How Setfsb works

[moral hazard]

Hi everyone,

My goal for this thread is to find out how setfsb works.

So here is my first question:

I have been able to overclock my FSB with setfsb (read more about how I did it in my sig) to 272mhz (I run it at 250mhz though).

The problem is that after I restart the notebook, the FSB goes back to 200mhz (stock).
Also if I tell the machine to hibernate, after I wake it up the FSB is back to 200mhz again. Even if I just put the notebook to sleep and wake it up the FSB is stock again.

I think I know how to fix this, but I would like someone to confirm that what I think is right.

My PLL is an ICS chip, more specifically ICS9LPR501SGLF.
Here is a link to the datasheet:
http://www.idt.com/products/getDoc.cfm?docID=18688128

Look on page 14 of the datasheet, bit 0 of byte 0.
The register value is called PD_restore,

1 = on Power Down de-assert return to last known
state
0 = clear all SMBus configurations as if cold
power-on and go to latches open state"

The value I have is 0. I think if I change it to 1, I will be able to put the notebook to sleep and then when it wakes up have the FSB still overclocked.

Request:
Can someone that is overclocking with setfsb and is able to put their machine to standby and then wake it up still overclocked tell me what value they have for their "PD_restore" bit.

Thanks in advance

[nando4]

My PD_Restore bit is not accurate I think because of TME mode being active. Though I'm not sure if it would set the custom clock dividers back to how they were on startup(??)

So the issue is how to get FSB to be automatically set after a resume-from-hibernate or resume-from-standby. My workaround is a vbs script that runs in the Windows Startup folder. It will setfsb on initial startup, then look for a resume event upon which it will re-run setfsb. It runs as an infinite loop, consuming very little resources.

resume-setfsb.vbs -> Edit with your unique system's parameters: PLL/FSB. Then put a shortcut to it in your Windows Startup Folder to have it call setfsb upon resume

' resume-setfsb.vbs
' Automatically calls your setfsb commandline  after a resume from hibernate or standby
'the script loops forever, so you will have to kill it using the task manager
'if you want to stop it, look for wscript.exe


Option Explicit
'[ User-edittable section ] ************************************************************************
' Put your commandline entry to call setfsb below as resume_script variable
' NOTE: Script doesn't like to refer to a directory name with spaces in it,eg C:\Program Files
'       so consider copying setfsb in some short directory name location OR use a batch file instead.
'       
'                              setfsb -w [wait(s)]-s[FSB] -i[incr] -cg[PLL]
Const resume_script="c:\setfsb\setfsb -w0         -s166   -i13      -cg[ICS9LPRS355BGLF]"
'Above line is for a HP 2510P, PLL=ICS9LPRS355BGLF, desired FSB=166, increment FSB up in 13Mhz increment
'*************************************************************************************************

Dim wmiPowerManagementEvent, oShell, strcomputer, objWMIService, colMonitoredEvents,strLatestEvent,run_it

Const cntEventEnteringWakeup = 7
set oShell=createobject("WScript.Shell")

strcomputer="." 
Set objWMIService = GetObject("winmgmts:" & "{impersonationLevel=impersonate}!\" & strComputer & 

"\root\cimv2")
Set colMonitoredEvents = objWMIService.ExecNotificationQuery ("Select * from Win32_PowerManagementEvent")

' Run setfsb initially on startup
set oShell=createobject("WScript.Shell")
oShell.Run resume_script,7, true

' Now loop looking for a resume-from-standby or resume-from-hibernate to re-rerun setfsb
Do
    set strLatestEvent = colMonitoredEvents.NextEvent()
    If strLatestEvent.EventType = cntEventEnteringWakeup Then
        oShell.Run resume_script,7, true
    End If
Loop

Above code was simplified from resume-faster.vbs which does the above as well as has AC<->DC power change event detection so could code up a custom solution. Code indicates a call to faster.bat that in my case has provision to call memset (to change memory timings) as well as rmclock profiles to change CPU voltage. Rmclock doesn't allow AC/DC profiles, but can capture the registry details and dynamically import the AC or DC rmclock settings prior to loading rmclock.

[moral hazard]

I think I'll have to swap hard disks to my hard disk with visual studio so I can run that.
Thanks for posting it

Does it use a lot of CPU though?
If so then it would decrease battery life

[nando4]

Originally Posted by moral hazard

Does it use a lot of CPU though?
If so then it would decrease battery life

Don't know much about VBS scripts, other than I put together code I found on the net.I didn't need any special software to run it other than XP. VBS scripts are to XP like batch files are to DOS. No decrease in battery life.. it's monitoring for a resume-from-standby/hibernate event.

[moral hazard]

Now I get an error saying:
line: 36
char: 1
error: The system cannot find the file specified.

[nando4]

Originally Posted by moral hazard

Now I get an error saying:
line: 36
char: 1
error: The system cannot find the file specified.

I've post #2 above. Maybe if you had used long filenames or directories with spaces you'd see something like this. I've altered the script to make it as simple as possible and elaborated more on what to do and how to set it up. I can't really explain how to check if doesn't work for you since it works fine for me.

[moral hazard]

in post #2 the link to the file is not working.

[nando4]

Originally Posted by moral hazard

in post #2 the link to the file is not working.

Updated post #2 with this link: resume-setfsb.vbs

[moral hazard]

Ok that link works.

But I already copied the code to notepad and it works great

I think the problem before was the encoding notepad used (I changed it from ANSI to Unicode).

Anyway thanks for the awsome code Nando, if I could give you rep I would

[moral hazard]

My next VERY IMPORTANT QUESTION:

Does the spread spectrum affect the maximum clockspeed?

read this first:
http://en.wikipedia.org/wiki/Spread_spectrum

Just read the part under the heading "Spread-spectrum clock signal generation".

Here is an important quote

...An ability to disable spread-spectrum clocking for computer systems is considered useful as the spread-spectrum techniques used can affect the maximum clockspeed achievable by the components involved due to clock skew, affecting overclocking efforts.

It's very easy to dissable the spread spectrum with setfsb.
For example with my PLL you can see on page 14.
Byte 4 "Output Enable and Spread Spectrum Disable Register", bits 1 and 0.

A value of 0 for bit 0 of byte 4 will dissable the spread for PLL3.
A value of 0 for bit 1 of byte 4 will dissable the spread for PLL1.

So I believe dissableing the spread spectrum will increase interference to other electronics, but will give greater performence.

Am I correct or is it a bad idea?


Edit: After reading more about the spread spectrum, I'v decided to dissable it.
It seems that having the spread spectrum enabled causes instability and a performence loss.