Can't restart/shutdown Lion or Mountain Lion.

[dutchcow]

It only happens on certain brands/types of HDD on this D630 it happens with my Hitachi and Seagates, I'm not alone here, hence this tool. Just though it could prevent people from thinking their drive is broken. I read that Apple flashes their own firmware on their drives that interprets the apm signals differently than stock drives. There's a lot of forum posts from people with the clicking on their Macs after upgrading the drive to a non-Apple.

[Hervé]

Now that you mention it, one of the D630 that I very rarely use does indeed have a faint but regular HDD click. It's a Seagate 160Go 7.2k rpm unit. I have Hitachi 80Go in others but I have not noticed anything that bothered me.

 

I'm using an SSD in my daily Hack, so no noise at all... 

[dutchcow]

I just had something odd. I installed SL with the myHack+bootpack and before installing EDP ran Xbench graphics parts only; scores over 150! Blazing fast. So I installed EDP, rebooted, re-ran the tests a few times, all scores well below 50, sometimes even 32. There is something going on there. Maybe later I will try the bootpacks as Extra on L/ML too just to see what happens. Any ideas so far what could possibly explain this difference?

 

Just to be sure I re-installed EDP rebooted and retsted and then re-installed the bootpack Extra, using the bootpack kexts is os much faster than EDP. I can't even start to guess at what could cause this or the differences between the final Extra folders. I will look at it tomorrow when I'm bored again.

[dutchcow]

Well the culprit is VoodooPState, once I remove that kext everything is dandy. I don't know where to go from here though.

 

Same thing happens on L/ML, once its removed the system is like a 100x faster, amazing. But howto get speedstep back?

 

http://imgur.com/f1iF01f

 

Left: Manually set to max frequency.

Right: Set to on demand performance.

 

The difference in user experience is huge, anyway to fix this? If I'd to believe htop the core syncing (VoodooTSync?) isn't working as well, unless some apps can actually only use one thread. I still don't understand how this only happens to my my D630 
 

Hrm, it only happens when I force the speed at the lowest two speeds, which don't really exist.

[Hervé]

Mmm, I don't understand most of all this, especially the part about the lowest frequencies "which don't really exist"... 

 

DId you check your BIOS settings and compare to those listed in dedicated thread?

 

Also, can you post a screendump of your Cham boot plist (opened with Cham wizard) and tell us what SMBIOS settings you chose?

 

You should also question the true meaning of benchmarking tools results: you're running graphics tests and the higher the CPU frequency, the better the results -> where's GPU involvement in there? I don't know how the benchmarking is done, but it seems to me that only CPU is tested and, should this be the case, it's not a true reflection of the computer's real graphics capabilities and/or performance.

 

I'd suggest you run 2 separate tests of GeekBench in the same way as well: first with on-demand performance then with frequency set to highest. I'd be interested to hear which one is better...

 

I guess that if you feel your laptop runs faster without Emulated SpeedStep, you should remove it.

[dutchcow]

The bios is fine, I tried the settings from the thread, and my own which are the same with just devices disabled like modem.

 

The lowest clockspeeds of the CPU that PState tries to set the CPU to is 595, does not exist for the CPU. I don't think 595mhz is a real speedstep supported by a T7x00. Anyways, its still strange that this only happens on my D630 and not D820. No matter how low I set the speed there, the GUI benches are always fast. Thats why I run those, I don't really care if it uses GPU or CPU, its exactly whats causing the sluggishness of the GUI on the D630.

 

The lowest clockspeed Windows/Unix will try on my T7100 is 1ghz, and then one or two steps more till the max of 1.8ghz. But never as low as 595. If only PState would use the real speedstep-steps in and not something it gets from I don't know where.. Still though, this shouldn't happen. Very annoying.

 

I will do the tests later. All PState really needs to get is a proper list of steps/frequencies supported by the actual CPU or a way to add/edit those values to the real values of the CPU. Finding documentation about the Info.plist and how to give it your own proper CPU speeds is unfindable to me.

 

Right now I put a new disk in the D630 and re-installing 10.8. Made the USB stick with myHack + bootpack. Will do a normal install, boot up, install EDP, use stock values in EDP, reboot and run the benches. I won't edit or touch any config files. Once I have the results I will also get the info you'd like to see. But I really think this is caused by PState running my CPU too slow, like the lowest speed should be something like 1.09ghz and not 595mhz.

 

Edit: Seems the solution is in patching SSDT tables so make sure PState will not try and set my CPU to a speed thats lower than the lowest supported speed, which I havent done, I did extract the tables as per wiki under Linux. But I need to learn about the patching I guess.

[dutchcow]

Fresh ML installed, ran the Xbench tests as before with the same results. Thought to try to set it to "conservative performance" and test that too. Guess what? As soon as the bench started the CPU speed (bars in PSTatemenu) went all full, something that didn't happen when using on-demand, not as fast anyways. As soon as the bench was over the bar slowly went down. It nearly looked pretty!

 

So it seems on my D630 on-demand and conservative are swapped   I get the expected slower response in delivering full CPU when using on-demand, and when set to conservative the speed is given much quicker to an application. This is one weird simple solution to this. I will continue to test and install some apps. And monitor the CPU temps. This has all been very enlightening, lol.

 

No wonder my CPU temps were in the low 30s, they are in the high 30s now the performance is given more often, higher load, higher temp all makes sense. Funny the modes got swapped. The same happens on my D820, setting it to conservative performance gives much higher bench results and even snappier UI experience. With the D820 still being twice as fast with the graphics stuff, but so be it. The system is snappier more usable than when using on-demand performance. Now if it would only remember it between reboots. Also the -v boot bug is back on a clean disk. Volume isn't remember between boots too. A lot of niggles to look into    

 

Does anyone else have the same when switching to 'conservative' mode? Or heard of this? Try it and see.

[Hervé]

The lowest clockspeeds of the CPU that PState tries to set the CPU to is 595, does not exist for the CPU. I don't think 595mhz is a real speedstep supported by a T7x00. Anyways, its still strange that this only happens on my D630 and not D820. No matter how low I set the speed there, the GUI benches are always fast. Thats why I run those, I don't really care if it uses GPU or CPU, its exactly whats causing the sluggishness of the GUI on the D630.

 

The lowest clockspeed Windows/Unix will try on my T7100 is 1ghz, and then one or two steps more till the max of 1.8ghz. But never as low as 595. If only PState would use the real speedstep-steps in and not something it gets from I don't know where.. Still though, this shouldn't happen. Very annoying.

 

I will do the tests later. All PState really needs to get is a proper list of steps/frequencies supported by the actual CPU or a way to add/edit those values to the real values of the CPU. Finding documentation about the Info.plist and how to give it your own proper CPU speeds is unfindable to me.

 

[...] But I really think this is caused by PState running my CPU too slow, like the lowest speed should be something like 1.09ghz and not 595mhz.

Oh, you need to get to know Intel CPUs a wee bit better...

 

PState does show/display the correct set of frequencies for your T7100, don't get mistaken. T7xxx CPUs do operate from a range of 600MHz to 2.8GHz for the fastest of the range in steps of 200MHz (ie. steps of CPU bus speed). It may not show up in Linux, but it certainly would in Windows. Look up the Intel datasheet if you need details or want to get familiar with IDA (a kinda of turbo/boost mode) or LFM/SuperLFM modes (low frequencies). Basically, a T7xxx can run at speeds of variable FID multipler x CPU bus speed, the lowest SuperLFM multiplier being x3 and the IDA multipler being x12 -> this gives a range from 600MHz to 1.8-2.8GHz... That's EIST for you (Enhanced Intel SpeedStep).

 

Now sometimes, the CPU FSB (which is quad-pumped Bus Speed) does not operate at the exact max. speed, but slightly below, so you could expect an 800MHz FSB (4 x 200MHz) CPU to actually run at 798MHz or 799MHz and, as such, it may report a frequency of say 595MHz instead of 600MHz or 2189MHz instead of 2200MHz. That's just perfectly normal.

 

Use tools like RmClock and CPUID in Windows and you'll see.

[dutchcow]

Yeah, it does actually run at the low speed of 600mhz, I didn't doubt that for a second judging by the cool cpu temps, haha. I did some benches and I'm done with it for now, will try and fix those niggles for now and see what todo with the D630. Here are the benches:

 

Lowest - highest cpu freq value: 

 

D820: 996, 1162, 1328, 1494, 1660, 1826

D630: 597, 796, 995, 1194, 1393, 1592, 1791

 

Geekbench 2 results:

 

D630 32bit: 2644 (conservative), 2635 (on-demand), 2635 (cpu max), 918 (cpu min)

D630 64bit: 2868 (conservative), 2895 (on-demand), 2892 (cpu max), 1008 (cpu min)

 

D820 32bit: 2577 (conservative), 2574 (on-demand), 2588 (cpu max), 1527 (cpu min)

D820 64bit: 2825 (conservative), 2824 (on-demand), 2834 (cpu max), 1680 (cpu min)

 

Xbench totals of graphics tests only:

 

D630: 109 (conservative), 38 (on-demand),  117 (cpu max),  29 (cpu min), 58 (@995mhz)

D820: 181 (conservative), 109 (on-demand), 208 (cpu max),  105 (cpu min)

 

The geek bench results show the slight advantage the T7100 has. The other Geekbench tests are pretty much useless as they both have the same disk and 4GB dual channel RAM. The same brands even. So to be the Geekbench results don't really matter much as you don't notice those differences. The scores are too close to call when it comes to some differences on the D630 between the frequency states. Those are really only noticed in daily usage of the OS itself. Sluggish feel.

 

Both systems use the same myHack+EDP install method and have nothing modified in /Extra other than the TJmax setting.

 

It is the GUI/user experience that matters the most to me. The D820 feels so much smoother and snappier. Probably due to that the GMA is a bit more optimized for OSX? Also the difference in lowest 'clockable' speed is quite a lot. As can be seen in the Geekbench cpu min scores. I guess the GMA chips just perform better than NVidia ones under OSX when it comes to GUI tasks at least. I don't think the screen resolution (a bit more work for the GPU on the D630) should cause the difference in GUI performance.

 

Anyways, it was fun testing all this, and with things set to conservative the D630 is a 100x more usable, but still a bit saddening it can't be as snappy as the D820, which other than the GMA are pretty much the same performance wise, so you'd elect the same result. Too bad I can't figure this out better, my knowledge is limited.

 

Herve, how do your GMA D630s vs. your D630 NVidia score on the Xbench graphics tests as big of a difference as here?

[dutchcow]

That isnt my issue though. But it shows the reason why the D630 is sluggish compared to the D820 in normal use. The GUI is just not snappy. And I (or anyone here) can seem to fix this. That is my main problem at this moment. I don't really care if it's to do with the CPU or GPU, its just frustrating that 2 nearly identical laptops perform so much differently when you just use it to surf and do your things. Does your NVidia score the same as mine?