Jump to content
323 posts in this topic

Recommended Posts

There are some options, boot priority, TMP etc but none CPU related. No wonder when BIOS code is 450 kB and for the other desktop computer is 3.5 MB.

 

btw CPU throttling works excellent on linux but not with speedstep module but acpi-cpufreq, which is weird in my opinion.

I'm downloading linux source right now and I'll try to figure this out.

Is it possible to add something like this to the kext?

 

/* Check to see if Enhanced SpeedStep is enabled, and try to
   enable it if not. */
rdmsr(MSR_IA32_MISC_ENABLE, l, h);

if (!(l & (1<<16))) {
	l |= (1<<16);
	dprintk("trying to enable Enhanced SpeedStep (%x)\n", l);
	wrmsr(MSR_IA32_MISC_ENABLE, l, h);

	/* check to see if it stuck */
	rdmsr(MSR_IA32_MISC_ENABLE, l, h);
	if (!(l & (1<<16))) {
		printk(KERN_INFO PFX "couldn't enable Enhanced SpeedStep\n");
		return -ENODEV;

Anybody got experiences with 10,4,10 ?

I installed Jas 10.4.8 afterwards i installed koolcal 10.4.10 and intel only Kernel on a Gigabyte P35 DS4.

Runs very good!

I dont want to screw up my system, but what i want is speedstep and S3 sleep if possible, my 4 cores are burning.

I only read about 10.4.9.

Satyr, I'll look into that code bit, but I'm pretty sure thats similar to how the kext does it right now. As for linux using the acpi module, as I was saying earlier, as of 10.4.8 osx has zero support for speedstep via acpi. I'm pretty sure the AppleACPIPlatform.kext had all the speestep acpi code stripped out of it. Its either directdrive or nothing.

 

Goofy, the sources have not been released for 10.4.10 so there is no speedstep support until that happens.

Thanks again. I have original kext source and I couldn't find anything similar inside although my c++ knowledge almost == 0. You probably have modified source though... I have now reinstalled windows and I'm dissecting power utility trying to figure out how it works.

Hey Satyr,

I took a look at the source again, I think I found something funny between whats in the source, and whats in your dmesg.

Can you confirm for me that in your dmesg you don't see a line like this:

ACPICPUThrottle: Adding CPU0

 

If not, I think I might have found something and I can build a kext tonight for you to try.

Your dmesg posted above had these lines instead:

ACPICPUThrottle: Adding CPU1

and

ACPICPUThrottle: Adding CPU2

 

Paulicat.

I noticed that too but looking at the IORegistry Explorer , CPUs are labeled exactly the same, 1 and 2. You mentioned you have Core Solo so that might explain the difference. btw, AppleIntelCPUpowr...kext identifies CPUs as well as 1 and 2. Can't see the logic behind it though..

Ok here it is Satyr.

chown -R root:wheel before you try to load it.

You dont need to put it in any special directory to see if it loads, but you do have to do the chown -R on it.

Let me know how it goes.

 

Do a sysctl -a | grep throt and post output please.

There will be other errors in this build as I lost my original sources so I have to remember a few things I changed.

Regardless, this has a fix for cpu1 in it and it should load and create the sysctl entries for throttle.

 

Paulicat

ACPICPUThrottle.kext.zip

Aug 22 00:40:57 -computer kernel[0]: ACPICPUThrottle: Adding CPU1

Aug 22 00:40:57 -computer kernel[0]: ACPICPUThrottle: Failed with code -536870212

Aug 22 00:40:57 -computer kernel[0]: ACPICPUThrottle: Adding CPU2

Aug 22 00:40:57 -computer kernel[0]: ACPICPUThrottle: Failed with code -536870212

Aug 22 00:40:57 s-computer kernel[0]: ACPICPUThrottle: No valid CPUs returned by ACPI! ....

 

-computer:~/Desktop $ sysctl -a | grep throt

kern.exec: unknown type returned

 

:<

 

Edit: tried with

 

1> 8.9.1 Darwin Kernel Version 8.9.1: Sun May 6 17:37:57 UZT 2007; made by ToH:xnu-792.18.15/BUILD/obj/RELEASE_I386 i386 i386

2> 8.9.1 Darwin Kernel Version 8.9.1: Wed Apr 18 21:20:01 EDT 2007; paulicat:xnu-792.18.15/BUILD/obj/RELEASE_I386 i386 i386

I have compiled Enhanced Speedstep (FreeBSD rip) driver to see if it would work and I get

 

Aug 22 02:42:11 s-computer kernel[0]: Processor claims to support Enhanced Speedstep, but is not recognized.

Aug 22 02:42:11 s-computer kernel[0]: Please update driver or contact the maintainer.

Aug 22 02:42:11 s-computer kernel[0]: cpu_vendor = GenuineIntel msr = 6130d2a06000d2a, BUSCLK = 64

 

Now, I'm in search for Core Duo datasheet so I can fill kext with appropriate info, if there is any in datasheet.

Is this possible or should I give up?

 

Where is ACPICPUThrottle.kext getting CPU powerstates from?

There is one other thing that I noticed. This particular CPU, T2250 is for OEMs only, it isn't mentioned anywhere in the datasheets and there is only one page on intel.com that mentions T2250.

 

And about the BIOS: there are 3 versions, 304, 305, 306. On 305 and 306, acpi-cpufreq on linux doesn't works so I'm using 304.

I also tried 305 and 306 with os x but to no avail. Well, screw the BIOS if I get this to work...somehow...

 

Now I have to reinstall windows to get correct FID and VID values :lol:

Just FYI, my Core Solo is also the OEM, its a T1350.

 

I've been looking into a few things, and I found one thing that bothers me, I don't think the C states are being used at all (at least on my notebook).

I loaded up Ubuntu on it and the temps are WAY less than under osx. I also ran Archlinux which had no cpu frequency scaling by default, so my cpu was running full speed, and the temps were STILL 12 degrees less than in OSX. :blink:

It could be linked to efi somehow, cause the kernel code doesn't really do any specific checking besides a cpuid check.

I'm not sure what to do about it right now though.

 

Paulicat

would like to get some advice from the experts here. I upgraded to 10.4.10 (with Pascal's) from 10.4.8 (Jas). I have an Intel Core Duo (Dell Inspiron E1505). I got the cputhrottle to work in 10.4.8. My sleep never worked. What are my chances to get sleep and/or speedstep to work under 10.4.10? I have looked through the posts and there are some conflicting results, some say sleep and speedstep still work, some say it shouldn't work.

 

any help appreciated

thanks!

Pauli: I give up for now. I just can't figure out some things. I tried to disassemble asus cpu clock freq utility and there are things that puzzle me.

Ubuntu recognizes 4 power states, one <1 GHz, 3 > 1 GHz, under windows there are at least 6 power states, the lowest being 269 MHz and with different voltages etc. I went through Intel datasheets, ACPI, linux kernel, os x kernel, I learned a lot but to no avail... I really appreciate your help, thanks for everything!

 

btw that cpu temp monitor is 10+ degrees C off but still, laptop runs hot and fan is blasting like a jet engine.

 

The great deceiv.. : There is now .10 kernel sources yet so speedstep can't be implemented (yet). Best thing is to experiment with .9 kerneles, there is plenty of them and you might get lucky.

I'm trying this on the uphuck 10.4.9i r3 install on a toshiba laptop, and still can't seem to get sleep to work. I had a couple questions I was hoping someone knowledgeable might be able to answer.

 

1) Would having a Celeron (which doesn't support SpeedStep) affect my ability to go to sleep mode? Obviously, I can do it in Windows, but I was wondering if the Mac version might be more tied together than that. I may eventually pop in a non-Celeron processor later, but not right now...

 

Edit: Actually, the CPU is somehow throttling down in Windows Vista, despite the fact that everything I've read (including the Intel docs) says that it does not support SpeedStep. Checking the processor speed using CPU-Z shows it throttling down to 800 Mhz-1 Ghz when I set the Vista Power scheme to Maximum Battery.

 

2) Would I possibly have better luck with sleep under 10.4.8, or should the compatability under 10.4.9 be the same as 10.4.8?

The Great Deceiv...: No speedstep at all in 10.4.10 until Apple releases the kernel sources.

Satyr: Sorry we couldn't get it working bro.

Clockworx: Which model celeron is it supposed to be? I've NEVER heard of a celeron of any kind have any sort of throttling. I don't trust Vista either. I didn't notice any difference between 4.8 and 4.9 regarding sleep, but I can't speak for everyone.

 

Paulicat

×
×
  • Create New...