Will the new AMD boards support ECC RAM?

My understanding is that the Ryzen mobile chips, much like the desktop chips, have ECC support in their controllers. However, mainboards aren’t required to route the additional pins and firmware may configure RAM in non-ECC mode. https://www.amd.com/en/product/13041 documents this for the 7840HS. “ECC Support: Yes (FP7r2 only; Requires platform support)”. FP7r2 is DDR5 as opposed to LPDDR5, which is only usable in a soldered RAM configuration. Since the Framework laptop uses SO-DIMMs, the chips will support this.

Will the new AMD boards support ECC RAM, either for the 13 or the 16?

10 Likes

I was wondering the same thing. The possibility to use ECC RAM would be a Unique Selling Point. Would instantly buy three of those boards if this is true.

8 Likes

Really like the presentation for the 16.

I want ECC for main system memory and ECC for the 16” GPU. !!!

3 Likes

See ECC support? - #5 by Fraoch

I am also very interested in knowing the answer to this. I posted a separate thread asking about ECC RAM for the Framework Laptop 16 ( ECC support? ) but that is currently under information embargo.

That said, I’m curious to know about what the verdict is for the Framework Laptop 13 as well.

According to the full specification of both Ryzen 7640U and Ryzen 7840U, ECC is supported by the U-series APUs found on the new motherboards, too:

https://www.amd.com/en/product/13186
https://www.amd.com/en/product/13196

So, it’s down to Framework themselves whether ECC will work.

10 Likes

DDR5 modules have on-die ECC support. Full ECC support is not currently confirmed as compatible with non-PRO Ryzen 7040 Series mobile processors.

Edit: nrp’s comment: ECC support? - #86 by nrp

1 Like

That’s a pretty uninspiring statement from them. That is always what AMD says, and they always support ECC on the right board with it enabled in firmware. It’s the same memory controller, after all. It’s just not guaranteed by the spec so they could in theory change it in a later silicon revision, though they won’t because designing a new memory controller isn’t worth it.

See e.g. https://www.hwcooling.net/en/amd-launches-ryzen-7040-4nm-apu-with-zen-4-cores-and-rdna-3-en/ for reporting that confirms this.

If Framework claims they didn’t get it working in a small test, then why does it say in the official AMD specification that ECC memory is supported by non-Pro models?

TBF, even the AM5 socket, which also advertised ECC support, took a while for the BIOS support to enable it. It’s dependent on the CPU, wiring, BIOS, AGESA, and finally the OS. It wasn’t until recently that AMD added ECC support and until version 6.5 of the Linux Kernel (released a few weeks ago). There are entire threads with crowdsourced posts following the slow confirmation of which motherboards and BIOS combinations support ECC.

3 Likes

Relevant: [RESPONDED] AMD Batch 1 Guild - #196 by TheTwistgibber

Edit: Personally, I’ll experiment with an ECC RAM kit because I can unless someone else does that before me.

4 Likes

The spec sites for 7640U and 7640U PRO do not differ at all with regard to ECC support. Both say: “Yes (FP7r2 only; Requires platform support)”.

I’m in the same boat as @resample and would risk buying ECC SO-DIMMS just to experiment with this since I would love ECC in my AMD Batch 1 FW 13. The only reason for me to not even try would be a definitive statement with hardware based reasons (other than “it’s the non-pro processor”) for ECC not being able to work, e.g. missing traces on the mainboard.

My desktop is an AM4 board with 3700X bought very close to release and the ECC support needed some UEFI/AGESA updates and patches to the linux kernel to actually work including error reporting. That took several months to come together but since then works perfectly. So if the hardware is at least theoretically capable I will try this. Probably with two of Kingston’s KSM56T46BS8KM-16HA. 5600MT JEDEC and ECC.

4 Likes

According to the last blog post the first AMD Frameworks will soon be on their way. Time to buy some memory :slight_smile:
Has anyone found some 5600MT ECC DDR5 other than KSM56T46BS8KM-16HA?

2 Likes

Short update: The memory itself works fine without parity. dmidecode -t memory shows 72 bit total width, that would be 64 bit data plus 8 bit parity. This alone does not mean anything in regards to potential for actually working ECC. It’s however at least better than dmidecode showing only the 64 data bits for total width. All other sources (edac-util, ras-mc-ctl, kernel boot log) I tried do not show anything that hints to working ECC.

6 Likes

Does this mean that the motherboard wiring has added ECC traces ?

No. This is read out from ID chips on the RAM.

2 Likes

No changes with BIOS 3.03 but I did not really expect any.

The modules delivered by framework are reported as 64 bit wide, and so the physical memory array description concludes: “Error Correction Type: None”. Does it say something else with your 72 bit modules?

# dmidecode 3.5
Getting SMBIOS data from sysfs.
SMBIOS 3.5.0 present.

Handle 0x0011, DMI type 16, 23 bytes
Physical Memory Array
        Location: System Board Or Motherboard
        Use: System Memory
        Error Correction Type: None
        Maximum Capacity: 64 GB
        Error Information Handle: 0x0014
        Number Of Devices: 2

Handle 0x0012, DMI type 17, 92 bytes
Memory Device
        Array Handle: 0x0011
        Error Information Handle: 0x0015
        Total Width: 64 bits
        Data Width: 64 bits
        Size: 16 GB
        Form Factor: SODIMM
        Set: None
        Locator: DIMM 0
        Bank Locator: P0 CHANNEL A
        Type: DDR5
        Type Detail: Synchronous Unbuffered (Unregistered)
        Speed: 5600 MT/s
        Manufacturer: A-DATA Technology
        Serial Number: 00013173
        Asset Tag: Not Specified
        Part Number: AD5S560016G-B       
        Rank: 1
        Configured Memory Speed: 5600 MT/s
        Minimum Voltage: 1.1 V
        Maximum Voltage: 1.1 V
        Configured Voltage: 1.1 V
        Memory Technology: DRAM
        Memory Operating Mode Capability: Volatile memory
        Firmware Version: Unknown
        Module Manufacturer ID: Bank 5, Hex 0xCB
        Module Product ID: Unknown
        Memory Subsystem Controller Manufacturer ID: Unknown
        Memory Subsystem Controller Product ID: Unknown
        Non-Volatile Size: None
        Volatile Size: 16 GB
        Cache Size: None
        Logical Size: None

Handle 0x0013, DMI type 17, 92 bytes
Memory Device
        Array Handle: 0x0011
        Error Information Handle: 0x0016
        Total Width: 64 bits
        Data Width: 64 bits
        Size: 16 GB
        Form Factor: SODIMM
        Set: None
        Locator: DIMM 0
        Bank Locator: P0 CHANNEL B
        Type: DDR5
        Type Detail: Synchronous Unbuffered (Unregistered)
        Speed: 5600 MT/s
        Manufacturer: A-DATA Technology
        Serial Number: 00236745
        Asset Tag: Not Specified
        Part Number: AD5S560016G-B       
        Rank: 1
        Configured Memory Speed: 5600 MT/s
        Minimum Voltage: 1.1 V
        Maximum Voltage: 1.1 V
        Configured Voltage: 1.1 V
        Memory Technology: DRAM
        Memory Operating Mode Capability: Volatile memory
        Firmware Version: Unknown
        Module Manufacturer ID: Bank 5, Hex 0xCB
        Module Product ID: Unknown
        Memory Subsystem Controller Manufacturer ID: Unknown
        Memory Subsystem Controller Product ID: Unknown
        Non-Volatile Size: None
        Volatile Size: 16 GB
        Cache Size: None
        Logical Size: None

It says Error Correction Type: None for me too. That is the same as my Zen 2 desktop if I disable ECC in the BIOS there. With ECC enabled my desktop shows Error Correction Type: Multi-bit ECC.

Sounds like you’re trying to test ECC in the new AMD frame.work? I heard a mere firmware update could be the difference between the motherboard supporting it or not… Hope might not be lost afterall! Unsure if that includes the APU, but it would be sweet if the integrated graphics could use ECC too!