overclocking & stability?

e36Adz

e36Adz

e36Adz

Soldato
Joined
28 Nov 2003
Posts
11,718
Location
Manchester, UK
I recently bought a new setup (in sig).

The cpu has been running at 4.0Ghz @ 1.3v for the past 2 days. I have been browsing the net, installing apps, gaming etc. The usual things I would do day to day on the system and I have had no issues or problems.

I thought I would try Prime95, and 1 core failed after 2 mins the others ran fine until I stopped after about 6 mins. The temps (in CoreTemp) shot up hitting as high as 80C! On idle it sits around the 40C mark.

So, would you regard the system as unstable?

I am quite happy to leave it as it is, unless it ever crashed on me, that would be my indication that something isn't quite right. Is it 'safe' to leave it at the current settings?

Thanks.
 
The problem is that it has to get a few calculations badly wrong to actually crash. It can be running with no visible signs of misbehaviour while repeatedly getting minor things wrong, with the results of data corruption (particularly if the ram is unstable) and the operating system gradually falling to pieces. Sod's law will lead to it going wrong at a time when you really don't want it to.
 
Yup. Whilst stress testing might be a pain and the machine may appear stable, it's always better to be sure.
 
Hmmm, so what would you advise bearing in mind windows and everything runs fine? Should I lower or increase the volts slightly? Should I try lowering the cpu speed abit?

I want 4.0Ghz prime stable! :D
 
^BumP^

Ok same system but just installed new hard drives which have given a nice boost. Also, added a second fan to the corsair h50. Now I definately want to get the most out of this system so 4.0Ghz is a must! :p (prime stable)

Currently sitting @ 200x20 = 4.0Ghz at ~1.35v

Idle temps ~ 35C (coretemp)
Load temps upto 80C (coretemp)

It runs prime95 fine for 10 minutes but as mentioned temps get as high as 80C.

Is this temp dangerously high? What temps can these cpus take at full load? What is the highest 'safe' voltage I should be thinking about for 24/7 use?

I might try for 4.1ghz or even 4.2ghz, but I need an idea first of what temps and voltages are considered 'safe'.
 
YES IT IS
for 4 on a 21 multi your looking for 191 fsb btw that is wat ive gone for and running stable at 1.25 volts fully stable which is nice i dont like pushing much higher
 
What's your working definition of safe?

Assuming it is similar to the i7 920 I'm using, I consider sub 1.45V vcore, sub 1.4V qpi safe if under 65 degrees. If it was running at 75, I'd revise these to 1.4V, 1.35V qpi. I wouldn't be happy running it over 80.

These are very approximate limits, derived from experience of others and from how my system reacts to voltage, rather than anything I've calculated.

Had a stab at calculation. The workings are pretty rough and ready, and it assumes death from electromigration, but the result derives neatly enough from Black's equation. Increase in temperatures decreases lifespan, decrease in voltages increases lifespan. As such the two can balance.

(decrease in temperature) is proportional to ln(increase in voltage) for lifespan to be unchanged

I can't get equality from this alone I'm afraid, one of the experimental constants won't drop out as it controls vulnerability to voltage. However lifespan is exponentially dependent on temperature increase and only a polynomial on increase in voltage. Keep it cold and hope that n is small :)

edit: might get further taking vid, maximum stated operating temperature and warranty duration from Intel specs to guess at lifespan while out of spec, but it's lunchtime now and I fear we still won't know what n is. Plus we don't know what failure % @3 years they're working to. Wonder if they'd tell one of us.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom