Putting more than 1.6 volts through a cpu (Air Cooled)

Soldato
Joined
18 Aug 2005
Posts
13,173
Location
Shropshire
Right, can someone please explain to me why i shouldnt put more than 1.6 volts through my cpu on air..

My chip is rated to 57oc using A64 Max temp and is being cooled with a scythe ninja with two 120mm fans strapped to it..

I thought that putting more volts through a cpu creates more heat , so therefore you have to cool it better,

So if i put 1.7 volts through my cpu, it gets hotter, but it doesnt overheat because its cooled well.

Please could someone put my mind at rest :)
 
Soldato
Joined
28 Dec 2004
Posts
7,621
Location
Derry
TaKeN said:
Right, can someone please explain to me why i shouldnt put more than 1.6 volts through my cpu on air..

My chip is rated to 57oc using A64 Max temp and is being cooled with a scythe ninja with two 120mm fans strapped to it..

I thought that putting more volts through a cpu creates more heat , so therefore you have to cool it better,

So if i put 1.7 volts through my cpu, it gets hotter, but it doesnt overheat because its cooled well.

Please could someone put my mind at rest :)

I don't think putting 1.7 volts will cause you any permanent damage, if it's cooled properly but I don't think it'll help you increase your clockspeed, I've found the best voltages are between 1.4 and 1.6 and regardless of me pumping more volts through it, I still can't get the system clocked any further and stable.
 

daz

daz

Soldato
Joined
18 Oct 2002
Posts
24,079
Location
Bucks
You shouldn't put much more through your CPU because you can kill it over time. Look at what happened to the people that put high voltages through their Northwoods when they came out (it was a new 0.13 micron process...) the smaller the process, the less volts you should put through them really.
 
Associate
Joined
18 Jan 2005
Posts
1,651
Location
127.0.0.1
If temps are fine more than 1.6v won't kill it. However, as said before, the extra volts will increase electro migration and therefore shorten the life of your CPU if run for any length of time.

I know from personal experience that 1.8v won't kill a 90nm CPU - it's temps that are you main enemy.
 
Gangster
Permabanned
Joined
11 Nov 2005
Posts
362
Kesnel said:
If temps are fine more than 1.6v won't kill it. However, as said before, the extra volts will increase electro migration and therefore shorten the life of your CPU if run for any length of time.

I know from personal experience that 1.8v won't kill a 90nm CPU - it's temps that are you main enemy.

i dont think you can say that yet as 90nm cpu's havent been around long enough to know for sure, 1.8v isnt likely to kill it instantly but over time its gonna do some damage, probably more so than an averagely high temperature. very high temps will kill it quick tho :p

to the OP, go over 1.6v for testing by all means, but dont go silly like keeping it at 1.7v just for another 50mhz
 
Associate
Joined
18 Jan 2005
Posts
1,651
Location
127.0.0.1
tickle me elmo said:
i dont think you can say that yet as 90nm cpu's havent been around long enough to know for sure, 1.8v isnt likely to kill it instantly but over time its gonna do some damage, probably more so than an averagely high temperature. very high temps will kill it quick tho :p

to the OP, go over 1.6v for testing by all means, but dont go silly like keeping it at 1.7v just for another 50mhz

Yep, as I said, higher volts will result in electro migration, thus diminishing the life of the cpu. :)

I was speaking more from a testing point of view.
 
Gangster
Permabanned
Joined
11 Nov 2005
Posts
362
Kesnel said:
Yep, as I said, higher volts will result in electro migration, thus diminishing the life of the cpu. :)

I was speaking more from a testing point of view.

wasnt clear, u dont want to give the impression that 1.8v is ok and someone uses that and his cpu pops quite quickly :p
 
Associate
Joined
18 Jan 2005
Posts
1,651
Location
127.0.0.1
You're right, I should've made that explicit.

1.8v on air or even water will probably run you into temp problems pretty quickly and perhaps damage the cpu.
 
Soldato
Joined
18 Jan 2005
Posts
4,171
Location
Northants
Opterons have been known to die at 1.6 i believe.
And i don't think its entirely true that heat is the only thing thats kills cpus.
It doesn't matter how cool it is, raising the voltage far too high will kill it even if its being cooled to -50.
 
Soldato
Joined
18 Jan 2005
Posts
4,171
Location
Northants
I'm pretty sure i read about this on eoc, but i can't find it.
Thats only a one off, they may run higher, but i wouldn't go near or above 1.6 myself. I assume it was a 1.35v one.
 
Last edited:
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
It's not the temperatures alone that does the damage. I was reading a thread on XS and a guy called Thorry stuck some in an oven.

Thorry said:
The CPU can physicly go up to 200 degrees without an trouble. The first things to fail are the soft metal connections used to connect the core to the PCB. The PCB itself can withstand temps up to 500 degrees without trouble, the core itself is baked at a much higher temp so that's not a problem. The soft metal connections are between 200 and 300 degrees.

We tested this some time ago using a couple of AXP's en P4's they all were fine @ 200 degrees but failed @ 250 degrees (only 1 survived, that was a Duron prolly because of the keramic PCB which deflected some of the heat). This was tested in an oven.

This is offcourse the max temp the CPU will survive, not the max temp on which it will run. We tested the max run temp to about 90-100 degrees, no CPU will run @ rated speed above 100 degrees. The Duron again did best by running stable for 30 mins @ 90-100 degrees.

http://www.xtremesystems.org/forums/showthread.php?p=991338&highlight=soft#post991338
 
Last edited:
Associate
Joined
12 Jun 2004
Posts
629
Location
Blackpool
i ran my xp2500+ at 1.85v just to get to xp3200 speeds
i ran it like this for over a year then one day it refused to clock
ive cleaned the heatsink and fan out and it does clock to xp3200 speeds again but it is no longer 100% stable
its almost two years old now though and my temps were around 45c idle and maxed at about 65c load with 1.85 up its ass so its my old fault for not improving on the stock hsf
lol
ive just got an opty 146 and this baby wont be subjected to the same cruel neglect
 
Associate
Joined
5 Jul 2005
Posts
1,734
Location
Edinburgh
Kesnel is right on this. The actual cause of cpu damage is electro-migration which means electrons have "jump from one wire to another" inside the cpu. This happens because when the voltage goes up they have more energy and so can jump the tiny gap. Also when volts go up temps go up so the electrons have more energy and so it is easier for them to jump the gap. If you have good cooling then they wont have enough energy to do this and so no problem. Remember amd give out rubbish heatsinks for stock volts so a good cooler and a few extra volts will be fine. Also there are 3 yr warrantys on amd so whats the problem - if it dies warranty it and forget to mention overclock. If it dies after 3 yrs then who cares as the cpus are worthless then and a replacement can be picked up for a fraction of the price you originally paid.
 
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
Its been awhile and I'm not a 100% but IIRC. EM is the displacement from momentum of metal interconnect atoms from high current densities. It causes open circuits and short circuits. Local imperfections in the conductors cause high resistances and the current density goes up. Making the electrons smash into the atoms moving them and eroding the metal. Any more heat and greater voltage potential increases the electron energy and momentum force. I vaguely remember an equation, I'll have to look it up. But basically keeping the temperatures down slows the migration, but a higher potential will still cause damage.

Excluding the extra wattage drawn from work done, i.e overclocking. Its worth noting that with assuming constant resistance e.g. a CPU core. That power consumed and thus heat, increases with the square of the voltage. So a small voltage increase can cause surprisingly high increases in temperatures as the thermal resistance of a HSF is constant.
 
Last edited:
Associate
Joined
5 Jul 2005
Posts
1,734
Location
Edinburgh
fornowagain said:
Its been awhile and I'm not a 100% but IIRC. EM is the displacement from momentum of metal interconnect atoms from high current densities. It causes open circuits and short circuits. Local imperfections in the conductors cause high resistances and the current density goes up. Making the electrons smash into the atoms moving them and eroding the metal. Any more heat and greater voltage potential increases the electron energy and momentum force. I vaguely remember an equation, I'll have to look it up. But basically keeping the temperatures down slows the migration, but a higher potential will still cause damage.

Excluding the extra wattage drawn from work done, i.e overclocking. Its worth noting that with constant resistance e.g. a CPU core. That power consumed and thus heat, increases with the square of the voltage. So a small voltage increase can cause surprisingly high increases in temperatures as the thermal resistance of a HSF is constant.


Yes thats right. Power = Current squared x resistance and voltage = current x resistance. So if voltage goes up current has to which means more power.
I think its about an extra 30w for an extra 0.2v or so on a amd dual core
 
Last edited:
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
You beat me to it, meh. ;)

P = V²/R = I²R = VI V=IR

Where R is a constant V~I

The Thermal resistance formula if anyone's looking:
cpu dT = (cpu load T) - (room ambient T)
thermal resistance (C/W) = (cpu dT) / (cpu watts)
 
Last edited:
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
Just to finish this bit. In reality the Resistivity of silicon changes with temperature. For a simpler model over the small temp ranges involved assuming a constant resistance holds. Applied to the CPU, this gives the formula that, from what I can tell, most of the online calculators use.

OC Watts = Default Watts * (Overclocked Mhz \ Default Mhz) * (OC Vcore \ Default Vcore)²

Based on V² and the ratio of work done (frequency).

Even though the Opteron 170 I use is rated at 110W it’s not using anything like that. From the wattage used at the socket, I think its more like 75W at stock volts/speed, or less if you look at TCaseMax/TDP. To get an idea at the rate the power can increase with voltage and clocks. The graph below is a simple example from 2GHz to 3GHz. Of course in the real world not each data point exists because you need the voltage to get the speed. And it’s just a rough estimate and each chip/type would need its own graph.



The thermal resistance of different coolers has quite an effect. Looking at a plot of the stock (2.8GHz) curve against typical HSF to see how the voltage changes the load temperatures. Again it’s just an example but the V² pushes up temps very quickly. Ambient=25C

 
Last edited:
Associate
Joined
5 Jul 2005
Posts
1,734
Location
Edinburgh
fornowagain said:
You beat me to it, meh. ;)

P = V²/R = I²R = VI V=IR

Where R is a constant V~I

The Thermal resistance formula if anyone's looking:
cpu dT = (cpu load T) - (room ambient T)
thermal resistance (C/W) = (cpu dT) / (cpu watts)


Lol - and i just had my electrical power exam which i know i failed
 
Back
Top Bottom