Voltage?

So I have read alot about over-volting CPUs, the pros and cons, etc. . .

I understand the effect that over-volting can have on the degradation of the CPu, but I also know the specs Intel has utilizes are very conservative to avoid warranty claims and also calculated to preserve the life of the CPU to the 3 year mark.

So my question is, what if anything have members read in respect to the shortening the life of a cpu by overvolting and do you have any evidence, such as tables showing relative degradation estimates.

My thought is that I usually replace my CPU every 1-2 years depending on the technology. At the present, I do not forsee upgrading for at least another year. Can I expect my 2.8c to survive that long at 1.8vcore.

And while we are on the subject, what in the hell is worng with these MOBO manufacturers. Do that not relaize that some mem can handle 3.1 within specs?

Personally I would not go over 1.7VCORE on a new P4. First, they don't seem to really need it, second after 1.7V you are entering the danger zone for SNDS. What cooling do you have?

On my 2.6C I run it with 1.675 set in the bios, MBM shows 1.66 or so and even if I push 1.8V to the CPU, I get no higher of an OC.

I am water cooling my rig and curently I am running at 1.7 settings, with MBM indicating 1.66 . . . but we all know the finicky nature of MOBO temp readings, and I assume that the relaibility factor with respect to voltages is on par.

I have pushed it to 1.775 with prime95 running stable at 252 FSB. If I lower it any, it fails after first test. I would rather not run over, but I can't resist event the 3 FSB increase I get as a result of overvolting. Yes, I am a freak

I have read that cooling does'nt make a difference when it comes to overvolting and that the damage will be done regardless. Intel reccomends nothing in excess of 1.75, but I do believe the 2.8c is only rated at 1.5-1.6.

My cpu temps only hit 40c under prime95 torture test despite the different voltages I use, even though it seems to effect idle temps by raising them approximately 4 degrees higher than without.

I guess I can fool around with it, of course accepting that I could fry my cpu at any time, and see what the results are. I would like to know how overvolting plays out in the long run (at least a 6 month period). I was thinking about digging up an old celeron I have laying in the dust bin and putting it to the test, but the circuitry, resistors, etc are so far removed from the P4 that I couldn't really make a comparison. It would be nice if someone has already done what I thinking of doing, so as to obviate the need to sacrifice my CPU on the alter of curiosity

Let me see if I can clear a few things up here...

First, nikhsub1, you probably mean 1.7V VID. VDIMM is the reference voltage for memory (hence the DIMM).

Second, Intel does _not_ recomment you run your chip up to 1.75V. In the datasheet it says the most Vcc can be is 1.75V when the clock is stopped (that means, the chip can't be doing anything at all). The recommended operating voltage is of course much lower.

Voltage does shorten the life of the chip. By how much varies from chip to chip. The average chip running at stock voltage might last 20+ years. But at 1.85V, there is a change (albeit not more than 10%) that the chip will fail outright within 8 weeks. At 1.7V you can probably expect it to last between 1 year at the low-end and 4 years if it's a good chip. It might not fail completely during that time, but it will get slower gradually.

Now, when I say 1.7V, I don't mean the actual operating voltage received by the chip; I mean VID. There _is_ a difference. A 1.7V VID should give about a 1.60V Vcore/Vcc level. This is normal and expected.

You're also right in that overvolting does damage regardless of temperature. But high temperatures can only make it worse. If you do not have excellent cooling, I wouldn't push it much if you're concerned about the wellbeing of your chip.

Thank you, thats what I was looking for.

I am assuming then that if I want my cpu to last that I need to set it at 1.6v VID, as opposed to relying upon the resulting vcore reading?

As an aside, do you think that a chip beocmes more stable with use as opposed to being newly introduced to the system? I have read many different opinions as to the value of burning in a system at defaults and then increasing after the CPU has familiarized itself with its environment so to speak. Other s say that burn in is a placebo and that it has no real world effect on ability to oc the system.

This will be the last pick of your brain today. I promise

You can set VID at 1.7V, or maybe 1.75V is your load temps are good (say, less than 50C). Then your Vcore reading in MBM, et al., should be around 1.6V. Basically, don't keep pushing up VID just because MBM says you're "only" getting 1.6V. And don't use any more voltage than you really need.

I personally don't think there's much to burning in. I know chip manufactures "burn in" to kill off the weak chips in a batch, so overclocks might have a misguided perception of what burn-in is for. I've heard people swear that burning-in for 3 days got them and extra 40 Mhz. But that's only a 2% variation. And, after all, Arctic Silver does set after about 2-3 days...

My opinion is that burn-in should be used to validate the stability of your system. If you've proven it's stable and want to try to push it another 50Mhz, go for it. But then you'll need to check your stability all over again.

NookieN said:

Let me see if I can clear a few things up here...

First, nikhsub1, you probably mean 1.7V VID. VDIMM is the reference voltage for memory (hence the DIMM).

Click to expand...

Err, yeah I meant to say VCORE, not VDIMM, will edit...

Thank you for the info, and I am of the same opinion re burn in and its funny that you mention the 72 hour set time of thermal compounds, as my thoughts were running along the same lines.

Luckily my Water rig gets the job done right and keeps me at a cool 40 - 41 degrees under load. In truth, I am able to run my [email protected]+, which is by an accounting of the database a good oc for the batch. I just hate almost being able to get a stable oc, and then have it fail by a gnats hair.

Again thanks for the info and tips, you have been a great help and saved me some time. . . and probably a CPU