Peltiers nessassary for anything?

I can understand people wanting below freezing CPU temperatures, but a peltier requires a good watercooling setup anyway. Wouldn't removing the pelt keep a system at admirable temps? I know this is overclockers.com, but isn't a 31°C CPU as stable as a -10°C one?

As I understand it, and I may be wrong, if you keep the CPU down around 0C you effectivly eliminate heat as an overclocking issue. Granted some of it probally has to do with my CPU is colder than yours, but I have a feeling that some of the people here are just interested in getting the best overclock they can while others enjoy the cooling aspect more than the actual overclock itself. I say to each is own and let everyone have there fun where they have there fun. It wouldn't be nearly as interesting a forum if everyone was happy with 31C, at least in my opinion.

PiSan314159 said:

I know this is overclockers.com, but isn't a 31°C CPU as stable as a -10°C one?

Click to expand...

No, the higher the overclock, the lower the temperature needs to be in order to reach system stability. It is all about getting those few extra MHzs..

If you drop your temps by 40C with a pelt you are doing really good - most people who try to can't.

Voltage is the ultimate determinant of maximum overclock.

The colder you keep your chip, the higher voltage that is safe to pump into it.

The higher your overclock.

WELCOME TO THE FORUMS!

IMOG said:

Voltage is the ultimate determinant of maximum overclock.

The colder you keep your chip, the higher voltage that is safe to pump into it.

The higher your overclock.

WELCOME TO THE FORUMS!

Click to expand...

Thats not correct!! You cant just pump 1.9 volts into a P4 even if you have it cooled at near zero temps.

You see the problem with max oc is not the temp as many people think, its a phenomenon called ellectrodialisis or something, this means that because the new P4 chips are so small , the transistors in them are also very tine, actually they are only 3-4 atoms thick, so when extreme electrical power ,extreme volts that is , pass through them there is a chanche that because of the strong magnetical fields the extreme volts have, the transistor atoms will desolve, the electrons from their atoms will be pulled from the magnetic fields and the transistors (cpu) will break up from the inside. So you should not let high volts pass through your cpu, high volts meaning very higher than what the cpu has for default. Now because the P4 Bs and Cs (nortwood) are made with 0.13 microns that means that the transistors are even smaller than the P4 As which were made with 0.18 i think. And if the transistors are smaller (lets say just 2 atoms!?) then they are more vurnerable to magnetic fields than their predecestors.) There were many people who had fried a P4A with overvolting, actually the whole think of frying (or actually electrodializing) a P4 A took great dimentions as more and more overclockers found their CPUs dying suddenly after 2-3 months of successfull overclocking at extremely high volts over 1.75-1.8. The whole thing waa named NSDS, that is Northwood Sudden Death Syndrome and at the begining noone knew what was causing it, until some people scientifically found out that it was electrodialisis and not overheating as some thought although the dead cpus run at low temps.
That is history and took place almost 2 years ago. Now we have P4 Cs (or even Bs) and people took their lessons, we are more carefull with overvolting the core and thats why we dont hear a lot of poeple crying about burned - electrodialized CPUs.
Anyway below 1.7 Volts is what we found out was safe for the P4As, as the Bs and Cs are theoritically more fond to electrodialise because of their smaller size then lets say that below 1.65 is is safe for our cpus. I personally try to run at 1.7 max.


So you see that the oc limit is not the temps, but the volts (for the cpu that is and specifically the P4) , AMD Cpus dont have problems with extreme overvolting and that is probably cause of deferent condtruction teqniques.

Any comments?

If you don't agree with that basic statement I don't know what to tell you. Take a look around the boards and look at the trend in OC's... Higher OC's will have higher Vcores on average. Usually when people move to TEC's or phase change, they bump their Vcore a notch or two higher, and they also get a considerably higher overclock. This is how the people with the HIGHEST OC's achieve them - they all run phase change with a LOT of voltage.

I clearly wasn't talking about SNDS as its something seperate that you have to look out for if it applies to your chip - the question was if peltiers are necessary, or useful, for anything. You are right, but we are talking about two seperate things.

First off the term you are looking for is electromigration, not electrodialisis. This can occur in any size device not just sub micron size, however it does become more of a problem the smaller you go in size. Second the need for overvolting comes from the fact that the gates in the cmos transistors has a small capacitance, which needs to charge up with some electrons in order to create the electric field to turn on the gate (hence the name “field effect transistors”). In order to to do this we need more current or electrons per second. An easy way to do this is to up the voltage a little, and the current goes up, and the gates charge faster which in turn means the CPU runs faster. However running faster means that we generate more heat. Heat causes failures on it’s own without going to higher voltages, but it also can increase electromigration, which will also cause a failure. Many people have conflicting opinions about electromigration and how much it shortens the life of your CPU. But one thing is for sure heat is the enemy and must be removed!
Overvolting also has an upper limit in which the gate oxide brakes down and acts like a big hole for electrons and the transistor no longer functions. At this point the smoke gets out and your CPU no longer functions like a CPU, but more like heater and then a paper weight.

That damn magic smoke that keeps a CPU working never does want to stay put.

Just keep that smoke in and it will work forever Ha Ha Ha!

The transistors are a LOT bigger than a few atoms. They're working on transistors for FUTURE tech that are just several molecules. This isn't even developed yet. .13micron is roughly 1000 atoms. Not that it changes your point... just dont want people getting the wrong idea.

I thought 2-3 atoms seemed like an exageration.

Maybe I'll consider a peltier for the next computer down the line. Thanks for the replies and info.

The size that is stated as 0.13 micron or what ever, is not the thickness of the transistor. It is the min. feature size on the chip and usually is the channel size for gates. However when the gates widths are scaled down so is the thickness of the gates as well. There are some design rules that must followed for things to work right as well as for the lithography thats used.

When I said 2-3 atoms i was exagerating, but you got the point.






IMOG ofcourse Volts are one of the main factors in oc capacity (how much you oc) but its not a danger free factor!!

read what you said:

Originally posted by IMOG

"Voltage is the ultimate determinant of maximum overclock.

The colder you keep your chip, the higher voltage that is safe to pump into it.

The higher your overclock."

Its not like that! there is a limit to the volts you can throw at your CPU, a limit of safety (overvolt and you will make the smoke come out as you guys said) and a limit of oc gain. Do you think that if you could safely put 3.0 V !!! at your P4 2.4C it would oc to 5Gz?? Ofcourse not!! Havent you noticed that there is a contruction-physical limit to how much oc each cpu can get?? Above that limit whatever volts you throw to it it doesnt go any higher!!

look at this:

Dont you guys agree on that?

Extra volts is a way to push your cpu to its limits but it is the cpu it self that determines the max oc it can get and how stable it will be. You just help it with the extra volts.

And the thing you said IMOG that "The colder you keep your chip, the higher voltage that is safe to pump into it."

thats not deeply true. There always is a max volt that you can throw at your CPU whatever the temp. And again there is a point in the volts you give above which you cant see any oc gain. Even if you have your cpu at -50 , if you throw 2.0 Volts at it it will sooner or later die..

Lower Temps do however minimize the risk of damage (electromigration) to your cpu when you feed it with extra volts.

Anyway the whole concept you wanted to pass IMOG about the original question of whether the pelts are any good, is that they do allow you to feed your cpu more volts with safety , thus increasing your oc (cause more volts mean more oc) . Isnt that what you wanted to point out? Well the safety part is true, not the oc though.

What if you have reached your cpu's limit? I mean if you feed your cpu at air 1.8 Volts. and it doesnt get any more oc even if you feed it 1.9 (it reached its oc limit) . By going to below ambient temps (or even below zero) you DO ,as you correctly pointed out , make it safer for your cpu to run with that volts but i dont think it will gain anything in terms of oc. Maybe more stability. or a few Hrz. You will get these few Hz but then you will find youself feeding your cpu 1.9+ volts. and that is something i wouldnt do to my cpu. Long term anyway.

But these are my thoughts. Maybe i am wrong , dont know. what do you people think?

Still, wouldn't normal watercooling and peltier setups both keep temperatures low enough to safely pump tons of volage through? Wouldn't anything under about 40°C be equally stable?

if you keep the CPU down around 0C you effectivly eliminate heat as an overclocking issue

Click to expand...

Nah, that won't happen until you hit 0K.

Wouldn't anything under 40C be stable?

No. This is why people do LN2 experiments and run DIY phase change - it gets you better overclocks. That is an extreme example but you get the idea.

Alienhack, I made the assumption that people would understand that you can't just put 50 volts through your core no matter how cold it is. That's a fairly safe assumption to make I think.

Anyway the whole concept you wanted to pass IMOG about the original question of whether the pelts are any good, is that they do allow you to feed your cpu more volts with safety , thus increasing your oc (cause more volts mean more oc) . Isnt that what you wanted to point out? Well the safety part is true, not the oc though.

Click to expand...

Right, thanks for trying to see my point.

Doc explained why the OC part of this relation is true. It's a better explanation than what you can get most anywhere else, and I won't bother trying to explain it better, as I couldn't.

The way I look at it... The best overclock a chip can get is with temporary LN2 cooling. You will never find the limit of any chip on air cooling or watercooling. Even if you have a crappy overclocker, if you get your chip cold enough, it will produce slightly better.

Construction-physical limit? Sure some chips seem to have a very hard time at a given Vcore and FSB/multi settings, but give it considerably better cooling and a little more juice and you might be suprised to find the limit just isn't really there.

If you are trying to say that higher voltages shouldn't be used, because you are rolling the dice with the life of your chip - you might be right. At the same time though, we are overclockers and there are plenty of people who have ran with considerably high vcore's for over a years worth of time. YMMV always plays into things though.

Not entirely True IMOg, about the colder temps and higher Vcores.

The lower the temps, generally speaking the farther you can get with the same Vcore. For example, when you hit say 1.9V on an athlon XP, with watercooling you may be at 2.4ghz, but if you put it on phase-change you might get 2700mhz. Cooler temps make the voltage more stable, however some have had P4's not even work at less than -150C when using Ln2 or liquid helium. Something about it increasing the resistance of the silicon or something I'm not quite sure on.

However, it is a fairly big investment for the last few mhz, but if you are looking to push your hardware to the max, it's for you.