Few errors but always on same CPU

[michte]

Hi,

Due to a few weird crashing on a virtual machine this week, I decided to run a quick Memtest to check if everything was fine. It did pick up a few errors, but what seems weird, is that they all came from tests running on CPU core #22 (I'm using a Ryzen 3900X, so 24 active cores available).

I ran another test the following night, same thing : just one error, but one the same CPU core. What are the odds that 4 errors would pop up on the same CPU 2 times in a row ? Would this rather indicate that the CPU is deficient ?

Thanks for your input.

Logs of the first run :

System Information

EFI Specifications	2.70
System
Manufacturer	Gigabyte Technology Co., Ltd.
Product Name	X570 AORUS ELITE
Version	-CF
Serial Number	Default string
BIOS
Vendor	American Megatrends Inc.
Version	F12f
Release Date	03/06/2020
Baseboard
Manufacturer	Gigabyte Technology Co., Ltd.
Product Name	X570 AORUS ELITE
Version	x.x
Serial Number	Default string
CPU Type	AMD Ryzen 9 3900X 12-Core
CPU Clock	3793 MHz [Turbo: 3793.3 MHz]
# Logical Processors	24
L1 Cache	24 x 64K (210902 MB/s)
L2 Cache	24 x 512K (78362 MB/s)
L3 Cache	1 x 65536K (19016 MB/s)
Memory	32733M (18115 MB/s)
DIMM Slot #0	16GB DDR4 XMP PC4-28800
	G Skill Intl / F4-3600C16-16GVKC
	16-19-19-39 / 3602 MHz / 1.350V
DIMM Slot #1	16GB DDR4 XMP PC4-28800
	G Skill Intl / F4-3600C16-16GVKC
	16-19-19-39 / 3602 MHz / 1.350V

Result summary

Test Start Time	2020-05-13 00:42:20
Elapsed Time	8:17:39
Memory Range Tested	0x0 - 81F300000 (33267MB)
CPU Selection Mode	Parallel (All CPUs)
ECC Polling	Enabled
# Tests Passed	46/48 (95%)

Lowest Error Address	0x6F5471D50 (28500MB)
Highest Error Address	0x6F7471E70 (28532MB)
Bits in Error Mask	0000000000004000
Bits in Error	1
Max Contiguous Errors	1

Test	# Tests Passed	Errors
Test 0 [Address test, walking ones, 1 CPU]	4/4 (100%)	0
Test 1 [Address test, own address, 1 CPU]	4/4 (100%)	0
Test 2 [Address test, own address]	4/4 (100%)	0
Test 3 [Moving inversions, ones & zeroes]	4/4 (100%)	0
Test 4 [Moving inversions, 8-bit pattern]	4/4 (100%)	0
Test 5 [Moving inversions, random pattern]	3/4 (75%)	1
Test 6 [Block move, 64-byte blocks]	3/4 (75%)	2
Test 7 [Moving inversions, 32-bit pattern]	4/4 (100%)	0
Test 8 [Random number sequence]	4/4 (100%)	0
Test 9 [Modulo 20, ones & zeros]	4/4 (100%)	0
Test 10 [Bit fade test, 2 patterns, 1 CPU]	4/4 (100%)	0
Test 13 [Hammer test]	4/4 (100%)	0

Last 10 Errors

2020-05-13 07:01:05 - [Data Error] Test: 6, CPU: 22, Address: 6F7471E70, Expected: 00001000, Actual: 00005000

2020-05-13 07:01:05 - [Data Error] Test: 6, CPU: 22, Address: 6F5471E90, Expected: 00001000, Actual: 00005000

2020-05-13 04:46:24 - [Data Error] Test: 5, CPU: 22, Address: 6F5471D50, Expected: ED112FAD, Actual: ED116FAD

Logs of the second run :

System Information

EFI Specifications	2.70
System
Manufacturer	Gigabyte Technology Co., Ltd.
Product Name	X570 AORUS ELITE
Version	-CF
Serial Number	Default string
BIOS
Vendor	American Megatrends Inc.
Version	F12f
Release Date	03/06/2020
Baseboard
Manufacturer	Gigabyte Technology Co., Ltd.
Product Name	X570 AORUS ELITE
Version	x.x
Serial Number	Default string
CPU Type	AMD Ryzen 9 3900X 12-Core
CPU Clock	3793 MHz [Turbo: 3793.4 MHz]
# Logical Processors	24
L1 Cache	24 x 64K (205025 MB/s)
L2 Cache	24 x 512K (79142 MB/s)
L3 Cache	1 x 65536K (19265 MB/s)
Memory	32733M (18376 MB/s)
DIMM Slot #0	16GB DDR4 XMP PC4-28800
	G Skill Intl / F4-3600C16-16GVKC
	16-19-19-39 / 3602 MHz / 1.350V
DIMM Slot #1	16GB DDR4 XMP PC4-28800
	G Skill Intl / F4-3600C16-16GVKC
	16-19-19-39 / 3602 MHz / 1.350V

Result summary

Test Start Time	2020-05-14 00:13:50
Elapsed Time	8:11:39
Memory Range Tested	0x0 - 81F300000 (33267MB)
CPU Selection Mode	Parallel (All CPUs)
ECC Polling	Enabled
# Tests Passed	47/48 (97%)

Lowest Error Address	0x6F5471D50 (28500MB)
Highest Error Address	0x6F5471D50 (28500MB)
Bits in Error Mask	0000000000004000
Bits in Error	1
Max Contiguous Errors	1

Test	# Tests Passed	Errors
Test 0 [Address test, walking ones, 1 CPU]	4/4 (100%)	0
Test 1 [Address test, own address, 1 CPU]	4/4 (100%)	0
Test 2 [Address test, own address]	4/4 (100%)	0
Test 3 [Moving inversions, ones & zeroes]	4/4 (100%)	0
Test 4 [Moving inversions, 8-bit pattern]	4/4 (100%)	0
Test 5 [Moving inversions, random pattern]	4/4 (100%)	0
Test 6 [Block move, 64-byte blocks]	4/4 (100%)	0
Test 7 [Moving inversions, 32-bit pattern]	4/4 (100%)	0
Test 8 [Random number sequence]	3/4 (75%)	1
Test 9 [Modulo 20, ones & zeros]	4/4 (100%)	0
Test 10 [Bit fade test, 2 patterns, 1 CPU]	4/4 (100%)	0
Test 13 [Hammer test]	4/4 (100%)	0

Last 10 Errors

2020-05-14 06:42:55 - [Data Error] Test: 8, CPU: 22, Address: 6F5471D50, Expected: 663DBEF6, Actual: 663DFEF6

[David (PassMark)]

More likely it is bad RAM.

In the current V8.4 release of Memtest86 each core tests memory blocks of max 64MB at a time.
| Core 0 - 64MB | Core 1 - 64MB | ..... | Core N - 64 MB |

So there is a fixed relationship between cores and addresses. So if you have a single bad memory address, it is no surprise that it pops up on the same core each time.
We have discussed this a bit internally, and likely we'll switch to some pseudo random allocation of cores in a future software release.

[michte]

Thanks for your input. Just noticed too that address of the error on the second run matches the address of one of the errors on the first one.