Jump to content

DSDT - Vanilla Speedstep - Generic Scope (_PR)

FKA

By FKA,
in Power Management

 Share
1949 posts in this topic

Recommended Posts

kdawg
kdawg
Posted
Posted
Can you attach your version of mark-i? All that i try don't work well on my series p5q.

Don't forget to install kext as well.

 

marki.zip

 

With Keeza's patch no overclock

post-16858-1277646934_thumb.png

 

With Keeza's patch overclocked to 3.8GHz

post-16858-1277646939_thumb.png

Smith@@™
Smith@@™
Posted
Posted
VoodooMonitor only shows the VID, on iStat I see real Vcore changing like this, idle:

Screen_shot_2010_06_27_at_16.37.57.png

 

On full load:

Screen_shot_2010_06_27_at_16.39.00.png

 

 

i don't know, istat don't show me the drop of voltage. Which plugin you're using? Winbound or ite? Or superio? Or acpimonitor with code in dsdt for fan, voltage, etc. etc.

mm67
mm67
Posted
Posted
@mm67 and Smith,

I'm not using the legacy kext just the DSDT patch. I use mark-i exclusively for voltage because the current releases of FakeSMC tend to break C-States.

 

I also tend to loose C-States when overclocking. Although until just recently overclocking and C-States were working. I need to find out what I did to loose them.

 

I also lose C-states as soon as I use a manually set Vcore value, using Normal as Vcore and setting LLC enabled allows me to use FSB as high as 460, on my MSI board there is no such restrictions.

Smith@@™
Smith@@™
Posted
Posted
Don't forget to install kext as well.

 

marki.zip

 

With Keeza's patch no overclock

marki_with_Keeza_CST_no_OC.png

 

With Keeza's patch overclocked to 3.8GHz

marki_with_Keeza_CST.png

 

Thanks :) The same result, i see always the voltage chosen in bios, not the drop..

Smith@@™
Smith@@™
Posted
Posted

ok, i've substitute my winbond plugin with super i/o but the result is the same. The winbound plugin is specifically for my mobo and it works better. I'll do others test.

 

Ehi, i've do tests from windows, and it's the same. Only coretemp show me the vid as in voodoomonitor, the rest show the core that i 've chosen in bios, with littles drop. I thinks that these mobo from p5q series work in this way. Also for c1e, these mobos support only c1e from bios. By the way, nothing change from windows;)

 

Now, i'm lost, i'm lost completely. Help me, what are we talking about before the excursus on vcore and him vdrop?

mm67
mm67
Posted
Posted
ok, i've substitute my winbond plugin with super i/o but the result is the same. The winbound plugin is specifically for my mobo and it works better. I'll do others test.

 

No idea why that don't work for, my other GB board works just the same way.

Idle:

post-375259-1277649541_thumb.png

 

And load:

post-375259-1277649637_thumb.png

 

You are sure that C-states actually work ? Easy way to test is to disable method CST and check temps, on my system temps jump about 5 degrees higher while idling.

mm67
mm67
Posted
Posted
yes the cstate work, if i delete the cstate code from dsdt i have the "famous" error in verbose and temperature are higher 10° about..

That error message doesn't tell nothing about C-states actually working, it only says that you have some kind of CST defined for all cores. But that 10 degree temp drop tells that C-states probably work. You really don't see Vcore changing even in Windows ? Tried Everest ? This is how Vcore behaves on Gigabyte running OCCT.

post-375259-1277650645_thumb.png

kdawg
kdawg
Posted
Posted
That error message doesn't tell nothing about C-states actually working, it only says that you have some kind of CST defined for all cores. But that 10 degree temp drop tells that C-states probably work. You really don't see Vcore changing even in Windows ? Tried Everest ? This is how Vcore behaves on Gigabyte running OCCT.

post-375259-1277650645_thumb.png

If C-States were working we'd see two things a drop in VCore voltages as well as lower temperatures. Correct?
Smith@@™
Smith@@™
Posted
Posted
ok, i've substitute my winbond plugin with super i/o but the result is the same. The winbound plugin is specifically for my mobo and it works better. I'll do others test.

 

Ehi, i've do tests from windows, and it's the same. Only coretemp show me the vid as in voodoomonitor, the rest show the core that i 've chosen in bios, with littles drop. I thinks that these mobo from p5q series work in this way. Also for c1e, these mobos support only c1e from bios. By the way, nothing change from windows;)

 

Now, i'm lost, i'm lost completely. Help me, what are we talking about before the excursus on vcore and him vdrop?

 

 

That error message doesn't tell nothing about C-states actually working, it only says that you have some kind of CST defined for all cores. But that 10 degree temp drop tells that C-states probably work. You really don't see Vcore changing even in Windows ? Tried Everest ? This is how Vcore behaves on Gigabyte running OCCT.

2010_06_27_17h54_VCore.png

 

 

Yes, i 've write in previous post. I don't see, also with everest. Only coretemp show me vid as voodoomonitor.

 

The action is the same. Only c1e can be this ? The asus mobo work in this way. I've try also on other configuration with only windows, p5q se, and with c1e enable and eist enable in bios, i see the same things

 

That error message doesn't tell nothing about C-states actually working, it only says that you have some kind of CST defined for all cores. But that 10 degree temp drop tells that C-states probably work. You really don't see Vcore changing even in Windows ? Tried Everest ? This is how Vcore behaves on Gigabyte running OCCT.

2010_06_27_17h54_VCore.png

 

 

When i have "that" error message in verbose it 's because there isn't that code in dsdt and cstate obviously doesn't work, i think. In fact the temperature are higher 10° about.

mm67
mm67
Posted
Posted
Yes, i 've write in previous post. I don't see, also with everest. Only coretemp show me vid as voodoomonitor.

 

The action is the same. Only c1e can be this ? The asus mobo work in this way. I've try also on other configuration with only windows, p5q se, and with c1e enable and eist enable in bios, i see the same things

 

It's possible that your SuperIO chip just isn't capable of showing that actual Vcore, but normal behavior for Vcore is to go up and down when C-states are working. Gigabyte's use ITE chip and MSI has Fintek SuperIO, both show Vcore changing with load.

 

BTW Hardware Monitor also works nice with new Fakesmc, don't really need to use Windows even for stress testing anymore since Prime95 and Linpack are available also for OS X, this is how a short Linpack run looks like on OS X

post-375259-1277653080_thumb.png

mm67
mm67
Posted
Posted
Maybe. Can you post a screen on everest from your configuration that show the drop vcore when cstate work?

 

Thanks

 

That's exactly what you see on that OCCT screenshot, Vcore is 1.06 V when idling, on load it goes to 1.25 V. And do remember that I have LLC enabled so there is no Vdroop, only a very small Vdrop. Without LLC graphics would look bit different.

Smith@@™
Smith@@™
Posted
Posted

Oh finally! Thanks so much for confirmation, thanks :( These screen are "amazing" for me now, because delete my doubts. In that screens on my series p5q i see always the same voltage in load and in idle :)

mm67
mm67
Posted
Posted
Oh finally! Thanks so much for confirmation, thanks ;) These screen are "amazing" for me now, because delete my doubts. In that screens on my series p5q i see always the same voltage in load and in idle ;)

 

It's still possible that your SuperIO chip just doesn't show actual Vcore, what happens on Windows if you disable C1E from bios, do temps go up ?

kdawg
kdawg
Posted
Posted
I also lose C-states as soon as I use a manually set Vcore value, using Normal as Vcore and setting LLC enabled allows me to use FSB as high as 460, on my MSI board there is no such restrictions.

Could that be because we leave the frequency and milliwatts in PStates at zero? Assuming that zero means we let the board calculate them. So if we leave the MOBO VCore voltage set to auto this isn't a problem. If we fill in the frequency and milliwatts then perhaps we can regain CStates while overclocking. Did that make any sense? I'll have to run some tests but I have no idea when.

       Method (PSS, 0, NotSerialized)
       {
           Return (Package (0x06)
           {
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x4820, 0 }, // 8.5x32 48 in this case means that 8 is the multiplier and the preceding 4 is the half
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x081E, 1 }, //   8x30 FIDxVID
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x471C, 2 }, // 7.5x28
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x071A, 3 }, //   7x26
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x4618, 4 }, // 6.5x24
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x0616, 5 }  //   6x22
           })
       }

mm67
mm67
Posted
Posted
Could that be because we leave the frequency and milliwatts in PStates at zero? Assuming that zero means we let the board calculate them. So if we leave the MOBO VCore voltage set to auto this isn't a problem. If we fill in the frequency and milliwatts then perhaps we can regain CStates while overclocking. Did that make any sense? I'll have to run some tests but I have no idea when.

       Method (PSS, 0, NotSerialized)
       {
           Return (Package (0x06)
           {
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x4820, 0 }, // 8.5x32 48 in this case means that 8 is the multiplier and the preceding 4 is the half
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x081E, 1 }, //   8x30 FIDxVID
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x471C, 2 }, // 7.5x28
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x071A, 3 }, //   7x26
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x4618, 4 }, // 6.5x24
               Package (0x06) { [color="#FF0000"]Zero[/color], [color="#0000FF"]Zero[/color], 0x0A, 0x0A, 0x0616, 5 }  //   6x22
           })
       }

 

No, that doesn't matter. Same thing happens on Windows and Linux also, with stock SSDT tables. Just seems to be one more Gigabyte "feature".

kdawg
kdawg
Posted
Posted
No, that doesn't matter. Same thing happens on Windows and Linux also, with stock SSDT tables. Just seems to be one more Gigabyte "feature".
OK I can confirm that if I set my VCore to "Normal" not "Auto" instead if inputting my own voltage settings, voltage is show as dynamic again in mark-i. I also want to add that my temps are now crazy low. Lower than I've even seen them. We'll see how stable it is at a 3.91GHz and 460 FSB. LLC is also enabled to BTW, otherwise I can get my ram to go over 1066.
mm67
mm67
Posted
Posted
OK I can confirm that if I set my VCore to "Normal" not "Auto" instead if inputting my own voltage settings, voltage is show as dynamic again in mark-i. I also want to add that my temps are now crazy low. Lower than I've even seen them. We'll see how stable it is at a 3.91GHz and 460 FSB. LLC is also enabled to BTW, otherwise I can get my ram to go over 1066.

 

It depends totally on your chip how high you can go with Vcore set to Normal, you have to do some stress testing and find the max FSB for your cpu. On this Q9550 max I can use is 460, on E7400 max is 343 which gives 3.6 GHz. If you have a very low VID chip then maybe you can even get higher than 460, my Q9550 has 1.2250 as VID.

kdawg
kdawg
Posted
Posted
It depends totally on your chip how high you can go with Vcore set to Normal, you have to do some stress testing and find the max FSB for your cpu. On this Q9550 max I can use is 460, on E7400 max is 343 which gives 3.6 GHz. If you have a very low VID chip then maybe you can even get higher than 460, my Q9550 has 1.2250 as VID.

How exactly did you determine the VID of your chip?

mm67
mm67
Posted
Posted
How exactly did you determine the VID of your chip?

 

That's the default value for Vcore in bios, you can also see it using Core Temp on Windows.

 

This is my bios voltage settings for 3.91 GHz:

******Motherboard Voltage Control******

Load-Line Calibration..........................[ Enabled ]
CPU Vcore [ 1.225].............................[ Normal  ]
CPU Termination [ 1.20 ].......................[ 1.22    ]
CPU PLL [ 1.50 ]...............................[ 1.57    ]
CPU Reference [ 0.760 ]........................[ Auto    ]

MCH Core [ 1.100 ].......................[ 1.30   ]
MCH Reference [ 0.760 ]..................[ Auto   ]
MCH/DRAM Refernce [ 0.900 ]..............[ Auto   ]
ICH I/O [ 1.500 ]........................[ 1.57   ]
ICH Core [ 1.100 ].......................[ 1.20   ]

DRAM Voltage [ 1.800 ]...................[1.90    ]
DRAM Termination [ 0.900 ]...............[ Auto   ]
Channel A Ref [ 0.900 ]..................[ Auto   ]
Channel B Ref [ 0.900 ]..................[ Auto   ]

 Share
Go to topic listing
×
×
  • Create New...