Asus P6T SE Retail Snow Leopard Installation Guide [FLASHDRIVE] [UPDATED JUN. 28]]

[SAKrisT]

new error with another DSDT.aml

 

 

Fixed!

add to boot list -pci0 and add new boot EFI from http://netkas.org/?p=212

 

The Snow Leo install was not modified at all apart of installing of 4850-fix from netkas.org:

http://netkas.org/?p=212

 

Important hint for all:

- please do the 1st start after you build your Snow-system using options "-v -f" in order to rebuild extensions.mkext. In case your graphic card is not working use the VESA mode by adding option "-x"

--

ATI HD4850

[Bchemist]

This flash issue is really strange, it occurs randomly with flash videos or flash web applications-games. New imacs late 2009, have this problem too. It seems that the osx version of flash player needs more cpu power. I have this problem with my laptop too and it's worst (75c when watching youtube video) and also my mac mini gets hot when i watch flash videos (70c). There is a 10.1 prerelease version of adobe's flash player that seems to be "lighter" but it is not stable yet.

 

I've held off on the update for the near future since an extensive bug-hunt didn't find anything unusual other than known issues that don't really affect my usage or I can easily work around them. I noticed that the temperatures climb to around 60 or so during .rar extractions but Logic plus the MOTU using core audio only pushes the cores to 47C. But then again I'm running it in 10.6.1 (P6 aka Rampage) and I held off on the 9.1 update since we use Reason, Symphobia, RA, Kontakt, and a Roland Fantom plus a Roland TD-20....none are 64 bit. That plus the 9.1 is still a beta and I'll wait until I see what happens with 10.6.3 before we make the move. One of those "if it ain't broke, don't fix it" kind of things.

[Bchemist]

I just did a RAID install on this. Two things confuse me:

1) Backing up the main RAID drive (striped) to the 'Time Machine' RAID drive (both RAID volumes are on the same pair of disks) causes Time Machine to cause a Kernel Panic at 2.49GB completed.

2) XBench scores for the computer are low... I mean very low. CPU returns 155-166. Eh?? The memory component isn't brilliant either. Any connection with the 'Bus Speed = 533MHz' in System Profiler?

 

Thanks to anyone who can help me.

 

I understand your configuration but I'm curious as to why you would chose such a configuration. At a minimum, if you are going to put the TM backup on one or both of those two disks, why not just partition a portion of one of the two disks and use that as your backup and reduce your chances of total loss of data due to a crash by 50%.

 

In my case I literally "waste" at least 50% of my hard drives when configuring a RAID for video and photo editing to ensure maximum read/write speed since nothing else comes even close to being as demanding/limiting as Photoshop on I/O, latency, etc. For example if I have a terabyte drive, I'll make 2 partitions, with the top 50% formatted as normal and the second half left as free space due to the sharp drop in I/O when accessing the bottom 50% of the disk. The net is 500gig of usable space. In addition, I don't go over 80% capacity on the RAID to ensure minimal fragmentation. That said, putting Time Machine on those disks is going to bring your data transfer to a crawl since TM tends to slow down the system as it does in the case of it being on a separate internal or external drive and probably even worse if it's on the same set of disks to the point of kernel panic since the system I/O is so bogged down.

 

So my setup has the boot files on a SSD but the scratch disk is a striped RAID that is separate from the boot drive. Obviously apps load fast on a SSD, but then file access is twice as fast on the RAID....as in adjusting Gaussian blur looks like it is happening in real-time.

 

The bus speed is cosmetic if it differs from the bus speed set in your mobo.

 

Another scenario is this (even though I am not saying that this is good data management practice to anyone who might stumble upon this thread):

 

Partition your two disks into 7 parts using this illustration assuming 1 terabyte drives as an example (size and respective ratios are important):

 

Disk 1:

 

Partition 1: 250 gig - boot disk

Partition 2: 125 gig - RAID striped vol 1

Partition 3: 125 gig - RAID striped vol 3

Partition 4: 500 gig - RAID mirror vol 1

 

Disk 2:

 

Partition 1: 125 gig - RAID striped vol 2

Partition 2: 125 gig - RAID striped vol 4

Partition 3: 500 gig - RAID mirror vol 2

 

note that 250 gig would need to remain as "free space"

 

Then create a striped RAID. Use the mirrored RAID volume to mirror the striped RAID....at least I think it is in this order but you may have to mess with it. Just keep the ratios the same in that if you want 4 volume RAID speed and redundancy based on the limitations of Apple's software RAID.

[wahwahman89]

I understand your configuration but I'm curious as to why you would chose such a configuration. At a minimum, if you are going to put the TM backup on one or both of those two disks, why not just partition a portion of one of the two disks and use that as your backup and reduce your chances of total loss of data due to a crash by 50%.

 

In my case I literally "waste" at least 50% of my hard drives when configuring a RAID for video and photo editing to ensure maximum read/write speed since nothing else comes even close to being as demanding/limiting as Photoshop on I/O, latency, etc. For example if I have a terabyte drive, I'll make 2 partitions, with the top 50% formatted as normal and the second half left as free space due to the sharp drop in I/O when accessing the bottom 50% of the disk. The net is 500gig of usable space. In addition, I don't go over 80% capacity on the RAID to ensure minimal fragmentation. That said, putting Time Machine on those disks is going to bring your data transfer to a crawl since TM tends to slow down the system as it does in the case of it being on a separate internal or external drive and probably even worse if it's on the same set of disks to the point of kernel panic since the system I/O is so bogged down.

 

So my setup has the boot files on a SSD but the scratch disk is a striped RAID that is separate from the boot drive. Obviously apps load fast on a SSD, but then file access is twice as fast on the RAID....as in adjusting Gaussian blur looks like it is happening in real-time.

 

The bus speed is cosmetic if it differs from the bus speed set in your mobo.

 

Another scenario is this (even though I am not saying that this is good data management practice to anyone who might stumble upon this thread):

 

Partition your two disks into 7 parts using this illustration assuming 1 terabyte drives as an example (size and respective ratios are important):

 

Disk 1:

 

Partition 1: 250 gig - boot disk

Partition 2: 125 gig - RAID striped vol 1

Partition 3: 125 gig - RAID striped vol 3

Partition 4: 500 gig - RAID mirror vol 1

 

Disk 2:

 

Partition 1: 125 gig - RAID striped vol 2

Partition 2: 125 gig - RAID striped vol 4

Partition 3: 500 gig - RAID mirror vol 2

 

note that 250 gig would need to remain as "free space"

 

Then create a striped RAID. Use the mirrored RAID volume to mirror the striped RAID....at least I think it is in this order but you may have to mess with it. Just keep the ratios the same in that if you want 4 volume RAID speed and redundancy based on the limitations of Apple's software RAID.

 

Thanks Bchemist. I ended up just going back to a normal install as I didn't find that there was really enough of a difference to warrant the risk of the disk failing. Works alright now and there isn't the Time Machine crash anymore - so figure you were completely right about the RAID + TM RAID setup and it bogging the system down - good note for anyone who makes the same mistake as me.

 

I'm still curious about the XBench scoring though - any ideas from anyone? What do all of you get?

 

Thanks for your help again Bchemist and nice one for all the work you've done on the DSDTs for this motherboard!

[tweak41]

I just ran GeekBench and scored 10598.

 

That's overclocked to 3.2 Ghz and 12 GB 1600 MHz DDR3.

[Bchemist]

I'm still curious about the XBench scoring though - any ideas from anyone? What do all of you get?

 

Thanks for your help again Bchemist and nice one for all the work you've done on the DSDTs for this motherboard!

 

My pleasure

 

For geekbench it's about 8700 on average with stock speeds. My overclock speeds give me about the same scores as tweak41. I plan to get some better RAM in the near future to see how much that can be improved.

 

I was playing with this last night and it's pretty cool...the Cinebench 4D benchmarking tool: http://www.maxon.net/downloads/downloads/c...ebench-115.html

 

It measures both CPU and graphics performance and, like Geekbench, it is up to date. Xbench, in contrast, is obsolete based on what I've read (plus it doesn't seem like they've updated their website in at least 2-3 years) in that it doesn't test for hyperthreading. I'm going to play with it today to see what happens.

[Bchemist]

I was messing with my DSDT last night, based on some ideas from d00d in a Gigabyte thread. I'm now down to an audio kext and fakesmc.kext...was able to ditch nullCPU and sleepenabler.

 

I modified HPET for ASUS specific mobos as well as added some code to handle speedstepping and turbo in addition to C-states. C-states we had working but since OSX doesn't support native speedstepping in hyperthreading CPU's we were "pseudo-stepping" as is the case with real Mac Pro 4,1's and iMacs. AppleIntelCPUPowerManagementClient.kext is working now. My temperatures still idle at around 45C (hotter than my Rampage) but my Geekbench is now ranging around 9200-9300 in stock speeds as before vs 8700 with no temperature increase.

 

Sleep still works also. I need to test this further before I release anything but this is definitely cool so far. It actually worked the first time I tried it...well not counting when I had sleepenabler still in the boot usb where it hung. But once I removed that last kext it roared to life. I tried a few overclocking scenarios and it worked also.

 

I'll post it in the next 24 hours or so when I have more time to test it.

 

EDIT: To clarify, I'm seeing the turboboost 112 notification immediately after the platform plugin shows that it has arrived and loaded. I had to make modifications to the Fakesmc.kext and had to modify the HPET on an educated guess. I'll be uploading the modified fakesmc later also.

 

I knew that you could get rid of the two kexts earlier but some of the feedback of hotter temperatures and slower performance made me hold off until I could dig deeper into it. I think all of those issues are now solved in that there is a very noticeable difference. Several other benchmarking apps have confirmed this. I'll be posting everything this time including boot files and smbios files that I've been tweaking.

[japhex]

Post #424, did some benchmarks a while back and also ended up modifying my original dsdt to include the mp4.1 speedstep fix found on another site.

 

Your increment in performance is related to the turbo mode, you can check cpu speed with msr_tools booting in 32bits. When testing at 4.2ghz I often see it going up to 4.31ghz.

 

You should also check the overclocked performance to see if the "speedstep" is always working and not defaulting to stock clocks (like some other wrong speedstep dsdt implementations that I've seen posting on this thread)

[KariNeko]

@ Bchemist and japhex

I asume that you are getting SpeedStep to work using Model Identifier: MacPro3,1 or as MacPro4,1 but copying the missing sections (PLimitDict and StepDataDict) from MacPro3_1.plist to MacPro4_1.plist inside IOPlatformPluginFamily.kext, right?

[Bchemist]

Post #424, did some benchmarks a while back and also ended up modifying my original dsdt to include the mp4.1 speedstep fix found on another site.

 

Your increment in performance is related to the turbo mode, you can check cpu speed with msr_tools booting in 32bits. When testing at 4.2ghz I often see it going up to 4.31ghz.

 

You should also check the overclocked performance to see if the "speedstep" is always working and not defaulting to stock clocks (like some other wrong speedstep dsdt implementations that I've seen posting on this thread)

 

Thanks for the advice as I forgot about MSR tools. I always booted in 32 bit mode in other systems but I could never get it to work where it would either crash or hang. This time it worked.

 

What you described is exactly what I saw...speedstep isn't enabled according to MSR but Turbo is enabled and I'm seeing a fluctuation of 100-105%. I googled around and the only thing I could find was by The King outside of these forums but it seems that they got about as far as I have (at least based on what I saw from the thread). Any advice or insight would be appreciated since I've been able to achieve a resting temperature of between 35-45C and my benchmarks as well as my system work perfectly, but I've never really been sure about speedstep other than I don't get C-state errors like before (before I knew anything about DSDT) and now I'm seeing Turboboost (don't remember if I ever did...maybe in 10.6).

 

I misread the part about overclocking and testing it. I'll do that a bit later and report back my results. Any other insights you may have would be appreciated since I based my perspective before on the performance of native Mac Pros but if there is a way to improve on that, I'd love to implement it to our chipset.

 

@ Bchemist and japhex

I asume that you are getting SpeedStep to work using Model Identifier: MacPro3,1 or as MacPro4,1 but copying the missing sections (PLimitDict and StepDataDict) from MacPro3_1.plist to MacPro4_1.plist inside IOPlatformPluginFamily.kext, right?

 

I used the model identifier of 4,1 but I didn't copy the MacPro4_1.plist info. What I did was change the fakesmc.kext data from "napa" to "t****" (I can't remember and it's on this system) and the version number in the .plist. I'll try to find the 4.1 plist but if you could post it also, please do

 

What I did was use the Gigabyte x58 method as a template and then made adjustments based on the ASUS version of the x58. It's definitely not a cut and paste job but I managed to extrapolate most of it to our P6*.* platform.

 

EDIT: I must have experienced a blind spot when reading this since it was right in front of my face. Overexcited on my part I guess Duh. http://www.insanelymac.com/forum/index.php...t=0&start=0 I assume this is it (4,1 plist portion) and it's the only thing I didn't do. Hopefully this will be the definitive fix.

 

EDIT2: Wow that was easy. I wonder why Mac Pro4,1 models don't have that.

[_Wolverine_]

i just tryed going 64bit

im all sotck

 

P6T SE

Intel i7 920 stock

3gb (3x1) 1033mhz Corsair stock

Radeon HD 4870 Sapphire stock

PCEFI 10.6

 

fakesmc.kext

openhaltrestart.kext

IOAHCI....kext

 

latest DSDT from bchemist

 

i got 9100 on geekbench

 

can post a screeny later

[chris wright]

Hi, i have tried to boot off the USB stick and it come up with the don't steal mac os x error. i then deleted the dsmos kext and tried to boot again, it now hangs at the ethernet bit where it tells you the address of the ethernet.

 

Any Ideas Guys, any help much appreciated.

[Bchemist]

Speedstep works now according to MSR Tools. Also the temperatures are lower by around 5C with a boost of 500 Geekbench points at stock speeds. Even when overclocked the idle CPU speeds cut in half and only one core is active. I modified the IOPlatformPlugin.kext to allow for the MacPro4,1 to use speedstep like the MacPro3,1. The modified kext is included in the folder and you'll need to install it in S/L/E. Technically that makes it "non-vanilla" but I don't think anybody would complain based on the expected results

 

Also you'll need to swap out the fakesmc.kext with the one I included in the folder. If you are using the bootloader I included you'll need to use the boot.plist file as well as the SMbios.plist. If you are using a different bootloader that doesn't autodetect memory, clock speed, etc, you'll need to add that to the SMbios.plist but all of the other data should stay the same. The MacPro4,1, 1.395F, etc are very important to get this to work. I suppose that this would work with iMac however so if you change your smbios to version 1.0, iMac11,1 and iMac for the family, it will work also since iMac has speedstepping enabled.

 

For this to work, you need to trash nullCPU and sleepenabler. Sleepenabler will hang the system. Sleep still works, however.

 

The DSDT has the Jmicron code that Caxio provided earlier. New to the code is the speedstep code. The C-state code at the bottom was merged with the P-States at the top. Also new is the HPET code that allows you to trash nullCPU.

 

So to review you have two options:

 

iMac11,1 - system stays vanilla meaning that you don't need to add the IOplatformplugin.kext. You still need to add the fakesmc and the DSDT. Your SMbios version should be 1.0 to match the vanilla IOplatformplugin.kext. NOTE: I have a theory about this with the fakesmc.kext. If you don't get speedstepping, further modifications to fakesmc might be necessary.

 

MacPro4,1 - add everything I attached in the package. Note, however, that with the bootloader and this smbios, your system will show up as an Intel Xeon CPU. With the improvements I'm seeing, I definitely don't mind

 

If this holds, Thundercat DSDT will be officially 2.0 since this solves every issue that I can think of other than audio that's out of our control with respect to DSDT unless we add an external card.

 

The only things that I have yet to figure out are mainly cosmetic or trivial issues but I think those are related to the bootloader (since boot speeds stink now vs before when it was as fast as 3 seconds).

SstepThundercat.zip

[jmhuff]

thanks for the guide.

[andrew79]

Speedstep works now according to MSR Tools. Also the temperatures are lower by around 5C with a boost of 500 Geekbench points at stock speeds. Even when overclocked the idle CPU speeds cut in half and only one core is active. I modified the IOPlatformPlugin.kext to allow for the MacPro4,1 to use speedstep like the MacPro3,1. The modified kext is included in the folder and you'll need to install it in S/L/E. Technically that makes it "non-vanilla" but I don't think anybody would complain based on the expected results

 

Also you'll need to swap out the fakesmc.kext with the one I included in the folder. If you are using the bootloader I included you'll need to use the boot.plist file as well as the SMbios.plist. If you are using a different bootloader that doesn't autodetect memory, clock speed, etc, you'll need to add that to the SMbios.plist but all of the other data should stay the same. The MacPro4,1, 1.395F, etc are very important to get this to work. I suppose that this would work with iMac however so if you change your smbios to version 1.0, iMac11,1 and iMac for the family, it will work also since iMac has speedstepping enabled.

 

For this to work, you need to trash nullCPU and sleepenabler. Sleepenabler will hang the system. Sleep still works, however.

 

The DSDT has the Jmicron code that Caxio provided earlier. New to the code is the speedstep code. The C-state code at the bottom was merged with the P-States at the top. Also new is the HPET code that allows you to trash nullCPU.

 

So to review you have two options:

 

iMac11,1 - system stays vanilla meaning that you don't need to add the IOplatformplugin.kext. You still need to add the fakesmc and the DSDT. Your SMbios version should be 1.0 to match the vanilla IOplatformplugin.kext. NOTE: I have a theory about this with the fakesmc.kext. If you don't get speedstepping, further modifications to fakesmc might be necessary.

 

MacPro4,1 - add everything I attached in the package. Note, however, that with the bootloader and this smbios, your system will show up as an Intel Xeon CPU. With the improvements I'm seeing, I definitely don't mind

 

If this holds, Thundercat DSDT will be officially 2.0 since this solves every issue that I can think of other than audio that's out of our control with respect to DSDT unless we add an external card.

 

The only things that I have yet to figure out are mainly cosmetic or trivial issues but I think those are related to the bootloader (since boot speeds stink now vs before when it was as fast as 3 seconds).

 

 

works perfect. So far no problems Just a question. You said you have only 2 kexts (audio and fakesmc). I have 3. If i remove IOAHCIBlockStorageInjector kext, all my drives are displayed as external. Is there another way to fix this?

[Bchemist]

works perfect. So far no problems Just a question. You said you have only 2 kexts (audio and fakesmc). I have 3. If i remove IOAHCIBlockStorageInjector kext, all my drives are displayed as external. Is there another way to fix this?

 

Sorry...that was a case of no sleep for 2 days while writing off of the top of my head, totally forgetting about SATA support Don't trash that one.

[_Wolverine_]

i just tryed your new dsdt with booter and everything...

i can see my cpu as xeon

replaced boot0 and boot1h and boot

replaced the kext

but i dont get speed improvements

rebooted with -f

 

im booting with 64bits and running geekbench 64bits

am i doing something wrong?

[akhenamenra]

woah ive been out of the office for a while and this is what you guys are on ....man tons of reading... ill have to have a look on the weekend good to see your still here bchemist lol whats with the thundercat title ha ha ....hope all is cool any way :D

[Bchemist]

i just tryed your new dsdt with booter and everything...

i can see my cpu as xeon

replaced boot0 and boot1h and boot

replaced the kext

but i dont get speed improvements

rebooted with -f

 

im booting with 64bits and running geekbench 64bits

am i doing something wrong?

 

If you are using my boot.plist and smbios.plist that I included you shouldn't be able to boot into 64 bit mode.

 

Try using the boot.plist as it is along with everything else. Even if you boot with the 32 bit kernel you can still run 64 bit apps and extensions. Look at Logic Pro 9.1 that is 64 bit but runs on the 32 bit kernel. Back when I was experimenting with 64 bit out of curiosity I actually got higher Geekbench scores when running the 64 bit app in 32 bit mode.

 

Bottom line to this fix is the speedstepping. The 5-7% gain in benchmarking scores is just a nice addition to the goal of reduced power consumption and reduced heat if we can get speed step to work.

 

You'll know if it's working if you use MSR tools to test....in fact that is the important test for this. You should see your voltage and processor speed back off their peaks and idle at around 50% of stock full speed. Next step is to overclock and you should still see speed stepping and power reduction at idle. Also you should be seeing Turbo Boost that would show up in improved Geekbench scores.

 

I really think the key here is to boot under 32 bit mode until every last kernel in OSX is 64 bit. But personally I don't see that happening until OSX 10.7.

[_Wolverine_]

i modded it so i can boot on 64 =P

ill change back and post results

[_Wolverine_]

i changed it back to 32bits

no speed change

no improve on geekbench and MSR tools crash (freeze on calculating i think)

[unarmedgunman]

Hi,

 

I have followed this guide from beginning to end and it was working perfectly in 10.6 with 1080p (no hdmi sound though) until i updated to 10.6.2, now my graphics drivers are outputting only 720p and graphics are very choppy

 

when i enabled graphics mode in com.applle.boot.plist i got no signal on my tv after the mac loading screen

 

in systempreferences/display the only option is 1280x720 when in 10.6 there was many different options

now it is set to false and update to 10.6.2 succeeded

 

might it have been because i removed arch=i386 to enable 64 bit?

 

any help would be greatly appreciated.

 

my specs

intel core i7 920 @2.7ghz

asus p6tse

nvidia geforce 9800 gtx

3x corsair 2gb ddr3 " 1333mhz

37 inch lg display

antec 900 case

[marc5]

Got a small problem guys,

 

i recently got 2x 4890 ati cards.

 

Works wonderfully in windows however in 10.6.2 the other card isn't recognised and the fan of the second one is on 100% all the time.

 

The first card works with dual dvi working.

 

heres a sample of the -v on boot

 

 

 

ATI VGA Controller [1002:9460] : : PciRoot(0x0)/Pci(0x3,0x0)/Pci(0x0,0x0)

old pci command - 7

boot display - 1

Not going to use bios image file

Found bios image file

adding binimage to card 9460 from legacy space with size f000

ATI VGA Controller [1002:9460] : : PciRoot(0x0)/Pci(0x7,0x0)/Pci(0x0,0x0)

old pci command - 0

new pci command - 6

boot display - 0

found bios image

wrong pci header signature ffffffff

 

after that the fan of the second card goes on full. They system doesn't recognise it even in system profiller but in windows, crossfire works

 

so any idea what is with the wrong pci header signature ffffffff ?

[_Wolverine_]

Unarmedgunman

try adding back the arch=i386 to com.Apple.boot.plist

if still dont work repair permissions and boot with -v

try using pcefi 10.6 or aserebln from bchemist a few posts up

btw leave the graphicsenabler on

 

mark5

as far as i kmow there isnt any support for crossfire

need to check with bchemist

[marc5]

well, i just want it to recognise 2 video cards.