ok, I get asked a lot to explain the GTL+ reference levels for intel core2 Cpu's
so, here are 2 links to some Long and technical explanations to what it is
http://www.thetechrepository.com/showthread.php?t=87
http://ocxtreme.org/main/?q=node/1
Understanding Intel Core2 GTL+ reference levels
ok, I get asked a lot to explain the GTL+ reference levels for intel core2 Cpu's
so, here are 2 links to some Long and technical explanations to what it is
http://www.thetechrepository.com/showthread.php?t=87
http://ocxtreme.org/main/?q=node/1
Last edited by Just Learnin'; 07-15-2008 at 10:50 AM. Reason: put an "s" in undertanding
I'm glad you started this thread, cause I have a few questions.Originally Posted by nutcase
ok, I get asked a lot to explain the GTL+ reference levels for intel core2 Cpu's
so, here are 2 links to some Long and technical explanations to what it is
A good explaination of what we are trying to achieve, although it only covers the CPU GTL's.
And can we assume from the fact that he never updated this thread, that it never actually worked, & his voltmodded P5W DH Deluxe ended up with no higher FSB than anyone elses?
I did email him at the time with the link he left for questions & comments, but he never replied.
A lot of technical stuff here, but for those of us who don't have an oscilloscope, perhaps it would have been better if someone had answered delirious's post.
If you are trying to find the highest stable FSB, when do you start playing with the CPU GTL voltage & the VTT, the NB voltage, or the NB GTL voltage.
Are these settings effected by changing the CPU PLL? If so do your optimum PLL settings work the same with the CPU @ 3gig as 5gig.
ie can you spend many days finding those "magic" settings with the CPU multiplier on 6X & then just put the multiplier on 11X?
And does memory frequency need to be part of the "magic" mix?
If so, is it only frequency & not NB workload & the best CPU GTL & NB GTL settings for 2000MHz @ 500FSB would be the same for cas#9 as cas#7?
Although this was written for the DFI P35, it is the best attempt at explaining how to find these settings that I have come across. Starting page 12.
Although my limited playing with the P5Q3 Deluxe, seemed to suggest fine tuning your NB GTL voltage with your memory on 1:1 & being Prime stable,
wouldn't post when I changed it to 1:4 in spite of my memory previously being stable above that speed.
So basically.... Which settings are completely independent of others, & which have to be in the CPU GTL / FSB Termination voltage & NB GTL / NB voltage mix?
P5Q is gonna be a bad example. its voltages are all over the place.
IE: what you set is not what you get.
its also a very flakey board
sorry it took so long to respond but been busy.
and here I get long winded
first, here I will try to explain everything:
first is the PLL voltage. PLL stands for 'Phase Locked Loop'. there is actually 2 PLL's in an intel system. The first PLL is the Clock generator on the Mobo which determines your FSB speed. The second is built into the Cpu and this PLL is determined by the Clock coming from the Mobo (clock generator) and by the CPU multiplier and it determines the frequency the cpu runs at. Basically when adjusting CPU PLL voltage, you increase the strength of the clock signal to prevent the cpu from stuttering or Halting from a too weak of clock signal.
VTT voltage is a reference voltage only. It is always going to be around 2/3rd's of the VCC voltage. VCC will be different depending on what area it is in. The Cpu Vcore is the Cpu's VCC voltage for example. VTT is just the maximum usable Voltage swing for the data signal. this is always pretty much set and non adjustable.
and here we go:
GTL reference Only affects the Data and address signals running between the cpu and the Northbridge. all it does is move the reference location where the signal transitions from a 0 to a 1.
when Overclocking, to get stability back they reccomend doing the following:
1. when hitting a stability problem. they recommend try adjusting GTL reference first. If this clears problem, then the data errors were being generated on the FSB between the Cpu and northbridge. the added benefit of this is you fixed the error without increasing Vcores which generates more heat
2. If it does not clear problem, then the errors are being generated inside the Cpu or in the Northbridge. and only adjusting the NB Voltage or cpu Vcore will clear problem.
3. adjusting PLL voltage is for when cpu starts to stutter or halt when overclocking.
I am still doing some research on the memory though. But the Memory settings and ratio's only affect the data between the memory and the northbridges Memory controller. when I am done with that, I will post it later.
Here's where it gets confusing (like it wasn't already)
When people use different terminology for the same thing, & worse the same terminology for different things.
In the link above Kristopher Broughton says Intel claim the maximum VTT for there desktop processors is 1.55v & that this is what Asus refer to in the bios as FSB Termination voltage.
Asus of course give you much more than this, because they think it will kill your processor before your motherboard.
Incidently, according to readings taken by some at XS, setting FSB Termination Voltage to auto on the P5Q Deluxe, has it on 1.53 (rather close to Intels maximum for their CPU's, when people who aren't overclockers think Auto would be stock 1.20v or 1.10v for Yorkfields.)
I still haven't got my P5Q3 Deluxe back from RMA, so I haven't been able to check mine on auto.
We now have CPU GTL reference voltage, & the NB GTL reference voltage, so presumably you would try all bootable settings in both directions for one & if no luck, put it back where it was & do the same with the other. If still no luck, put that back & MR VOLTAGEGTL reference Only affects the Data and address signals running between the cpu and the Northbridge. all it does is move the reference location where the signal transitions from a 0 to a 1.
when Overclocking, to get stability back they reccomend doing the following:
1. when hitting a stability problem. they recommend try adjusting GTL reference first. If this clears problem, then the data errors were being generated on the FSB between the Cpu and northbridge. the added benefit of this is you fixed the error without increasing Vcores which generates more heat
What form does this stuttering take when your running OCCT?3. adjusting PLL voltage is for when cpu starts to stutter or halt when overclocking.
So you know you need PLL voltage as apposed to just more vcore?
Keep em coming mate, I'll get the hang of this stuff yet.I am still doing some research on the memory though. But the Memory settings and ratio's only affect the data between the memory and the northbridges Memory controller. when I am done with that, I will post it later.
Saw quite a bit of halting in my Pi 1M runs.adjusting PLL voltage is for when cpu starts to stutter or halt when overclocking.
Surely that is just that program halting, If your CPU halted, wouldn't the system reboot.
Or is that a sign you need more PLL voltage if it freezes, as apposed to giving an error?
I had a little trouble wording this.
The people at XS are going a little overboard with this. All Asus does is set the VTT to the max allowable to help the average overclocker when raising the FSB. gigabyte does the same thing. The one thing the people at XS are not taking into account is that Intel has specific safety margins for error accounted for in their specs and most specs I see with Intel are set at +/- 5%. so if the VTT max is 1.55V, The margin will allow up to 1.62V without possible damage to the transistors in the cpu.
the reason for this is to account for margins for external components. Resistors can vary as much as 10% from their rated value. Capacitors are even worse as they can vary as much as 25%. InteL Usually will give this margin in the notes for the specs.
Problems with the Clock (pll output) are going to show up as Many different problems.
All functions in a digital circuit (cpu for our use) are done based on the clock. All data movement, calculations, etc are done based on the clock cycles. OCCT will not catch about 60% of the problems with clock signals.
I am going to use an extreme example here:
one cpu instruction is to move the data from Register A to Register B. (a register is just a temporary storage area built into the cpu). If while moving the data, register B misses a clock, the Data in register A and Register B will not Match due to an incorrect transfer.
If the rest of the Cpu catches the error, it now must determine if the data in register A or register B has the incorrect data and flush the incorrect data,and redo the move causing a performance loss. this may not show up in any tests though when you think your clock cycle at 3Ghz is only 3.3 microseconds in time (or is it .33 microseconds?).
now, if the cpu does not catch the error, and it's next instruction is to add the Data in register B to the data in register C, now you have a calculation error and we know what happens then
Now Back to asus motherboards. Asus has preset values inside it's bios when the voltage's are set to Auto. and easy way to show it is set your cpu Vcore to auto and start raising your FSB. you will notice at certain FSB's, the Cpu Vcore automatically has increased.
if you have somebody complaining about a cpu overheating at a low cpu clock on an Asus Mobo (gigabyte also) have them check the vcore Setting. Some Asus Mobo's I have seen the Vcore as high as 1.55V at 350 FSB when set to auto!
i really wish Asus would stop screwing around with the bios files.
you should get what you set, period.
I did some research and found out the following:
FSB termination voltage adjusts the VCC termination voltage. Not the actual VTT voltage.
ok,
I am going to expand on this in seperate threads.
I have asus, gigabyte and EVGA (XFX) mobo's at my disposal and I will setup threads that explain what each bios function does.
Posting Permissions
Reply With Quote
Bookmarks