[Leaplist] Only seeing 3G memory

[Bryan J. Smith]

From: Gray Frost <grayf327 at gmail.com>
> Checking aperture...
> Node 0: aperture @ 3b26000000 size 32 MB
> Aperture beyond 4GB. Ignoring.
> No AGP bridge found
> Memory: 2830784k/2882496k available (2699k kernel code,
> 51324k reserved, 1485k d

The first part of that, it's something else (AGPgart).
There should be many more lines than this in Fedora.

You should have a BIOS memory map of usable/reserved/etc...
That should actually be at the very beginning.

E.g., (NOTE: this is actually from i686, not x86_64, and
       only a 2GiB system, although it's still similar
       on a 4GiB system with and without PAE on i686,
       let alone x86_64)

BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009f000 (usable)
 BIOS-e820: 000000000009f000 - 00000000000a0000 (reserved)
 BIOS-e820: 00000000000d2000 - 00000000000d4000 (reserved)
 BIOS-e820: 00000000000dc000 - 0000000000100000 (reserved)
 BIOS-e820: 0000000000100000 - 000000007f6d0000 (usable)
 BIOS-e820: 000000007f6d0000 - 000000007f6df000 (ACPI data)
 BIOS-e820: 000000007f6df000 - 000000007f700000 (ACPI NVS)
 BIOS-e820: 000000007f700000 - 0000000080000000 (reserved)
 BIOS-e820: 00000000f0000000 - 00000000f4000000 (reserved)
 BIOS-e820: 00000000fec00000 - 00000000fec10000 (reserved)
 BIOS-e820: 00000000fed00000 - 00000000fed00400 (reserved)
 BIOS-e820: 00000000fed14000 - 00000000fed1a000 (reserved)
 BIOS-e820: 00000000fed1c000 - 00000000fed90000 (reserved)
 BIOS-e820: 00000000fee00000 - 00000000fee01000 (reserved)
 BIOS-e820: 00000000ff800000 - 0000000100000000 (reserved)

On i686, it will be followed by HIGHMEM/LOWMEM and other lines,
then eventualy the ACPI and various, key descriptors.

On x86_64, it should be followed by the ACPI and various, key
descriptors, none of the other paging schtuff.

> and this is the only error I found in the dmesg listing: 
> [@localhost ~]$ dmesg |grep -i error
> ACPI Error (dsobject-0501): Package List length (5) larger
> than NumElements
> count (3), truncated

It won't be an error.  After all, your BIOS thinks you only
have 3GiB.  I'd like to see the map, although that may still
not tell me your issue.  ;)

> It is a 939 processor.  I have four each sticks of one gig
> (4 gig total) modules DDR 400 loaded into the only 4 slots
> (two blue and two black) provided.

By JEDEC specifications, your mainboard _should_ slow down
the timing to DDR-333/PC-2700 for two (2) DIMMs per channel.
Since you have two (2) channels and two (2) DIMMs per channel,
they should not signal at DDR-400/PC-3200.

If it does not, all bets are off.  Some mainboards are
rather cheap and pathetic and will not do this automagically.
They will happily cause some DIMMs to not even work and/or
act unstable.  You may wish to go into the BIOS and see if
you can manually reduce the speed of your memory from
DDR-400 (200MHz DDR) to DDR-333 (166MHz DDR).

I've had both inexpensive consumer (Award BIOS - nForce 4/600)
and more expensive server (AMI/Phoenix ServerBIOS - nForce
Pro 2000/3000) BIOS-chipset setups do that automatically when
I used more than one (1) DDR-400/PC-3200 DIMM per channel.

> I was a little unsure about how they were to be arranged.

That can be important.

> They came two in a package and so I put them in pairs per
> package into the blue then the black slots according to the
> mother board manual which I have attached the memory
> installation page.  The screenshot is taken from the
> page for the motherboard I have.

Says populate 1-2, then 3-4, for dual-channel and proper
pairing.  So your original should be 1-2, then 3-4.  That
is sequential, when going left-to-right.

> The Machine is a "emachine  T6212"

Several sites show that the support DIMM IC config is 128x64
for 1GiB, which is typical (and maximum) for 184-pin DDR SDRAM
DIMMs.  So nothing unusual there.

BTW, you had 2GiB originally, correct?  Not 1GiB?  Correct?
I know, dumb question, but I had to ask.  ;)

This is an ATI Xpress 200.  I've never, ever use one of those.
It was ATI's first AMD chipset (before AMD bought them ;).


From: Steve Litt <slitt at troubleshooters.com>
> I had no idea this was a problem.

I only saw it back in the original Opteron server prototypes
(circa 2003-2004) from Tyan, SuperMicro, a few others which
were addressed by a BIOS fix (2004).  But the nice thing about
these early server configurations is that they were almost
always 4-16GiB, so they addressed it in quick order.

I saw it more on the consumer mainboards in the following
years.  Why?  Two reasons:  

1)  Most consumers had only 512MiB-2GiB on them because the
    available, affordable DDR DIMMs were 512MiB or less.

2)  Most users were booting Windows XP, and both Home and
    Pro cannot use more than 3GiB, period (only Server 2003
    can)

But soon consumers ran into the issue, so vendors were still
issuing BIOS updates as late as 2007 to deal with their lack
of even thinking of such.

> I guess I got lucky on my new computer:
> [slitt at mydesk ~]$ head /proc/meminfo
> MemTotal:      8302156 kB

Depends.  I assume this is a consumer mainboard.  If so,
you clearly have 240-pin DDR2 SDRAM DIMM slots with either
4x2GiB (with slower signaling) or 2x4GiB (faster signaling)
DIMMs.  If you have a server mainboard with registered DDR2,
that changes things though.

I've never seen a newer, AMD Socket-AM2 with 240-pin DDR2
slots that had this issue.  That's because most support
at least 8GiB, if not 16GiB, using consumer 2GiB or 4GiB
DDR2 DIMMs, respectively.  Consumer DDR DIMMs used to top
out at 1GiB.

For Intel, it's been a long while since I've seen one of
the older LGA-775 with 240-pin DDR2 slots that had the 4GiB
issue.  Most are i945 or similar with no Yonah (Core) logic
support.  Basically anything i946 or i965 should not have
issues, but some mainboard vendors do "cut corners."


-- 
Bryan J Smith        Professional, Technical Annoyance
b.j.smith at ieee.org  http://www.linkedin.com/in/bjsmith
------------------------------------------------------
I'm a PC, but Linux -- Windows: Life Without Firewalls


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.