Re: [gentoo-user] New CPU, what needs to be done?
On Freitag, 10. August 2007, Zac wrote: The reason I asked is that I have been running this on my AMDX2 systems since it has been available with no noticeable degradation, and possibly a little better responsiveness. I thought maybe you had some experience with it that would contradict my experiences. I always assumed that the If unsure say no/yes in the kernel was related to being unfamiliar with the hardware you where configuring the kernel for. That said, all I can really say is that it works for me. I had it 'in' once too. And yes, it worked. But I did not see any improvements, so I took it out. I deactivate everything I do not need ... The big improvements came, when I started using Ingo Molnar's cfs ;) -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
On Freitag, 10. August 2007, Florian Philipp wrote: Am Freitag 10 August 2007 16:57 schrieb Volker Armin Hemmann: On Freitag, 10. August 2007, Zac wrote: Why don't you want mulit-core scheduler support? read the help text. CONFIG_SCHED_MC: Multi-core scheduler support improves the CPU scheduler's decision making when dealing with multi-core CPU chips at a cost of slightly increased overhead in some places. If unsure say N here. so you have an option that improves the decisions, but penalize you with more overhead Who guarantees you, that the cost is not higher than the improvements? + 'if unsure say N here'. multi-core CPU chip sounds like an appropriate description for AMD X2 processors. Are you sure you don't mean SMT or NUMA? no. -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
On Freitag, 10. August 2007, Zac wrote: Why don't you want mulit-core scheduler support? read the help text. -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
Why don't you want mulit-core scheduler support? On 8/10/07, Volker Armin Hemmann [EMAIL PROTECTED] wrote: On Freitag, 10. August 2007, Grant wrote: I just upgraded from an AMD64 Sempron to an AMD64 Athlon X2. What do I need to do in software to accommodate the new CPU? Do I need to enable something in the kernel to get the dual cores working? I've read something about support for AMD's Cool and Quiet technology too. make menuconfig Processor Type and Features you don't need numa or multi core scheduler support. Just choose Symmetric multi-processing support. And cool'n'quiet. That is something the sempron can do too its under the power managment topic. Just use menuconfig. -- [EMAIL PROTECTED] mailing list -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
On Freitag, 10. August 2007, Grant wrote: I just upgraded from an AMD64 Sempron to an AMD64 Athlon X2. What do I need to do in software to accommodate the new CPU? Do I need to enable something in the kernel to get the dual cores working? I've read something about support for AMD's Cool and Quiet technology too. make menuconfig Processor Type and Features you don't need numa or multi core scheduler support. Just choose Symmetric multi-processing support. And cool'n'quiet. That is something the sempron can do too its under the power managment topic. Just use menuconfig. -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
The reason I asked is that I have been running this on my AMDX2 systems since it has been available with no noticeable degradation, and possibly a little better responsiveness. I thought maybe you had some experience with it that would contradict my experiences. I always assumed that the If unsure say no/yes in the kernel was related to being unfamiliar with the hardware you where configuring the kernel for. That said, all I can really say is that it works for me. -Zac On 8/10/07, Volker Armin Hemmann [EMAIL PROTECTED] wrote: On Freitag, 10. August 2007, Florian Philipp wrote: Am Freitag 10 August 2007 16:57 schrieb Volker Armin Hemmann: On Freitag, 10. August 2007, Zac wrote: Why don't you want mulit-core scheduler support? read the help text. CONFIG_SCHED_MC: Multi-core scheduler support improves the CPU scheduler's decision making when dealing with multi-core CPU chips at a cost of slightly increased overhead in some places. If unsure say N here. so you have an option that improves the decisions, but penalize you with more overhead Who guarantees you, that the cost is not higher than the improvements? + 'if unsure say N here'. multi-core CPU chip sounds like an appropriate description for AMD X2 processors. Are you sure you don't mean SMT or NUMA? no. -- [EMAIL PROTECTED] mailing list -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
Am Freitag 10 August 2007 16:57 schrieb Volker Armin Hemmann: On Freitag, 10. August 2007, Zac wrote: Why don't you want mulit-core scheduler support? read the help text. CONFIG_SCHED_MC: Multi-core scheduler support improves the CPU scheduler's decision making when dealing with multi-core CPU chips at a cost of slightly increased overhead in some places. If unsure say N here. Symbol: SCHED_MC [=y] Prompt: Multi-core scheduler support Defined at arch/x86_64/Kconfig:324 Depends on: SMP Location: - Processor type and features - Symmetric multi-processing support (SMP [=y]) multi-core CPU chip sounds like an appropriate description for AMD X2 processors. Are you sure you don't mean SMT or NUMA? pgpd7Xpv3Sxaw.pgp Description: PGP signature
Re: [gentoo-user] New CPU, what needs to be done?
The reason I asked is that I have been running this on my AMDX2 systems since it has been available with no noticeable degradation, and possibly a little better responsiveness. I thought maybe you had some experience with it that would contradict my experiences. I always assumed that the If unsure say no/yes in the kernel was related to being unfamiliar with the hardware you where configuring the kernel for. That said, all I can really say is that it works for me. I had it 'in' once too. And yes, it worked. But I did not see any improvements, so I took it out. I deactivate everything I do not need ... The big improvements came, when I started using Ingo Molnar's cfs ;) Thanks for the info. Running great. CFS sounds interesting. I'm reading that will be in the main kernel for 2.6.23. Still with hardened-sources-2.6.20 here, but looking forward to it. Big difference? - Grant -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
On Freitag, 10. August 2007, Grant wrote: The reason I asked is that I have been running this on my AMDX2 systems since it has been available with no noticeable degradation, and possibly a little better responsiveness. I thought maybe you had some experience with it that would contradict my experiences. I always assumed that the If unsure say no/yes in the kernel was related to being unfamiliar with the hardware you where configuring the kernel for. That said, all I can really say is that it works for me. I had it 'in' once too. And yes, it worked. But I did not see any improvements, so I took it out. I deactivate everything I do not need ... The big improvements came, when I started using Ingo Molnar's cfs ;) Thanks for the info. Running great. CFS sounds interesting. I'm reading that will be in the main kernel for 2.6.23. Still with hardened-sources-2.6.20 here, but looking forward to it. Big difference? short: yes long answer: not in gaming (the difference in ut2004 is not very big. Maximum FPS are a little bit lower, minimum FPS a bit higher ... ) but everywhere else is. Desktop is snappier than the 'old' scheduler. Compiling stuff in the background? Even with PORTAGE_NICENESS=19 it was very obvious that something was happening. with cfs I did an emerge -e world (after switching to gcc 4.2) some days ago - emerge running with a niceness of 0 and most of the time I did not even realize that something was happening. I needed to switch to the vt (or use htop) to check that everything was working as it should. Really, from my POV cfs is just great. It does not help with slow-as-hell-swap but that is a completly different story. -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
The reason I asked is that I have been running this on my AMDX2 systems since it has been available with no noticeable degradation, and possibly a little better responsiveness. I thought maybe you had some experience with it that would contradict my experiences. I always assumed that the If unsure say no/yes in the kernel was related to being unfamiliar with the hardware you where configuring the kernel for. That said, all I can really say is that it works for me. I had it 'in' once too. And yes, it worked. But I did not see any improvements, so I took it out. I deactivate everything I do not need ... The big improvements came, when I started using Ingo Molnar's cfs ;) Thanks for the info. Running great. CFS sounds interesting. I'm reading that will be in the main kernel for 2.6.23. Still with hardened-sources-2.6.20 here, but looking forward to it. Big difference? short: yes long answer: not in gaming (the difference in ut2004 is not very big. Maximum FPS are a little bit lower, minimum FPS a bit higher ... ) but everywhere else is. Desktop is snappier than the 'old' scheduler. Compiling stuff in the background? Even with PORTAGE_NICENESS=19 it was very obvious that something was happening. with cfs I did an emerge -e world (after switching to gcc 4.2) some days ago - emerge running with a niceness of 0 and most of the time I did not even realize that something was happening. I needed to switch to the vt (or use htop) to check that everything was working as it should. Really, from my POV cfs is just great. It does not help with slow-as-hell-swap but that is a completly different story. Nice. I game not at all so that's no problem. This new CPU has (I think) really revealed how slow my hard disk is. It's a 320GB Seagate SATA2 (3GB/s) so it's not slow as HDs go, but it can't keep up with the CPU and memory (2GB DDR800). Whenever a video stutters or something takes a little longer than it should, the xfce4 CPU graph shows it wasn't even close to maxed out, and the memory usage is at like 300MB. I'm sure neither of those are incredibly accurate, but I think they do reveal something. Does that sounds like an IO problem to you? I could get another HD and set up RAID at some point I guess. Hopefully the CFS scheduler you're talking about will help. - Grant -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] New CPU, what needs to be done?
On Samstag, 11. August 2007, Grant wrote: The reason I asked is that I have been running this on my AMDX2 systems since it has been available with no noticeable degradation, and possibly a little better responsiveness. I thought maybe you had some experience with it that would contradict my experiences. I always assumed that the If unsure say no/yes in the kernel was related to being unfamiliar with the hardware you where configuring the kernel for. That said, all I can really say is that it works for me. I had it 'in' once too. And yes, it worked. But I did not see any improvements, so I took it out. I deactivate everything I do not need ... The big improvements came, when I started using Ingo Molnar's cfs ;) Thanks for the info. Running great. CFS sounds interesting. I'm reading that will be in the main kernel for 2.6.23. Still with hardened-sources-2.6.20 here, but looking forward to it. Big difference? short: yes long answer: not in gaming (the difference in ut2004 is not very big. Maximum FPS are a little bit lower, minimum FPS a bit higher ... ) but everywhere else is. Desktop is snappier than the 'old' scheduler. Compiling stuff in the background? Even with PORTAGE_NICENESS=19 it was very obvious that something was happening. with cfs I did an emerge -e world (after switching to gcc 4.2) some days ago - emerge running with a niceness of 0 and most of the time I did not even realize that something was happening. I needed to switch to the vt (or use htop) to check that everything was working as it should. Really, from my POV cfs is just great. It does not help with slow-as-hell-swap but that is a completly different story. Nice. I game not at all so that's no problem. This new CPU has (I think) really revealed how slow my hard disk is. It's a 320GB Seagate SATA2 (3GB/s) so it's not slow as HDs go, but it can't keep up with the CPU and memory (2GB DDR800). Whenever a video stutters I only see stutters, when swap is involved. -- [EMAIL PROTECTED] mailing list