Why under clockers should be shot! Under Volting isn't easy! :)

[Lupiron]

Alright, I am way tired of seeing the sad arguments from the supposed under clockers.

A/GTL +

Thats the signaling design method used currently by intel!

Depending on your chip type, under volting it can in effect over volt the thing!

VTT = 1.2000 Standard.

The "A" in A GTL + stands for assisted!
A device within your chip is designed to always pull your voltage up to VTT when transiting from a low level logical state to a high level logic. It assists the signal by quickly pulling it up to VTT then releasing its effect.

Now, if the pull up device always pulls the voltage up to VTT. 1.2000 normal, and you have it set for lower, you still get 1.2000 volts to the chip every low to high logic transition.

So, can you really under volt a chip if it always sees 1.2000 when it changes state?

So, all you under clockers, eat VTT anyways!

Hahahaha!

(q6x00 series for sure, but since all the chips use it... the new 45 nm ones get a lower VTT, remember? Now you know why. It always brings the voltage up to VTT. So they obviously had to change VTT to allow for the lesser voltage use of the new 45 nm chips!)

You may also note that if VTT min is 1.2000, that also means that 1.2000 is the Lowest VID chip in the q6x00 series, and I have one!!!
The VID of a chip must be equal to or more than the VTT because the pull up device will always bring the voltage up to VTT. So in effect, if the VID was under VTT, then that chip would always get over volted to the VTT value.

--Lupi

 

[Andrius]

Talk about dumb luck. My Q6600 needs 1.225V(BIOS) for 3.0GHz.
When I ran it at stock and 1.075V it was running 10C cooler but since you say it gets hit by 1.2V anyway ...

I still think the bus interface is electrically seperated from the core in some way but I don't really know enough to be sure. It just runs way cooler at 1.1V than it does at 1.2V or 1.3V so ...

Anyway where can I read some more about A/GTL+? Have a good link?

 

[Lupiron]

I didnt mean it "gets" 1.2000 anyways, I mean it peaks at it. It will only pulse to 1.2000 and then return to its allocated voltage each time the processor switches from a low level logic to a high level logic.

And yes, it is separate, but your chip still hits 1.2000!

Though it may DRAW less than 1.2000 for operational use, it will HIT 1.2000 each transition.

--Lupi

 

[Andrius]

Yes, but if you think about it. At 6GB/s (an 80% transition rate) it's basicly always(at least as far as we can tell) at 1.2V.
The currents for such rapid changes must be mindblowing ...

 

[Lupiron]

Ahh, but you see my friend. Intel decided that they'd bypass that whole thing with needing actual high current! They are relying on the fact that electrons pass between minimal barriers at a given rate.

The 65 NM design size makes the electrons pass BETWEEN the individual 500 million transistors! Because the distance is so small between them. So intel isnt using anything in your chip but a single thing. Electricity.

While VTT is independent as far as its ability to distinguish between two states, Resisted electricity, or non resisted electricity, it is also tied to the primary thing. Electricity in the form of VCore. It uses the same VCore source to distinguish the signals, and set a reference point, min and max ref, and min and max vtt.

So OCing the chip and making it use more VCore to sustain its operation as all electro-mechanical devices must, it has the side effect of weakening this needed 1.2000 volt level, and its ability to use VCore to tell between the states.

--Lupi

 

[Lupiron]

Damn Double up!

--Lupi

 

[Andrius]

Okey. I remmember something about optimization on QP FSB during the P4 days at the University and transfering FULL WORDS without transfering anything (or that's how it sounded). I also remmember (from a different lecture) that for any state change of a register you need some current. The shorter the time it takes to change that state the larger the current has to be. But since I have no idea how the current chips understand different voltages and what the voltage difference is between logical low and logical high state I'll have to take your word for it.

So the simple way to improve reliability and/or stability would be to make VTT = Vcore but that kills chips does it not?

 

[Lupiron]

VID can never be over VTT. You never want VTT higher unless in EXCEPTIONAL circumstances. Although your chip can take it, your settings are set to limit the voltage in the chip to the VID as its MAXIMUM. Thats why VDrop is in place. Select 1.2000 in your BIOS and its droped to some odd 1.1650 so in case of spikes, it NEVER crosses the 1.2000.

In VTTs case, VTT is the termination point, so it ALWAYS can terminate a signal up to its rating. Thats for your chips assisted pull up device. the pull up device will pull over VTT as fast as possible, but because VTT is in place, it terminates the voltage at 1.2000, just as it is supposed to do!

Now if your VID was 1.1900 and not 1.2000, the pull up device would ALWAYS pull it up past it, to VTT. 1.2000

That may not be good for certain chips. Most can easily take 1.2000. But if you think you are running your chip lower than VTT, you are half right only!

This post is designed to make you search for the answers!

--Lupi

 

[Andrius]

My Vdrop is only 0.05V (IE 1.225V BIOS to 1.22V CPU-Z, idle and 50% load are the same and 1.20V with Prime95 for 100% load).

So since my chip is over 1.2V VID it can clearly take the 1.2V VTT.
But since I never changed VTT when running below 1.2V VCore the chip is still getting 1.2V for every change on the bus interface.

Now if only this would explain why my chip only goes from 54C max at 3.0GHz and 1.225V to 55C max at 3.3GHz and 1.2975V I'd be happy with that.

I'll really have too look into this a bit more...

 

[Lupiron]

I am not the Temp man. A few hot cores don't scare me. I could care less, lol! But I do have one weird chip. A VID 1.2750 chip. Its core are ALWAYS the same when loaded. Its amazing! Is it a perfectly placed set of cores that they are all read properly??

It freaks me out, man, and yeah, its best to read up on A GTL +. I am keeping tight lipped for now because I have managed to use it to eek out some gains! Once you all test yer own and realize what happens to stability at the higher speeds and more VCore need/used.

--Lupi

 

[Lupiron]

I even made the VCore high for its needed 3.6 and they still stayed the same!

--Lupi

 

[Andrius]

My Q6600 is my primary workhorse rig so I don't want to risk much. My temps are 10C appart at times. The 55C is the hottest core.
I'm terrified of voltage. When it didn't run stable at 1.4V BIOS I just stopped trying.
115% voltage for 150% clock speed. I really should try 1.5V BIOS and see if it runs at 3.6GHz

 

[Lupiron]

Don't worry about voltage so much, because its designed to take 1.5 volts. Supposedly the voltage doesnt void the warranty, changing the speed of the chip does. Its made to opporate with that as its maximum spec voltage. We all know it can go higher, but so what?

So what? You should be fine with 1.5 volts as its maximum with each chip.

*Disclaimer* If yours blows up, I didn't say that!

--Lupi

 

[Andrius]

That disclaimer is what scares me.
I've built a few computer over the past 2 years with Intel's chips for my friends. I'm not scared of overclocking those. All they do is game and read mail. I however earn money with mine and can't take risks like that.

I'll get another build going in July and I'll tweak it over then ...
I just have to hold on and fight the urge to do it before that.

 

[Lupiron]

::chuckles.:: Yeah, I tried to fight it as well, and look where it got me!

But I guess I do have three operational computers. If one blows up I will be sad, and bury it with full honors! But can can move over to number 2 if I had too.

--Lupi

 

[Andrius]

I have 3 as well. But 2 were around when Lara Croft still had triangle shaped boobs.

 

[roadrunner197069]

^lol

 

[Lupiron]

Thats great, lol!

Hehehe~

--Lupi

 

[Andrius]

^That's why they are omitted from my configuration.
If this thing(PC1) goes south I'd have to work on either my notebook(P4M) or PC2 that's getting a makeover in July (spending tax returns on usless not really needed upgrades is always fun).