How do I tell if my Memory is ECC or Non-ECC?

Question

I need to replace the memory in my system and I'm not quite sure how to determine if the memory is ECC or Non-ECC. How do I determine this?

Is there anyting in the handbook of your PC, Usually it is specified there. Otherwise download and install [Speccy](http://www.piriform.com/speccy). This program tells you exactly which hardware you have in your PC, then you can check it. — Wernfried Domscheit, Mar 24 '15 at 21:23
Look up the motherboard manual, or the system manual for your system, and see what is required for your hardware? Use the memory suggestion tools on the various RAM vendors web sites? — Zoredache, Mar 24 '15 at 21:24
Related: [How to tell whether RAM ECC is working?](http://unix.stackexchange.com/q/139319/2465) on [unix.se]. — user, Mar 25 '15 at 13:18
As a general answer to the question - if you're doing home computing (i.e. not building enterprise servers) you are (and should be) using non-ECC. Also helpful: http://forum.crucial.com/t5/Crucial-memory-for-PC-systems/ECC-vs-non-ECC-What-do-I-Have-and-Can-I-Mix/ta-p/150486 — WannabeCoder, Mar 25 '15 at 13:22
@MichaelKjörling Data integrity at that level is not critical in home-use applications. Parent applications (or the OS) can handle that easily, and ECC memory comes with a small performance overhead. Additionally, virtually all consumer CPUs do not support ECC memory (thus the "should"). — WannabeCoder, Mar 25 '15 at 15:36
@WannabeCoder All AMD CPUs, some desktop Intel CPUs, and all server Intel CPUs, support ECC RAM. The performance impact of ECC RAM is negligible at most. Come back after you have experienced a critical failure due to RAM breakdown, or for that matter after a number of impossible-to-explain software crashes. (It's the same thing as the saying that there are people who have suffered a hard disk crash, and the people who aren't backing up their data; the sets are largely complementary.) — user, Mar 25 '15 at 15:52
@MichaelKjörling Neat fact about AMDs I didn't know. I like your comparison about using hard drive backups - I would very much apply that to this situation and remind you that the vast, vast majority of home users do not consider (serious) data backups necessary for their use-case. ECC memory is a step beyond even that; so I have a hard time seeing it as a requisite or even a valid recommendation for the average home user. The debate gets very technical on its effective worth. Given this user does not understand ECC at all - I would suspect that debate is out of scope; thus: "should" — WannabeCoder, Mar 25 '15 at 17:32
@WannabeCoder Let us [continue this discussion in chat](http://chat.stackexchange.com/rooms/22279/discussion-between-michael-kjorling-and-wannabecoder). — user, Mar 25 '15 at 17:58
Years ago, [Logic Parity](https://en.wikipedia.org/wiki/RAM_parity#History) (or "Fake Parity") RAM was available that allowed the use of non-parity RAM in a motherboard that requires ECC. I don't think fake partiy is common now in legitimate RAM, but off-brand or counterfeit RAM may still use fake parity to cut costs. If your ram did have this fake-parity, your operating system wouldn't know, it would think it's real parity. — Johnny, Mar 25 '15 at 21:01
ECC is used as a marketing tool: it is technically very cheap to provide, but it is something which some people want to have at any cost (kind of like brakes on a car), while the vast majority of users is willing to live without, it is used to segregate the market and impose higher margins on those people willing to pay for it. — Stefan, Mar 26 '15 at 01:13

score 82 · Answer 1 · edited Mar 08 '23 at 13:45

82

For Windows 7 you can run the following command in command prompt:

wmic MemoryChip get DataWidth,TotalWidth

If the TotalWidth value is larger than the DataWidth value you have ECC memory.

Example output:

//ECC Memory
DataWidth  TotalWidth
64         72

//Non-ECC Memory
DataWidth  TotalWidth
64         64

A better way to determine is via the following command:

wmic MemPhysical get MemoryErrorCorrection

This will return a code based on the type of memory installed:

Value	Meaning
0 (0x0)	Reserved
1 (0x1)	Other
2 (0x2)	Unknown
3 (0x3)	None
4 (0x4)	Parity
5 (0x5)	Single-bit ECC
6 (0x6)	Multi-bit ECC
7 (0x7)	CRC

edited Mar 08 '23 at 13:45

Cristian Ciupitu

5,513
2
37
47

answered Mar 24 '15 at 21:19

James Mertz

26,224
41
111
163

@DanNeely My system (which I know has 8 strips of 4GB ECC) gives 2 values from that command: 6 and 3 on a second line of output. The other command shows 8 lines with 64 and 72 (as expected) and 1 extra line showing 2 and 2. I have never seen that before. Any idea what that means ? For the record: It is a HP XW8600 workstation (Intel 5400/6311 server motherboard.) – Tonny Mar 25 '15 at 19:47
@Tonny which command are you using? – James Mertz Mar 25 '15 at 20:04
What does code 3 (none) mean? I have no memory installed? – David says Reinstate Monica Mar 25 '15 at 20:51
@DavidGrinberg it means your memory doesn't have ECC. Outside of servers and some high end work stations very few computers use it. – Dan Is Fiddling By Firelight Mar 25 '15 at 21:15
@KronoS wmic memphysical get memoryerrorcorrection gives 6 and 3. (Windows 7 x64) then wmic memorychip get datawidth, totalwidth shows the 9 lines of output. 8x "64 72" and once "2 2". – Tonny Mar 26 '15 at 10:48
I can also confirm that these commands are not a reliable indicator. I get the same values as KronoS for Corsair Vengeance DDR3 (CMZ8GX3M2A1600C9B). As with his experience, HWInfo64 correctly reports it as non-ECC. – nextgentech Oct 24 '16 at 18:44

score 54 · Answer 2 · edited Mar 08 '23 at 13:11

54

For FreeBSD (and probably most Unix like platforms):

dmidecode -t 17

Example output:

# dmidecode 2.12
SMBIOS 2.5 present.

Handle 0x1100, DMI type 17, 28 bytes
Memory Device
        Array Handle: 0x1000
        Error Information Handle: Not Provided
        Total Width: 72 bits
        Data Width: 64 bits
        Size: 2048 MB
        Form Factor: DIMM
        Set: 1
        Locator: DIMM1
        Bank Locator: Not Specified
        Type: DDR2
        Type Detail: Synchronous
        Speed: 667 MHz
        Manufacturer: AD00000000000000
        Serial Number: 00002062
        Asset Tag: 010839
        Part Number: HYMP125P72CP8-Y5
        Rank: 2

The Total Width: 72 bits is the part you are looking for.

More information in detecting this in Linux can be found on our sister site Unix & Linux Stack Exchange.

More information on how error correcting code works can be found in this simple post where I showed how you could use extra bits to detect and correct errors. This is why we have DIMMs which are 64 bits wide (8 bytes of data wide) or 72 bits wide (64 data plus extra bits to store redundant information).

19.07.2022 - Minor update now that ECC DDR5 is out. DDR5 is not 64 bit normal or 64+8 for ECC. It instead has two 32 bit channels, and more bits are needed for ECC. So expect 80 bits for ECC on DDR5.

edited Mar 08 '23 at 13:11

Cristian Ciupitu

5,513
2
37
47

answered Mar 24 '15 at 21:30

Hennes

64,768
7
111
168

4

dmidecode is standard on Linux, too. I think error correction will happen by default with ECC RAM installed, but if your kernel is monitoring the memory controller to track the memory error rate, that would be another confirmation that you have ECC RAM. (check the kernel log.) – Peter Cordes Mar 25 '15 at 08:49
How do you check if the memory controller is tracking that? mcelog? – Hennes Mar 25 '15 at 10:52
I don't have access to any ECC-equipped servers anymore, but I'd look for a kernel log message about it. Maybe there isn't one. There's a link to something about it on http://buttersideup.com/ (best domain name ever for a project, IMO). – Peter Cordes Mar 25 '15 at 11:14
Works on Fedora as well. – l0b0 Mar 25 '15 at 12:44
@Hennes See http://unix.stackexchange.com/a/139526/2465 for example log output with ECC and non-ECC RAM on Linux. – user Mar 25 '15 at 13:16
Oddly, checking one of my known-ECC boxes (and I'm sure it works, as I've gotten single-bit notices before) data width is 72, total is 64—clearly, they're labeled backwards. – derobert Mar 25 '15 at 20:39
1

I think you should explain in your answer what to conclude about ECC if `Total Width` is `72 bits` or `64 bits`. – A.L Mar 26 '15 at 15:02
2

"The Total Width: 72 bits is the part you are looking for." And what do I do after I've found it? – endolith Oct 15 '16 at 21:53
dmidecode is not available on esxi hosts since version 5 I think. I used `smbiosDump` and page through to memory section – DoTheEvo May 04 '22 at 14:08

score 25 · Answer 3 · edited Apr 02 '15 at 06:53

25

If you look at the physical memory module, ECC will usually have 9 (sometimes more) chips. Non-ECC will have only 8 (or rarely, 8x2=16).

ECC vs non-ECC (Image courtesy of Puget Systems)

edited Apr 02 '15 at 06:53

Mokubai

89,133
25
207
233

answered Mar 25 '15 at 20:24

BlueRaja - Danny Pflughoeft

7,667
6
43
54

7

@Hennes: You can have 10, but 9 is far more common. The memory is typically split into chunks of 72-bits *(8 bits read at once from each of 9 data-chips)*, with 64-bits of data + 8-bits of error-correction in each chunk. See [here](http://en.wikipedia.org/wiki/Hamming_code#Hamming_codes_with_additional_parity_.28SECDED.29) for more technical info. – BlueRaja - Danny Pflughoeft Mar 25 '15 at 21:38
2

As a more general rule, if the number of memory chips are divisible by 3 then the module is ECC. – James Mertz Mar 26 '15 at 15:00
Count them on both sides. – Leo Mar 29 '20 at 04:46

Jonas Stein · Answer 4 · 2023-04-09T20:42:45.390

inxi can do that:

# On this desktop PC non ECC-RAM modules are used, so it outputs `EC: None`:

$ sudo inxi -m -xxx
Memory:    Array-1 capacity: 32 GB devices: 4 EC: None
           Device-1: ChannelA-DIMM0 size: 4 GB speed: 1333 MHz type: DDR3 (Synchronous)
           bus width: 64 bits manufacturer: Kingston part: KHX1600C9D3/4GX serial: B7ED5A53

# This server has 16 GB ECC RAM:
$ sudo inxi -m -xxx
Memory:
  RAM: total: 15.49 GiB used: 592.6 MiB (3.7%)
  Array-1: capacity: 32 GiB slots: 2 EC: Multi-bit ECC max-module-size: 16 GiB note: est.
  Device-1: PROC 1 DIMM 1 type: no module installed
  Device-2: PROC 1 DIMM 2 type: DDR4 detail: synchronous unbuffered (unregistered) size: 16 GiB
    speed: spec: 2666 MT/s actual: 2667 MT/s volts: 1.2 width (bits): data: 64 total: 72
    manufacturer: HPE part-no: 879527-091 serial: N/A

Just the `EC: None`. The `4` is part of the previous field (`devices: 4`). — , Mar 25 '15 at 21:25

MrDaniel · Answer 5 · 2015-03-27T01:26:02.043

10

On a Mac you can look in the System Information application to determine ECC status of the Computer and each RAM module.

/Applications/Utilities/System Information

Select Memory on the right pane, under the hardware section.

Then with the "Memory Slots" listing selected. The window below should show an ECC status.

See picture below:

System Information App window

Also as an FYI each RAM slot will have a Status field if there is a problem detected in a RAM module the status will be a value other than "OK"

edited Mar 27 '15 at 01:26

answered Mar 25 '15 at 23:44

MrDaniel

1,352
4
14
28

Does "Disabled" mean the ram is non ECC ? it's not quite clear from your answer – mounaim Nov 14 '16 at 16:27
That's correct a status of "Enabled" would means that he RAM is ECC. – MrDaniel Nov 24 '16 at 19:28

htm11h · Answer 6 · 2015-03-26T16:34:41.057

0

You might also try a free app like Belarc. Gives a bunch more info about your system also...

I have used the software a few times. I have no affiliation with this software, but I do know that it is used by some universities.

I can not provided a screen shot of the results because it contains confidential data, but the web site should provide some examples. Its pretty straight forward (and fast), download the file, run it and the results are presented.

edited Mar 26 '15 at 16:34

answered Mar 25 '15 at 20:09

htm11h

107
6

Please give more detail as to how this work. For reference see [this meta post](http://meta.superuser.com/a/5330/47225) – James Mertz Mar 26 '15 at 15:05

How do I tell if my Memory is ECC or Non-ECC?

6 Answers6