Centos 7 not seeing memory upgrade
 

Getting into Linux, installed Centos 7 on an old workstation that had Vista on it. System has four memory slots where the system had 2 x 1 gig sticks, and 2 x 512 meg sticks, for a total of 3 gigs (DDR 2).

I took three sticks out and put 3 x 2 gig memory sticks, and left one of the original 1 gig sticks. This should equal 7 gigs.

Centos still sees the system as having 3 gigs.

The thing is, I do not know if the motherboard supports more than 3 gigs of memory.

I was playing around with this until bedtime last night and will get back on it when I get home.

When I get home I will get into the bios and see if the motherboard is seeing the 7 gigs of memory.

Until I get home this evening, the question is:

Will Centos 7 see added memory, or do I need to change something on the OS?

This is an older Gateway FX desktop with a core 2 quad cpu. I do not remember who made the motherboard.

Yes. If the computer (the BIOS) knows about it, so does normally Centos. You could check for error or warning messages in the message buffer at early boot time. Use the dmesg command.

No, at most 6 GB as "pairs" of memory sticks should be identical, so AT MOST it is seen as 2x2G plus 2x1G (if the motherboard accepts those 2 GB sticks).
Also make sure the (better keep both 1 GB sticks in the system) first two slots are filled with the larger sticks, so slot 0 and 1 with 2 GB, 2 and 3 with 1 GB to reach that total of 6 GB
Or, IF the motherboard can handle 2 GB sticks: buy a 4th one for 8 GB total (at least all the sticks then are identical).

Seems you like to go by the book. Yes, that is the way things are "supposed" to go.

In real life I have been mixing memory on computers for close to two decades and rarely if ever have an issue.

Motherboards usually do good at sensing the memory and getting things to work.

And does with an Intel Core 2 cpu, as that is a two-way interleave processor, so needs its memory in matched sizes.

You think using the single one gig stick is preventing the system from seeing the other 4 gigs? Maybe it would be better to take that single stick out with the matching stick on the other slot and see if I can at least get 4 gigs.

I found the system on an old Cnet page, which says the computer supports 8 gigs of memory.

I would at least start the troubleshooting there. Make sure it sees 2x2 first, then see if you can add an additional 2x1.

Sounds like a good idea.

Will replace the 2 gig module that matches with the 1 gig with a 1 gig, this will be 2 - 1 - 2 - 1 . Hopefully get 6 gigs. Go from there and see what happens.

It was late last night when I put the memory in and did not have a lot of time to play with it.

Got home, ate dinner, switched one of the 2 gig memory modules out with a one gig, 2 - 1- 2- 1, system still shows three gigs of memory.


Rebooting, going to check bios... go from there

Took 2 chips out, left two of the two gig. 2 - empty - 2 - empty. Centos still showed three gigs of memory. Put memory back in with the two 2 gig sticks along with the two 1 gig. Rebooted

Rebooted, system showed the bios screen for a split second, and I was able to hit pause. Motherboard shows 6 gigs of memory. I got my camera and took a picture if yall want to see it.

So what would cause the OS to only see 3 gigs. Its almost as if a 32 bit os was installed.

My kernel version - 3.10.0-1062.1.1.el7.x86_64

~~~~~ EDIT ~~~~~~

Maybe I could reinstall with the new memory? This is a fresh install so nothing would be lost.

~~ EDIT 2 ~~

[root@unknown00E0B8E6C968 ~]# sudo lshw -class memory
*-firmware
description: BIOS
vendor: American Megatrends Inc.
physical id: 0
version: 080014
date: 01/14/2008
size: 64KiB
capacity: 512KiB
capabilities: isa pci pnp apm upgrade shadowing escd cdboot bootselect socketedrom edd int13floppy1200 int13floppy720 int13floppy2880 int5printscreen int9keyboard int14serial int17printer int10video acpi usb ls120boot zipboot biosbootspecification
*-cache:0
description: L1 cache
physical id: 5
slot: L1-Cache
size: 128KiB
capacity: 128KiB
capabilities: internal write-back data
configuration: level=1
*-cache:1
description: L2 cache
physical id: 6
slot: L2-Cache
size: 8MiB
capacity: 8MiB
capabilities: internal write-back unified
configuration: level=2
*-memory
description: System Memory
physical id: 25
slot: System board or motherboard
size: 6GiB

But free command shows 3 gigs

[root@unknown00E0B8E6C968 ~]# free
total used free shared buff/cache available
Mem: 3172048 1280780 1025480 89900 865788 1642960
Swap: 3145724 0 3145724

System monitor shows three gigs of memory.

Does that machine have the ability to boot from USB? If so, can you create a CentOS 7 install USB just to see if the live image sees the full available ram? Or ANY live linux really? Just to test that the mobo is passing the information properly to the OS (I have seen this happen with old, early generation 64-bit motherboards).

Thank you and yes, I installed the OS with a bootable thumbdrive.

What do I need to do?

Reboot the system with the thumbdrive, then what? I am not sure what you meant by "live image."

Live image - Any OS that can run from a USB (or cd/dvd but obviously that's a lot slower) and doesn't just go straight into an installer. You're not needing to actually do anything, just boot to a USB environment with whatever OS you can, verify how much ram said environment sees.

This step is strictly to rule out the bios/bootloader communication as the issue.

Thank you. Last night I did a fresh install of Centos 7. During the install I selected reclaim the drive space. This was to check if Cent was not seeing the memory upgrade.

Same issue with a fresh install, the OS sees only 3 gigs of memory.

This morning I downloaded Ubuntu, created a bootable flashdrive, booted into Ubbuntu, brought up terminal and ran sudo lshw -class memory

memory
description: System Memory
physical id: 25
slot: System board or motherboard
size: 6GiB

The memory display shows the correct size size memory module in the correct slots.


Top command says KiB Mem : 3328596 total.

System monitor - resources - memory - 3.2 gigs.

The motherboard shows to have 6 gigs of memory, but for some reason neither ubuntu or Centos can see no more than 3 gigs.

ubuntu@ubuntu:~$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 18.04.3 LTS
Release: 18.04
Codename: bionic

I don't suppose there's a BIOs update for that board? Sounds like it's a BIOS issue. Some of the early generation Athlon 64 boards I saw the same thing in.

I'm noticing that the OP is reporting chip configurations as
2-1-2-1 or 2-empty-2-empty

Have you tried 2-2-1-1 or 2-2-empty-empty?

My understanding is that not only do the chips need to be installed in pairs, but that the matching chips need to be adjacent.

And yes, all that is a relatively new requirement...back in the day, the MBs didn't care what got plugged in where, but match pairing has been required for the last several years.

Also when I run the lshw command on my CentOS 7 desktop I get:
The OP didn't post the memory info from that command
(and please use code tags to post output...makes it much more readable)

Just :twocents:

That is a good question.

The picture I took last night says:

American Megatrends
680-IT-GB V1.00F

A quick google search does not turn up any results.

Give me a little bit, will shut the system down see if I can find motherboard make and model.

~~~~~~~~~~~~~~~~~~

Good suggestion.

When I shut the system down to look at the MB I will try a different combination of slots.

Might take a nap on this lazy saturday, so it may be a little while before I post again.

THAT is motherboard specific, there are motherboards too that pair even (0 and 2) cq odd (1 and 3) memory slots. Often then they will be a different color, i.e. 0 and 2 are blue connectors and 1 and 3 black. But again, that still is motherbord specific!

Point taken. The one 4-slot motherboard I have handy is blue-blue black-black, with matching colors adjacent.

Matching sticks should be in matching colors.

I understand the thinking, but a lot of systems will boot even if the modules are not in the matching colors.

The system has two modules in the two black slots and boots to 3 gigs of memory.

Looks like this is a Gigabyte board built for Gateway - 680-IT GB. A search for 680-IT on Gigabytes website does not return any results. Still looking.

In all honesty I do not understand why people are in a tissy about where the memory modules are at in the board. It's as if yall have no hardware experience and just quoting guidelines printed in some book.

I have been building systems since around 1996, and rarely, if ever, have I had an issue where the memory modules are at in the MB. I bought my first computer in 1994 with Windows 3.11, and upgraded it in 1995 to play Quake. From there I built my own computers from 1995 to 2018. I took a hardware and advanced hardware class at the local college.

The system I am working on was given to my close to a decade ago by a customer. They did not want it anymore and they gave it to me instead of payment.

Typically:

Matching colors = dual channel
Not matching colors = single channel

Most motherboards will sense where the modules are at and adjust accordingly.

Not in a “tissy” Just offering suggestions in an attempt to help.
I’ll bow out now.

Yeah, and if you're not using the matching colors, then having matching (or very nearly) dimms you're purposely leaving a decent amount of performance off your build. SOME system won't work (here's looking at a fair amount of the late Pentium 4's) if the dimms don't match. Some boards also would ONLY support their max in dual channel. So again, not having them in the matching slots is a possible cause of the issue, and leaving performance in the trash. Although I still think it's a BIOS issue.

Hopefully this is a dumb question: Is it possible that you're not running the "pae" kernel?

I once screwed up by installing the non-pae kernel (apparently, thinking "Nah... I'll never need that much RAM on this system") on an old Core2 Duo m'board and was unable to pull off a memory addition like you're attempting until I re-installed with the correct kernel.

HTH...

Please excuse my ignorance, I do not know what a pae kernel is.


I installed a windows 64 bit OS and it is having the same issues. The OS shows 6 gigs installed, but only 3.2 gigs usable. So this is not a Linux issue per say.

In short:

• Older motherboard is unable to relay memory above 3 gigs to the OS.
• I am unable to find bios update.
• Thinking about setting this system to the side and booting up a more recent computer I have in storage.

Thoughts, suggestions? I feel like I am spinning my wheels.

If anyone has suggestions, please post them.

It is getting late Saturday evening and I want to relax before bedtime, so let's continue this conversation on Sunday.

Well, one thing to try before you set it to the side is reset cmos through jumper with all the memory in it. Maybe it'll clear something that's preventing it from reading properly.

From the Wikipedia entry for PAE:

``It defines a page table hierarchy of three levels (instead of two), with table entries of 64 bits each instead of 32, allowing these CPUs to directly access a physical address space larger than 4 gigabytes (2^32 bytes).''

It could very well be a Linux issue. Or, rather, might have been in the past. CentOS 7 seems to have had 32-bit support at one point and if that kernel didn't have the PAE hooks (See the PAE's Wikipedia entry for the gory details) that would likely be why the memory is recognized. But I also read that CentOS went PAE-only as a default somewhere along the may in the 7.x releases. I'm not currently a CentOS user so I can't speak to what they're using as their kernel these days.

Windows showing only 3.2GB available sounds like something I've heard of before as being an OS limitation---I think that 0.8GB is/was reserved by Windows. (I sat next to a guy at work who had to explain to a perturbed project manager why the system design he wanted to implement was not going result in a 4GB-of-visible-RAM system he'd promised to his departmental client and the QA people were going to squawk about 3.2 seen != 4 specced.) There's always the chance that it's a chipset limitation. I still have an ancient Compaq EN that's only capable of accessing 768MB despite having four memory slots: 3x256 works, 4x256 does not. 4x128, of course, works. If this turns out to be a chipset limitation, obviously, all you can do is install the maximum allowed by the hardware.

Interested in seeing your motherboard information... that'll tell us a lot.

Later...

There WILL be a certain amount reserved by the hardware, and therefore that certain amount of RAM will be unavailable to the OS - I think it depends on the OS as to exactly how much it "won't be able to see/address", but I could be wrong about the last part of what I just said (the first part of what I said is true AFAIK). But that said, the BIOS should still report the full amount of RAM installed, so if the OP has installed 6GiB of RAM all up, then the BIOS should still report 6GiB of RAM being present. So if the BIOS isn't reporting the full amount of RAM installed; then that would mean a hardware issue, and not a OS issue.

For example, my machine has 8GiB of RAM, but if I look at KDE system monitor, it reports 7.7GiB of total RAM available that the OS "sees" and can address. It was the same story when I was still using CentOS 7 as my main OS. It also says in the motherboard PDF for my machine's board that a small amount of RAM won't be available to the OS - so that's not limited to Linux.

If I was using my board's on-board Intel integrated graphics, then even less RAM would be available to the OS for "general purposes", because some of it would also be shared for graphics memory.

A pae (Physical Address Extensions) kernel is a 32-bit kernel that can use multiple 4 GB pages for different address spaces. So the RAM is one page, the screen memory ON the video adaptor is another, etc. It is the "CONFIG_HIGHMEM64G=y" option in the kernel config file, which allows the KERNEL to make use of up to 64 GB of memory. Again: this is for the 32-bit kernel only.

You (and applications) still cannot break the 32-bit (4 GB) barrier, but at least every application now has its own 4GB page (and the kernel switches between them as needed).

Of course pae is never needed when the cpu is using full 64-bit mode.

Normally an amount of memory of 3.2 GB means the cpu is in 32-bit mode without pae, because then the screen memory (etc) has to be within a single 4 GB address space too.

It just may mean that the machine isn't correctly switching to 64-bit (long pointers) mode, which means it only addresses UP to the 4 GB limit of a 32-bit pointer.
This may be a BIOS limitation.

Thank you everyone for your help and suggestions. I took Sunday off from this project and just laid around the house.

Let's pick this back up Monday evening.

Even if I can not get past the 3 gig issue, I would still like to use this machine for a project. It has hardware raid built into the motherboard which I would love to play around with.

If nothing else, I will get one of my newer machines going for another project. Maybe use the one with the three gig limit as a backup server or something.

Suggestions would be appreciated.

Guys and gals, I decided to give up on using this computer for a test platform. Instead, I took my retired AMD 8350 (8 core) cpu with a motherboard that supports 32 gigs of memory and use it as a test platform.

Thank you everyone for your help, but I feel using this old core 2 quad that will only see 3 gigs of memory is a waste of time.

Honestly, can't blame you.

Thank you.

I was wanting to set up a lab with multiple systems, hardware configs, and OSs. Now its looking more like one system using virtualization to run multiple OSs.

With the way things are going in the tech world using virtualization may be a better option than dedicated computers to a single OS.