We just got a hold of a Dell Poweredge 2850 with dual EMT64's in it. I thought I would load of FC3 and give it a shot...Are there any issues that the list knows about that I should watch out for?
Try the FC3 X86_64 realease candidate (just look at archive on list) and see if everything works out. I wouldn't mind seeing how much slower the EMT64 is compared to a true X86_64 CPU like the AMD Opteron :-)
On Wed, 27 Oct 2004 17:21:08 -0700, Sean Bruno sean.bruno@dsl-only.net wrote:
We just got a hold of a Dell Poweredge 2850 with dual EMT64's in it. I thought I would load of FC3 and give it a shot...Are there any issues that the list knows about that I should watch out for?
-- Sean Bruno - TELECOM sean.bruno@metro1.com --
-- fedora-test-list mailing list fedora-test-list@redhat.com To unsubscribe: http://www.redhat.com/mailman/listinfo/fedora-test-list
On Thu, 28 Oct 2004 14:16, Jerone Young jerone@gmail.com wrote:
Try the FC3 X86_64 realease candidate (just look at archive on list) and see if everything works out. I wouldn't mind seeing how much slower the EMT64 is compared to a true X86_64 CPU like the AMD Opteron
How is an EMT64 not a "true X86_64"? They both run the same code.
On Thu, 2004-10-28 at 19:18 +1000, Russell Coker wrote:
On Thu, 28 Oct 2004 14:16, Jerone Young jerone@gmail.com wrote:
Try the FC3 X86_64 realease candidate (just look at archive on list) and see if everything works out. I wouldn't mind seeing how much slower the EMT64 is compared to a true X86_64 CPU like the AMD Opteron
How is an EMT64 not a "true X86_64"? They both run the same code.
not entirely; em64t is missing a few instructions, has one difference in behavior (which is unspecified so not a big deal) and has a few new instructions compared to AMD64/Opterons. It also lacks an iommu which is a bit of a pain if you have more than 4Gb of ram.
I'm not saying "worse" or "better" (both cpu platforms are quite competitive), but stating that there is no difference also isn't correct.
On Thu, Oct 28, 2004 at 07:18:36PM +1000, Russell Coker wrote:
On Thu, 28 Oct 2004 14:16, Jerone Young jerone@gmail.com wrote:
Try the FC3 X86_64 realease candidate (just look at archive on list) and see if everything works out. I wouldn't mind seeing how much slower the EMT64 is compared to a true X86_64 CPU like the AMD Opteron
How is an EMT64 not a "true X86_64"? They both run the same code.
Almost the same code. EM64T lacks prefetchw() and I believe also 3Dnow. It also has no iommu which may be real consideration if using > 3Gb of RAM.
Alan
On Wed, 27 Oct 2004, Jerone Young wrote:
Try the FC3 X86_64 realease candidate (just look at archive on list) and see if everything works out. I wouldn't mind seeing how much slower the EMT64 is compared to a true X86_64 CPU like the AMD Opteron :-)
I have both a dual 2.2 GHz Opteron and a dual 3.6 GHz EM64T (an HP xw8200). On all of the codes that we have tested, the Opteron is between 30% and 50% faster than the EM64T.
-s
On Wed, 27 Oct 2004, Sean Bruno wrote:
We just got a hold of a Dell Poweredge 2850 with dual EMT64's in it. I thought I would load of FC3 and give it a shot...Are there any issues that the list knows about that I should watch out for?
We have several here. I've not dinked with them in a couple of weeks, but the PERC card in them is a new revision, so you'll have to go with very recent (FC3 T3 or later) builds. I also ran into some issues where the megaraid drivers weren't getting loaded by anaconda / getting put into the initrd, so you had to do a bit of handholding of the install and first boot to get it to go. I think Jeremy got all those issues straightened out now though.
FC2 and older and Debian you can only install on them by building your own install kernel (because of the updated megaraid driver requirement). RHEL, SUSE 9.x, and SLES all have update disks available which support them.
I also noticed some oddities with the Fedora x86_64 kernel at the time on them (memory leaks) that went away when I switched to using kernel.org kernels. I've not had time to look into that further or to see if current Fedora x86_64 kernels behave better on them....
At any rate, no insurmountable issues, but perhaps not 100% "start the kickstart and walk away" (it may have reached that point by now, but it wasn't there 3 or so weeks ago).
later, chris
On Thu, 2004-10-28 at 12:09 -0400, Chris Ricker wrote:
I also noticed some oddities with the Fedora x86_64 kernel at the time on them (memory leaks) that went away when I switched to using kernel.org kernels. I've not had time to look into that further or to see if current Fedora x86_64 kernels behave better on them....
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=132947
A patch for this issue is known and should be applied to FC3 before it ships.
Dan
Can someone confirm this patch has been applied to RC5? I am dld'ing RC5 right now to load it up on my EMT64 machine for testing.
On Thu, 2004-10-28 at 09:17, Dan Williams wrote:
On Thu, 2004-10-28 at 12:09 -0400, Chris Ricker wrote:
I also noticed some oddities with the Fedora x86_64 kernel at the time on them (memory leaks) that went away when I switched to using kernel.org kernels. I've not had time to look into that further or to see if current Fedora x86_64 kernels behave better on them....
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=132947
A patch for this issue is known and should be applied to FC3 before it ships.
Dan
On Oct 30, 2004, Sean Bruno sean.bruno@dsl-only.net wrote:
Can someone confirm this patch has been applied to RC5?
It didn't make RC5. Last I heard, Ingo is still looking into the problem. The patch works for me, but I've no clue on what other implications the change may have, so I thought I wouldn't push it too hard.
A patch for this issue is known and should be applied to FC3 before it ships.
On Wed, 2004-10-27 at 17:21 -0700, Sean Bruno wrote:
We just got a hold of a Dell Poweredge 2850 with dual EMT64's in it. I thought I would load of FC3 and give it a shot...Are there any issues that the list knows about that I should watch out for?
There is a yum multilib arch bug which is fixed in CVS. Workaround - change /etc/rpm/platform to:
x86_64-redhat-linux
Paul
On Thu, 2004-10-28 at 15:11, Paul Nasrat wrote:
On Wed, 2004-10-27 at 17:21 -0700, Sean Bruno wrote:
We just got a hold of a Dell Poweredge 2850 with dual EMT64's in it. I thought I would load of FC3 and give it a shot...Are there any issues that the list knows about that I should watch out for?
There is a yum multilib arch bug which is fixed in CVS. Workaround - change /etc/rpm/platform to:
x86_64-redhat-linux
Working on getting this applied for fc3 final. If it can't get in it will be out as an update for fc3 soon after the release.
-sv
On Thu, 2004-10-28 at 20:11 +0100, Paul Nasrat wrote:
On Wed, 2004-10-27 at 17:21 -0700, Sean Bruno wrote:
We just got a hold of a Dell Poweredge 2850 with dual EMT64's in it. I thought I would load of FC3 and give it a shot...Are there any issues that the list knows about that I should watch out for?
There is a yum multilib arch bug which is fixed in CVS. Workaround - change /etc/rpm/platform to:
x86_64-redhat-linux
Paul
I just did an FC3RC2 and am experiencing this bug. One question - is my install fundamentally broken in some way? Like for instance, why do I have two versions of hal installed:
scatterbrain> alias rpmwhat rpm -q --qf '%{name}-%{version}-%{release}.%{arch}.rpm\n' scatterbrain> rpmwhat hal hal-0.4.0-5.i386.rpm hal-0.4.0-5.x86_64.rpm scatterbrain>
Thanks,
tjb
On Thu, 2004-10-28 at 15:16 -0400, Thomas J. Baker wrote:
scatterbrain> alias rpmwhat rpm -q --qf '%{name}-%{version}-%{release}.%{arch}.rpm\n' scatterbrain> rpmwhat hal hal-0.4.0-5.i386.rpm hal-0.4.0-5.x86_64.rpm scatterbrain>
If there is some i386-only package that depends on hal, and you've got that package installed, then you're going to get two copies of HAL, one for each arch.
For example, since openoffice.org isn't 64-bit yet, and since it links to GTK, you're going to have both an x86_64 and an i386 GTK installed in parallel. Welcome to multilib :)
Dan
On Thu, 2004-10-28 at 15:21 -0400, Dan Williams wrote:
On Thu, 2004-10-28 at 15:16 -0400, Thomas J. Baker wrote:
scatterbrain> alias rpmwhat rpm -q --qf '%{name}-%{version}-%{release}.%{arch}.rpm\n' scatterbrain> rpmwhat hal hal-0.4.0-5.i386.rpm hal-0.4.0-5.x86_64.rpm scatterbrain>
If there is some i386-only package that depends on hal, and you've got that package installed, then you're going to get two copies of HAL, one for each arch.
For example, since openoffice.org isn't 64-bit yet, and since it links to GTK, you're going to have both an x86_64 and an i386 GTK installed in parallel. Welcome to multilib :)
Dan
Thanks. That makes sense. I was thinking that I might have mis-clicked during the install and essentially installed both 32 and 64bit versions of everything. (Is that possible?) I did go for all compatibility options.
tjb