Distribution: Debian Sarge Hardware Environment: HP nx7000 notebook (Centrino-based), US model is called "x1000" Software Environment: Custom kernels compiled from vanilla sources (no patches) Problem Description: When using the acpi processor module, the notebook makes a very high-pitched sound. The problem seems to occur on all notebooks of that series, not only my own one. Couldn't check other centrino/pentium-m based notebooks. The sound stops as soon as frequency scaling is being used. This may, of course, be connected to some hardware-specialities. But the reason for this bug-report is: - The noise doesn't occur under "that other OS"[tm] - The noise really doesn't sound very healthy Maybe one of the kernel developers has such a notebook (seems to be a very popular model) and can help to investigate on this issue. Steps to reproduce: Use a 2.6 kernel with acpi processor support enabled and boot it on a nx7000/x1000 machine. Some log messages: Apr 6 20:26:12 nx7000 kernel: Initializing CPU#0 Apr 6 20:26:12 nx7000 kernel: Detected 1495.175 MHz processor. Apr 6 20:26:12 nx7000 kernel: Console: colour VGA+ 80x25 Apr 6 20:26:12 nx7000 kernel: Calibrating delay loop... 2981.88 BogoMIPS Apr 6 20:26:12 nx7000 kernel: Memory: 512552k/524096k available (1783k kernel code, 11156k reserved, 549k data, 280k init, 0k highmem) Apr 6 20:26:12 nx7000 kernel: Dentry-cache hash table entries: 65536 (order: 7, 524288 bytes) Apr 6 20:26:12 nx7000 kernel: Inode-cache hash table entries: 32768 (order: 6, 262144 bytes) Apr 6 20:26:12 nx7000 kernel: Mount-cache hash table entries: 8192 (order: 4, 65536 bytes) Apr 6 20:26:12 nx7000 kernel: Buffer-cache hash table entries: 32768 (order: 5, 131072 bytes) Apr 6 20:26:12 nx7000 kernel: Page-cache hash table entries: 131072 (order: 7, 524288 bytes) Apr 6 20:26:12 nx7000 kernel: CPU: L2 cache: 8192K Apr 6 20:26:12 nx7000 kernel: CPU: L3 cache: 5632K Apr 6 20:26:12 nx7000 kernel: Intel machine check architecture supported. Apr 6 20:26:12 nx7000 kernel: Intel machine check reporting enabled on CPU#0. Apr 6 20:26:12 nx7000 kernel: CPU: Intel(R) Pentium(R) M processor 1500MHz stepping 05 Apr 6 20:26:12 nx7000 kernel: Enabling fast FPU save and restore... done. Apr 6 20:26:12 nx7000 kernel: Enabling unmasked SIMD FPU exception support... done. Apr 6 20:26:12 nx7000 kernel: Checking 'hlt' instruction... OK. Apr 6 20:26:12 nx7000 kernel: Checking for popad bug... OK. Apr 6 20:26:12 nx7000 kernel: POSIX conformance testing by UNIFIX Apr 6 20:26:12 nx7000 kernel: enabled ExtINT on CPU#0 Apr 6 20:26:12 nx7000 kernel: ESR value before enabling vector: 00000000 Apr 6 20:26:12 nx7000 kernel: ESR value after enabling vector: 00000000 Apr 6 20:26:12 nx7000 kernel: Using local APIC timer interrupts. Apr 6 20:26:12 nx7000 kernel: calibrating APIC timer ... Apr 6 20:26:12 nx7000 kernel: ..... CPU clock speed is 1495.2057 MHz. Apr 6 20:26:12 nx7000 kernel: ..... host bus clock speed is 99.6803 MHz. Apr 6 20:26:12 nx7000 kernel: cpu: 0, clocks: 996803, slice: 498401 Apr 6 20:26:12 nx7000 kernel: CPU0<T0:996800,T1:498384,D:15,S:498401,C:996803> Apr 6 20:26:12 nx7000 kernel: mtrr: v1.40 (20010327) Richard Gooch (rgooch@atnf.csiro.au) Apr 6 20:26:12 nx7000 kernel: mtrr: detected mtrr type: Intel Apr 6 20:26:12 nx7000 kernel: PCI: PCI BIOS revision 2.10 entry at 0xf031f, last bus=3 Apr 6 20:26:12 nx7000 kernel: PCI: Using configuration type 1 Apr 6 20:26:12 nx7000 kernel: PCI: Probing PCI hardware Apr 6 20:26:12 nx7000 kernel: PCI: Using IRQ router default [8086/24cc] at 00:1f.0
Do you know whether the sound is from the CPU fan? Does it also happen on recent 2.6.6 (2.6.7-rc2) kernel? Does it happen if you stop Intel SpeedStep software in "the other OS"?
I can confirm a high pitched noise using the 2.6.6-1.397 Arjan kernel under ACPI using a Dell Inspiron 4150 with a Pentium 4 Mobile 1.6 GHz. This is not coming from the fan, rather the noise seems identical to the APM idle noise. According to the kernel mailing lists, this is produced by the new 1kHz clock speed of the 2.6 kernel. If I boot using apm=on apm=idle-threshold=100 the noise disappears. I have checked this makes no difference to power consumption and thermal properties of my system. Likewise the noise was not present under the 2.4 kernels. I don't know how to turn the noise off under ACPI.
> Do you know whether the sound is from the CPU fan? It's *definitely* not the CPU fan. It sounds a bit like the high-pitched tone, some old TVs and CRTs make. > Does it also happen on recent 2.6.6 (2.6.7-rc2) kernel? I'm now using a 2.6.6 kernel but with the C3 powerstate disabled because the centrino wlan driver doesn't like it (using the small patch from ipw2100.sf.net). No noise with that configuration! I'll try a vanilla 2.6.7 kernel when it's released. > Does it happen if you stop Intel SpeedStep software in "the other OS"? No
2.6.7 was released. How is it going?
Still happens on a vanilla 2.6.7 kernel.
Does set HZ to 100 help? --- linux-2.6.7/include/asm-i386/param.h 2004-06-17 09:34:28.000000000 +0800 +++ linux-2.6.7-new/include/asm-i386/param.h 2004-07-19 13:30:54.207125536 +0800 @@ -2,7 +2,7 @@ #define _ASMi386_PARAM_H #ifdef __KERNEL__ -# define HZ 1000 /* Internal kernel timer frequency */ +# define HZ 100 /* Internal kernel timer frequency */ # define USER_HZ 100 /* .. some user interfaces are in "ticks" */ # define CLOCKS_PER_SEC (USER_HZ) /* like times() */ #endif
This seems to help. The noise disappeared. David Lowe: Can you confirm this?
I am no longer able to get my machine to make the same noise under ACPI. I recall previously my machine loaded the "thermal" and "processor" kernel modules. However after an update (I'm now using Fedora Core 2), these modules are no longer there, and seem to have been replaced by asus_acpi, for me. There is still a faint continuous high-pitched whine, but it is also present under Windows XP, so I assume its something different, and regard this as a fix. Under APM the noise is still there, turning on and off dependent on processor load, so I am still using the apm=idle-threshold=100 boot parameter.
I also get the same noise on an HP nx7000 1500MHz Centrino, it is still there with 2.6.8.1. The noise seems to be connected to the cpu frequency and system load. If the load is high, and the cpu is working at a high frequency, the noise stops, when the load gets low again, the noise comes back. Under low system load, lowering the cpu speed with cpufreq also stops the noise. Setting HZ to 100 as in comment 6 solves the problem altogether.
note: FC2 includes the processor module as part of the kernel, which is why you can't unload it. Others have reported noise when their 2.6 system enters C3 power saving state in idle. It is good to know that HZ=100 also addresses the issue. You can observe C3 usage in /proc/acpi/processor/CPU0/power You can prevent C3 usage by plugging in a USB mouse, as this causes bus master activity that prevents entering C3. You could also run a low priority "cycle soaker" to make your system never enter idle, and see if that makes the noise go away. There are also a couple of ways to disable just C3 in the latest kernel -- see bug 3549. closing as a duplicate of 3406 "c3 causes noise". Please re-open if disabling C3 doesn't fix it. *** This bug has been marked as a duplicate of 3406 ***