CPU Electromigration/Degregation Chart

[TechyInAZ]

Thread will be closed for an undetermined amount of time. It is impossible to tell if I can get enough information (accurate information) about specific architectures and how fast they desegregate.

Please TAKE ALL THE TEMPS DOWN BELOW WITH A GRAIN OF SALT. Only use these voltages once you've backed it up with other forums and other threads with reliable and experienced overclockers. Thank you!

Hello Community! For you newer/more worried overclockers out there, this thread should give you a good idea of what voltage is good for your CPU before CPU degregation starts. (ie. The BEST voltages for 24/7 use.)

CPU Electromigration (or CPU degregation), is one of a few brick walls you can face when overclocking. Even if you have very good temperatures, pushing too many volts will still kill a CPU. There is no way to stop it.

In real basic terms, electromigration is the wear and tear of CPUs. Just like how tires wear out, the constant supply of voltage to the transistors wears the CPU out. Eventually making the CPU unstable at a certain voltage it used to be stable at. However, if you don't overvolt like a mad man, it takes a LOT of years before a CPU degrades to be noticeable.

When overclocking, typically electromigration is accelerated because many overvolt the CPU to allow for higher clock speeds. However, just like I said above, the amount of acceleration depends on how high or how low you overvolt. (The same logic applies to underclocking/undervolting, lower the voltage below spec, and your CPU has even less electromigration at stock speeds.)

The general rule is, if you keep your voltages lower than 10% beyond the stock vcore of your CPU lifespan will remain en-tacked. (However, keep in mind, that temperatures also affect degregation aswell, if you run a super low temp of say 40-60C, you can run at 15ish% higher voltage than stock, this is because temperatures affect the resistance of the transistors too.

There's only one problem left, the internet as a whole can't seem to figure out which voltages are safe and unsafe, even your cooling is really really REALLY good. It's more of a "what makes you feel safe" kind of thing. Now, if your like me, "feeling" something out is absolute nonsense. I want hard facts and evidence. (To clarify, the overclocking community DOES agree on a voltage you should not go over, but it's the voltage just below that max voltage that gets everybody butting heads.)

With this thread I hope to help everybody get to the truth about their CPU's degregation as much as possible. So lets start:

**************************************************************************************
Haswell/Devils Canyon:
During all my research, haswell is probably one of the most annoying architectures when it comes to safe 24/7 voltages. There are MANY people who say that 1.35v is a NONO no matter what, then there are MANY other people who say 1.35v is fine if you get low temps. Here is my chart of what I believe is safe and unsafe for haswell after hours of looking up threads and forums on the web about this topic.

**************************************************************************************
100% Safe Voltages:

1.2-1.3v below 90C
1.32v below 85C
1.35v below 70C

Unsafe Voltages:

1.2-1.3v above 90C
1.32 above 85C
1.35v above 75C
1.4v
1.5v

Again these are voltages for 24/7 uses. If you want to make your chip melt faster, more power too you.

**************************************************************************************


More coming soon!

I would LOVE feedback on this topic, please share, discuss and argue (civily that is).

 

[Dark Lord of Tech]

http://www.tomshardware.com/community/profile-111321.htm
Ask HIM TO ASSIST YOU.

 

[chenuki]

What being described as a lack of consensus on hard voltage numbers is because of silicon lottery. The best data we have comes from Intel's datasheet. Even the numbers from a large community sample size are at best approximations.

https://www-ssl.intel.com/content/www/us/en/processors/core/7th-gen-core-family-desktop-s-processor-lines-datasheet-vol-1.html

 

[Robert Cook]

I'll copy pasta this here:
Actually hardcore overclocking on youtube does a good (albeit round about)job of explaining high voltages on CPUs, the Amps are what really matter. (after all electron migration is a result of electron flow which is what the amp is (coulombs per second (or units of charge (electrons) per second)) That is also why LN2 voltages can be so high, as the CPU draws way less current per given voltage at those temps.

So temps are the killer, but only because higher temp results in greater current draw through the chip (the silicon is just as stable chemically at that higher temp, and really is not moving that much faster at the atomic level)

https://www.youtube.com/watch?v=bXOu3hseXRg
https://www.youtube.com/watch?v=eZbBj9bC7wU

 

[TJ Hooker]

@Robert Cook he doesn't really explain much in those videos as far as I'm concerned. It's mostly armchair expertise and speculation. Which is kind of my problem with most of his videos that I've seen.

Also, when claiming that amps kill CPUs, not voltage, he states that the Iccmax listed on the data sheet you'll kill your CPU. Well, the data sheet lists 100 A as the max for Skylake quad core unlocked CPUs, and if you look at the power consumption results for the 6700k/7700k, you can clearly see that they must be drawing 100+ A in some cases (by using I=P/V), so it would seem he's wrong there.

So temps are the killer, but only because higher temp results in greater current draw through the chip (the silicon is just as stable chemically at that higher temp, and really is not moving that much faster at the atomic level)

I'm not so sure about that. He states that you can use higher voltages with LN2 because it reduces current. That may be true, but I think the temperature reduction is more important, as electromigration is dependent on temperature as well as current. Black's Equation states that:

mean time to failure = A/(j^2)exp(Q/kT)

j is current density, T is temperature.