Core i7 Overclock - CPU Life

[MChim]

I bought a Core i7 920 which I intend to push to 3.8ghz when I pickup the cooler today. I was wondering, what would the consequences be on CPU life if I pushed it to the limit, but did not use harmful voltages and I made sure it's a rock stable overclock?

Thanks,
Chim

 

[Mr Fox]

Originally posted by: MChim
I bought a Core i7 920 which I intend to push to 3.8ghz when I pickup the cooler today. I was wondering, what would the consequences be on CPU life if I pushed it to the limit, but did not use harmful voltages and I made sure it's a rock stable overclock?

Thanks,
Chim

It all depends upon how experienced that you are in jamming electrons thru Silicon/Hafnium, and recognizing trouble quickly.

It also depends upon you comfort level for having a very expensive ($300.00) Key Fob...

Because of this being days after launch, you can be guaranteed that every RMA will be root caused for their initial Pareto Data. And this will continue until the management decides to cut the sample size, based upon supporting data.

In Initial Testing there is a Failure Mode, and Effect Analysis done that creates failures by various means, and then the chips are examined to understand the weak aspects of their process, and to test the design limits.

 

[Idontcare]

Originally posted by: MChim
what would the consequences be on CPU life if I...

The consequence would be that the CPU lifetime would be reduced.

But that isn't really your question, and so the response won't really satify you either.

The question you mean to ask is "how much" will the CPU be reduced. Years? Months? Days?

Only Intel has the Weibull distributions for fail rates that are needed to answer that question. (and naturally they aren't about to tell us mere point-of-use consumers)

 

[MChim]

Well how about those guys that clocked the Core 2 Duos to 4ghz. How are they holding up. Have you seen anything about their chips dying out?

 

[ch33zw1z]

Originally posted by: MChim
Well how about those guys that clocked the Core 2 Duos to 4ghz. How are they holding up. Have you seen anything about their chips dying out?

Those are not the chip in question.

 

[Idontcare]

I have only ever heard of one (1) CPU dying in recent times from excessive voltage/temps and that was one by Aigomorla in which he was intentionally trying to see what would kill it.

No doubt there are more chipkill stories out there, but we just don't seem to hear about them for whatever reason.

Anyone here kill a HKMG 45nm CPU from Intel and want to admit to it?

 

[exar333]

Originally posted by: ch33zw1z

Originally posted by: MChim
Well how about those guys that clocked the Core 2 Duos to 4ghz. How are they holding up. Have you seen anything about their chips dying out?

Those are not the chip in question.

Yes, and C2D chips share nothing in common with i7...45nm, similar fabrication technology, voltage, etc...

 

[MChim]

Are you comfortable keeping it at 3.0ghz or are you going to try and hit higher on the 920?

 

[MChim]

Also, do you think the Antec tp3-650 would be efficient to achieve at least 3.2ghz overclock.

(Sorry, but I haven't overclocked in ages. Since the AMD Athlons where I'd buy them depending on what serial number they were because certain batches were underclocked. Then I'd solder them to 900 or so mhz.)

 

[exar333]

Originally posted by: MChim
Are you comfortable keeping it at 3.0ghz or are you going to try and hit higher on the 920?

Supposedly i7 can handle more voltage and their C2Q counterparts; that tells me have a good chance to live a long and "good" life OC'd.

No one here has had one running at 30-40% OC for 6 months yet, so only time will tell.

 

[evilbix]

i've been running my pentium d @ 3.8ghz on 1.525v for about three years solid. runs about 16 hours a day on average.. it's still running fine.

normally overclocking won't kill a chip until it's beyond obsolete unless you really bump the volts to extreme levels.

 

[Markfw]

Originally posted by: ExarKun333

Originally posted by: MChim
Are you comfortable keeping it at 3.0ghz or are you going to try and hit higher on the 920?

Supposedly i7 can handle more voltage and their C2Q counterparts; that tells me have a good chance to live a long and "good" life OC'd.

No one here has had one running at 30-40% OC for 6 months yet, so only time will tell.

I think my X3350(Q9450) has been over 6 months, but only at 3.5 ghz.

And my E8400 ES has already had the OC go down to 3.96 stable after only a few months.

Both are at 70c max temp, and about 1.375 vcore or less.

 

[JohnVM]

Originally posted by: Idontcare
I have only ever heard of one (1) CPU dying in recent times from excessive voltage/temps and that was one by Aigomorla in which he was intentionally trying to see what would kill it.

No doubt there are more chipkill stories out there, but we just don't seem to hear about them for whatever reason.

Anyone here kill a HKMG 45nm CPU from Intel and want to admit to it?

You can make that two. We blew out a QX9770 in April. You may remember our "Overclocking QX9770" thread -- probably one of the biggest "epic fail" threads ever. The flaming has already been done for that, so please no rekindling the fire.

 

[Idontcare]

Originally posted by: JohnVM

Originally posted by: Idontcare
I have only ever heard of one (1) CPU dying in recent times from excessive voltage/temps and that was one by Aigomorla in which he was intentionally trying to see what would kill it.

No doubt there are more chipkill stories out there, but we just don't seem to hear about them for whatever reason.

Anyone here kill a HKMG 45nm CPU from Intel and want to admit to it?

You can make that two. We blew out a QX9770 in April. You may remember our "Overclocking QX9770" thread -- probably one of the biggest "epic fail" threads ever. The flaming has already been done for that, so please no rekindling the fire.

Oh yeah, had totally forgotten about that.

For value to the OP, can you remind us the voltage you were using? IIRC you had a vapochill rig but had overloaded the cooling capacity and all the refigerant gasified so your temps really went ballistic (worse than air cooling type situation) but the voltage is relevant.

 

[JohnVM]

We took it up at least to 1.79V, maybe higher -- can't really remember. Wasn't for very long, but the damage was fast and brutal. We're talking minutes of stability before totally fubar. Once we knew we had blown the chip we took it to the max possible to just see what it could do. Obviously nobody here should ever set a voltage that high if they ever want to use the CPU again.

I think we had it stable at around ~1.5-1.65V, but again, we didn't leave it at that voltage and just tried to crank the chip as far as it would go, so I have no possible word as to the long term viability of such a voltage. I think it lasted at least a few days at that voltage.

In retrospect, the entire thing was retarded. Just as a "post-mortem" follow up, that rig never ran stable, and we went through ~10 motherboards and a couple CPU's to get it decent (which it has been since). The Vapochill backplate shorted out all the boards, and in the process of trying to make it not short, the computer was ran with the Vapochill not entirely secured, which resulted in the CPU being operated in a literal pool of water. You can imagine how that turned out.

Anyways, to the OP: A reasonable overclock will probably be fine/not reduce your CPU life too much. You may want to read our thread, which can be used as an example of on how NOT to overclock your PC, before you do it though: http://forums.anandtech.com/me...id=28&threadid=2173204

Don't do anything we did.

lol.

 

[Extelleron]

As Idontcare says, any overclock and certainly any overvoltage is going to reduce the life of a CPU. The reason so many people overclock without noticing any ill effects is that normal CPU life expectency is so long, reducing a bit or even a lot has little effect on anyone who upgrades frequently.

Under air cooling I would not exceed 1.4V if you want the chip to last long without any damage or degrading; on Penryn that is the rule, with Nehalem maybe a tiny bit more (although I'm skeptical of that.)

Under air cooling with, say 1.45V on a Penryn CPU, you aren't going to wake up to a dead CPU most likely..... but over time the CPU will degrade and require more voltage for the same speeds. Sudden death is going to occur when you have the CPU at 1.6V+ on air and are going crazy.

 

[Markfw]

Wow, looking back at that thread, once of the biggest mistakes that I saw (other than rediculously high vcore) was no adjustments made to the northbridge, and this is always needed for a good OC.

The old adage of Not too much vcore, and good temps still holds true. Now the addition of extra voltages that can also damage a cpu are coming into play (vdimm). Based on well within acceptable limits for all voltages and good temps, I would say your odds of have a cpu last 5 years or more when OC'ed are very good. YMMV.

 

[DrMrLordX]

Just an aside, but the amount of vdimm you push on a Core i7 setup may also affect its lifespan. Pushing more than 1.6v vdimm is allegedly not good for the integrated memory controller.

 

[MChim]

Thanks for all the feedback. I was looking around and I found

http://www.bit-tech.net/hardwa...ng-intel-core-i7-920/1

and I'm going to shoot for 3.8ghz at 20 190. I have to go pick up that cooler that guy had though... Thermalright Ultra 120 eXtreme 1366-RT. Can't find it right now... Only been looking for a minute or two.

I can't wait to post this overclock.. keep the input coming.. going into something without knowing all is stupid and I trust you members more than anyone on the net.

Thanks!

 

[Zap]

A "reasonable" i7 overclock can last so long that you'd toss the CPU away as worthless before it died.

=== DISCLAIMER: OPINION ONLY, NOT RESPONSIBLE FOR ANY DEAD PARTS ===

What is "reasonable?" Most likely, stock voltages and within 10% of highest Intel stock clocks is reasonable. This means maybe around 3.4-3.5GHz.

The next step is "calculated risk" meaning voltage boosts of 10%, with memory at 1.65v. I'm guessing this will net around 3.6-4.0GHz, and a lifespan of a few years, perhaps with decreasing clocks after a year or two.

Beyond that is keychain territory. This means high voltages, ESPECIALLY VDIMM, and higher than 4.0GHz clocks.

http://www.anandtech.com/weblog/showpost.aspx?i=521

Intel's guidance continues to be not setting VCore past 1.55V, VDimm past 1.65V and VQPI (uncore to us) past 1.315V when overclocking. Kris will go over this in detail shortly, but in my experiences so far, getting these three voltages too far out of sync will quickly cause problems as one of our i965 processors can attest to now.

This tells me that the intrepid Anandtech reviewers have already killed a Core i7. I think such "wrong" settings will kill a Core i7 within several months at the longest.

BTW, I think this might be a result of the memory controller being on the CPU. If anyone remembers... back in the socket 939 days some enthusiasts discovered that setting CPU VCORE too low and VDIMM too high will kill Athlon 64 chips very quickly. I recall discussions about this at the now defunct DFI Street.