Overclocking Using Default Settings

[ccokeman]

It was a QX9650. VTT is the quick and easy way to get there but adjusting the GTL ref voltages is the more time consuming way to get a Quad overclocked. The duals are just much easier to clock. fewer "rules" . Not to demean Praz's results in any way they are nice especially on auto settings. I would like to see the tweaked results as well.

[dr_bowtie]

That sure looks nice Praz....

 

Didnt we tell you this guy was good....?

 

 

I am sure if Praz says it wont gain much by changing the settings his word is holy grail....

 

This guy test more boards before 6am then most people do all week....so he outta know..

[kingdingeling]

This guy test more boards before 6am then most people do all week....so he outta know..

Sig worthy material right there

[Praz]

I copied these screenshots from a post I did elsewhere. Thought some might like to see the capability of the board. They in no way demonstrate the ability or inability of a user's capability of setting up the board. The intent of the original post was to give users that may be having difficulties a baseline of the board's capabilities.

 

If this was a board I was actually using as a 24/7 setup all values would be set manually. Not only for whatever gain there may be in performance but to also lock the values in. Once I have a setup configured I prefer not to use auto settings as there is always a chance that one or more values could change on their own at boot-up.

 

With the BIOS configured manually there will definitely be a performance increase. Although the increase would be not as much as one would think. Oskar has done a good job engineering the default values of the BIOS. Quite a few of the settings default to values users would be satisfied with if the values were being input manually. The downside is with inferior components some users may experience a no-boot condition at times using default values.

 

Manually tightening up the settings to see the difference in performance is an idea worth exploring. However its not something I can pursue at this time. Users are wanting help with the holy grail of overclocking. 500 FSB and 1200MHz memory. Not sure why though. 500 FSB is not the x38/48 chipset's strong point as it pertains to needed voltages or performance. Not to mention most users will never have 2x2GB memory capable of running at 1200MHz.

 

When overclocking a board once it is set up fine tuning GTLREF values, if available, should always be attempted. Usually the result will allow the lowering of one or more of the voltage settings.

 

Unofficially, Intel is very specific on max voltages for PLL and VTT as they apply to 45nm processors. For 24/7 use they caution that VTT should not exceed 1.40V with 1.45V the max for short benching sessions. Unlike VTT, PLL will not kill the CPU but an excessive amount will degrade the FSB potential of the processor. For 24/7 use a max of 1.6V is suggested with 1.70V being the max for benching. I haven't killed any processors with VTT voltage but have ruined (FSB) several 45nm and 65nm processors with high PLL.

[Puck]

I copied these screenshots from a post I did elsewhere. Thought some might like to see the capability of the board. They in no way demonstrate the ability or inability of a user's capability of setting up the board. The intent of the original post was to give users that may be having difficulties a baseline of the board's capabilities.

 

If this was a board I was actually using as a 24/7 setup all values would be set manually. Not only for whatever gain there may be in performance but to also lock the values in. Once I have a setup configured I prefer not to use auto settings as there is always a chance that one or more values could change on their own at boot-up.

 

With the BIOS configured manually there will definitely be a performance increase. Although the increase would be not as much as one would think. Oskar has done a good job engineering the default values of the BIOS. Quite a few of the settings default to values users would be satisfied with if the values were being input manually. The downside is with inferior components some users may experience a no-boot condition at times using default values.

 

Manually tightening up the settings to see the difference in performance is an idea worth exploring. However its not something I can pursue at this time. Users are wanting help with the holy grail of overclocking. 500 FSB and 1200MHz memory. Not sure why though. 500 FSB is not the x38/48 chipset's strong point as it pertains to needed voltages or performance. Not to mention most users will never have 2x2GB memory capable of running at 1200MHz.

 

When overclocking a board once it is set up fine tuning GTLREF values, if available, should always be attempted. Usually the result will allow the lowering of one or more of the voltage settings.

 

Unofficially, Intel is very specific on max voltages for PLL and VTT as they apply to 45nm processors. For 24/7 use they caution that VTT should not exceed 1.40V with 1.45V the max for short benching sessions. Unlike VTT, PLL will not kill the CPU but an excessive amount will degrade the FSB potential of the processor. For 24/7 use a max of 1.6V is suggested with 1.70V being the max for benching. I haven't killed any processors with VTT voltage but have ruined (FSB) several 45nm and 65nm processors with high PLL.

Interesting. A few points of relevance floating around that may or may not be true...

 

X38 boards overvolt the VTT by ~.1, a healthy margin, which is why I think it commonly causes degredation since most people are running quite a bit more then they set in BIOS. I would be more worried about PLL or vcore then VTT assuming you keep VTT at 1.45 or lower...seems like they have killed more chips over at XS then the fsb termi volts, and your personal experiences back that up as well...

 

Also, straight vcore can kill the 45nms overtime even with all other specs in range...over 1.4v on air has caused degredation to chips whose VTT and PLL were all within recommended spec. They clock very high so it is a shame they really are pretty fragile :'(

 

Depending on who you ask, some "degredation" may be false - there are some that say the 45nm chips themselves need a burn-in period. For example they may hit a certain clock at 1.3vcore brand new, but after two weeks they might need 1.35 for the same exact clockspeed keeping everything else equal. From then on they are good to go. An increase in voltage requirements has become quite common after a few weeks of running, but from then on after the initial drop the chip is then fine at that speed/clocks (?). According to some if a .025-.05v increase in voltage needed happens just a few weeks after purchase it may just be the chip has burned in, but after a few months it is most likely degredation. Quite confusing to the enthusiast!

 

Once again, interesting post and thanks for taking the time out for it. There are quite a few topics and theories floating around the web of the same thing...I think the whole community in general is still slowly finding out what works and what doesnt work with these 45nm chips. I can only imagine the amount of dead or damaged chips we will see with bloomfield as everyone tests different voltages and settings...

[Praz]

I guess there will always be confusion for users as far as BIOS settings go. GTLREF settings is a good example. A basic design that's been around and used for years in signaling circuits. One of the best tools we have available to us for overclocking yet so there are so many that can't get their head wrapped around it.

[kitfit1]

Praz is right of course. If you look around the net at peeps claimed GTL settings and VTT, it's patently obvious that they have, a)just copied someone else's or b)hav'nt really got a clue what to set so just use one of the published charts. I often wonder how on earth a claimed clock can be stable when i look at peeps published settings. A point in question would be someone using a VTT of 1.4v+ on a dualcore at say 500fsb. Totally unneaded and probably unstable as well, too much VTT is just as bad as not enough.

[saint-francis]

When overclocking a board once it is set up fine tuning GTLREF values, if available, should always be attempted. Usually the result will allow the lowering of one or more of the voltage settings.

 

Always? Even if the OC isn't very high? GTL tuning is an arduous (at least for me) process and I haven't been able to ring any more from my current settings by adjusting GTL values. Maybe I jut haven't tried long or hard enough.

[Praz]

If voltages are within a user's acceptable range of use then there would be no need to fine tune GTLREF. And yes, for some this set of adjustments can be a tedious task.