Hi there, has anyone run into any Desktop/Workstation Intel brand motherboards that work with CentOS 5.1 (4.6/3.9 out of box compatibility is a very large plus) and work with the 1333MHZ cpus such as the 8400/9450 and accept 8GB of RAM? The closest we've found so far is the DG31PR which is alright but it has the limitation of 4GB of RAM and an unsupported Realtek NIC. The DQ35JO is also ok, but requires quite a bit of 'jostling' to get to work such as pci=nommconf and other tweaks, we need something that will work out of box, anyone found anything recently?
Thanks in advance if anyone has any hints.
-Drew
Drew Weaver wrote:
Hi there, has anyone run into any Desktop/Workstation Intel brand motherboards that work with CentOS 5.1 (4.6/3.9 out of box compatibility is a very large plus) and work with the 1333MHZ cpus such as the 8400/9450 and accept 8GB of RAM? The closest we’ve found so far is the DG31PR which is alright but it has the limitation of 4GB of RAM and an unsupported Realtek NIC. The DQ35JO is also ok, but requires quite a bit of ‘jostling’ to get to work such as pci=nommconf and other tweaks, we need something that will work out of box, anyone found anything recently?
Thanks in advance if anyone has any hints.
-Drew
--
Join OSCC MAMPU Mailing List
*http://lists.oscc.org.my/mailman/listinfo/oscc-discuss*
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Drew,
Typically Linux hackers can hack the kernel to use very large RAM. Current applications are built to access up to 4Gb RAM at a time only. For machines with RAM of more than 4Gb you need to verify that the CPU comes with Physical Address Extensions (PAE). Call Intel to verify this CPU is with PAE. So, now its the Linux kernel that you need to turn to.
CentOS 5.1 64 bit should not have this limitation. However, the following method allows up to 64Gb out of box for CentOS 5.1 32bit (u can try this if the 64bit didnt work out).
Do the following in runlevel 3 as root user;
# yum install kernel-PAE # reboot # free
But all this needs to come with distributed swap and use of appropriate file system type to ensure performance.
On Tue, May 20, 2008 at 8:13 PM, Nicholas nicholas@oscc.org.my wrote:
Typically Linux hackers can hack the kernel to use very large RAM. Current applications are built to access up to 4Gb RAM at a time only. For machines with RAM of more than 4Gb you need to verify that the CPU comes with Physical Address Extensions (PAE). Call Intel to verify this CPU is with PAE. So, now its the Linux kernel that you need to turn to.
CentOS 5.1 64 bit should not have this limitation. However, the following method allows up to 64Gb out of box for CentOS 5.1 32bit (u can try this if the 64bit didnt work out).
You are missing the point about the bug in some Intel chipsets. http://www.google.com/search?q=intel+chipset+kernel+4gb
Linux wrote:
You are missing the point about the bug in some Intel chipsets. http://www.google.com/search?q=intel+chipset+kernel+4gb
or the lack of support in 5.1 for ICH9 in non-AHCI mode which means the OS won't even install, never mind 4GB+ memory issues.
so, really there's four completely separate issues here...
A) Intel desktop chipsets don't support 'memory remapping' so somewhere between 750MB and 1GB of RAM is physically inaccessible on a machine with 4GB+ memory, no matter WHAT you do in software. This 750MB-1GB of inaccessible memory is at the same address as the PCI and PCI-E busses. In a pure 32bit environment, this all has to fit in the first 4GB.
B) many Intel desktop chipsets have a maximum 4GB physical memory limitation anyways, which when combined with A) means 3-3.25GB usable memory.
C) 32bit systems with > 4GB memory require PAE support both in the CPU and in the operating system... /all/ Intel CPUs since about Pentium Pro have this. For systems with > 4GB ram, you probably want to run 64bit native rather than mess with PAE anyways.
D) out of box support for the chipset and its peripherals so the OS can be installed without having to dink around with special flags or add-on drivers. The newest ICH9 and ICH9-R chips ARE supported in 5.1 if they are in ACHI mode, but apparently Intel has leaned on motherboard makers to not support AHCI mode on ICH9 in the BIOS, only on ICH9-R (which is the same basic backend chip with fake-raid enabled).
your 'bug' is issue A). I think its more a design feature or side effect, not an outright bug, but thats a nomenclature thing.
Ralph Angenendt wrote:
John R Pierce wrote:
C) 32bit systems with > 4GB memory require PAE support both in the CPU and in the operating system... /all/ Intel CPUs since about Pentium Pro have this.
(At least) The first Pentium M generation doesn't support PAE.
I should have qualified that. All intel CPUs that are supported by chipsets that can handle 36 bit addressing and > 4GB memory had PAE... Pentium-M was a laptop processor, and its supporting chipsets had a 2GB max memory anyways, so PAE support wouldn't have mattered.
your 'bug' is issue A). I think its more a design feature or side effect, not an outright bug, but thats a nomenclature thing.
Seems like it is the way Intel chips are designed. One solution would be to ensure that 64bit is used. However, due to cost I am sure many would have gone for the 32bit machines with AHCI mode.
For the Realtek drive maybe you can try using http://sourceforge.net/projects/realtekr1000
On Tue, May 20, 2008 at 5:59 PM, Nicholas nicholas@oscc.org.my wrote:
For the Realtek drive maybe you can try using http://sourceforge.net/projects/realtekr1000
Realtek drivers can be found in the CentOS wiki:
http://wiki.centos.org/HardwareList#head-851e245f4f537add3de9c3c6a6d686771fb...
Akemi
Realtek drivers can be found in the CentOS wiki:
http://wiki.centos.org/HardwareList#head-851e245f4f537add3de9c3c6a6d686771fb...
Sweet thanks. Now I know about dkms-enabled driver package which rebuilds the driver automatically for each kernel upgrade!
Shawn
Shawn wrote:
Realtek drivers can be found in the CentOS wiki:
http://wiki.centos.org/HardwareList#head-851e245f4f537add3de9c3c6a6d686771fb...
Sweet thanks. Now I know about dkms-enabled driver package which rebuilds the driver automatically for each kernel upgrade!
ofcouse, you dont need that on CentOS :D
- KB
Karanbir Singh wrote:
Shawn wrote:
Realtek drivers can be found in the CentOS wiki:
http://wiki.centos.org/HardwareList#head-851e245f4f537add3de9c3c6a6d686771fb...
Sweet thanks. Now I know about dkms-enabled driver package which rebuilds the driver automatically for each kernel upgrade!
ofcouse, you dont need that on CentOS :D
rhel 5.2 contains updated drivers. so as centos 5.2 will be release these problems will vanish.
Farkas Levente wrote:
Sweet thanks. Now I know about dkms-enabled driver package which rebuilds the driver automatically for each kernel upgrade!
ofcouse, you dont need that on CentOS :D
rhel 5.2 contains updated drivers. so as centos 5.2 will be release these problems will vanish.
Perhaps I was not clear in my original email, the point being that you dont need to rebuild drivers when kernels update ( in 99% of the cases )
On Thu, May 22, 2008 at 5:42 AM, Karanbir Singh mail-lists@karan.org wrote:
Perhaps I was not clear in my original email, the point being that you dont need to rebuild drivers when kernels update ( in 99% of the cases )
Is that now true also of the nvidia driver(s)? I haven't seen anything so to indicate.
Thanks.
mhr
MHR wrote:
Perhaps I was not clear in my original email, the point being that you dont need to rebuild drivers when kernels update ( in 99% of the cases )
Is that now true also of the nvidia driver(s)? I haven't seen anything so to indicate.
The nvidia driver, for me, built against 2.6.18-8.el5) has worked fine upto the latest released kernel ( 2.6.18-92.el5 )
try running the vmware-server installed on centos-5, irrespective of what kernel version you use / run - the vmware binary modules used are from an early kernel as well...
------ [kbsingh@koala misc]$ /sbin/modinfo vmmon.o | grep verm vermagic: 2.6.18-8.el5 SMP mod_unload gcc-4.1 [kbsingh@koala misc]$ uname -r 2.6.18-53.1.4.el5 -------
On Sun, May 25, 2008 at 4:23 PM, Karanbir Singh mail-lists@karan.org wrote:
MHR wrote:
Is that now true also of the nvidia driver(s)? I haven't seen anything so to indicate.
The nvidia driver, for me, built against 2.6.18-8.el5) has worked fine upto the latest released kernel ( 2.6.18-92.el5 )
That's what I'm using, too. I was just wondering if I still needed to use that one.
try running the vmware-server installed on centos-5, irrespective of what kernel version you use / run - the vmware binary modules used are from an early kernel as well...
Likewise.
On Wed, May 21, 2008 at 3:59 AM, Nicholas nicholas@oscc.org.my wrote:
Seems like it is the way Intel chips are designed. One solution would be to ensure that 64bit is used. However, due to cost I am sure many would have gone for the 32bit machines with AHCI mode.
Well, 64bit is the solution to address over-4GB memory. But since now, whenever I decided to upgrade a intel chipset mobo over 4GB, performance has significantly decreased. This is the BUG with latest chipsets and bios; not 32bit/64bit addressing issue.
-
Akemi Yagi -
Drew Weaver -
Farkas Levente -
John R Pierce -
Karanbir Singh -
Linux -
MHR -
Nicholas -
Ralph Angenendt -
Shawn