how do you actually damage your cpu by overclocking?

S

screwallyall

n00b
Joined
Sep 1, 2004
Messages
42
i was just wondering, is it safe to give my athlon 64 3200+ 1.7 volts as long as i keep the temps down? or is there still the chance i could do some damage to it either way.
 
I wouldn't push 1.7v to it unless you were on really good water or phase cooling, and even then, I wouldn't push 1.7v to it for very long. I know there are people here that don't have a doubt about 1.7v....so it can really go either way :p

Heat and excessive voltage are what will kill it. Even if you keep the heat down, there is still the chance to damage it from excessive voltage.

I can hit 2.4ghz with my 3000+ winchester at stock voltage. However, to get to 2.6ghz, it takes 1.65v. Not really worth it IMO.
 
In overclocking, you have to ask yourself if the payoff is worthwhile.

In the past, payoffs were huge. A Celeron 300A @509ghz on a BX board was killer. I had a Duron 900 at 1.2ghz for a year before a 1ghz chip even existed. That chip was overvolted for 2 years. It retired to stock speed, and is still running in a nephew's box. On the other hand, more than a few chips got fried by folks, usually by bad installs.

I can't remember what the exact numbers were, but the lifespan of those chips was something like 10 years of continous 24/7/365 use. Trading a few years for the increase in performance was a no brainer.

What degrades chips is electron migration. As voltage, and subsequently heat increase, the "signal" if you will, has a tendency to jump the tracks. The physical proccessor is designed to cope with this phenomenon. It's a normal occurance. The danger is that if it becomes a continous process, the errant electrons begin to etch new pathways. You overwhelm the proccessors' ability to cope with errors. Performance degrades, or the CPU takes a total shit.

Once again, I don't remember the exact numbers, but a 300A was around 30- 45million transistors. Modern processors are what? upwards of 165 million transistors? They are more tightly packed, smaller, and thinner than in the past. Modern CPU's are much more susceptable to electron migration, and damage.

I still tweak the box to see what it will do, but spend 99% of my time at stock speeds/voltage. I'm running a 4200 X2, and I just can't get it to lag anywhere at stock speed.
 
Is electromigration a concern anymore? I read that modern processors are not prone to this as much as old ones.
 
oh noes, im gonna kill my 2.6ghz venice 3000+ cuz i use 1.59 volts instead of 1.36-1.39 :eek:
 
Don't panic. Overvolting is almost essential to get stable overclocks. Understand the potential downside, and fire away!

I rarely have a coherent system in place for more than 2 years. Do I give a rat's ass if I take 5 years off the projected 10 year lifespan of a CPU? Not really. I have yet to fry a CPU. I do spend more for good cooling solutions, and I understand that it's my problem if I ever do manage to fry a chip. I'm a hobbiest, not a pro. I did spend a few years reading every brain numbing white paper or technichal document AMD or Intel published regarding processor design. My theory on electron migration is an extension on previously published articles, nothing more. It seems plausible to me that the tighter you pack the transistors, with the incumbent shrink in physical architecture, the greater potential for electron migration. Neither Intel or AMD have been able to suspend the laws of physics yet. To me, it's a plausable explanation of why overclocking in general doesn't reap the same extreme benefits it did in the old days. Modern processors are torqued pretty tight out the door.
 
Overheating is more dangerous than overvolting imo. I have run a number of CPUs at 1.8v+ (under phase) and never killed one yet. However I did fry my old fx-55 due to overheating (I was out and the block slipped off since I didnt tighten it properly).
 
josanai said:
current kills, although it requires voltage
Click to expand...
Well someone is fresh from A+, pat on the back?

But on topic... is there a chart anywhere with the max suggested voltages for each cpu? Ive heard a few times its on the manufacturer's website, but all I can find is the stock voltage.
 
paintb4707 said:
Well someone is fresh from A+, pat on the back?

But on topic... is there a chart anywhere with the max suggested voltages for each cpu? Ive heard a few times its on the manufacturer's website, but all I can find is the stock voltage.
Click to expand...
130nm = 1.5v
90nm = 1.4v

Max suggested.
 
Borgschulze said:
Read what I quoted.
Click to expand...
You quoted my question then said 1.4 is max suggested vcore for 90 nm cpus. Which is the stock vcore for my x2 3800. Am I missing something?
 
paintb4707 said:
But on topic... is there a chart anywhere with the max suggested voltages for each cpu? Ive heard a few times its on the manufacturer's website, but all I can find is the stock voltage.
Click to expand...
Those voltages I listed are AMD's maximum suggested, because anything more is overvolting.

90nm CPU's default at 1.35v
 
could be that winchesters use a slightly higher voltage since they were the first 90nm parts (to my knowledge anyway).

I can run it as low as 1.17v at stock speed though...keeps it nice and cool :p
 
I run my 3200+ Venice at 2300mhz and 1.538 volts 24/7. I've never had a problem. It's under a 1 year warranty from TigerDirect, so if it dies anytime soon I get a new one. This isn't a reason to do anything crazy too it, but I'm not a afraid to run it up to .2V above stock. This is all on air.
 
Oline61 said:
I run my 3200+ Venice at 2300mhz and 1.538 volts 24/7. I've never had a problem. It's under a 1 year warranty from TigerDirect, so if it dies anytime soon I get a new one. This isn't a reason to do anything crazy too it, but I'm not a afraid to run it up to .2V above stock. This is all on air.
Click to expand...
Uh, I believe if you OC your cpu it instantly voids your warranty. How they could find out I have no clue.
 
I have always thought of overclocking a cpu like a light bulb... you want more light? Put an extra battery in the circuit... it will be brighter but run hotter and if you put another battery on the light will burn out.
 
paintb4707 said:
Uh, I believe if you OC your cpu it instantly voids your warranty. How they could find out I have no clue.
Click to expand...

They could have a register in the cpu record a voltage if it goes past a certain point. There are other ways, and it would be easy to do actually.
 
paintb4707 said:
Uh, I believe if you OC your cpu it instantly voids your warranty. How they could find out I have no clue.
Click to expand...
This isn't an AMD warranty, it's a dealer warranty. I bought an OEM chip and a heatsink/fan at the same place and got a 1 year warranty. The salesperson said that I could overclock it and still get it replaced if it crashes and burns.

And what other ways are you talking about josanai?
 
Once I couldnt OC it as much as I wanted, so I took that FX-57 and ripped it out, threw it in the street, and took a sledgehammer to it. Thats how overclocking damages my CPUs...

now I have a 3000+ :(
 
1.6 seems a little high for a 90nm...I try to stick to 1.55 or less.

I'm not saying you are wrong though...im just not comfortable with 1.6 on air.
 
im on 1.4 volts on my winchester 3500 and im at 2.6ghz atm...i dnt think ill increase the voltage tho...i want to hit around 2.7ghz because its a kwl number :p
 
Oline61 said:
And what other ways are you talking about josanai?
Click to expand...

I was just stating a way that AMD etc. could implement detection of overvolting (I am sure they already know of ways, they probably just don't care to implement it), not that they have it implemented now. As far as them telling if it has been overvolted, I am sure it could be done, but the cost associated with testing would not be worth it.
 
You must log in or register to reply here.
Back
Top