Announcement

**David (PassMark)** · Aug-18-2023, 09:07 PM

For the ECC errors on cold boot:
There is likely a bug in your EDAC/BIOS. Your ECC RAM is OK, but was not initialized properly by the BIOS on boot. In order to initialize ECC, memory has to be written before it can be used. Usually this is done by BIOS, but with some motherboards this step is skipped if "Quick Boot" is enabled. Possible Solution: If your system allows for it, try disabling Quick Boot in the BIOS, some error messages should disappear. The boot process may taker 30-60 seconds longer, but the EDAC error messages disappear due to the RAM check by the BIOS when booting.
BIOS already known to experience this issue: KONTRON AMI BIOS

So test 0 on the birl screen corresponds to test: 1 in the report.

No. I don't think this is the case. There is a Test 0 result in results you pasted above.

**UrRa_346798ze** · Aug-19-2023, 10:05 AM

Hi,

thank you for your reply. I agree that the RAM is ok.

And it seems that the BIOS has a bug.
Unfortunately the Kontron AMI Bios has no option to disable quick boot. Therefore I can not check whether that helps.

In order to narrow it down i made the following tests:

1. Running memtest86 after cold boot

Shows correctable eecc errors in test 0 and test 1 (not in test 2-13 though)

2. Running memtest86 after initial run (full 4 passes) and reset by reset button

shows no errors.

3. Running memtest86 after cold start and reset during post showing A2 on boot screen
Same as 1 above

4. Running memtest86 after cold start and reset during start page of memtest86
Same as 1 above

5. Running memtest86 after cold start and reset during test 0 (and showing first ecc error)
Same as 1 above

6. Running memtest86 after cold start and reset during test 1 (and showing first ecc error in test 1)
Same as 1 above

7.6. Running memtest86 after cold start and reset during test 2
Same as 2 above

Conclusion: It seems that test 2 somehow initializes the ECC RAM properly in a way that a rerun of memtest after reset shows no ecc errors (the board somehow seems to memorize that ECC has been initialized until power off)

However there is something which might contradict the thesis:
If memtest86 runs several passes it shows in each pass (in test 0 and test 1) the identical ecc errors. That should not happen if the first run of test 2 initilializes the RAM properly. Then test 0 and test 1 of pass 2-4 should not show the errors again. Unless some function after test 2 de-initializes the ECC RAM again to its original state, in a way that test 0 and 1 find the ECC RAm in the same state as after cold boot.

Does this make sense?

Best regards

(sorry for the confusion regarding my report of non existing inconsitencies between screen and report in the original post. Double chedcked it and you are right. I can not reproduce what led me to the wrong assumption.)

**UrRa_346798ze** · Aug-19-2023, 07:25 PM

Further Tests:

8. Cold boot configure memtest86 (active only test 2 and 3 all other tests deactivated)

This time memtest finds 7 ECC RAM erors in test 2 and no in test 3 (in every pass).

9. Cold boot configure memtest86 (active only test 3 and 4 all other tests deactivated)

This time memtest finds 23 ECC RAM erors in test 3 and no in test 4 (in every pass).

Conclusion:

It seems that the first test (independent from which test) initializes the ECC RAM. And in the next test no error no error is found [one exeception though in test 0 1 error and test 1 6 errors]. Now I understand your sentence ("In order to initialize ECC, memory has to be written before it can be used.") better.

From that I would conclude, that even if kontron does not update its bios/implements option to deactivate quick boot, I do no t have to worry. Its seems that ECC RAM is initilized at the first time it is written to. That can be done at boot itme or later. It might be that at the firtst time it is written to an ecc Error is happening. At the second time no error will be happen again.

Therefore it might be that the operating system loggs few corrected ECC Errors after cold boot - that`s it. Right?

Best regards

**David (PassMark)** · Aug-20-2023, 11:06 PM

Therefore it might be that the operating system loggs few corrected ECC Errors after cold boot - that`s it. Right?

That is our understanding of the issue yes. But it depends on what comes first in the boot process. Writing to the RAM, or reading errors from the memory controller.

Ultimately it is a BIOS bug however and the vendor (Kontron / Fujitsu) should fix it.

**UrRa_346798ze** · Sep-08-2023, 05:47 AM

Originally posted by David (PassMark) View Post

That is our understanding of the issue yes. But it depends on what comes first in the boot process. Writing to the RAM, or reading errors from the memory controller.

Ultimately it is a BIOS bug however and the vendor (Kontron / Fujitsu) should fix it.

Hi,

thank you for your assessment.

I repeated the test with an I3-8100 (as user datacollector wrote over here https://forums.unraid.net/topic/1362...fehlung-d3644/ that he had no issues with this cpu) but with the same results. So the BIOS bug seems no to be dependant on the CPU.

I pointed the bug out to Kontron. But they seemed to be not interested (Even though the board has extended lifetime) and but have referred me to the retailer, as if he would be able to do a BIOS update.

Best regards

**David (PassMark)** · Sep-08-2023, 05:52 AM

I pointed the bug out to Kontron. But they seemed to be not interested

We are not surprised.
Is nearly impossible to get vendors to fix their motherboard bugs.

**vlorentz** · 2024-12-20T23:56:37

Hello,

I am just wondering about the following: In the specs of the Fujitsu/Kontron D3644-B board, we can find the following: "Note: 128 GByte memory configuration (4x 32 GByte modules) not supported". This does not give any information if 1x or 2x 32GB modules are supported or not. However, in the User Manual, it is written: "Module size 1, 2, 4, 8 or 16 GB per module".

I am therefore wondering if the problem that you encountered is also appearing if you are using modules with 16GB (or less)? If yes, is there any way to find a workaround? The background of my question is the following statement, which indicates that 4x 32GB is feasible on the Kontron D3644-B board, but it seems limited to 2400 MT/s: https://www.phs-memory.com/computer-...-sp420331.html

My questions are:
(1) Is the problem you have found out specific to the 32GB modules in general, or specific to the 32GB of Kingston only?
(2) Have you the possibility to test with smaller modules?

Best regards

**David (PassMark)** · 2024-12-21T00:10:24

Kontron should know the answers to all these questions.

We don't have access to this motherboard (and even if we, it would feel a bit like we are doing free technical support for Kontron, which isn't right, as they should support their own products).

Previous customer who reported this posted a year ago. They might not be around anymore and might not have a range of different memory sticks to test with.

**vlorentz** · 2024-12-21T01:26:13

I totally agree, David. As you said yourself, chances are low that Kontron will update anything on this old board. I was hoping for your critical view on this topic, especially, if it could be theoretically possible, that the BIOS is not initializing properly the RAM of 32GB modules. To me, it seems at least to be a possibility.

**David (PassMark)** · 2024-12-21T01:36:28

If you have the motherboard you are probably in the best position to try it out.

Report Date	2023-08-18 09:09:39
Generated by	MemTest86 V10.6 Free (64-bit) Visit MemTest86.com to Upgrade to Pro
Result	INCOMPLETE PASS

EFI Specifications	2.70
System
Manufacturer	FUJITSU
Product Name
Version
Serial Number
BIOS
Vendor	FUJITSU // American Megatrends Inc.
Version	V5.0.0.13 R1.28.0 for D3644-B1x
Release Date	12/01/2022
Baseboard
Manufacturer	FUJITSU
Product Name	D3644-B1

CPU Type	Intel Core i3-9100 @ 3.60GHz
CPU Clock	3692 MHz [Turbo: 4102.6 MHz]
# Logical Processors	4
L1 Cache	4 x 64K (212227 MB/s)
L2 Cache	4 x 256K (90814 MB/s)
L3 Cache	6144K (53806 MB/s)
Memory	32516M (14151 MB/s)
RAM Configuration	DDR4 ECC 2400MT/s / 1.200V
Number of RAM SPDs detected	1
SPD #0	32GB DDR4 2Rx8 ECC PC4-25600
Vendor Part Info	Kingston / 9965745-039.A00G /
JEDEC Profile	3200MT/s 22-22-22-52 1.2V
Number of RAM slots	4
Number of RAM modules	1
DIMM A1	32GB DDR4 2Rx8 ECC PC4-25600
Vendor Part Info	Kingston / 9965745-039.A00G /
SMBIOS Profile	3200MT/s 1.2V
DIMM A2	Empty slot
DIMM B1	Empty slot
DIMM B2	Empty slot

Test Start Time	2023-08-18 09:07:15
Elapsed Time	0:02:17
Memory Range Tested	0x0 - 86C800000 (34504MB)
CPU Selection Mode	Parallel (All CPUs)
CPU Temperature Min/Max/Ave	49C/55C/52C
ECC Polling	Enabled
# Tests Completed	4/48 (8%)
# Tests Passed	4/4 (100%)

Test	# Tests Passed	Errors
Test 0 [Address test, walking ones, 1 CPU]	1/1 (100%)	0
Test 1 [Address test, own address, 1 CPU]	1/1 (100%)	0
Test 2 [Address test, own address]	1/1 (100%)	0
Test 3 [Moving inversions, ones & zeroes]	1/1 (100%)	0
Test 4 [Moving inversions, 8-bit pattern]	0/0 (0%)	0
Test 5 [Moving inversions, random pattern]	0/0 (0%)	0
Test 6 [Block move, 64-byte blocks]	0/0 (0%)	0
Test 7 [Moving inversions, 32-bit pattern]	0/0 (0%)	0
Test 8 [Random number sequence]	0/0 (0%)	0
Test 9 [Modulo 20, ones & zeros]	0/0 (0%)	0
Test 10 [Bit fade test, 2 patterns, 1 CPU]	0/0 (0%)	0
Test 13 [Hammer test]	0/0 (0%)	0

Announcement

Test (test 0 and test 1) ECC error only after cold boot (not after repeat/reset)

Test (test 0 and test 1) ECC error only after cold boot (not after repeat/reset)

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment