I have two machines (home/work) both running FC4
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2
The lower speced machine is far,far,far quicker (gnome etc).
Anyone any ideas how this could be?
On Fri, 2005-06-24 at 11:48 +0100, Mike old wrote:
I have two machines (home/work) both running FC4
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2
The lower speced machine is far,far,far quicker (gnome etc).
Anyone any ideas how this could be?
Did you install the binary nvidia drivers on one of them? X performance looks a lot snappier using these drivers instead of the nv driver from xorg.
You can also check with hdparm if ultradma is enabled for the disk in the celeron. If it's not the machine will react a lot slower since the cpu is tied up by the disk i/o...
Klaasjan
On Fri, 2005-06-24 at 14:06 +0200, Klaasjan Brand wrote:
On Fri, 2005-06-24 at 11:48 +0100, Mike old wrote:
I have two machines (home/work) both running FC4
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2
The lower speced machine is far,far,far quicker (gnome etc).
Anyone any ideas how this could be?
Did you install the binary nvidia drivers on one of them? X performance looks a lot snappier using these drivers instead of the nv driver from xorg.
not installed on either
You can also check with hdparm if ultradma is enabled for the disk in the celeron. If it's not the machine will react a lot slower since the cpu is tied up by the disk i/o...
yep working on "slower" machine, ie: the higher specced one
Klaasjan
Klaasjan Brand writes:
On Fri, 2005-06-24 at 11:48 +0100, Mike old wrote:
I have two machines (home/work) both running FC4
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2
The lower speced machine is far,far,far quicker (gnome etc).
I'm also wondering the same thing.
My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM
The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
DMA is on on both.
Anyone any ideas how this could be?
Did you install the binary nvidia drivers on one of them? X performance looks a lot snappier using these drivers instead of the nv driver from xorg.
You can also check with hdparm if ultradma is enabled for the disk in the celeron. If it's not the machine will react a lot slower since the cpu is tied up by the disk i/o...
Klaasjan
-- fedora-test-list mailing list fedora-test-list@redhat.com To unsubscribe: http://www.redhat.com/mailman/listinfo/fedora-test-list
On Mon, Jun 27, 2005 at 09:16:35AM -0400, Janina Sajka wrote:
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2 The lower speced machine is far,far,far quicker (gnome etc).
I'm also wondering the same thing. My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
How are you measuring fast? For some workloads, that's completely to be expected.
(I'm still wondering which the original poster thinks is a "higher speced" machine....)
Matthew Miller writes:
On Mon, Jun 27, 2005 at 09:16:35AM -0400, Janina Sajka wrote:
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2 The lower speced machine is far,far,far quicker (gnome etc).
I'm also wondering the same thing. My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
How are you measuring fast? For some workloads, that's completely to be expected.
Latency of keystrokes. I press the key and the feedback is oo so slow.
In my case it's TTS feedback, not on screen, but same principle. Both machines have the exact same TTS board (ISA) and software drivers, both running FC4, but this was also true FC 3.
The 348Mb box is snappy. I press [char], I hear [char] echoed out the TTS. On the slow machine, I would warrant there's a delay of hundreds of ms. Not measured, but I'd warrant on the order of 400-700 ms.
(I'm still wondering which the original poster thinks is a "higher speced" machine....)
-- Matthew Miller mattdm@mattdm.org http://www.mattdm.org/ Boston University Linux ------> http://linux.bu.edu/ Current office temperature: 90 degrees Fahrenheit.
-- fedora-test-list mailing list fedora-test-list@redhat.com To unsubscribe: http://www.redhat.com/mailman/listinfo/fedora-test-list
On Mon, Jun 27, 2005 at 09:51:51AM -0400, Janina Sajka wrote:
My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
How are you measuring fast? For some workloads, that's completely to be expected.
Latency of keystrokes. I press the key and the feedback is oo so slow. In my case it's TTS feedback, not on screen, but same principle. Both machines have the exact same TTS board (ISA) and software drivers, both running FC4, but this was also true FC 3.
Hmmm. That should be next to instant on either system. I wonder if something is simply configured badly and introducing a delay.
I'm going to make the assumption that you're not running a heavyweight GUI environment such as GNOME or KDE. In that case, the greater RAM is probably not usually a factor at all. And the Celeron 700 is a very low-end chip with pretty low overall performance.
The 348Mb box is snappy. I press [char], I hear [char] echoed out the TTS. On the slow machine, I would warrant there's a delay of hundreds of ms. Not measured, but I'd warrant on the order of 400-700 ms.
On the other hand, either machine ought to be able to do better than that.
Matthew Miller writes:
On Mon, Jun 27, 2005 at 09:51:51AM -0400, Janina Sajka wrote:
My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
How are you measuring fast? For some workloads, that's completely to be expected.
Latency of keystrokes. I press the key and the feedback is oo so slow. In my case it's TTS feedback, not on screen, but same principle. Both machines have the exact same TTS board (ISA) and software drivers, both running FC4, but this was also true FC 3.
Hmmm. That should be next to instant on either system. I wonder if something is simply configured badly and introducing a delay.
I'm going to make the assumption that you're not running a heavyweight GUI environment such as GNOME or KDE. In that case, the greater RAM is probably not usually a factor at all. And the Celeron 700 is a very low-end chip with pretty low overall performance.
The 348Mb box is snappy. I press [char], I hear [char] echoed out the TTS. On the slow machine, I would warrant there's a delay of hundreds of ms. Not measured, but I'd warrant on the order of 400-700 ms.
On the other hand, either machine ought to be able to do better than that.
Yes, I think so.
The 348 box is runlevel 3 by default and is very snappy.
The 700 Mhz Celeron slowpoke is runlevel 5 by default, but I'm talking console in the above, and it's the same slowness regardless.
On the other hand, the slow poke responds very snappy over console ssh. It's only on its own consoles that it's bad. I can't test at the moment because it is setup as my pristine FC4 box as we try to put the Speakup Modified for FC4 together. So no speech there right now. I will try re display response when someone sighted shows up here.
Matthew Miller mattdm@mattdm.org http://www.mattdm.org/ Boston University Linux ------> http://linux.bu.edu/ Current office temperature: 89 degrees Fahrenheit.
-- fedora-test-list mailing list fedora-test-list@redhat.com To unsubscribe: http://www.redhat.com/mailman/listinfo/fedora-test-list
Janina Sajka janina@rednote.net wrote:
Matthew Miller writes:
On Mon, Jun 27, 2005 at 09:51:51AM -0400, Janina Sajka wrote:
My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
How are you measuring fast? For some workloads, that's completely to be expected.
Latency of keystrokes. I press the key and the feedback is oo so slow. In my case it's TTS feedback, not on screen, but same principle. Both machines have the exact same TTS board (ISA) and software drivers, both running FC4, but this was also true FC 3.
Hmmm. That should be next to instant on either system. I wonder if something is simply configured badly and introducing a delay.
Note that in my experience a good or bad motherboard can make worlds of difference between machines. I.e., an i386/25 being quite a bit faster than an i486dx4/100 when compiling kernels... one a Novell certified mobo (quite expensive in its time), the other a el-cheapo mobo from the PC store at the next corner.
Matthew Miller wrote:
On Mon, Jun 27, 2005 at 09:16:35AM -0400, Janina Sajka wrote:
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2 The lower speced machine is far,far,far quicker (gnome etc).
I'm also wondering the same thing. My fast machine is a Pentium 2 clocking at 348 Mhz with 192 Mb RAM The Celeron slowpoke clocks at 701 and sports 512 Mb RAM
How are you measuring fast? For some workloads, that's completely to be expected.
(I'm still wondering which the original poster thinks is a "higher speced" machine....)
To add to this thread.
On another list, there is discussion about FC4 being slower than Windows on the same hardware. Default installation and text entry being to slow to refresh the screen when typing in OOo.
Slow enough to be a show stopper.
On Fri, Jun 24, 2005 at 11:48:54AM +0100, Mike old wrote:
I have two machines (home/work) both running FC4 machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2 The lower speced machine is far,far,far quicker (gnome etc). Anyone any ideas how this could be?
Which of these machines is the "lower speced" one? And how are you measuring quicker? Startup time, "responsiveness", something else?
On Friday 24 June 2005 05:48, Mike old wrote:
I have two machines (home/work) both running FC4
machine 1 k6/2 500mhz 256Mb geforce4 machine 1 celeron 933mhz 256Mb geforce2
The lower speced machine is far,far,far quicker (gnome etc).
Anyone any ideas how this could be?
If you are building your own kernels, make sure that you don't accidentally turn on CONFIG_DEBUG_PAGEALLOC in your .config. You can check this by doing a 'grep CONFIG_DEBUG_PAGEALLOC /usr/src/linux/.config'. It should not be set.
I speak from recent experience with this matter; I enabled debug in my config so that I could use the sysrq key. Unknowingly, I turned on pagealloc debug also (the default if debug is enabled) - and it took me a little while and some effort to track the problem down. I checked the stock FC4 kernel, and it of course does not have this config set. But be wary if building your own kernel. -Dan
-
Dan A. Dickey -
Horst H. von Brand -
Janina Sajka -
Klaasjan Brand -
Matthew Miller -
mike -
Mike old -
Robin Laing