Why do I have to underclock my CPU?
 

Hi,

Since I got this machine, I had to underclock it for Linux. It works just fine with Windows at normal speed.
On Linux, the system freezes or programs suddenly close.

note: I have a Pentium 4 3 Ghz with Hyper Threading underclocked to 2.47 Ghz. I added the kernel argument acpismp=force. Some website advised me to do so with HT technologies.

Specs generated by sysinfo:

SYSTEM INFORMATION
Running Ubuntu Linux, the Kubuntu 8.10 (intrepid) release.
GNOME: unknown (unknown)
Kernel version: 2.6.27-9-generic (#1 SMP Thu Nov 20 21:57:00 UTC
2008)
GCC: 4.3.2 (i486-linux-gnu)
Xorg: unknown (24 October 2008 08:00:16AM)
Hostname: robert-desktop
Uptime: 0 days 0 h 8 min

CPU INFORMATION
GenuineIntel, Intel(R) Pentium(R) 4 CPU 3.00GHz
Number of CPUs: 2
CPU clock currently at 2474.994 MHz with 512 KB cache
Numbering: family(15) model(2) stepping(9)
Bogomips: 4949.98
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe pebs bts cid xtpr

MEMORY INFORMATION
Total memory: 1770 MB
Total swap: 2863 MB

STORAGE INFORMATION
SCSI device - scsi1
Vendor: AOPEN
Model: DUW1608/ARR
SCSI device - scsi1
Vendor: TSSTcorp
Model: CDW/DVD TS-H492A
SCSI device - scsi2
Vendor: ATA
Model: MAXTOR STM380215
SCSI device - scsi3
Vendor: ATA
Model: SAMSUNG HD160JJ/

HARDWARE INFORMATION
MOTHERBOARD
Host bridge
Intel Corporation 82865G/PE/P DRAM Controller/Host-Hub Interface (rev 02)
Subsystem: ASUSTeK Computer Inc. Device 80f2
PCI bridge(s)
Intel Corporation 82865G/PE/P PCI to AGP Controller (rev 02)
Intel Corporation 82801 PCI Bridge (rev c2)
USB controller(s)
Intel Corporation 82801EB/ER (ICH5/ICH5R) USB UHCI Controller #1 (rev 02)
Intel Corporation 82801EB/ER (ICH5/ICH5R) USB UHCI Controller #2 (rev 02)
Intel Corporation 82801EB/ER (ICH5/ICH5R) USB UHCI Controller #3 (rev 02)
Intel Corporation 82801EB/ER (ICH5/ICH5R) USB UHCI Controller #4 (rev 02)
Intel Corporation 82801EB/ER (ICH5/ICH5R) USB2 EHCI Controller (rev 02) (prog-if 20)
ISA bridge
Intel Corporation 82801EB/ER (ICH5/ICH5R) LPC Interface Bridge (rev 02)
IDE interface
Intel Corporation 82801EB (ICH5) SATA Controller (rev 02) (prog-if 8f [Master SecP SecO PriP PriO])
Subsystem: ASUSTeK Computer Inc. Device 80a6

GRAPHIC CARD
VGA controller
nVidia Corporation NV18 [GeForce4 MX 440 AGP 8x] (rev c1)
Subsystem: ASUSTeK Computer Inc. Device 810b

SOUND CARD

NETWORK
Network controller
RaLink RT2561/RT61 802.11g PCI
Subsystem: Linksys Device 0055
Ethernet controller
Marvell Technology Group Ltd. 88E8001 Gigabit Ethernet Controller (rev 13)
Subsystem: ASUSTeK Computer Inc. Device 811a

NVIDIA GRAPHIC CARD INFORMATION
Model name: GeForce4 MX 440 with AGP8X
Card Type: AGP 8x
Video RAM: 64 MB
GPU Frequency: 274 MHz
Driver version: NVIDIA Linux x86 Kernel Module 96.43.09 Mon Oct 27 14:23:30 PST 2008


Thanks in advance,
A Linux fan

oh, I forgot to mention: The temperature is just fine. I checked the hardware monitor and the CPU was about 40 degrees Celcius. When running at 2.47Ghz, it was about 38 degrees Celsius.

That is a bit odd.

I have almost the same CPU in this desktop box.
Mine is the 1024k cache version:

model name : Intel(R) Pentium(R) 4 CPU 3.00GHz
stepping : 1
cpu MHz : 2995.083
cache size : 1024 KB
bogomips : 5993.52

At default FSB speed this motherboard is clocked a few mhz below
3000 as you can see, but it is one of the few SuperMicro(P4SCE) boards that
actually has overclocking options in the bios, and I have had no
problems when running overclocked a bit.

What I would look at next if I was you, is the system memory.
Run memtest86 or similar test on it to see if any errors.

I have found linux to be more sensitive to memory errors than winblows OS's.

Good luck.

Thanks for your quick reply,

I ran a test a few days ago, no errors found. When I "overclock" my cpu to default manually, it doesn't even show a boot screen and just turns off. When I set it back to default, non-manual, it starts normally but runs very unstable.

Does the problem go away if you remove this? Is this a recently added kernel parameter? I do not see it listed in "Documentation/kernel-parameters.txt" for the 2.6.25.5 kernel source I have here. It is not in the latest kernel source on the Linux Cross-Reference site, either.

The only option I see that might apply to your situation is "acpi=ht", which executes only enough ACPI to enable "Hyper Threading".

Remove your WiFi card and try increasing to 3 GHz. You might also need to over volt the processor a little. Also turn off automatic over clock in the BIOS if it has such option. Over volt the memory helps too to keep the system reliable and stable 24/7. Buying a new quality power supply and upgrading the BIOS might have to be done.

Just tried to over volt my cpu and memory. My bios doesnt allow me to go above 2.6Ghz manually. I just get an error and everything is set back to defaults.

My power supply is also pretty new, I recently replaced it so I guess it's not the problem.

I think the problem lies somewhere in my BIOS. How do I upgrade my bios if I have to? Can I just replace it?

I remember running fedora very stable at 3Ghz.. This was an i686 (is it possible?) kernel or something like that. They said this would be OK for pentium 4 or higher. It did often crash at boot time however.
Maybe I need a different kernel version?

A new power supply does not mean it is a quality unit. Power supply brands that I recommend are Seasonic and Enermax.

If you are unsure how to reliable update the BIOS, replacing the BIOS could be done. ASUS does not make this easy to do since you have to dig through their site to choose the BIOS for your motherboard model. Removing the BIOS needs a special tool which makes it easy, but a flat head screw driver and patience is all you need. You will have to note where is pin one for either DIP, millipede like chip, or PLCC, a square chip with pins for each side. Disconnect the system from the AC and remove the battery before replacing the BIOS chip.

You should remove the WiFi adapter and try again. Not all WiFi adapters complies to specifications on how much EMI and RFI they generate. Removing these adapters can bring back stability when clocking the system at high speed. If the processor can be clocked at 3 GHz and Windows shows this, then something is causing an instability. Windows is not an efficient OS, so it only uses a fraction of the processor resources to run it and takes longer to finish its tasks. Linux is more efficient, so it uses more resources to do its tasks in shorter amount of time.

From your limited information that you have given, I can not suggest anything else. The motherboard model is not posted and the temperature the BIOS displays for the processor is also not posted.

An i686 code is just an instruction model. A Pentium 4 is an i786. It can handle executing any instruction model up to i786. A Pentium Pro, Pentium MMX, Pentium II, Pentium III, Core 2 are i686.

Are you still passing the "acpismp=force" kernel parameter? I googled it, and it seems to only apply to 2.4 (and 2.5) kernels.

Did you try the "acpi=ht" kernel parameter? How about "acpi=noht"?

If Windows is able to run without any problems, I would assume the problem is with Linux somewhere. It should be easy enough to try a different kernel. Building your own kernel using the latest stable release from kernel.org would be the best solution.

Electo is on the right track.

If it is not faulty memory, motherboard or PSU is next to look at.

Post bios hardware monitor values if you can. Output from sensors also
if you have that installed.

I got several P4 HT cpu systems running,(478 socket), and no extra kernel parameter
values have been needed for full smp operation at either default or
overclocked speeds.

uname -a
Linux p43000 2.6.26.8-57.fc8 #1 SMP Thu Dec 18 19:19:45 EST 2008 i686 i686 i386 GNU/Linux

As David suggested, if the hardware seems ok, try a reinstall with newer kernel.

The hardware monitor is showing me a lot of fans in the first place, all running fine.

My CPU is at about 40 degrees celcius, the same as when running at 3 Ghz.
My motherboard is about 35 degrees celcius.

I'll try installing a new kernel. Which kernel would you suggest?

I tried the acpi=ht, with no results.

I always download the latest stable release from kernel.org and build it from source.

I just installed the gnome desktop environment, which seems to run much more stable at 3Ghz. Only one firefox crash occured in a few days. Maybe this fixed my problem. I'll stick to gnome anyway, it just feels better :)

Thanks for help anyway, I really appreciate it.