Does undervolting a Haswell CPU kill it?

[Dabombinable]

I'm asking this because over the last 2 days I've been running my i5 4440 at 1.4V VCCIN instead of 1.8V VCCIN, and a voltage offset of -0.1V on the core, and had my computer suddenly lock up and refuse to post. I tried the usual standard of resetting the bios, and that didn't work. Since on previous occasions I've had my LGA775 rigs refuse to post with dead CPU (eg. Core 2 Duo E8500, Pentium 630), I switched the i5 4440 for my i7 4790K and the computer is now booted up and running just fine. I haven't tried the i5 in my Z97 Sabertooth yet (don't have a USB keyboard or USB-PS/2 adapter) so I can't confirm, but since the 4790K is running just fine with no issues (albeit with a downlcock to 3.8GHz since I'm using an Arctic Freezer Pro 7 Rev 2.0 with this motherboard) can I safely assume that the i5 is dead?

[Dabombinable]

Look, I'm being serious. I think I killed my i5 4440 (paired with my H87M Pro) by undervolting it.

[BrightCandle]

I would test the i5 in another motherboard since you have one but its certainly sounding likely that you killed it.

[Nena Trinity]

Really it should not AMD and Intel gives the CPUs out with more voltage than needed not just barely what they need...

[ApolloX75]

I really don't see how 1.4v on the VCCIN could kill it. Lock up under load because of vdroop? Sure, I can see that.

Seriously odd set of circumstances.

[Patrick Bateman]

snip

 sounds like you killed it.. but its really weird tho i wouldnt expect undevolting to kill it.

[Fetzie]

Isn't the VCCIN one of the IMC input voltages? It isn't the same as the core voltage for the CPU (that's the one you set to 1.2-1.4V to overclock).

 

Does the memory controller have over-current protection? Maybe reducing the voltage by that amount caused the IMC to request more (milli)amps to run POST and that fried it?

[incarnate]

Isn't the VCCIN one of the IMC input voltages? It isn't the same as the core voltage for the CPU (that's the one you set to 1.2-1.4V to overclock).

Does the memory controller have over-current protection? Maybe reducing the voltage by that amount caused the IMC to request more (milli)amps to run POST and that fried it?

VCCIN is the voltage that is supplied to the voltage regulator, which you want higher than the vcore and other voltages.

[Dabombinable]

Really it should not AMD and Intel gives the CPUs out with more voltage than needed not just barely what they need...

Which is why I was undervolting it so far-it seemed perfectly stable with a 12 hour run of Aida 64 not crashing.

 

I really don't see how 1.4v on the VCCIN could kill it. Lock up under load because of vdroop? Sure, I can see that.

Seriously odd set of circumstances.

I'd thought it had just locked up as well, but when I had to install my i7 to get this rig to post.....

 

 sounds like you killed it.. but its really weird tho i wouldnt expect undevolting to kill it.

Same here, none of my LGA775, s370 and Slot 1 CPU have ever had any issues with undervolting which is why I'm wondering if this is Haswell specific.

 

Isn't the VCCIN one of the IMC input voltages? It isn't the same as the core voltage for the CPU (that's the one you set to 1.2-1.4V to overclock).

 

Does the memory controller have over-current protection? Maybe reducing the voltage by that amount caused the IMC to request more (milli)amps to run POST and that fried it?

 

VCCIN is the voltage that is supplied to the voltage regulator, which you want higher than the vcore and other voltages.

If the VCCIN is too low the computer just locks up while booting Windows-same as a failed overclock with my 4790k. 1.3V VCCIN was a no go which is why I used 1.4V

[ApolloX75]

Which is why I was undervolting it so far-it seemed perfectly stable with a 12 hour run of Aida 64 not crashing.

I'd thought it had just locked up as well, but when I had to install my i7 to get this rig to post.....

Same here, none of my LGA775, s370 and Slot 1 CPU have ever had any issues with undervolting which is why I'm wondering if this is Haswell specific.

If the VCCIN is too low the computer just locks up while booting Windows-same as a failed overclock with my 4790k. 1.3V VCCIN was a no go which is why I used 1.4V

You know now that I think about, since the VRM is located on the CPU and not the board then that's likely the reason it acts differently. As was pointed out the VCCIN is the overall voltage supplied, which includes the memory controller and likely the iGPU side.

Dropping voltage stresses components about as much as overvoltage, just in a little different way. It's possible the VRM failed trying to over compensate and distribute.

Just a theory.

[Dabombinable]

You know now that I think about, since the VRM is located on the CPU and not the board then that's likely the reason it acts differently. As was pointed out the VCCIN is the overall voltage supplied, which includes the memory controller and likely the iGPU side.

Dropping voltage stresses components about as much as overvoltage, just in a little different way. It's possible the VRM failed trying to over compensate and distribute.

Just a theory.

If that's the case, I'll be the first person to have killed a CPU by undervotling it. Though from my understanding the motherboard has 4 phases for the initial step down, then the CPU has another 320 phases to step it down further.

[ApolloX75]

If that's the case, I'll be the first person to have killed a CPU by undervotling it. Though from my understanding the motherboard has 4 phases for the initial step down, then the CPU has another 320 phases to step it down further.

Yeah which is why I think my theory is bunk. But hard to say.

Have you been able to test the i5 in another board?

[Dabombinable]

Yeah which is why I think my theory is bunk. But hard to say.

Have you been able to test the i5 in another board?

Haven't gotten around to it yet as my 4790k is running just fine. I've also noticed that the short freezes-with the last sound repeating- that I was getting with my i5 4440 at stock are no longer happening with my 4790K stock or underclocked, so I think there was something seriously wrong with my i5 4440 before it died. Luckily however I have the receipt for it so I should be able to return it under Intel's warranty (if it really dead that is).

[ApolloX75]

Haven't gotten around to it yet as my 4790k is running just fine. I've also noticed that the short freezes-with the last sound repeating- that I was getting with my i5 4440 at stock are no longer happening with my 4790K stock or underclocked, so I think there was something seriously wrong with my i5 4440 before it died. Luckily however I have the receipt for it so I should be able to return it under Intel's warranty (if it really dead that is).

Interesting! Certainly points to a sick CPU, that's for sure.