Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

Reply
 
Thread Tools
Old 08-15-07, 07:17 AM   #1
prudy
Registered User
 
Join Date: Aug 2007
Posts: 14
Default 100.14.11 reboots if non-symmetric dual-channel RAM

Hi
I am configuring asus laptop f3sv (nVIDIA GeForce 8600M GS) for fedora 7 (x86_64) using livna repository.
I have tested latest (kmod-nvidia-100.14.11-1.2.6.22.1_41.fc7.x86_64.rpm) and testing version (kmod-nvidia-100.14.11-2.2.6.22.1_41.fc7.x86_64.rpm) of the driver.

Both worked perfectly fine until the laptop owner decided to upgrade RAM from 1GB up to 3GB, adding additional 2GB. Now when xserver starts the machine reboots. I tested the following configurations:
a) 1G - OK
b) 2G - OK
c) 1G+2G - ERR
d) 2G+1G - ERR

I will borrow some RAM to test 2G+2G and 1G+1G - this should tell if the problem happens if both RAM socket are used or only if non-symmetric RAM configuration is used.

Note that when only one memory slot is used bios reports "Single-Channel Mode". For two memory slots it reports "Dual-Cannel L-Shaped Mode". I presume the "L-Shaped" means that RAM sockets use different memory size each.

Why laptop enables dual-channel for 1G+2G configuration at all?
Shouldn't it work in single mode and for dual mode only if both sockets use the same RAM size?
Why RAM size and/or single/dual mode has an impact here at all? Is it possible that 'something' assumes that if dual-channel is enabled it must be symmetric, so non physically existing 1GB is used anyway (as if it was there giving 4GB in total)?

Note that vista works using 1G+2G.

The Xorg log shows the following problem:
-----------------------------------------------
(II) Found conflict at: 0x1000bffff
(II) Found conflict at: 0x1000b7fff
(II) Found conflict at: 0x1000affff
(WW) ****INVALID MEM ALLOCATION**** b: 0x100000000 e: 0x10fffffff correcting
(II) window:
[0] -1 0 0x00000000 - 0x0fffffff (0x10000000) MX[b]
(II) resSize:
(II) window fixed:
[0] -1 0 0x00000000 - 0x0fffffff (0x10000000) MX[b]
Requesting insufficient memory window!: start: 0xf7f00000 end: 0xfdffffff size 0x10000000
(EE) Cannot find a replacement memory range
(WW) ****INVALID IO ALLOCATION**** b: 0x10000bc00 e: 0x10000bc7f correcting
(EE) end of block range 0xffffffff < begin 0x100000000
(II) window:
[0] -1 0 0x0000b000 - 0x0000bfff (0x1000) IX[b]
(II) resSize:
(II) window fixed:
[0] -1 0 0x0000b000 - 0x0000bfff (0x1000) IX[b]
Requesting insufficient memory window!: start: 0xb000 end: 0xbfff size 0x100000080
(EE) Cannot find a replacement memory range
-----------------------------------------------

Anyone hit similar problem and can give me some hints?
prudy is offline   Reply With Quote
Old 08-16-07, 11:45 AM   #2
prudy
Registered User
 
Join Date: Aug 2007
Posts: 14
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Tested now with 2G+2G RAM - no success. X starts but there is a total mess on the screen - artifacts and strange colors. The same problem in the log.

Anyone with f3sv/f3sc laptop to test it?
prudy is offline   Reply With Quote
Old 08-16-07, 11:55 AM   #3
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Quote:
Originally Posted by prudy
Tested now with 2G+2G RAM - no success. X starts but there is a total mess on the screen - artifacts and strange colors. The same problem in the log.

Anyone with f3sv/f3sc laptop to test it?
No, I don't have a f3sv/f3sc but a G1S with 4 Gigs...

If your screen looks a bit like this
you have "change" your pci mappings...
chunkey is offline   Reply With Quote
Old 08-16-07, 04:38 PM   #4
prudy
Registered User
 
Join Date: Aug 2007
Posts: 14
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

The mess is like in the picture.

Is this the problem mentioned in the link below under "64-Bit BARs (Base Address Registers)" section?
http://us.download.nvidia.com/XFree8...ppendix-l.html

In fact the memory region in the log is from above 32b address space but the article is about "I/O" regions, not memory regions.

Anyway I be glad to know the solution.
prudy is offline   Reply With Quote
Old 08-16-07, 04:45 PM   #5
prudy
Registered User
 
Join Date: Aug 2007
Posts: 14
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Below is the 'lspci' diff for bad (2X2G=4G RAM) and good (2G RAM) case:

Region 1: Memory at 140000000 (64-bit, prefetchable) [size=256M] <-- if err
Region 1: Memory at c0000000 (64-bit, prefetchable) [size=256M] <-- if ok

Expansion ROM at fdfe0000 [disabled] [size=128K] <-- if err
[virtual] Expansion ROM at fdfe0000 [disabled] [size=128K] <-- if ok
prudy is offline   Reply With Quote
Old 08-16-07, 05:20 PM   #6
prudy
Registered User
 
Join Date: Aug 2007
Posts: 14
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Adding reference:
http://www.nvnews.net/vbulletin/show....php?p=1343905
prudy is offline   Reply With Quote
Old 08-16-07, 06:49 PM   #7
zander
NVIDIA Corporation
 
zander's Avatar
 
Join Date: Aug 2002
Posts: 3,740
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Quote:
Originally Posted by prudy
Is this the problem mentioned in the link below under "64-Bit BARs (Base Address Registers)" section?
http://us.download.nvidia.com/XFree8...ppendix-l.html

In fact the memory region in the log is from above 32b address space but the article is about "I/O" regions, not memory regions.
Yes, BARs mapped above 4GB are problematic. In some cases, the SBIOS provides an option that gives control over whether or not an I/O window is provided below the 4GB boundary.
zander is offline   Reply With Quote
Old 08-16-07, 07:01 PM   #8
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Quote:
Originally Posted by zander
Yes, BARs mapped above 4GB are problematic. In some cases, the SBIOS provides an option that gives control over whether or not an I/O window is provided below the 4GB boundary.
Well, the problem is not the mapping window itself.

This is the situation on my G1S, but it's probably the "same" on any ASUS:
The SBIOS sets the window actually "in" the low 4gb range, however the window is too big (0xc0000000 - 0xffffffff => 1GB!) and overlaps all other devices. So Linux tries to solve to conflict and remap the brigde & gfx bars
into the _free_ > 4GB range.

And there isn't any option for this in the Bios... =>
"It isn't Nvidia's fault!"
chunkey is offline   Reply With Quote

Old 08-16-07, 07:07 PM   #9
zander
NVIDIA Corporation
 
zander's Avatar
 
Join Date: Aug 2002
Posts: 3,740
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Interesting, I hadn't been aware of this problem. prudy, can you attach the output of `dmesg` after the system has first booted, as well as an nvidia-bug-report.log file?
zander is offline   Reply With Quote
Old 08-16-07, 07:20 PM   #10
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

in the mean time:

here is my old thread with a nvidia-bug-report.log

Code:
/proc/iomem
140000000-14fffffff : PCI Bus #01
  140000000-14fffffff : 0000:01:00.0

xorg.log
	[14] -1	0	0x140000000 - 0x14fffffff (0x10000000) MX[b](B)

however:
/usr/bin/lspci -d "10de:*" -v -xxx

01:00.0 VGA compatible controller: nVidia Corporation Unknown device 0407 (rev a1) (prog-if 00 [VGA])
	Subsystem: ASUSTeK Computer Inc. Unknown device 1515
	Flags: bus master, fast devsel, latency 0, IRQ 16
	Memory at fc000000 (32-bit, non-prefetchable) [size=16M]
	Memory at 140000000 (64-bit, prefetchable) [size=256M]
	Memory at fa000000 (64-bit, non-prefetchable) [size=32M]
...
only the brigde window is "wrong", the BARs of the card behind the bridge are OK (atleast size wise)!

To get the card working:
change the window values before Linux verifies them (a quick hack needs only ~4 lines )

EDIT: little correction: the prefetch window _covered_ the ACPI tables. The 1GB is also a Vista thing...

Last edited by chunkey; 08-16-07 at 07:36 PM. Reason: edit
chunkey is offline   Reply With Quote
Old 08-16-07, 07:47 PM   #11
dkulp
Registered User
 
Join Date: Apr 2005
Posts: 6
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

I have a similar problem. With 3GB of memory, Linux just boots into a black screen. With 2GB, all works fine. I tried the pci=nommconf flag but that didn't help.

I've attached the nvidia-bug-report.log.
Attached Files
File Type: log nvidia-bug-report.log (87.1 KB, 246 views)
dkulp is offline   Reply With Quote
Old 08-17-07, 06:25 AM   #12
prudy
Registered User
 
Join Date: Aug 2007
Posts: 14
Default Re: 100.14.11 reboots if non-symmetric dual-channel RAM

Hi,
As requested, attaching 'dmesg' as a first thing after boot and bug-log.
There are two series of them, '2G-ok' is for boot with 2G RAM when everything is ok, '4G.err' stands for 4G RAM when there is a problem.
Regards.
Attached Files
File Type: gz dmesg-4G.err.log.gz (8.6 KB, 182 views)
File Type: gz nvidia-bug-report-4G.err.log.gz (33.7 KB, 178 views)
File Type: gz dmesg-2G.ok.log.gz (8.4 KB, 245 views)
File Type: gz nvidia-bug-report-2G.ok.log.gz (32.8 KB, 187 views)
prudy is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


Similar Threads
Thread Thread Starter Forum Replies Last Post
(PR) ADATA Adds DDR3-2133X 8 GB and 16 GB Dual Channel Kits to XPG Xtreme Series News Archived News Items 0 05-16-12 10:40 AM
Dual Channel DDR ryank1979 CPUs, Motherboards And Memory 3 08-13-02 06:37 AM

All times are GMT -5. The time now is 10:59 AM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.