PDA

View Full Version : P4 overclocking


netape
12-30-02, 10:45 AM
Hi, I'm thinking about overclocking my CPU a little more but I'm not sure about some things.

I have my P4 (1.6a) clocked @ 2Ghz (125x16 @ 1.45v), rockstable.

1) I'm thinking about raising my fsb to 133 so it would be a 2.13Ghz @ 1.45v. Is that possible, or should I raise my volts aswell?

2) What do I notice when I'm overclocking my CPU too much? Artifacts? BSODs?

Thanks :D

pelly
12-30-02, 12:09 PM
Raising your voltage would certainly be a good idea...do it in small increments...

You'll know when you're near the end of the road when:

the system will not boot into windows 100%
Benchmarks crash and do not finish
You notice a DECREASE in scores ( corrupt data )
BSOD
Any other strange behavior...

So long as you have good cooling ( and be sure to use some kind of thermal paste instead of the thermal pad on the retail cooler ), you should still have some life out of the 1.6A. I've had one beyond 2.4 with the retail cooler, and many have gone even further...

Keep an eye on the temps...and be sure to stress-test with Prime or 3dMark loops...

Good luck!

:D

RM. Andersson
12-30-02, 12:16 PM
I dont think you should increase the voltage on a P4.
A P4 cant handle that. It can kill the cpu.
If you increase to perhaps 1.65V you will be able to overclock it more and perhaps it will remain cool if you have a good heatsink and fan.
But after perhaps 3 or 4 months you could get a bluescreen and the cpu will be dead.
So my advice is to get better cooling and see if you can overclock it more. But dont change the voltage.



http://forum.oc-forums.com/vb/showthread.php?threadid=115166


http://www.xbitlabs.com/community/?board=f06;action=display;num=1039026648

http://www.aceshardware.com/forum?read=80051005

A quote from Ace's:

"Electromigration is an exponential effect. The life is cut in half for every 0.03V or so on a 0.18u process with a certain gate thickness. Given a shrink to 0.13u which is about 70.7% in all three dimensions, now the effect is a halving of life every 0.21V or about 31 times for a 0.1V increase. So a stock CPU at designed specifications might last 10 years or more. Add 0.1V to the Vcore and you just shrunk it to 1 to 2 years (some low level destruction may be due to other factors which are more linear with respect to voltages or temperatures). Add another 0.1V and you are down to a few months (now almost all due to electromigration). Add another 0.1V and it is now a few days."

Regards!

netape
12-30-02, 12:34 PM
Thanks for the replies :D

So should I try 133 fsb @ 1.45v first and then if it doesn't work raise the voltage a little bit?

RM. Andersson
12-30-02, 12:50 PM
I think you will have a good chance with some better cooling only.
At least never increase the voltage more than 0.1V.
That will kill the cpu sooner or later.

Regards!

netape
12-30-02, 02:08 PM
Thanks

My temps are rather high @ the moment (34'c mb, 48'c cpu IDLE) because of my case, it doesn't have any slot for rear fan. But it's stable and that's what I care about :D

So I'll buy better cooling then :D

pelly
12-30-02, 05:15 PM
At least never increase the voltage more than 0.1V.

Anand ( from Anandtech ) states that raising the core voltage more than 10% will start to effect the life of the processor...However, he goes on to state that this is not really an issue as most people will not be using the same CPU for 5-7 years. As such, it makes no difference whether the CPU lasts the standard amount of time ( say, 10 yrs ) versus the overclocked time ( say 8 years )

To say that you can't raise the voltage of a CPU in order to overclock is blasphemy! :p

Honestly, take a look at any good overclock and you'll see a healthy dose of voltage in every case...

:D

LORD-eX-Bu
12-30-02, 07:18 PM
yep gotta agree with pelly on this one. :D

RM. Andersson
12-30-02, 07:47 PM
New cpu.s using the 0.13 micron process are very sensitive.
Much more than older cpu:s based on 0.18 and 0.25.
Running them with the voltage out of spec is very risky.

The cpu will last 1-2 years with +0.1V and a few months at +0.2V.
And cooling will not help in this case because itīs the voltage that kills the cpu. Not the heat.

Because what will happen is that different problems that are under control and can be predicted when the cpu is running within spec will become very serious.
Such as "Hot carrier damage", TDDB(time dependent dielectric breakdown) and perhaps even electromigration.

Some of the more extreme overclocking freaks buy 5 cpu:s every year. If it stops working they just go and buy a new one.
Or they want something new to play with and buy a new cpu because of that.
There is of course nothing wrong with that if they have fun.

But for ordinary users that wants to overclock a bit to get more performance running the voltage out of spec is a bad idea.

Regards!

pelly
12-30-02, 08:01 PM
Honestly RM. Andersson, I have to question where you get your information from...

At the risk of sounding arrogant, I should tell you that I have a BS in Electrical Engineering ( with a focus in ElectoMagnetics and dielectric-based electrogravitics [controversial] ) ....I've been overclocking systems for a number of years and know several "extreme" overclockers who are always pushing the envelope with systems...( Never have they gone through multiple CPU's unless they were upgrading )

I have never seen anything you've mentioned appear in enough magnitude to adversely effect any component. ( albeit, that CPU might "only" last 7 years instead of 10...a fact that doesn't bother me b/c I'll likely upgrade by then )

It looks as though a bit more research into overclocking is warranted....Here are a few great sites that are rich in information and knowledge...take a look and ask lots of questions...there are some great people in these places...

Xtreme Systems (http://www.xtremesystems.org/forums)

Overclockers.com (http://forum.oc-forums.com/vb/)

[H]ard OCP (http://www.hardforums.com/forumdisplay.php?forumid=2) :D

I want to emphasize that I am not trying to point fingers or make you feel bad...I honestly believe that there is some more information out there which you should check out...

Regardless, have a good one...

:D

RM. Andersson
12-30-02, 09:04 PM
Pelly,
If you have a good education I can of course respect that.
However, I dont think anyone can claim that a oveclocked 0.13 cpu with the voltage at perhaps 1,65V will last 7-8 years.
They have been on the market for about one year at this time.

Im sure you know that the problems I was talking about exists.
In my previous reply I had linked several disussions about dead cpu:s.
At overclockers.com, aces hardware and xbitlabs.
Check them out!

Here is one link with some related info.

http://www.mosis.org/Faqs/tech_cmos_rel.pdf

Regards!

pelly
12-30-02, 09:11 PM
There was a brief rash of reports jumping around the net a few weeks ago which talked of this "Sudden Northwood Death" syndrome...

My main gripe with this ( and I assume others share my thoughts as the topic has seemingly vanished )...is that there are thousands of users out there who are overclocking these processors. In the case of the 1.6A NW alone, there are countless people overclocking the CPU without error and have done so for many months. Were there a serious issue with the NW's ( aside from a bad batch or something ), wouldn't the news be having a field day?

Like all things, we have to keep an open view...Limiting yourself to a few select forums will not give you "the big picture". As such, explore the various hardware communities...see what people say...and then make a decision...

:D

darkmiasma
12-31-02, 08:34 AM
Originally posted by pelly
There was a brief rash of reports jumping around the net a few weeks ago which talked of this "Sudden Northwood Death" syndrome...

Pelly is right -

All of there were attibuted to people using the P4 Vid Pin Wire Wrap ... they were all using extremely high Vcores on their chips 1.85-2.1 ... thats what caused it ...

- Mike

RM. Andersson
12-31-02, 10:52 AM
Originally posted by darkmiasma
Pelly is right -

All of there were attibuted to people using the P4 Vid Pin Wire Wrap ... they were all using extremely high Vcores on their chips 1.85-2.1 ... thats what caused it ...

- Mike

Thank you for the info!
Then it is at least true that extremely high voltages(1.85-2.1) will result in problems like I was talking about.

I would like to add that I have good intentions if I read about overclockers on several forums that have destroyed cpu:s with to high voltage and after that advice people to be careful and not increase the voltage to much.

Regards!

DaveW
12-31-02, 11:08 AM
"Electromigration is an exponential effect. The life is cut in half for every 0.03V or so on a 0.18u process with a certain gate thickness. Given a shrink to 0.13u which is about 70.7% in all three dimensions, now the effect is a halving of life every 0.21V or about 31 times for a 0.1V increase. So a stock CPU at designed specifications might last 10 years or more. Add 0.1V to the Vcore and you just shrunk it to 1 to 2 years (some low level destruction may be due to other factors which are more linear with respect to voltages or temperatures). Add another 0.1V and you are down to a few months (now almost all due to electromigration). Add another 0.1V and it is now a few days."



That seems hard to believe. Those lousy PSUs you get in most OEM PCs like Dell and Compaq aren't that accurate anyway. I've seen many of them have voltages .2V or more higher than what they should be. Why aren't Dell and Compaq PCs blowing up and being return by the millions?

nin_fragile14
12-31-02, 01:00 PM
I'm running a 1.6a P4 right now, at a 144 fsb with the voltage set to 1.55. It's been running like this for the past 6 months without a single lockup or problem. I'm running the stock HSF with some arctic silver paste. I don't think there would be any harm raising your CPU to at least 2.13 ghz.