[6850k] Killed my first CPU in 15 years!

[Charlie22911]

Well killed may be a strong word, severely degraded is probably more appropriate.

My CPU is a i7 6850k @ 4.2GHz core 1.275v \ 3.2GHz cache 1.15v installed in an MSI Godlike Gaming Carbon will a full custom hardline water setup. Last night I was reading some threads on what clocks and voltages others were running and decided to give 4.4GHz @ 1.32v core a try, I left it for a couple hours to run 6 threads of Linx which failed after an hour and a half or so; core temperatures peaked at about 78°C, fluid temperature was ~31°C depending on how reliable the XSPC thermistor is (pre radiator).

Set it back to my 10 hour stable small FFT prime profile, only to instacrash with a stop code on boot.
So I'm thinking cool, no worries; that's why I do a system image before tinkering in case I corrupt something...
Except it is now crashing even when booting from a Windows 10 DVD. I'm not an expert but I do have basic troubleshooting down, even a clean cleared CMOS is crashing.

I finally managed to get it back stable by clocking it down to 2GHz at 1.28v, but it looks like she's done for.
I also decided this is a good excuse to get a 6900K (I know, yes I do threaded work too), in the mean time I'm going to see if Intel will swap my CPU and sell the replacement should I be so lucky. No I don't have their overclockers warranty and I don't plan on being dishonest about the circumstances.

I'm not looking for advice so much as I wanted to share my experience with the community, mostly as an example in case anyone else was thinking of or is running voltages in this range.

TL;DR: I killed Broadwell-E with only ~1.32v (give or take a few mv for load line calibration and such).

 

[LTC8K6]

When I got my 4790K replaced, Intel didn't ask me anything about what I'd done with it. If they had, I would have told them. But I actually wasn't able to overclock it much as all I had for it was a Hyper212 cooler. It just died early for no apparent reason. I think all they asked about was the board, the purchase date, and the numbers off the chip.

 

[Justinus]

I don't think that should have killed your chip unless it was marginal in the first place. 1.32v isn't anything extreme and 78C is totally acceptable.

Keep in mind you always have the option to purchase the Performance Tuning Protection Plan and get it replaced after the mandatory wait period as a backup plan. It would suck if you didn't already plan to get a 6900K but since you do, waiting wouldn't be that bad.

Funny that I actually did purchase the tuning plan for my 5960X and I can't seem to hurt it no matter what I do. I have kept the volts at or under 1.4 since I don't have a custom water loop.

 

[VirtualLarry]

Sure that you didn't put in "Offset" at 1.32V? Just checking...

 

[Charlie22911]

When I got my 4790K replaced, Intel didn't ask me anything about what I'd done with it. If they had, I would have told them. But I actually wasn't able to overclock it much as all I had for it was a Hyper212 cooler. It just died early for no apparent reason. I think all they asked about was the board, the purchase date, and the numbers off the chip.

Thanks, that's good to know. I'll have to drain the loop (ugh) and pull the CPU when I get home from work and start that process.

I don't think that should have killed your chip unless it was marginal in the first place. 1.32v isn't anything extreme and 78C is totally acceptable.

Keep in mind you always have the option to purchase the Performance Protection Tuning Plan and get it replaced after the mandatory wait period as a backup plan. It would suck if you didn't already plan to get a 6900K but since you do, waiting wouldn't be that bad.

It didn't give any indication that it was marginal from my experience with overclocking, more average than anything.
The 6900k will be in my hands Tuesday or Wednesday and I'll get the "protection plan" for it, I don't usually go for those sorts of things but I can see the use for it in this case.

Sure that you didn't put in "Offset" at 1.32V? Just checking...

I use voltage offsets for cache\ring thanks to the microcode bug with Broadwell-E, but not vcore. I doubt this thing would even post at 2.32v, poor chip.

 

[Justinus]

It didn't give any indication that it was marginal from my experience with overclocking, more average than anything.
The 6900k will be in my hands Tuesday or Wednesday and I'll get the "protection plan" for it, I don't usually go for those sorts of things but I can see the use for it in this case.

I don't normally go for it, either, but for a $1000 CPU, $35 is a super cheap free replacement guarantee.

I mean that either your motherboard did something very, very bad with the voltage when it was stress testing and it ruined your perfectly good chip or there was some part of the silicon that was marginal and the upped volts broke/degraded it during the stress test.

It's not like you went on a suicide run with major volts.

 

[Charlie22911]

I don't normally go for it, either, but for a $1000 CPU, $35 is a super cheap free replacement guarantee.

I mean that either your motherboard did something very, very bad with the voltage when it was stress testing and it ruined your perfectly good chip or there was some part of the silicon that was marginal and the upped volts broke/degraded it during the stress test.

It's not like you went on a suicide run with major volts.

The more I think of it, the more I am leaning on the motherboard possibly playing with voltage. In the MSI bios CPU and Ring voltage modes are linked, I had it set to adaptive+offset so I could adjust vring without losing idle voltage throttling.
CPU vcore offset was set to auto, and voltage set to 1.32v; it is very possible the board added some offset in addition to the 1.32v specified. I'll be sure to keep an eye out for this behavior with the new chip.

I'm sure some would scold me for leaving some things on auto, but I didn't expect this behavior and haven't encountered it in the past. Granted this is my first Intel build since my ancient Pentium MMX, I've built with AMD from the K6-2 until now. Also, I like to live life in the danger zone (insert obligatory Archer photo here).

 

[Justinus]

The more I think of it, the more I am leaning on the motherboard possibly playing with voltage. In the MSI bios CPU and Ring voltage modes are linked, I had it set to adaptive+offset so I could adjust vring without losing idle voltage throttling.
CPU vcore offset was set to auto, and voltage set to 1.32v; it is very possible the board added some offset in addition to the 1.32v specified. I'll be sure to keep an eye out for this behavior with the new chip.

I'm sure some would scold me for leaving some things on auto, but I didn't expect this behavior and haven't encountered it in the past. Granted this is my first Intel build since my ancient Pentium MMX, I've built with AMD from the K6-2 until now. Also, I like to live life in the danger zone (insert obligatory Archer photo here).

Manually setting the voltages is a good idea. I started doing that after I realized my MB was upping the PCH voltages. It was pushing the 1.05v -> 1.3+ and the 1.5 -> 1.75+. I manually set them to 1.05 and 1.5 and they don't affect my stability at all.

 

[Burpo]

I wouldn't even buy the protection plan.. I've never seen intel refuse a warranty, even on a 3 yr old cpu. You pay enough for these, they can afford the replacement, several times over..

 

[crashtech]

It had to be way more than 1.32V, or the chip was just defective in the first place. I think that board has DVM pinouts, you might want to use them.

 

[superstition]

My UD3P board got a corrupted BIOS and was feeding my 8320E 1.55V, at least, no matter what — until I flashed with the newer BIOS.

That's one danger of overclocking without sticking with known safe voltages ranges. Too little and your BIOS can get corrupted.

 

[LTC8K6]

I wouldn't even buy the protection plan.. I've never seen intel refuse a warranty, even on a 3 yr old cpu. You pay enough for these, they can afford the replacement, several times over..

I think Intel just gives you the first one without too much hassle.

I think it's only if you are a frequent flyer that they start to ask questions.

 

[FlanK3r]

I think,m your motherboard is "crap" and dead in VRM...If you can, try this CPOU in another board.

 

[Arachnotronic]

Last CPU I murdered in cold blood was a Lynnfield

 

[LTC8K6]

I think,m your motherboard is "crap" and dead in VRM...If you can, try this CPOU in another board.

Not sure if you are correct, but you make a good point.

Need to verify that the mobo is okay before putting a new $1K chip in it.

 

[ehume]

Not sure if you are correct, but you make a good point.

Need to verify that the mobo is okay before putting a new $1K chip in it.

Wouldn't it be nice if he is correct. Then you would have two good CPU's.

 

[2blzd]

my broadwell-e died too. after 10 days of use. I suspect the motherboard took it out when it went down.

 

[DrMrLordX]

my broadwell-e died too. after 10 days of use. I suspect the motherboard took it out when it went down.

Was yours an MSI mobo?

 

[BonzaiDuck]

Was yours an MSI mobo?

It might be helpful to know which MSI board was used by the OP.

I can also offer an experience I had with my 6700K Skylake at 4.6. I had read a guide or forum thread in which it was advised to twist up the PCH, CPU-Standby and PLL voltages. Adjustments to the first two had no effect on my observations after a normal boot-up, except that CPU temperatures were a tad warmer. I was more cautious about the PLL voltage, advised to be raised to 1.45V from a default of around 0.930V.

Earlier chipsets and boards -- at least my Z68 -- offered cooler temperatures and more stability for a Sandy CPU by bringing the PLL voltage down from 1.8 to 1.65 <= vPLL <= 1.71V. I noticed the color-coded voltage steppings of the motherboard, going from white to yellow to red. So I chose to raise the PLL voltage by only 0.3V, or about two or three clicks into the yellow range. System would freeze at the Windows startup logo; drive light would stay on.

I set them all back to auto and called it a day.

Don't think I damaged anything. I could imagine, though, for someone tweaking a new-gen motherboard with old-gen BIOS familiarity, you could make any number of mistakes. My Z68's had an "Extra Voltage for Turbo" which represented only an addition to voltage. Adaptive mode with "Voltage for Turbo mode" includes everything but the offset. I kept raising the voltage by increments from 0.00V and couldn't figure out why it wasn't increasing voltage over stock values.

There are sure a lot of scary CPU-death stories going around as we move toward Halloween. So what do they mean, exactly?

 

[Smoblikat]

I had an E5200 that I burnt, literally burnt. I ran it at extremely high voltages to get a stable 3.8ghz OC. During stress testing I would let it get up to 105C, and if it didnt crash, it was stable. When I pulled the chip there wer large patches of black and basically the entire rest of the chip was burnt copper looking. The thing still ran fine and does to this day OC'd.

 

[ehume]

I had an E5200 that I burnt, literally burnt. I ran it at extremely high voltages to get a stable 3.8ghz OC. During stress testing I would let it get up to 105C, and if it didnt crash, it was stable. When I pulled the chip there wer large patches of black and basically the entire rest of the chip was burnt copper looking. The thing still ran fine and does to this day OC'd.

Impressive.

 

[DrMrLordX]

It might be helpful to know which MSI board was used by the OP.

According to the OP, it was this one:

https://us.msi.com/Motherboard/X99A-GODLIKE-GAMING-CARBON.html#hero-overview

 

[BonzaiDuck]

According to the OP, it was this one:

https://us.msi.com/Motherboard/X99A-GODLIKE-GAMING-CARBON.html#hero-overview

Gee . . . six Franklins . . . and the customer reviews in comparison to the scatter you see for all motherboards combined are fairly stellar.

I'm more inclined to the probability -- or chances -- that this arose from some human error.

 

[superstition]

There are sure a lot of scary CPU-death stories going around as we move toward Halloween. So what do they mean, exactly?

The long-awaited Death of Overclocking is upon us?

 

[BonzaiDuck]

The long-awaited Death of Overclocking is upon us?

So far, I've only heard the OP's story firsthand, and Wingman in another thread posted a link to an Intel forum in which a 6600K owner was bemoaning the death of his chip. Purportedly, he only clocked it up to 4.4 Ghz, but there are many reasons this might happen.

If you can't overclock your chip with some reasonable expectations, it takes the fun out of building computers.

So I will mourn, if that is the case.