powernow-k7 (again) on 2.4.20

DISCUSSION emmel 2 participants 3 messages Jul 21, 2003 – Jul 24, 2003 2 messages not in the archive

Share & tools

emmel

author

2003-07-21 06:30 UTC #0 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I've posted that before, but I've serious trouble with cpufreq. It returns once loaded it makes the CPU speed drop from 1.4GHz to 800MHz (666 when throttled). If the kde info thingy is somewhat relyable at all that's for real - the bogomips drop as well. Returning to 1.4GHz seems impossible without restarting (I've tried unloading the module). Somehow cpufreq gets the stupid idea that the CPU is scaleable between 500 and 600MHz which has to be the reason for those odd results. We are talking about a Mobile Athlon XP 1600+ here, BTW (though it's reported by linux as a 1500+). If you ask me that's a serious problem. The odd thing, however, is that it didn't happen with the same kernel version before. Maybe it depends on some intel code to work correct (I used to have the CPU in the kernel set to PIII, now it's correctly at Athlon) or maybe it has something to do with the swsp patch that I *don't* have applied to my actual kernel. HELP!
The same odd behavior(sp?) appears with clean 2.5.72 and 2.6.0-test1 kernels.
- --
emmel <[email protected]>

Official AGC feedback maniac

"God is playing creatures - and we're the norns."

"A hundred dead are a tragedy - a hundred thousand are statistics."

"I guess you can call yourself lucky." - "I could, but Linda suits me a little better... :) Things called lucky tend to get hit by trucks."

GPG 1.2.1 signed
Fingerprint: 60B4 D8E3 9617 900C 6726 168F D677 5AAD D40F CCE7 Certserver : ldap://certserver.pgp.com

Hi, I'm a .sig virus. Just copy me to your .signature. And don't worry. -----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE/G4h91ndardQPzOcRAiOjAKCVDYOIMqAxMfRg6KzP10gJqbpIFwCg76XC H2bqYbCAIlW+8mX1VvkoquI=
=sYyM
-----END PGP SIGNATURE-----

Dominik Brodowski

2003-07-22 11:58 UTC #1 lore raw

Hi "emmel",

This is not a cpufreq bug, but a bug in the BIOS. It reports wrong values in the "PST" which should be dumped in "dmesg" when you load the powernow module [or once it is started, in case you built it into the kernel]. You might want to try up- or downgrading your BIOS or to override the BIOS table; some patches for this are in the list archives, Dave Jones intends to add a generic capability to override the PST to the powernow module in future.

Dominik

Quoted 17 lines · emmel

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
I've posted that before, but I've serious trouble with cpufreq. It returns once loaded it makes the CPU speed drop from 1.4GHz to 800MHz (666 when throttled). If the kde info thingy is somewhat relyable at all that's for real - the bogomips drop as well. Returning to 1.4GHz seems impossible without restarting (I've tried unloading the module). Somehow cpufreq gets the stupid idea that the CPU is scaleable between 500 and 600MHz which has to be the reason for those odd results. We are talking about a Mobile Athlon XP 1600+ here, BTW (though it's reported by linux as a 1500+). If you ask me that's a serious problem. The odd thing, however, is that it didn't happen with the same kernel version before. Maybe it depends on some intel code to work correct (I used to have the CPU in the kernel set to PIII, now it's correctly at Athlon) or maybe it has something to do with the swsp patch that I *don't* have applied to my actual kernel. HELP!
The same odd behavior(sp?) appears with clean 2.5.72 and 2.6.0-test1 kernels.

emmel

author

2003-07-24 05:10 UTC #4 lore raw

In reply to a message not in the archive

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 11 lines · you

Hi,
Quoted 8 lines
But I *had* it working correctly once. I guess it has something to do - the only things I changed are some kernel options and I didn't apply the swsp patch to the kernel I'm running now (it doesn't work anyway). However, I had it working correctly before and with a 2.4.20 kernel. It wouldn't strike me as odd if it hadn't. Well, in any case I have to know if loading the module really slows the comp down, or if the data is just a bit weird. Where's that bogomips tool that has been mentioned here now and again?

Quoted 8 lines

Unfortunately, I don't know where that bogomips tool is. Sorry. Never used it myself... I just keep asking others about it :-)
Now to the BUG() - have you upgraded your BIOS in the meantime? Had you patched the previous (working) kernel with the out-of-tree rtdvs patch which uses a different (and partly wrong) implementation of powernow-k7? Or did you patch it with one of the "override" patches available?
Dominik

As I said the only difference is that I didn't patch it with swsp (software suspemd) this time; I did before. And then there's still the possibility that some obscure kernel configuration was the cause. The kernel I'm running now has been completely reconfigured since I managed to lose my old .config for it :-( And AFAIR I applied exactly the same cpufreq patch (and acpi patch) to it. Still had them lying around in my source directory.
- --
emmel <[email protected]>