5820K Overclocking - How Is My Voltage and Other Settings

[Lays]

2 minutes ago, done12many2 said:

 

That sir is exactly what happens.  Haswell-E chips simply don't scale 600 MHz over the span of 100mv after 4.4GHz.

@MCCOPRA You may want to turn your VCCIN down a bit man.  That's a lot of CPU input voltage for that level of VCore.

It's a huge reason why tons of dickheads on hwbot were selling 4.7-4.8 ghz 5960x and 5820k's that did it at like 1.2v, but what they didn't tell you is that 4.9 or 5 needed 1.4-1.5v lol.

[Simon771]

6 minutes ago, Lays said:

It's a huge reason why tons of dickheads on hwbot were selling 4.7-4.8 ghz 5960x and 5820k's that did it at like 1.2v, but what they didn't tell you is that 4.9 or 5 needed 1.4-1.5v lol.

But to be fair, 4,8GHz at 1,2V is awesome. At least I think so.

And those extra 200MHz isn't worth it in my opinion.

[done12many2]

1 minute ago, Lays said:

It's a huge reason why tons of dickheads on hwbot were selling 4.7-4.8 ghz 5960x and 5820k's that did it at like 1.2v, but what they didn't tell you is that 4.9 or 5 needed 1.4-1.5v lol.

 

...and of those running 4.9 to 5 GHz at 1.4v to 1.5v, they were benchmark usable but FAR from stable.  I don't get it man.  This need to convince the world that you were luckier than everyone else in the "Silicon Lottery" is funny.

 

What's even funnier is that when these claims are made, they are always complete with so few screenshots of what really matters.

[MCCOPRA]

1 hour ago, done12many2 said:

 

Your CPU clock speed and voltage didn't scale this smoothly.  It just does not happen like that in Haswell-E.

 

If you can pass AIDA64 for 30 minutes at 5 GHz, then you should have no problem running Cinebench R15 at 5 GHz.  I bet it will either crash instantly or your score wont scale or match your 5 GHz clock speed. 

Maybe I am that lucky or maybe I am lying. Since i dont have decent cooling yet to get on with 5ghz clock speed we may never know i will try to post as may validation and test then ever my h115i arrives and if it can coop with the clocks and voltages.

 

but first when i get to 5 will try cinebecnh with cool processor

 

Thanks for your input.

[done12many2]

2 minutes ago, MCCOPRA said:

Maybe I am that lucky or maybe I am lying. Since i dont have decent cooling yet to get on with 5ghz clock speed we may never know i will try to post as may validation and test then ever my h115i arrives and if it can coop with the clocks and voltages.

 

but first when i get to 5 will try cinebecnh with cool processor

 

Thanks for your input.

 

Um, you ran it for 30 minutes at 5 GHz in AIDA64, right?  I think it can handle one 10 second pass on Cinebench R15.  

[MCCOPRA]

1 hour ago, done12many2 said:

@MCCOPRA You may want to turn your VCCIN down a bit man.  That's a lot of CPU input voltage for that level of VCore.

 

what i read from some forums I shoud be safe up to 2.050v on vccin but i can try.

 

I tried with 1.825 and 1.925v but was getting constant memory corrupt errors so i upped it to 1.975v testing it will try to lover to 1.950 and see if still stable with 4.6ghz - I think my main issue is running SLI 1070 maybe can use some input on this issue

[done12many2]

12 minutes ago, MCCOPRA said:

what i read from some forums I shoud be safe up to 2.050v on vccin but i can try.

 

I tried with 1.825 and 1.925v but was getting constant memory corrupt errors so i upped it to 1.975v testing it will try to lover to 1.950 and see if still stable with 4.6ghz - I think my main issue is running SLI 1070 maybe can use some input on this issue

 

VCCIN is the CPU input voltage, not the voltage for the CPU's Integrated Memory Controller (IMC).  If you were getting constant memory corruption and adjusting VCCIN fixed it, that means that your overall CPU overclock wasn't stable.  Not your IMC or memory.

 

In order to fix IMC related issues, you'll adjust VCCSA (System Agent) and maybe come VCCIO if necessary.  Memory issues can be related to CPU Cache voltage (VCache), VCCA, and VCCIO.

 

The reason why increasing VCCIN fixed your issues is that CPU input can stabilize a small lack of VCore.  This is to say that increasing VCCIN will allow you to get away with less VCore to a degree.  Some guys jack their VCCIN very high so they can run slightly lower VCore (due to lack of cooling), but you don't want to run VCCIN high constantly. I'd personally rather run high VCore than high VCCIN as too high for too long on the VCCIN will kill the chip eventually.

[jjohnthedon1]

6 hours ago, Megah3rtz said:

uhm why is turbo boost on 

Because turbo boost USB the name of the multiplier in some setups

because instead of changing all cores to 4.6 u could have 2 at 4.8 for dual thread loads but when the pinned more cores the rest run at 4.6!to save power and heat

[MCCOPRA]

4 hours ago, done12many2 said:

 

VCCIN is the CPU input voltage, not the voltage for the CPU's Integrated Memory Controller (IMC).  If you were getting constant memory corruption and adjusting VCCIN fixed it, that means that your overall CPU overclock wasn't stable.  Not your IMC or memory.

 

In order to fix IMC related issues, you'll adjust VCCSA (System Agent) and maybe come VCCIO if necessary.  Memory issues can be related to CPU Cache voltage (VCache), VCCA, and VCCIO.

 

The reason why increasing VCCIN fixed your issues is that CPU input can stabilize a small lack of VCore.  This is to say that increasing VCCIN will allow you to get away with less VCore to a degree.  Some guys jack their VCCIN very high so they can run slightly lower VCore (due to lack of cooling), but you don't want to run VCCIN high constantly. I'd personally rather run high VCore than high VCCIN as too high for too long on the VCCIN will kill the chip eventually.

Some Websites say 1.95V is max VCCIN and some others say up to 2.2 to 2.4 volts what are the max safe voltages for VCCIN - VCCSA - VCCIO ?

[done12many2]

12 minutes ago, MCCOPRA said:

Some Websites say 1.95V is max VCCIN and some others say up to 2.2 to 2.4 volts

 

Intel also says that the max VCore for HWE is 1.52v, but would you do that crazy shit?    

 

I've run over 1.52v on my 5960x, but it was benchmarking purposes and my cooling supported what I was doing.

 

12 minutes ago, MCCOPRA said:

 

what are the max safe voltages for VCCIN - VCCSA - VCCIO ?

 

The max VCCIN you should use for daily use is right around the 1.95v mark or less.  Once again, I've taken my VCCIN clear past 2v during benchmarking, but it's damn sure not something I do for long periods of time.  If you have a link to the articles outlining up to 2.4v as a safe VCCIN for Haswell-E, I'd love to see them.  Over 2v with Haswell was common for a lot of people, but not for Haswell-E.

 

VCCSA - no need to run more than 1.2v.  VCCSA is one of those touchy voltages so increase and decrease in small increments.  Too little or too much and you may have the same instability problems.

 

VCCIO - leave it on auto.  No need to mess with it unless you know you need to mess with it.

[MCCOPRA]

44 minutes ago, done12many2 said:

 

Intel also says that the max VCore for HWE is 1.52v, but would you do that crazy shit?    

 

I've run over 1.52v on my 5960x, but it was benchmarking purposes and my cooling supported what I was doing.

 

 

The max VCCIN you should use for daily use is right around the 1.95v mark or less.  Once again, I've taken my VCCIN clear past 2v during benchmarking, but it's damn sure not something I do for long periods of time.  If you have a link to the articles outlining up to 2.4v as a safe VCCIN for Haswell-E, I'd love to see them.  Over 2v with Haswell was common for a lot of people, but not for Haswell-E.

 

VCCSA - no need to run more than 1.2v.  VCCSA is one of those touchy voltages so increase and decrease in small increments.  Too little or too much and you may have the same instability problems.

 

VCCIO - leave it on auto.  No need to mess with it unless you know you need to mess with it.

Thx will try them - here is article link Article