5930K Overclock Help

[DrJones]

Hi everyone, I can't can't get my 5930K to be stable in RealBench. System specs are as follows (brand new build, everything is updated):

 

Rampage V Extreme

Intel i7 5930K

Corsair Dominator Platinum DDR4 64GB

GTX1070 x2

 

I've tried 4.5Ghz down to 4.2Ghz at 1.3V (I don't want to go higher than 1.3, I'll go 1.31 if I have to for stability).

ive run multiple stress tests in RealBench, and only got it to pass one time at 4.2Ghz at 1.30V, didn't ever get it to pass a second time.

Ive been reading reviews online, people are able to hit 4.5Ghz sometimes at less than 1.3, and easily 4.2 at less than that.

Should I continue to decrease the clock and see where I can pass a stress test at 1.3V?

What's the nex step I should take?

 

If it helps, my stress tests failed within 5 minutes at 4.3, 4.4, and 4.5 Ghz. I passed one 15 test at 4.2Ghz, then failed the second and third test at around 9 minutes consistently. 1.30V was used at each of these settings except I tried 1.31V at 4.5 and 4.4Ghz

[LionSpeck]

15 minutes ago, DrJones said:

Hi everyone, I can't can't get my 5930K to be stable in RealBench. System specs are as follows (brand new build, everything is updated):

 

Rampage V Extreme

Intel i7 5930K

Corsair Dominator Platinum DDR4 64GB

GTX1070 x2

 

I've tried 4.5Ghz down to 4.2Ghz at 1.3V (I don't want to go higher than 1.3, I'll go 1.31 if I have to for stability).

ive run multiple stress tests in RealBench, and only got it to pass one time at 4.2Ghz at 1.30V, didn't ever get it to pass a second time.

Ive been reading reviews online, people are able to hit 4.5Ghz sometimes at less than 1.3, and easily 4.2 at less than that.

Should I continue to decrease the clock and see where I can pass a stress test at 1.3V?

What's the nex step I should take?

 

If it helps, my stress tests failed within 5 minutes at 4.3, 4.4, and 4.5 Ghz. I passed one 15 test at 4.2Ghz, then failed the second and third test at around 9 minutes consistently. 1.30V was used at each of these settings except I tried 1.31V at 4.5 and 4.4Ghz

Firstly, the silicon lottery plays a huge role in OC, so keep that in mind. Secondly, you should rely on different stress tests, you shouldn't decide if an OC is stable just because that program said it (or said it's not): stress tests are exaggerations of real-world workflows, so it's expected that you may fail a stress test, but your PC may not even crash once in years under full load. Because these are synthetics, not real-world. Try AIDA64's stress test (with at least CPU, cache and RAM enabled) and Cinebench for instance. I don't suggest you to try Prime95, because with my Haswell (4.4GHz fully stable 80°C with AIDA after 1+ hour) it got to 90° just by running one second, so I judge it a little bit dangerous. Also in other OC forums, they say that Prime95 is powerful in general, but with Haswells is extremely taxing.

[matrix07012]

Looks like you lost the silicon lottery.

[DrJones]

17 minutes ago, LionSpeck said:

-Snip-

Will the trial version of AIDA64 give me everything I need? I'm running RealBench with all 64GB of RAM enabled. Yeah, I'm definitely staying away from Prime95, I know it can over-volt CPU's. I'll also be running an XMP Profile for my RAM (2800MHz). It increases Base clock to 127 not 125, is this ok or should I manually set it to 125? 

[done12many2]

So how about we get that RealBench to pass and talk about all the other stuff that you didn't ask about after that.  

 

You're more than likely having an issue related to CPU Voltage Input (also known as VCCIN).  Very common on HWE chips and not that big of a deal.  Increasing VCore will not really help this issue much.  

 

Go into BIOS and raise your CPU Input Voltage to 1.95v.  Start with 4.5 GHz and set VCore to 1.3v provided that your cooling setup can handle the thermal load.  Retest for 15 minutes and start dropping voltage until you start crashing again.

 

I very much doubt that you lost the "silicon lottery" that badly.  I'm 95% sure that we just need to get you dialed in.  No biggie.

 

**I should add, I've run RealBench for hours on my 5960x at 4.7 GHz, but I can tell you that it won't even do 5 minutes at 4.5 without a little extra Input Voltage.**

[LionSpeck]

12 minutes ago, DrJones said:

Will the trial version of AIDA64 give me everything I need? I'm running RealBench with all 64GB of RAM enabled. Yeah, I'm definitely staying away from Prime95, I know it can over-volt CPU's. I'll also be running an XMP Profile for my RAM (2800MHz). It increases Base clock to 127 not 125, is this ok or should I manually set it to 125? 

1- Yes, the trial of AIDA should give you 100% of the functionalities
2- No programs should be able to over-volt CPUs, you probably meant throttle or something else, because only the motherboard is able to adjust the CPU's voltage
3- The XMP profile should not affect the actual CPU OC, but the fact that the memory controller draws more power from the CPU input, that could be the issue. Try turning it off, restoring the BCLK to defaults. If this solves, than you should put the CPU Input Voltage to a higher value and after that, re-enable XMP and put the BCLK back to the OC'd value.

[DrJones]

17 minutes ago, done12many2 said:

So how about we get that RealBench to pass and talk about all the other stuff that you didn't ask about after that.  

 

You're more than likely having an issue related to CPU Voltage Input (also known as VCCIN).  Very common on HWE chips and not that big of a deal.  Increasing VCore will not really help this issue much.  

 

Go into BIOS and raise your CPU Input Voltage to 1.95v.  Start with 4.5 GHz and set VCore to 1.3v provided that your cooling setup can handle the thermal load.  Retest for 15 minutes and start dropping voltage until you start crashing again.

 

I very much doubt that you lost the "silicon lottery" that badly.  I'm 95% sure that we just need to get you dialed in.  No biggie.

 

**I should add, I've run RealBench for hours on my 5960x at 4.7 GHz, but I can tell you that it won't even do 5 minutes at 4.5 without a little extra Input Voltage.**

That sounds like a plan, I'll give it a go when I get home. Yeah I know people can lose the silicon lottery but not being able to hit 4.2 at 1.30V? Come on lol.

EDIT: Temps are not a problem, running a custom loop, 30 at idle, 43 max on GPU's under 100% load, hit 71 max CPU during overclock stress testing.

@LionSpeck thanks for the input, I've been trying to overclock with XMP off, won't turn it on until I get the OC dialed in, but I did just look at it in the BIOS.  So I'll OC then turn on XMP and run more tests after that.

[LastThoughts]

Like LionSpeck said the sillicon lottery play a big part in overclocking. But 1.3V seems a little high for 4.2GHz. Are you sure that the CPU are actually on that voltage? 

Maybe something else could be causing the instability. Try setting everything to default (Optimize Default), and only changing the Vcore and CPU Frequency (also set to the correct Memory Profile). 

And, yeah, stay clear from Prime95, it can fry your chip (it overvolt your CPU even further and draw too much power, I think Linus explains why on some OC Guide) 

[done12many2]

19 minutes ago, DrJones said:

That sounds like a plan, I'll give it a go when I get home. Yeah I know people can lose the silicon lottery but not being able to hit 4.2 at 1.30V? Come on lol.

EDIT: Temps are not a problem, running a custom loop, 30 at idle, 43 max on GPU's under 100% load, hit 71 max CPU during overclock stress testing.

 

Trust me, we'll get that bad boy dialed in.  You've just got to approach it with that Texan attitude.  Haha.

 

Demonstrating my point, I reran RealBench at a higher clock rate and grabbed a screen shot right before the 15 minute run ended.  It'll run all day long, but you get my point.  On top of that, it's 8 cores.

 

I've had a bunch of HWE chips and a couple of BWE chips and I've yet to have one not do 4.5 or better.  I attribute most of the stories out there of low overclocks to improper configurations.  I'm not hating, but these chips aren't exactly "plug and play" when it comes to the higher clocks.

 

 

[DrJones]

3 minutes ago, done12many2 said:

 

Trust me, we'll get that bad boy dialed in.  You've just got to approach it with that Texan attitude.  Haha.

 

Demonstrating my point, I reran RealBench at a higher clock rate and grabbed a screen shot right before the 15 minute run ended.  It'll run all day long, but you get my point.  On top of that, it's 8 cores.

 

I've had a bunch of HWE chips and a couple of BWE chips and I've yet to have one not do 4.5 or better.  I attribute most of the stories out there of low overclocks to improper configurations.  I'm not hating, but these chips aren't exactly "plug and play" when it comes to the higher clocks.

 

 

Haha thanks for the help, I'll keep you all posted. What temperature monitor is that you're using? I'm currently using realtemp GT

[done12many2]

1 minute ago, DrJones said:

Haha thanks for the help, I'll keep you all posted. What temperature monitor is that you're using? I'm currently using realtemp GT

 

AIDA64.  It's an on screen display that sits on my desktop 24/7.  Very handy for learning how your system behaves and responds to things when  you can see what's happening all the time. 

[DrJones]

16 minutes ago, done12many2 said:

 

AIDA64.  It's an on screen display that sits on my desktop 24/7.  Very handy for learning how your system behaves and responds to things when  you can see what's happening all the time. 

I may end up putting down the $40 for that program then. From what it seems, it's among the best stress tests and monitoring softwares. I know Linus uses nothing but that program. I got my other programs because they were free.

[LionSpeck]

2 minutes ago, DrJones said:

I may end up putting down the $40 for that program then. From what it seems, it's among the best stress tests and monitoring softwares. I know Linus uses nothing but that program. I got my other programs because they were free.

For monitoring only, HWinfo64 is a solid free alternative. As for benchmarks and stress tests, there's Cinebench, LuxMark and so many others... Stil, AIDA is one of the best in this and more, but if it's not needed, I wouldn't suggest buying it.

[LionSpeck]

2 minutes ago, DrJones said:

I may end up putting down the $40 for that program then. From what it seems, it's among the best stress tests and monitoring softwares. I know Linus uses nothing but that program. I got my other programs because they were free.

For monitoring only, HWinfo64 is a solid free alternative. As for benchmarks and stress tests, there's Cinebench, LuxMark and so many others... Stil, AIDA is one of the best in this and more, but if it's not needed, I wouldn't suggest buying it.

[done12many2]

8 minutes ago, DrJones said:

I may end up putting down the $40 for that program then. From what it seems, it's among the best stress tests and monitoring softwares. I know Linus uses nothing but that program. I got my other programs because they were free.

 

It's great for monitoring and really handy for measuring progress in memory/CPU cache overclocks.  As far as stability testing, not so much.  It's okay, but far from great.  Just because you can run and pass AIDA64 (CPU, FPU, memory) for hours doesn't mean that your really stable.  I simply use it as one of many stability tests that I run before I label an overclock solid. 

 

Heavy encoding runs are what I really like to use as the finale of testing.  HWBOT x265 Benchmark v1.2.0 with Priority : Real time, P.Mode: checked, Preset: 4k, and Overkill mode: 8x or higher will exposed a weak overclock really fast.

[DrJones]

@done12many2 I found this link/overclocking guide.

https://us.hardware.info/reviews/5775/3/haswell-e-overklocking-workshop-how-to-get-the-most-out-of-the-core-i7-5960x-5930k-and-4820k-basic-knowledge-bclk-and-multipliers

 

It says to go 0.4V above your Vcore for VCCIN (1.3V Vcore=1.7V VCCIN). Is this what you are referring to? Should I try 1.7 before a more aggressive 1.95V? It also reminded me of UnCore/Cache frequency and voltage to tweak with at the end.

 

[done12many2]

1 minute ago, DrJones said:

@done12many2 I found this link/overclocking guide.

https://us.hardware.info/reviews/5775/3/haswell-e-overklocking-workshop-how-to-get-the-most-out-of-the-core-i7-5960x-5930k-and-4820k-basic-knowledge-bclk-and-multipliers

 

It says to go 0.4V above your Vcore for VCCIN (1.3V Vcore=1.7V VCCIN). Is this what you are referring to? Should I try 1.7 before a more aggressive 1.95V? It also reminded me of UnCore/Cache frequency and voltage to tweak with at the end.

 

 

Start with 1.95v and you'll trim that once you confirm that you can pass the test. 

 

Worry about the memory, CPU cache and anything else after you dial in your CPU overclock.  This will aid you in trouble shooting once you start pushing the other areas of your overclock. 

[DrJones]

5 minutes ago, done12many2 said:

 

Start with 1.95v and you'll trim that once you confirm that you can pass the test.

 

Worry about the memory, CPU cache and anything else after you dial in your CPU overclock.  This will aid you in trouble shooting once you start pushing the other areas of your overclock.

Ok I get it, So it's almost like VCore, find where it works, then tone it down and fine tune. And yes, exactly, dial in the overclock, then worry about the others one at a time so I'm not troubleshooting 4 things at once

[done12many2]

1 minute ago, DrJones said:

Ok I get it, So it's almost like VCore, find where it works, then tone it down and fine tune. And yes, exactly, dial in the overclock, then worry about the others one at a time so I'm not troubleshooting 4 things at once

 

You're on it man. 

[DrJones]

9 hours ago, done12many2 said:

 

You're on it man. 

So I got it stable at 4.4Ghz at 1.30V with an input voltage of 1.92V. I tried 4.5 at 1.95, but it failed pretty quickly, within a little under two minutes. So what's next? Do I decrease VCCIN or Vcore? Or do I overclock RAM and everything else THEN work on VCCIN or Vcore? What are the steps to take in order?

 

UPDATE: Thank you so much! I ran 3 tests in a row stable at 4.4Ghz with a Vcore of 1.30V and Input Voltage of 1.92V. With XMP on, I'm at 4.45Ghz, with RAM at 2800Mhz. Passed another test with XMP on, running a second right now, I'll then run a third perhaps and any other stress tests I can throw at it.

I don't know why people don't mention Input Voltage more often, fixed all my problems.

[done12many2]

1 hour ago, DrJones said:

So I got it stable at 4.4Ghz at 1.30V with an input voltage of 1.92V. I tried 4.5 at 1.95, but it failed pretty quickly, within a little under two minutes. So what's next? Do I decrease VCCIN or Vcore? Or do I overclock RAM and everything else THEN work on VCCIN or Vcore? What are the steps to take in order?

 

UPDATE: Thank you so much! I ran 3 tests in a row stable at 4.4Ghz with a Vcore of 1.30V and Input Voltage of 1.92V. With XMP on, I'm at 4.45Ghz, with RAM at 2800Mhz. Passed another test with XMP on, running a second right now, I'll then run a third perhaps and any other stress tests I can throw at it.

I don't know why people don't mention Input Voltage more often, fixed all my problems.

 

Great to hear bud.  I had a hunch that all would be well.  

 

Don't forget to look into a CPU cache overclock as well.  Some CPU workloads benefit very nicely from cache overclocks and since you have an Asus board (a damn great one at that) you should have no problems taking advantage of it.  Cache speed also directly impacts memory bandwidth especially on x99 with its quad channel option.

 

When and if you decide to push cache a bit, you'll be concerned primarily with cache voltage obviously, but may need to tweak VCCSA and VCCIO voltages just a tad.  Save the last two if cache voltage does't help on its own.  Cache on HWE chips can easily go into the mid 4 GHz range (1:1 with the CPU clock) with some excellent examples reaching up past 4.7+ GHz.  This generally only happens on Asus boards.   

 

It's good to see people pushing their HWE and BWE chips.  Good luck and hit us up if you need anything.

 

**One last thing to put your mind at ease if it wasn't already.  VCCIN of 1.95v and lower is more that safe for daily use on HWE chips.  You're sailing easy at 1.92v.  Nice! 

 

[DrJones]

1 minute ago, done12many2 said:

 

Great to hear bud.  I had a hunch that all would be well.  

 

Don't forget to look into a CPU cache overclock as well.  Some CPU workloads benefit very nicely from cache overclocks and since you have an Asus board (a damn great one at that) you should have no problems taking advantage of it.  Cache speed also directly impacts memory bandwidth especially on x99 with its quad channel option.

 

When and if you decide to push cache a bit, you'll be concerned primarily with cache voltage obviously, but may need to tweak VCCSA and VCCIO voltages just a tad.  Save the last two if cache voltage does't help on its own.  Cache on HWE chips can easily go into the mid 4 GHz range (1:1 with the CPU clock) with some excellent examples reaching up past 4.7+ GHz.  This generally only happens on Asus boards.   

 

It's good to see people pushing their HWE and BWE chips.  Good luck and hit us up if you need anything.

 

Okay, I'll look into that. My second and third tests with XMP enabled failed. Tried increasing RAM voltage, didn't help. I think the extra bump to 4.45Ghz did it. So I just decreased my multiplier from 35 to 34 giving me 4.34Ghz. I'm about to test to see how it does.

[done12many2]

2 minutes ago, DrJones said:

Okay, I'll look into that. My second and third tests with XMP enabled failed. Tried increasing RAM voltage, didn't help. I think the extra bump to 4.45Ghz did it. So I just decreased my multiplier from 35 to 34 giving me 4.34Ghz. I'm about to test to see how it does.

 

Oh, 2800 RAM flips you over to the 125 strap.  That sucks.

 

Never mind, you probably bumped the BLCK.

 

 

 

[DrJones]

Just now, done12many2 said:

 

Oh, 2800 RAM flips you over to the 125 strap.  That sucks.

Yeah not only that but to 127.3

So I'm going to see if it passes at the lower multiplier. If it does I may bump it back up and try 1.31 on Vcore

[done12many2]

3 minutes ago, DrJones said:

Yeah not only that but to 127.3

So I'm going to see if it passes at the lower multiplier. If it does I may bump it back up and try 1.31 on Vcore

 

Try a little extra VCore and maybe a bit of positive System Agent voltage (VCCSA) offset.

 

Sparingly with the VCCSA increase because to much can be as unstable as too little.  You're basically trying to help your CPU's memory controller just a bit.