How to figure out if the memory is bad?

Question

I recently had a problem where the computer (Thinkpad E430) hang while booting and I suspect that it is a bad memory issue. Checkbox test fail on memory, but memtest86+ passed. After removing the memory and re-installing I managed to boot the computer. However, the memory information looks odd:

1. I have 4GB or RAM and running ubuntu 16.04 64 bit

   sudo lshw -class memory:
   *-memory
          description: System Memory
          physical id: a
          slot: System board or motherboard
          size: 4GiB
        *-bank:0
             description: SODIMM DDR3 Synchronous 1333 MHz (0.8 ns)
             product: HMT351S6CFR8C-PB
             vendor: Hynix/Hyundai
             physical id: 0
             serial: 0A545936
             slot: ChannelA-DIMM0
             **size: 4GiB**
             width: 64 bits
             clock: 1333MHz (0.8ns)
        *-bank:1
             description: DIMM [empty]
             physical id: 1
             slot: ChannelB-DIMM0
   

2. System monitor shows that there is only 2.6GB available and more than 50% is used even though I don't run any program: System Monitor image after boot without running any program

   grep Memory /var/log/kern.log:
   
   kernel: [    0.000000] Memory: 2562464K/**2730856K available** (8432K kernel code, 1291K rwdata, 3960K rodata, 1484K init, 1316K bss, 168392K reserved, 0K cma-reserved)
   

How can I figure out if there is a problem in the RAM or elsewhere?

Answer 1

I'd suggest booting from USB memory stick (e.g. Ubuntu 18.04 LTS). Also make sure to use 64 bit variant if you still can find 32 bit version somewhere.

The fact that lshw lists memory funny is not unexpected. The system I'm currently using has 4x8 GB of DDR3 memory and lshw listing looks basically like this:

  *-memory:0 UNCLAIMED
       physical id: 1
     *-bank UNCLAIMED
          description: DIMM DDR3 Synchronous 1333 MHz (0.8 ns)
          product: 99U5471-036.A00LF
          vendor: Kingston
          physical id: 0
          serial: 6B2B875D
          slot: ChannelA-DIMM0
          size: 8GiB
          width: 64 bits
          clock: 1333MHz (0.8ns)
  *-memory:1
       description: System Memory
       physical id: 5e
       slot: System board or motherboard
     *-bank:0
          description: DIMM DDR3 Synchronous 1333 MHz (0.8 ns)
          product: KHX1600C10D3/8GX
          vendor: Kingston
          physical id: 0
          serial: B804123E
          slot: ChannelA-DIMM1
          size: 8GiB
          width: 64 bits
          clock: 1333MHz (0.8ns)
     *-bank:1
          description: DIMM DDR3 Synchronous 1333 MHz (0.8 ns)
          product: 99U5471-036.A00LF
          vendor: Kingston
          physical id: 1
          serial: 692B865D
          slot: ChannelB-DIMM0
          size: 8GiB
          width: 64 bits
          clock: 1333MHz (0.8ns)
     *-bank:2
          description: DIMM DDR3 Synchronous 1333 MHz (0.8 ns)
          product: KHX1600C10D3/8GX
          vendor: Kingston
          physical id: 2
          serial: B704D03D
          slot: ChannelB-DIMM1
          size: 8GiB
          width: 64 bits
          clock: 1333MHz (0.8ns)
  *-memory:2 UNCLAIMED
       physical id: 2
  *-memory:3 UNCLAIMED
       physical id: 3

Note how memory:0 seems to have one UNCLAIMED bank which has one stick and memory:1 has three sticks and memory:2 and memory:3 are empty.

The slot: names seem to be okay so I would only trust those.

However, looking through sudo dmidecode output, it seems that this may be due BIOS bug because DMI seems to tell similar configuration for memory.

In the end, if grep "MemTotal" /proc/meminfo does not match the actual memory (minus GPU memory reserved for integrated graphics), then Linux cannot use all of your actual memory. If that is the case, I'd try to look for BIOS updates or try faking OS for ACPI: https://wiki.archlinux.org/index.php/DSDT. In case of broken bios, you could also try noacpi kernel flag which will disable e.g. power management as a side-effect but will help with most ACPI/BIOS problems.