Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Processor Speed (Actual vs Reported)
- Thread starter crazzyeddie
- Start date
- Sort by reaction score
OK, just so we're all on the same page, I too was able to gain full processor speed by resetting the PMU on my 12" 867 MHz PB.
However a clean reboot now leaves me with this:
sysctl hw.cpufrequency
hw.cpufrequency: 533333332
WTF ?
Does anybody know what the heck is going on?
Please, let's continue this thread. If anyone learns an answer or a reason why, absolutely do us all a favor and post your findings.
Thanks in advance,
Logicat
However a clean reboot now leaves me with this:
sysctl hw.cpufrequency
hw.cpufrequency: 533333332
WTF ?
Does anybody know what the heck is going on?
Please, let's continue this thread. If anyone learns an answer or a reason why, absolutely do us all a favor and post your findings.
Thanks in advance,
Logicat
Try it again, i found that after re-setting it the 2nd time, it seems to have stuck..
As for the guy who said it didn't make any effect, are you sure you resetted it?your computer should have its date set back to 1969 (apple's epoch it seems). If the date didn't need to be set, you didn't reset your PMU actually.
If my system goes back to 667, i'm going to an xbench off a fresh boot and then re-set it and run xbench again, and let you guys know if it actually affects performance any.
As for the guy who said it didn't make any effect, are you sure you resetted it?your computer should have its date set back to 1969 (apple's epoch it seems). If the date didn't need to be set, you didn't reset your PMU actually.
If my system goes back to 667, i'm going to an xbench off a fresh boot and then re-set it and run xbench again, and let you guys know if it actually affects performance any.
I've been searching around online for other reports of this. Seems like there's quite a buzz in certain circles (see some URLs below).
In my case, I'm convinced that there's a noticable difference, most especially w/regards to the windowing system, GUI animations and menus.
I ran one XBench run, shutdown and reset my PMU, started up and ran 30 more XBench tests. The tests weren't so far off. However, as I've read around the 'net, some people are noticing speed degradation with particular apps, and some are not.
Others say it's a bug in the code that references that CPU frequency value.
Still others say that regardless of where the bug is, that particular applications and processes will reference that value when running. Hence, some things WILL in fact, run more slowly.
I don't know where this is going, but I wish someone who knew something would speak up. I'm tired of all the closet hardware designers spouting nonesense.
Where are the hard-core CPU engineers that can answer this question for us?
Check some of these links below for more information:
http://www.macmegasite.com/modules.php?name=News&file=article&sid=1093
http://fsteele.dyndns.org/archive/cat_apple_powerbooks.html
http://spiny.com/forum/viewtopic.php?t=86
http://discussions.info.apple.com/WebX?13@208.asPXaYxukuQ.0@.599a4e72/42
http://3650anda12inch.blogspot.com/
http://laughingmeme.org/archives/001432.html#001432
In my case, I'm convinced that there's a noticable difference, most especially w/regards to the windowing system, GUI animations and menus.
I ran one XBench run, shutdown and reset my PMU, started up and ran 30 more XBench tests. The tests weren't so far off. However, as I've read around the 'net, some people are noticing speed degradation with particular apps, and some are not.
Others say it's a bug in the code that references that CPU frequency value.
Still others say that regardless of where the bug is, that particular applications and processes will reference that value when running. Hence, some things WILL in fact, run more slowly.
I don't know where this is going, but I wish someone who knew something would speak up. I'm tired of all the closet hardware designers spouting nonesense.
Where are the hard-core CPU engineers that can answer this question for us?
Check some of these links below for more information:
http://www.macmegasite.com/modules.php?name=News&file=article&sid=1093
http://fsteele.dyndns.org/archive/cat_apple_powerbooks.html
http://spiny.com/forum/viewtopic.php?t=86
http://discussions.info.apple.com/WebX?13@208.asPXaYxukuQ.0@.599a4e72/42
http://3650anda12inch.blogspot.com/
http://laughingmeme.org/archives/001432.html#001432
Perhaps it is a conspiracy between Apple and the Federal government to hide evidence of the existence of aliens. Too many Mac users are running Seti clients and now we are close to discovering the truth, so Apple must regulate CPU speed to slow down our progress to keep us masses in the dark from the truth.
I just reset the pmu and it didnt work. It even reset the date and it still reports 533mhz. 
I'd just like to know whats going on.
I'd just like to know whats going on.I've just put some feelers out for more enlightened commentary. Hopefully I can get in contact with an engineer who won't have to speculate, but simply will explain the facts.
When, and if, I succeed I will be sure to add to this thread.
Best,
Logicat
When, and if, I succeed I will be sure to add to this thread.
Best,
Logicat
A question about duallies
So if I have one app running on my dual 1.25ghz Powermac, say an encoder, and that is the only app I will be using during an encoding, will it be running at exactly the same speed as a single 1.25ghz powermac? Can't an application optimised for dual processors make use of the two processors? It's just that my powerbook 1.33ghz is noticably slower when encoding quicktimes in Cleaner than on my dual system.
Off topic, is there an alternative to Cleaner? Something a little faster. And I don't mean Quicktime Pro. I need something with more settings like Cleaner.
Terminal Result:
hw.model: PowerBook5,3
1333333328
I'm also short of 5hz!!
So if I have one app running on my dual 1.25ghz Powermac, say an encoder, and that is the only app I will be using during an encoding, will it be running at exactly the same speed as a single 1.25ghz powermac? Can't an application optimised for dual processors make use of the two processors? It's just that my powerbook 1.33ghz is noticably slower when encoding quicktimes in Cleaner than on my dual system.
Off topic, is there an alternative to Cleaner? Something a little faster. And I don't mean Quicktime Pro. I need something with more settings like Cleaner.
Terminal Result:
hw.model: PowerBook5,3
1333333328
I'm also short of 5hz!!
Re: A question about duallies
1.) Most likely not, and 2.) of course! There wouldn't be much of a point in selling two processors in one machine if you couldn't use them, they would be better off in another computer.
Mac OS X is MP savvy, so that helps to give an edge overall to tasks with dual processors, but to really get the benefits, you need those programs that are optimized to exploit duals to get the real performance potential from your system.
Xbench results from my TiBook, pre and post PMU reset
Ok, here goes. I'm not much of a techie, so I don't know what exactly this means but it looks like things got a little faster after the PMU reset according to xbench.
I closed all apps except finder and xbench and ran it, then reset the PMU, and checked the hw.cpufrequency. It now reports correctly, although my clock was not reset possibly due to the fact that I have it set to check the network server for the time which I assume it does when I log in. I then quit Terminal and ran xbench again. Here are the results.
Ok, here goes. I'm not much of a techie, so I don't know what exactly this means but it looks like things got a little faster after the PMU reset according to xbench.
I closed all apps except finder and xbench and ran it, then reset the PMU, and checked the hw.cpufrequency. It now reports correctly, although my clock was not reset possibly due to the fact that I have it set to check the network server for the time which I assume it does when I log in. I then quit Terminal and ran xbench again. Here are the results.
Attachments
Mactastic, the only problem with XBench is that single results aren't very useful. Only by repeatedly running the test can you arrive a a useful measure of performance.
Try running the XBench tests five to ten more times (without powering down, just run them consecutively).
You'll find that the results may vary quite a bit.
Best,
Logicat
Try running the XBench tests five to ten more times (without powering down, just run them consecutively).
You'll find that the results may vary quite a bit.
Best,
Logicat
Ok. I'll give it a try. I don't really notice it as feeling any faster, but I am curious to know where the problem is coming from and whether it has any affect on the system.
hw.cpumonitor is still reporting correctly though, I guess I'll have to monitor it and see when or if it reverts to 667.
hw.cpumonitor is still reporting correctly though, I guess I'll have to monitor it and see when or if it reverts to 667.
In my case, it will drop back to 533 whenever I boot up the system. This is true whether I cold-boot or warm-boot.
Oh well. I'm going to stop thinking about it and just get my work done.
Best,
Logicat
Oh well. I'm going to stop thinking about it and just get my work done.
Best,
Logicat
You people wanna stop screwing up with hardware settings if you don't know what the hell you are doing!!!? Now I suggest you bring it to apple store and have them take a look at it...
Play nice Rezet... did you get up on the wrong side of the bed this morning?
I've been a macintosh technology consultant, systems engineer and more since '89. I think I'm qualified to try rebooting my own hardware.
I'm posting up my activity because I want to help get to the bottom of this by posting helpful comments.
Maybe you'd like to do the same.
BTW, I don't believe a retail store is going to solve my problem.
I've been a macintosh technology consultant, systems engineer and more since '89. I think I'm qualified to try rebooting my own hardware.
I'm posting up my activity because I want to help get to the bottom of this by posting helpful comments.
Maybe you'd like to do the same.
BTW, I don't believe a retail store is going to solve my problem.
Alright then. I could care less really. Mine shows exactly the right numbers. It's really your problem.. Pffffhhhh!
rezet dont be an azz..
i did several xbench benchmarks before/after the numbers were the wrong/right ones.. as far as i can tell, the system runs nearly the same.. if anything it seems to run slightly faster with the lower numbers, but that may be because i did those tests first..
so in conclusion, the numbers don't really mean anything.. your computer is running the same speed even with wrong numbers (at least on mine, benchmarking your own system is an exercise left to theh reader).
i did several xbench benchmarks before/after the numbers were the wrong/right ones.. as far as i can tell, the system runs nearly the same.. if anything it seems to run slightly faster with the lower numbers, but that may be because i did those tests first..
so in conclusion, the numbers don't really mean anything.. your computer is running the same speed even with wrong numbers (at least on mine, benchmarking your own system is an exercise left to theh reader).
As another data point, I recently installed YellowDog Linux on my TiBook. The YDL site mentioned a problem concerning the cpu frequency being lower than the actual machine was capable of and gave a command to manually correct the value. My TiBook reported 667 MHz, but when I tried the command to correct it to 1 GHz, my machine *immediately* powered off. Disconcerting, to say the least. I haven't had time to try a PM reset to see what impact that has wrt YDL.
It sure would be interesting to get an answer to this.
My uneducated guess is that there is some problem in the sysctl query. Whether that bit of code is used by other pieces of software, and whether that has any performance issues is another matter. It doesn't seem to affect my machines usability or my perception of its speed, so I don't really care to much right now. I agree it would be nice to find out for sure though.
some realworld results
i decided to take the initiative and provide some quatifyable results, as it seems xbench is too controversial for some here
hardware: 12"pb 867/640mb Ram/60Gb HD/APExt/Superdrive
software: osx 10.3/photoshop 7.0
setup: safeboot (hold shift) and login holding shift (disable startup items).
procedure: launch photoshop and create a 20" x 20" 600 DPI RGB image. apply a 4-color cone blend from the center of the image to the upperleft corner. apply a gausian blur of 20 to the image.
i ran this procedure 3 separate times for each state (533 and 867). below are the results. also included are the screen redraw times from creating the gradient. this should be a nice constant because most of the work *should* be handled by the GPU rather than the CPU in question.
At sysctl hw.cpufrequency reporting 533:
pass 1)
gradient 2:15 (minutes:seconds)
redraw 2:07
blur 2:27
pass 2)
gradient 2:32
redraw 1:52
blur 2:34
pass 3)
gradient 2:16
redraw 1:56
blur 2:29
At sysctl hw.cpufrequency reporting 867:
pass 1)
gradient 2:14
redraw 1:51
blur 2:30
pass 2)
gradient 2:35
redraw 1:51
blur 2:35
pass 3)
gradient 2:18
redraw 1:53
blur 2:28
i hope this eases everyone's mind. i'm leaving my PMU alone now.
i decided to take the initiative and provide some quatifyable results, as it seems xbench is too controversial for some here
hardware: 12"pb 867/640mb Ram/60Gb HD/APExt/Superdrive
software: osx 10.3/photoshop 7.0
setup: safeboot (hold shift) and login holding shift (disable startup items).
procedure: launch photoshop and create a 20" x 20" 600 DPI RGB image. apply a 4-color cone blend from the center of the image to the upperleft corner. apply a gausian blur of 20 to the image.
i ran this procedure 3 separate times for each state (533 and 867). below are the results. also included are the screen redraw times from creating the gradient. this should be a nice constant because most of the work *should* be handled by the GPU rather than the CPU in question.
At sysctl hw.cpufrequency reporting 533:
pass 1)
gradient 2:15 (minutes:seconds)
redraw 2:07
blur 2:27
pass 2)
gradient 2:32
redraw 1:52
blur 2:34
pass 3)
gradient 2:16
redraw 1:56
blur 2:29
At sysctl hw.cpufrequency reporting 867:
pass 1)
gradient 2:14
redraw 1:51
blur 2:30
pass 2)
gradient 2:35
redraw 1:51
blur 2:35
pass 3)
gradient 2:18
redraw 1:53
blur 2:28
i hope this eases everyone's mind. i'm leaving my PMU alone now.