Is there any consensus on safe voltages for Sandy Bridge yet?

[pantsaregood]

Intel's spec says 1.52v, but that's just the maximum VID value available.

I have an i5-2500K, and I'd like to push it a little more aggressively. It ran at 4.5 GHz/1.35v for a while, then my PC started BSODing. I assumed the OC was bad, but it was a stick of RAM that went bad. Since then, it has been running at stock.

Temperature isn't really an issue. I can monitor temperatures and tell when my overclock is thermally limited. I just don't want some crazy electromigration issue that kills my CPU in a month.

I do leave EIST and C1E enabled, so my PC usually runs at 1.6 GHz anyway.

 

[BonzaiDuck]

Intel's spec says 1.52v, but that's just the maximum VID value available.

I have an i5-2500K, and I'd like to push it a little more aggressively. It ran at 4.5 GHz/1.35v for a while, then my PC started BSODing. I assumed the OC was bad, but it was a stick of RAM that went bad. Since then, it has been running at stock.

Temperature isn't really an issue. I can monitor temperatures and tell when my overclock is thermally limited. I just don't want some crazy electromigration issue that kills my CPU in a month.

I do leave EIST and C1E enabled, so my PC usually runs at 1.6 GHz anyway.

There's more or less a consensus -- among people here who built Sandy machines. There were some threads last year that provided a lot of input from veteran "push-the-edge-of-that-envelope" members as well as the more cautious. But -- as far as I know [and that too may have changed] -- there has never been an official pronouncement from Intel.

Based on the progression from LGA-775 Wolfdale/Yorkfield 45nm to Nehalem and then to the 32nm shrink of the Sandy Bridge, we'd deduced a "safe-range" upper bound of 1.35V and some had suggested that a 5% overage was insignificant. Others among us -- cautious types like myself -- had stopped where 100% load voltage was between 1.32 and 1.33V. the low-load (and un-drooped) value would bounce up to around 1.36+ with some over-clock settings.

Another "forum guru" suggested a table of expected lifespans under different voltage settings -- the ideal being something like 10 years for an Intel processor run to spec.

At this point, I'm guessing . . . that there is a consensual "guess" -- that you would get reasonable service over several years within the voltage choices put forward above . . .

 

[moonbogg]

People have had their chips OCd for well over a year now, and I have yet to hear of them dying. I have heard 1.4 is the max you would want to ever use, but with my own 2600K I was comfortable keeping it under 1.3 so it sat at 4.3ghz @ 1.28 just until I sold it. I run the 3930k at 1.36 @ 4.5 and feel fine with it, but these do require more voltage and a little more is normal for them.

 

[lamedude]

Ark says .6-1.35 for SNB-E.

 

[2is]

I'm sure there's a consensus that stock voltages are safe.

 

[BonzaiDuck]

Ark says .6-1.35 for SNB-E.

Well . . . it's a Sandy made to a 32nm spec . . . Shouldn't be different than the OP's 2500K . . .

 

[Rvenger]

1.35v is the safe limit.

 

[Smoblikat]

No more than 1.4 + 5%

 

[Idontcare]

Is there any consensus on safe voltages for Sandy Bridge yet?

Define what "safe" means to you. 10yrs safe is different than 2yrs safe.

I just don't want some crazy electromigration issue that kills my CPU in a month.

If by "safe" you are ok with a 1month-to-many month lifespan (but ok with it being decidely <1yr) then you can probably push 1.6V through it.

People who have been pushing 1.5V through theirs to get 5GHz stable clocks have not been reporting degredation or chipkill, and those have been out there now for ~16months.

If you want safe for 10yrs then 1.35V max, it'll be the same for SNB as it is for SNB-E.

 

[Kenmitch]

I'm sure there's a consensus that stock voltages are safe.

How do you define stock voltage on a K chip? Its multiplier based and influenced by amount of cores loaded topped off with what type of load even.

I say if your worried about it get the protection plan..

 

[2is]

How do you define stock voltage on a K chip? Its multiplier based and influenced by amount of cores loaded topped off with what type of load even.

I say if your worried about it get the protection plan..

I define stock voltages by the peak default voltage used at the highest default multiplier.

 

[samboy]

My 2600k died after 17 months 4.5GHZ@1.36v
However, I believe it was the Asus P8P67 MB that died and took the processor out.

In any case, both Asus and Intel RMA'ed a replacement for me.........

 

[chimaxi83]

One of my 2600K's has been sucking up 1.4-1.6V for a little over a year. Its still operating the way it did out of the box. If you're worried, you can get the Intel Tuning Plan and give it as much as you want. If you want it to last 10 years, leave it at stock.

Also, I don't understand where some people here get their max safe voltage, and state it as fact

 

[jacktesterson]

I've been pushing 1.35v to 1.45v to my 2500k since launch, depending on what I'm doing.

No issues at all.

 

[Rvenger]

My 2600k died after 17 months 4.5GHZ@1.36v
However, I believe it was the Asus P8P67 MB that died and took the processor out.

In any case, both Asus and Intel RMA'ed a replacement for me.........

Yep, those boards are known to overvolt the CPU to 2v.

 

[samboy]

Yep, those boards are known to overvolt the CPU to 2v.

Any further info on this? I'd read of a few other sudden death of CPU + MB for the P8P67 but had not seen anything about feeding 2V to the CPU.

I was running with Offset and had been monitoring the voltage, on occasion, and never saw it go past 1.36v with the monitoring tools.

Are you saying that it had a problem of spiking to 2V when it was switching frequencies? (i.e. a transient spike that only a scope could measure).

In any case, this may be good news for everyone if it can last 17months with frequent spiking to 2v.............

I'm now using a P8Z77 Pro and the replacement P8P67 is staying in a box as a spare (it's not worth much to sell anymore - maybe with good reason)

 

[Lex Luger]

Asus said the max safe 24/7 always under load voltage was 1.425.

The engineer who posted this number knows more about stress testing sandy bridge cpu's than anyone posting on these forums, but still it didn't come from Intel so take with grain of salt.

For all we know, a sandy bridge cpu kept at 1.425 with good cooling could last 10 years without degradation. Everything is still speculation at this point.

Personally I wouldnt worry about degradation unless you are pushing 1.45+ and stress testing constantly.

I've been running around 1.45 for 6 months, no problems. I was running 1.425 before that since sandy bridge release, no problems.

I also have my ram voltage at 1.625 since sandy bridge release, no problems.

 

[BonzaiDuck]

Define what "safe" means to you. 10yrs safe is different than 2yrs safe.



If by "safe" you are ok with a 1month-to-many month lifespan (but ok with it being decidely <1yr) then you can probably push 1.6V through it.

People who have been pushing 1.5V through theirs to get 5GHz stable clocks have not been reporting degredation or chipkill, and those have been out there now for ~16months.

If you want safe for 10yrs then 1.35V max, it'll be the same for SNB as it is for SNB-ENo others -- mostly overclocked and remaining or passed along within the last five years -- have failed or proved unreliable..

:thumbsup: :thumbsup: :thumbsup: and a toe . . .

We want longevity and reliability. Voltage is the biggest threat. Over some range at the end of the expected time under or beyond (3yr) warranty conditions, it wouldn't matter to you. I've kept one computer almost ten years; one for eight; two for maybe seven.

I've had one machine, fitted with an OCZ "VDIMM Boost," gadget fail after maybe two+ years. [The device just went "south," because I didn't fiddle with it much. Somehow, everything that coudn't be replaced had failed.] In addition to the OCZ disaster, I've had one motherboard fail in eight years (run stock forever), another after five (also stock). Now I give away the best machines after five, and I use, junk or sell the rest -- but mostly discard.

But so far I get the impression that those of us who concur about this are only making logical deductions from what went before, because Intel has been a lot less specific about "safe" versus "operational" ranges.

 

[moonbogg]

Clearly no consensus.

within intel specs, the max was reported to be 1.35 (non OC I believe) so 1.4 sounds about right. Jacking this chart makes me an expert.
For sandy-E, more voltage is needed or same speed as non E variants.