Jump to content

Any X2 users experience stutter like syndroms?

shoman24v

By shoman24v,
in OSx86 10.4 (Tiger)

 Share
208 posts in this topic

Recommended Posts

macgirl
macgirl
Posted
Posted
CHUD Tools caused my system to crash at the end of the program install and now won't boot OS X.
Beware of CHUD kexts on Mac OS 10.4.1 they crash on AMD.

 

But you can boot in single user mode and detele them with rm /System/Library/Extensions/CHUD*

 

EDIT:

I just installed the most recent XCode (V 2.2.1) on 10.4.3 and as some of you the installer crashed, then after rebooted the system crashed, so I just removed the CHUD kexts and everything is good again.

 

When you install XCode you can select not to install CHUD utils.

Link to comment
Share on other sites
sigxcpu
sigxcpu
Posted
Posted
So with the chud tools, my system will run properly?

Yay!

 

Anyway, I keep getting kernel panics always when booting natively, any ideas why? I've tried searching the forums, but i didn't find anything :)

I somewhere saw a topic telling me that it's only the x2's that have this issue, correct?

 

No, with CHUD tools you ca disable a core.

Link to comment
Share on other sites
wesley
wesley
Posted
Posted

I had this problem posted very long time ago in Win2OSX forums. X2 3800+ @ 2.4GHz here. It's been a persistent problem for me, and it looks like there's enough X2 users in OSX86Project to see this kind of thread. Very interesting. I know from the experience that an identical setup with just single core A64 would not stutter at all. It's definitely a unique problem to AMD dual core setup and I'm quite intrigued to see that you can use CHUD to 'simulate' a single core setup by disabling a core and work around the stutter.

 

It raises an interesting point, though. Is this one of the reasons why Apple is using Intel chips instead of AMD? Or is it just Apple being ignorant of AMD chips and not having tested this configuration thoroughly? I mean, look at it. The very first Intel Macs are dual core equipped. Heh. If Apple shipped them with AMD chips this problem would not have occured or fixed quite early on.

Link to comment
Share on other sites
sigxcpu
sigxcpu
Posted
Posted (edited)

wesley, it's the same problem that crashes windows from time to time and generates "time goes back" on linux, I think: different clocked cores.

 

 

edit: "time goes back" is seen in mac os x as well. check your logs in /var/log

Edited by sigxcpu
Link to comment
Share on other sites
Bug
Bug
Posted
Posted

This sounds like a problem on the northbridge and the memory controller for the two cores. I was going to say DMA, but the fact turning off one of the processors resolves it means it's got to be something on the board or the chip itself not being driven right.

Link to comment
Share on other sites
sigxcpu
sigxcpu
Posted
Posted (edited)

once again. it has to do with the unsynched TSC in the dual core chips, for sure.

 

somebody with dual core, please try to use "idlehalt=0" as a boot parameter (along with -v or Graphics mode blah blah). Use F8 to boot to make it temporary and see if there is stuterring.

 

also, for boot-time disabling one core you can use cpus=1 parameter, I think. Try it and come back with the results, please. I don't have my dual core with my at the moment.

Edited by sigxcpu
Link to comment
Share on other sites
wesley
wesley
Posted
Posted

Neither of those options seem to particularly work. Would the disabling of core still work after removing the CHUD kexts? The latest version, as it is installed, crashes the system and the system won't boot. I probably have to manually delete the related kexts, as macgirl said.

Link to comment
Share on other sites
sigxcpu
sigxcpu
Posted
Posted

strange enough. I have an X2 3800+ and CHUD does not crash my system (even the latest 2.2 version of Xcode).

idlehalt and cpus are kernel parameters found in the sources.

idlehalt=0 does not send CPU to power save mode when there is no job to be done and cpus overrides detected number of CPUs.

 

in osfmk/i386/machine_routines.c:
if (idlehalt || others_active) {
			DBGLOG(cpu_handle, cpu_number(), MP_IDLE);
			cpu_idle_handler();
			DBGLOG(cpu_handle, cpu_number(), MP_UNIDLE);
	} else {
			__asm__ volatile("sti");
	}

 

and

 

in osfmk/i386/i386_init.c:

if (PE_parse_boot_arg("cpus", &cpus)) {
			if ((0 < cpus) && (cpus < max_ncpus))
					max_ncpus = cpus;
	}

Link to comment
Share on other sites
Careless
Careless
Posted
Posted (edited)
strange enough. I have an X2 3800+ and CHUD does not crash my system (even the latest 2.2 version of Xcode).

idlehalt and cpus are kernel parameters found in the sources.

idlehalt=0 does not send CPU to power save mode when there is no job to be done and cpus overrides detected number of CPUs.

so you're saying that if i run idlehalt=0 it should work without stuttering and without disabling one processor?

EDIT: wow... i actually did that and i am able to run consistent xbench runs without it stuttering.

 

how do i enable this in the com.apple.Boot.plist

 

do i just add it to the same line as the platform=XXXX string?

or how do i go about making a new line for it?

Edited by Careless
Link to comment
Share on other sites
Careless
Careless
Posted
Posted
somebody with dual core, please try to use "idlehalt=0" as a boot parameter

 

 

thanks sigxcpu

 

still seems to be working perfect on my side.

no kernel panics, both processors running strong

no stuttering. checked the cpu graph. says 2 cpu's active. and both are reporting somewhat the same usage. there's the off graph spike here and there, but its nothing like no idlehalt parameter active.

 

overall great.

i even xbenched at 100!!!

 

so for my com.apple.Boot.plist...

shall i change this

 

<string>platform=ACPI</string>

to

<string>platform=ACPI idlehalt=0</string>

 

or do i have to enter another kernel flag <key>?

if so, what shall i type there?

Link to comment
Share on other sites
sigxcpu
sigxcpu
Posted
Posted

<key>idlehalt</key>

<string>0</string>

 

or as you've said, better. it's a bsd parameter, not an mach parameter.

 

Please reboot many times and say if it's working, because I will switch back my P4 with the X2 (that means windows reinstall :angry: )

Link to comment
Share on other sites
Careless
Careless
Posted
Posted
<key>idlehalt</key>

<string>0</string>

 

or as you've said, better. it's a bsd parameter, not an mach parameter.

 

Please reboot many times and say if it's working, because I will switch back my P4 with the X2 (that means windows reinstall ;) )

 

so should i do it that way, or my way?

im confused. lol

this bsd/mach stuff is not my type of thing. sorry. :angry:

 

 

either works?

 

and right now i am running it fine, but as soon as iLife 06 stops "transfering" i will be sure to restart and check again =)

Link to comment
Share on other sites
Careless
Careless
Posted
Posted
Do it your way. I think it's better. Dunno if that XML file is parsed at the very boot of the system or it's loaded by the System.kext.

 

Anyway, test more and if you are OK, I will put back my X2 :)

 

i just installed iLife 06... but then i was forced to restart the computer. but my friend had picked me up from the house so i had to just shut it off. I will be turning her on when i get home with the variables in place in the com.apple.Boot.plist file, and i will see if it is working as planned. if not, i will try your method. and if that doesnt work, i will try from darwin itself.

 

i'll let you know!

 

long live le AMD! ;)

Link to comment
Share on other sites
Careless
Careless
Posted
Posted (edited)

open terminal

type: nano /library/preferences/systemconfiguration/com.apple.Boot.plist

type: password

 

enter nano

 

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
	<key>Kernel</key>
	<string>mach_kernel</string>
	<key>Kernel Flags</key>
	<string>platform=ACPI idlehalt=0</string>
	<key>Boot Graphics</key>
	<string>Yes</string>
	<key>Graphics Mode</key>
	<string>2048x1536x32</string>
</dict>
</plist>

 

WriteOut file (save)

 

type: exit

type: exit

 

done.

 

 

working good so far. im back home and running xBench all the while!

Edited by Careless
Link to comment
Share on other sites
temis
temis
Posted
Posted

I just tested the platform=ACPI idlehalt=0 ...

 

without this flag xbench gave me a 82 mark and stuttering problem

with this flag only xbench gave me a 42 mark still stuttering and very slow

with this flag and fn=3 gave me a 84 mark back AND NO STUTTERING little improvment in speed

 

Athlon 64 X2 4400+, 6600GT, 1Go RAM here

 

Careless, can you give me a simple procedure to turn on OpenGL

Link to comment
Share on other sites
shoman24v
shoman24v
Posted
  • Author
Posted (edited)
I just tested the platform=ACPI idlehalt=0 ...

 

without this flag xbench gave me a 82 mark and stuttering problem

with this flag only xbench gave me a 42 mark still stuttering and very slow

with this flag and fn=3 gave me a 84 mark back AND NO STUTTERING little improvment in speed

 

Athlon 64 X2 4400+, 6600GT, 1Go RAM here

 

Careless, can you give me a simple procedure to turn on OpenGL

You did what? did you set it like this

platform=ACPI fn=3?

 

What does fn=3 do?

Edited by shoman24v
Link to comment
Share on other sites
Careless
Careless
Posted
Posted
I just tested the platform=ACPI idlehalt=0 ...

 

without this flag xbench gave me a 82 mark and stuttering problem

with this flag only xbench gave me a 42 mark still stuttering and very slow

with this flag and fn=3 gave me a 84 mark back AND NO STUTTERING little improvment in speed

 

Athlon 64 X2 4400+, 6600GT, 1Go RAM here

 

Careless, can you give me a simple procedure to turn on OpenGL

 

 

when i said i turned openGL on, i meant in the xbench test.

OpenGL is not yet supported for nVidia cards.

 

with the OpenGL test on.. it scores lower, due to me getting a 50 (somehow... without even support) in that test, bringing down my average.

 

with the OpenGL test off... it scores at about 20 marks higher, due to it bypassing one of the tests that cause it to be lowered.

 

also.. whats this FN=3 thing? what's that do?

 

and so far, i have had amazing results with this idlehalt

its working fine for me. perhaps it has something to do with the kernels aswell

 

i only have the CPUID patch on my kernel, as that is all i need.

 

Hmm that didn't seem to make things that much better....

 

try removing the platform part. or changing ACPI to X86PC

 

case sensitive

Link to comment
Share on other sites
temis
temis
Posted
Posted (edited)

Here is my com.apple.Boot.plist

 

?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">

<plist version="1.0">

<dict>

<key>Kernel</key>

<string>mach_kernel</string>

<key>Kernel Flags</key>

<string>fn=3 platform=ACPI idlehalt=0</string>

<key>Boot Graphics</key>

<string>Yes</string>

<key>Graphics Mode</key>

<string>1280x1024x32@60</string>

<key>Quiet Boot</key>

<string>No</string>

<key>Timeout</key>

<string>8</string>

</dict>

</plist>

 

 

with openGL off in xbench I also have a better mark = 94 instead of 84 with OGL on

 

informations about fn=3 here :EDIT: Removed content violating DMCA don't know if fn thing is relevant to the stuttering problem since it's related to the cpu fan but it's working here (only rebooted once between com.apple.Boot.plist changes)

Edited by Metrogirl
Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...