report, I'm confident that the modules are ok.

I will keep it up and started 10 passes for the night. The Mac Pro doubles as a space heater.

UPDATE: Rember did 1,5 passes without errors. Found some writeable DVDs stuffed away a decade ago, burned Memtest86 on it, it's running right now.

Booting to Memtest required a reboot, and it came back with all 64GB visible. A good sign.

UPDATE: memcache did 3 passes without any error. These sticks work just fine in the 2009=>2010 model.

 

[MRisberg]

One of the main reason of having ECC is allow the user replace that “start to have problem” DIMM ASAP. If the error is not logged or report. Then how can a user know the DIMM is failing?

I can say for sure, that if you get ECC Errors, system profiler will report them under Memory

status OK, will change to "ECC errors"

Makes sense. Just wondering if these kind of error messages get flushed when rebooting or resetting things like nvram. Can't see that I have an ECC error at the moment. Could this mean that I have not had one error during the thousands of hours I've used this machine? That would be nice.

 

[Kon_Kipa]

1 more week now & my 8 x 32GB Ram modules will arrive for the cDMP

 

[Macschrauber]

Update: I installed the sticks, I went with these: DDR3 16GB 1600 MHz REG ECC HP 672612-081 HMT42GR7MFR4C-PB

It was pretty wild.

1. Normal boot, OSX comes on, only 32 GB recognized in slot 1 & 2
2. PRAM reset
3. Normal boot, OSX comes on, only 32 GB recognized in slot 1 & 2
4. PRAM reset
5. Normal boot, folder with a question mark

6. Removed stick from slot 4
7. Normal boot, OSX comes on, only 48 GB recognized in slot 1, 2 & 3
8. Added stick from previous slot 4 to slot 3 (to test if the stick was bad)
9. Normal boot, OSX comes on, only 48 GB recognized in slot 1, 2 & 3

10. removed all sticks, installed all four
11. PRAM reset
12. Boot fails, slot 1 led red
13. Switched the sticks in slot 1 and 2 (to see if the stick in slot 1 was bad)
14. Normal boot, 64 GB of ram recognized. No red leds.

Running Rember now.

What Xeon do you use X or W ?

You can copy memtest to /usr/local and run it in single user mode
[doublepost=1551631468][/doublepost]

Makes sense. Just wondering if these kind of error messages get flushed when rebooting or resetting things like nvram. Can't see that I have an ECC error at the moment. Could this mean that I have not had one error during the thousands of hours I've used this machine? That would be nice.

I have seen ECC error reports in ASD test boot (Apple Service Diagnose) on Mac Pro 1.1 - red light off at this moment, also OS reports ok in System Profiler. Not sure where it got stored. After smc reset / power off it disappeared. Removed this stick of course.

 

[HunPro]

What Xeon do you use X or W ?

It's an X5670, I did write a bootable Memcache DVD and it passed multiple times without an error. It must have been a contact problem, or just quirk. I had memory modules behave like this in the old PC days.

It's working flawlessly now with 4x16GB, speed is the same with the 2x8GB modules it came with (dual channel 1333), as expected. 48GB would have been a trade of precious memory space for a dubious 50% memory bandwidth increase. This CPU has a 12 meg L3 cache and no integrated graphics to compete with, so it seemed unnecessary. Zero swap is waaay more important.

 

[Kon_Kipa]

8 x 32GB Ram has arrived for cDMP. now i just need to update the firmware to 141.0.0.0

 

[w1z]

8 x 32GB Ram has arrived for cDMP. now i just need to update the firmware to 141.0.0.0

If you do get these to work with the cMP, the memory sync speeds will drop to PC3-6400 which is 800MHz... There is no way around it.

 

[Kon_Kipa]

If you do get these to work with the cMP, the memory sync speeds will drop to PC3-6400 which is 800MHz... There is no way around it.

i've put in 4 x 32GB & successfully booted into 10.10.5 - speed @ 1066MHz however im thinking thats due to the fact that its a 4,1 --> 5,1 2 x 2.26GHz model.

I've jut received a dual tray with 2 x 2.4GHz Xeons thats going into my 2010 5,1

Ill see what happens on that machine before i start posting results

I've also got a pair of X5690's for the 5,1 mid 2010 model.

I'll install them as well & run the tests off of the mid 2010 5,1

 

[Kon_Kipa]

OK, just a quick update regarding the 32GB Ram modules...

1. I've updated the firmware on my 4,1 --> 5,1 to 140.0.0.0 & it went well & updated
2. I've cloned installed 10.13.6 on 1 partition of a 640GB HDD
3. I'm about to clone install 10.14.2 on the other partition of the same 640GB HDD
4. I've installed Windows X Pro x64 on a seperate/different 1TB HDD

Ram currently in the machine is 4 x 32GB + 2 x 16GB = 160GB Total

When i was running 10.10.5 ram was showing as 160GB in main "about this mac" window however when i selected the "memory" tab only 2 x 16GB ram modules showing & the rest of the slots were all empty - DDR3 Speed detected is 1066MHz (this might be due to me running 2 x 2.26GHz)

When i run 10.13.6 ram is showing as 32GB in the main "about this mac" window however when i select the "memory" tab it still only shows 2 x 16GB ram modules as occupying the slots - - DDR3 Speed still detected at 1066MHz

Windows 10 Pro x64, ram is showing at 160GB & working perfectly... (ill see if i can work out where i can get a DDR Ram speed viewer of some sorts - maybe CPU'z or something)

Now i have experienced a few shut downs/restarts - but i am not sure if it's the nVidia GT 660 Graphics card (non mac firmware rom) or if i have a faulty module - i'll report back on this a little later on - for now windows X is now updating with security patches etc - so i'm calling it a night for now & will post more info as it develops ;-)

 

[LightBulbFun]

very interesting stuff

in 10.13.6, what does Activity monitor or the hostinfo command in terminal say about memory?

that will tell us exactly what the OS Can see, because the About this Mac stuff is only technically cosmetic

CPU Z will indeed tell you info about the RAM in windows

(BTW in your earlier posts you said you ordered 8 of them, what happened to the other 4?)

 

[Kon_Kipa]

very interesting stuff

in 10.13.6, what does Activity monitor or the hostinfo command in terminal say about memory?

that will tell us exactly what the OS Can see, because the About this Mac stuff is only technically cosmetic

CPU Z will indeed tell you info about the RAM in windows

(BTW in your earlier posts you said you ordered 8 of them, what happened to the other 4?)

I've got all 8 here mate, yeh just been hesitant putting all of them in, in one go (not until i build up some drama first LOL)...

Yeh i'm quite stoked about the results so far & I'm really pleased...

I've been doing all the tests on my 4,1 --> 5,1 but like i mentioned earlier on i have a proper 5,1 which i've just purchased a dual tray for (yes the tray has arrived & i have a couple of X5690's as well that i want to put right into it first)

So for now its just a lazy Sunday curiosity type testing going on - once i'm happy/satisfied with results etc, i will start moving everything into the 5,1 Mid 2010... I just need some time to remove the heatsinks etc & install the CPU's

I just bought a Radeon HD 7950 from eBay as well (Proper Mac Edition), so i can't wait to get that installed as well.
[doublepost=1552212853][/doublepost]Tomorrow is a public holiday, once i finish off up from the gym, ill post some more results etc

 

[VaZ]

Looking forward to seeing if all 8 slots will see your 32gb@1333MHz !

I ended up finding a sweet deal locally on 4 Micron Technology 16gb ecc sticks. Confirmed working great at expected speeds! Specs below:

Micron MT36KSF2G72PZ-1G4E1HF 16GB 2RX4 PC3L-10600R HP PN 647653-081

 

[Le_Cactus]

I've just received one stick of Samsung M393B4G70BMO-YH9 32GB 4Rx4 PC3L-10600R, it doesn't boot with a single X5650 (High Sierra 10.13.6, firmware 140.0.0.0.0).

 

[tommy chen]

4Rx4 are wrong

must be a 2Rx4 to work in a cMP

 

[h9826790]

4Rx4 are wrong

must be a 2Rx4 to work in a cMP

AFAIK, so far, all reported "able to work" 32GB DIMM are 4Rx4, no one ever report anything about 2Rx4 32GB DIMM yet.

And 2Rx4 is not a mandatory requirement, but best compatibility.

 

[Le_Cactus]

What could be incompatible with the single processor?