Did some more tweaking and testing. I am now at1.31v (fixed) in the BIOS with essentially stock speeds (manually set to 42x). Level four LLC was dipping too much so I tried level three which seems to keep the minimum voltage close to the BIOS voltage when under load. At level three it still goes beyond the BIOS voltage a bit under certain conditions, but at 1.31v the maximum reaches a little under 1.35v.

Did Intel Burn Test on very high for 25 passes and it succeeded with the above settings. I will probably do a custom 8-12GB 25+ pass test when I have more time. I was previously getting crashes in IBT and/or Prime95 with vcore anywhere from 1.31v to 1.34v at 4.4GHz so still not sure what my problem is. Perhaps memory speeds/voltage? My motherboard seems to detect my 3200MHz memory via the XMP profile without issue but maybe it isn't playing nice with my overclock settings at 4.3GHz+

 

You should use a medium to high level of LLC. Linpack (IBT uses an out of date version of it) and Prime with certain settings on these CPU's will draw way more power than almost all non-stress-test programs and that causes a lot of voltage droop because droop is somewhat proportional to power draw.

^Most loads use a similar amount of power to x264, cinebench etc at the bottom
Quote:

still not sure what my problem is. Perhaps memory speeds/voltage?

Set your cache/uncore to ~500mhz below the core and leave memory at 2133 until you're sure about the settings required for Core stability

OC one thing at a time so that you don't have to ask these questions (and while you're at it, start with a pass at stock settings

)

 

Quote:

Originally Posted by Cyro999

You should use a medium to high level of LLC. Linpack (IBT uses an out of date version of it) and Prime with certain settings on these CPU's will draw way more power than almost all non-stress-test programs and that causes a lot of voltage droop because droop is somewhat proportional to power draw.

^Most loads use a similar amount of power to x264, cinebench etc at the bottom
Set your cache/uncore to ~500mhz below the core and leave memory at 2133 until you're sure about the settings required for Core stability. OC one thing at a time so that you don't have to ask these questions (and while you're at it, start with a pass at stock settings

)

Thank you for the info!

Do you think I would benefit from switching back to level four LLC even when stress testing with those applications? Even if it is stable when gaming/general use I feel like a "stable OC" should pass IBT and/or Prime95 despite the abnormally high amount of droop but I am obviously far from being an expert.

Regarding the cache/uncore, the only seemingly related setting that I see in my BIOS is the CPU cache clock which I've tried at auto, 40x, and 42x (depending on the CPU clocks and the results I had when stress testing them) so far. Are you referring to a different field?

I will try disabling the XMP profile before I make any additional changes so that it runs at 2133, probably for the best...couldn't help myself and was determined to keep it at 3200MHz throughout the process

 

IBT is just an out of date version (~2012?) of the Linpack test that doesn't include some newer optimizations or instruction sets. If you want to pass everything then you can go download the latest one and do 24hr primes etc but many people have moved away from this in the last 3-4 years because of the widening gap between worst-case stress tests and real world applications (you might see 90c while running Linpack but only 50-55c while running an x264 video encode at 100% CPU load across all cores, for example)

see http://www.overclock.net/t/1570313/skylake-overclocking-guide-with-statistics/
Quote:

Regarding the cache/uncore, the only seemingly related setting that I see in my BIOS is the CPU cache clock which I've tried at auto, 40x, and 42x (depending on the CPU clocks and the results I had when stress testing them) so far. Are you referring to a different field?

That's probably it. It goes by the same voltage as the CPU core but might not be stable at as high of a clock speed

 

Quote:

Originally Posted by Cyro999

IBT is just an out of date version (~2012?) of the Linpack test that doesn't include some newer optimizations or instruction sets. If you want to pass everything then you can go download the latest one and do 24hr primes etc but many people have moved away from this in the last 3-4 years because of the widening gap between worst-case stress tests and real world applications (you might see 90c while running Linpack but only 50-55c while running an x264 video encode at 100% CPU load across all cores, for example)

see http://www.overclock.net/t/1570313/skylake-overclocking-guide-with-statistics/
That's probably it. It goes by the same voltage as the CPU core but might not be stable at as high of a clock speed

Huh. I had no idea RE: IBT being outdated and/or inaccurate. Does this also apply to Prime95 in terms of being excessive? Sounds like I should just test stability via normal use of the computer (in my case being media hosting, media playback, gaming, web browsing, etc.).

Would you recommend keeping the CPU cache clock on auto, equal to the CPU clock, or slightly below the CPU clock? I know not to go over the CPU clock, but if I wanted say a 4.4GHz OC should 40x cache clock be more stable than 44x? (just trying to determine whether to start high or low when tweaking).

Thanks again

 

Try the following steps and provided your 6700K is decent overclocker, it should then be stable -

Firstly Reset all your bios Settings to Default, then change only the following settings below and leave the rest at AUTO or whatever default setting they have already.

CPU Overclocking Steps

1)Set CPU Ratio to ALL Cores - then set a multiplier of 45 for CPU Ratio, CPU Cache Ratio and Minimum CPU Cache Ratio. Leave BCLK Frequency at 100.00

2)Set Load Line Calibration to Level 1

3)Use fixed voltage and start off at 1.26 volts and see if it is stable. If not increase voltage to 1.27 and try again. Keep doing this increase until it becomes stable. If your cpu is half decent, then you should get it stable with no more than 1.3 volts. If it needs more than 1.3 volts set in the bios, that is still fine, but it just means that if you then try overclock further to say 4.7 or 4.8ghz for example, it will need a lot more volts and that will cause it to run quite hot under load.

Ram Overclocking Steps :

1)Usually turning on XMP Profile should be enough to run your ram at the rated XMP speed and timings. If after setting XMP it fails to boot, then set the following bios settings manually in addition to leaving XMP Profile ON and it should then boot up fine :

a) Set VCCIO Voltage to 1.2 volts
b)Set VCCSA Voltage to 1.2 volts
c)Set DRAM Voltage to 1.35 volts.

Answers to your questions :

1)No you can run higher than 1.35 volts as a 24/7 setup, all it depends on your cooling setup and what temperatures your cpu is producing at a given frequency and voltage. It is preferable to always keep your CPU Temps below 80 degrees max for a 24/7 setup.

2)Fixed Voltage or Offset Voltage comes down to individual preference. I prefer using Fixed Voltage, with the highest LLC Setting and then finding the lowest volts your cpu needs to maintain stability, which is the method I have described above. I have found that using this method produces the lowest cpu temps at a given frequency. As long as you have your power plan in Windows set to Balanced, then your cpu will downclock to 800mhz and use about 0.855 volts at idle.

3)You can use LLC 1,2 ,3 or 4, its the vcore voltage that will determine how high an LLC you will require. As I said above, I prefer having the least amount of vdroop ( so highest LLC), combined with running the lowest vcore voltage, as opposed to running a higher vcore with a lower LLC.

4)VCCIO and VCCSA voltages, depends on how good the IMC is on your 6700K. The better the IMC the lower those voltages can be set. For your default XMP timings of 16-18-18-36 at 3200mhz, 1.2 volts for each one is more than plenty.

5) Leave VCC PLL voltage and CPU internal PLL voltage set to AUTO. You do not need to touch these at all.

6)XMP ON, will not have any bearing on your CPU Overclocking. Your CPU stability issues are coming from too low an LLC Level and Vcore Voltage. If your XMP settings, memory voltages or any other memory timings were not correct, it would fail to train your memory and the system would not boot up.

7)FCLK Frequency has no impact on overclocking your cpu and its stability. It basically controls the speed at which your CPU communicates with the PCI bus, so your GPU would see a slight increase in performance by setting it from 800mhz to 1ghz, but this would only be noticeable in graphic benchmarking programs such as 3d Mark Firestrike for example.

 

Quote:

Would you recommend keeping the CPU cache clock on auto, equal to the CPU clock, or slightly below the CPU clock?

Well below the CPU core clock, like 500mhz below until you're sure of the CPU core settings and then bring it up afterwards. Overclocking two different clocks at the same time that rely on the same voltage and crash in similar ways is always a bad idea

 

Quote:

Originally Posted by tknight

Try the following steps and provided your 6700K is decent overclocker, it should then be stable -

Firstly Reset all your bios Settings to Default, then change only the following settings below and leave the rest at AUTO or whatever default setting they have already.

CPU Overclocking Steps

1)Set CPU Ratio to ALL Cores - then set a multiplier of 45 for CPU Ratio, CPU Cache Ratio and Minimum CPU Cache Ratio. Leave BCLK Frequency at 100.00

2)Set Load Line Calibration to Level 1

3)Use fixed voltage and start off at 1.26 volts and see if it is stable. If not increase voltage to 1.27 and try again. Keep doing this increase until it becomes stable. If your cpu is half decent, then you should get it stable with no more than 1.3 volts. If it needs more than 1.3 volts set in the bios, that is still fine, but it just means that if you then try overclock further to say 4.7 or 4.8ghz for example, it will need a lot more volts and that will cause it to run quite hot under load.

Ram Overclocking Steps :

1)Usually turning on XMP Profile should be enough to run your ram at the rated XMP speed and timings. If after setting XMP it fails to boot, then set the following bios settings manually in addition to leaving XMP Profile ON and it should then boot up fine :

a) Set VCCIO Voltage to 1.2 volts
b)Set VCCSA Voltage to 1.2 volts
c)Set DRAM Voltage to 1.35 volts.

Answers to your questions :

1)No you can run higher than 1.35 volts as a 24/7 setup, all it depends on your cooling setup and what temperatures your cpu is producing at a given frequency and voltage. It is preferable to always keep your CPU Temps below 80 degrees max for a 24/7 setup.

2)Fixed Voltage or Offset Voltage comes down to individual preference. I prefer using Fixed Voltage, with the highest LLC Setting and then finding the lowest volts your cpu needs to maintain stability, which is the method I have described above. I have found that using this method produces the lowest cpu temps at a given frequency. As long as you have your power plan in Windows set to Balanced, then your cpu will downclock to 800mhz and use about 0.855 volts at idle.

3)You can use LLC 1,2 ,3 or 4, its the vcore voltage that will determine how high an LLC you will require. As I said above, I prefer having the least amount of vdroop ( so highest LLC), combined with running the lowest vcore voltage, as opposed to running a higher vcore with a lower LLC.

4)VCCIO and VCCSA voltages, depends on how good the IMC is on your 6700K. The better the IMC the lower those voltages can be set. For your default XMP timings of 16-18-18-36 at 3200mhz, 1.2 volts for each one is more than plenty.

5) Leave VCC PLL voltage and CPU internal PLL voltage set to AUTO. You do not need to touch these at all.

6)XMP ON, will not have any bearing on your CPU Overclocking. Your CPU stability issues are coming from too low an LLC Level and Vcore Voltage. If your XMP settings, memory voltages or any other memory timings were not correct, it would fail to train your memory and the system would not boot up.

7)FCLK Frequency has no impact on overclocking your cpu and its stability. It basically controls the speed at which your CPU communicates with the PCI bus, so your GPU would see a slight increase in performance by setting it from 800mhz to 1ghz, but this would only be noticeable in graphic benchmarking programs such as 3d Mark Firestrike for example.

Many thanks! I will reference this and other posts when I go to do more tweaking tonight. In all of my previous testing I have gone no higher than 1.35v in the BIOS and I get max temps around 68-72C during IBT tests, so I may be able to go higher. Shouldn't have to though, the highest I'd want to go is 4.6GHz.

Initially you said to start at level one LLC but it sounds like I'm looking for what you prefer (least amount of vdroop combined with a low vcore). Would I be better starting off at level four or should I start at level one and increase one level at a time if the voltage range is too wide and causing crashes?

 

Quote:

Originally Posted by TMallory

Many thanks! I will reference this and other posts when I go to do more tweaking tonight. In all of my previous testing I have gone no higher than 1.35v in the BIOS and I get max temps around 68-72C during IBT tests, so I may be able to go higher. Shouldn't have to though, the highest I'd want to go is 4.6GHz.

Initially you said to start at level one LLC but it sounds like I'm looking for what you prefer (least amount of vdroop combined with a low vcore). Would I be better starting off at level four or should I start at level one and increase one level at a time if the voltage range is too wide and causing crashes?

LLC Level 4 is simply too low for 4.5ghz and above. Your main goal is to determine the lowest vcore voltage required for a given frequency, in this case 4.5ghz.
The best way to do that is to set the LLC to Level 1, which basically means no vdroop, that way you don't have to worry about vdroop, while you are focusing on achieving the correct voltage.
Then you start off with the lowest voltage, which for an above average 6700k at 4.5ghz, its about 1.25-1.26 vcore volts.
Then see if it boots into Windows and try run Prime95 or ROG RealBench Stress Test, or AIDA Extreme Stress Test for an hour or two at most and if it passes then it means you are stable at that voltage.

With regards to CPU Cache, ideally you want to be setting your cache ratio to be the same as your cpu ratio for maximum performance. However increasing the CPU Cache does require more vcore voltage, than just increasing the cpu ratio alone. Try set both the same to start off with and see what vcore voltage you get to, for it to be stable.

If it is no more than 1.3 volts than you can leave both the same. If your vcore volts are considerably higher than 1.3 volts, such as 1.35, then set your cache back to default/AUTO and then try find the lowest vcore voltage for just the cpu core being stable at 4.5ghz.

What is your system setup, what are your components and what cooling setup do you have, air or water ( AIO cooler or custom loop) ?

 

Quote:

Originally Posted by tknight

LLC Level 4 is simply too low for 4.5ghz and above. Your main goal is to determine the lowest vcore voltage required for a given frequency, in this case 4.5ghz.
The best way to do that is to set the LLC to Level 1, which basically means no vdroop, that way you don't have to worry about vdroop, while you are focusing on achieving the correct voltage.
Then you start off with the lowest voltage, which for an above average 6700k at 4.5ghz, its about 1.25-1.26 vcore volts.
Then see if it boots into Windows and try run Prime95 or ROG RealBench Stress Test, or AIDA Extreme Stress Test for an hour or two at most and if it passes then it means you are stable at that voltage.

With regards to CPU Cache, ideally you want to be setting your cache ratio to be the same as your cpu ratio for maximum performance. However increasing the CPU Cache does require more vcore voltage, than just increasing the cpu ratio alone. Try set both the same to start off with and see what vcore voltage you get to, for it to be stable.

If it is no more than 1.3 volts than you can leave both the same. If your vcore volts are considerably higher than 1.3 volts, such as 1.35, then set your cache back to default/AUTO and then try find the lowest vcore voltage for just the cpu core being stable at 4.5ghz.

What is your system setup, what are your components and what cooling setup do you have, air or water ( AIO cooler or custom loop) ?

Full specs are in my sig but I am currently using a Noctua NH-D15 with only one of the two fans installed at the moment (temps haven't gone beyond 68-72C regardless of settings when running IBT and games).

Regarding LLC, when I had it at level one it seemed to fluctuate quite a bit...when trying levels one through three I found that the voltage went too low or too high under load. Level four was where vdroop was the lowest in my case, kept the minimum somewhat steady though still found that the maximum was too high (noticeably beyond the BIOS voltage but still not beyond what I would consider to be a safe max voltage for 24/7 use). I've yet to go below 1.3v or above 1.35v in BIOS so far but I will try dropping it and messing with LLC a bit more. Thanks again!

 

Also, I am currently on the high performance Windows 10 power profile. I think I will try switching it back to balanced prior to making additional changes in hopes that it'll be cooler/quieter and less power hungry when not under load. I work full time so I'm only using the computer for 4-6 hours on weekdays.

I will then then reset my BIOS to default settings and start trying out some of the info provided in this thread.

 

Alrighty, so I just did a ton of testing. The following results all use the following settings: CPU/cache/min all set to 44. XMP is on. VCCIO and VCCSA are set to 1.2 volts. My memory is set to 1.35 volts. I am using the balanced Windows 10 power profile. Still using IBT for testing since it was quicker than firing up Battlefield 1 on each attempt. I am using HWMonitor and CPU-Z to monitor voltages. Idle clocks are 800MHz and it runs at full clock speed under load. These results are listed in the order that I tried them. I was able to log into Windows on every single attempt, but the majority of attempts resulted in freezes before/during the IBT test. I ran the IBT test within one minute of logging into Windows. I ran IBT on very high which uses 4GB of RAM (of 16GB). I bailed on configurations that resulted in voltages over 1.35v even if they passed IBT. I only did five passes on IBT but the majority of the freezes occur pretty much immediately in the first run.

BIOS vcore: 1.28v
LLC level: one
Startup/idle voltage: up to 1.328v
Maximum voltage: 1.36v
Result: freezes before/during IBT test

BIOS vcore: 1.28v
LLC level: two
Startup/idle voltage: n/a
Maximum voltage: n/a
Result: freezes before IBT test

BIOS vcore: 1.3v
LLC level: two
Startup/idle voltage: 1.312v
Maximum voltage: 1.36v to 1.37v
Result: freezes before/during IBT test

BIOS vcore: 1.3v
LLC level: three
Startup/idle voltage: 1.312v
Maximum voltage: n/a
Result: freezes before/during IBT test

BIOS vcore: 1.31v
LLC level: three
Startup/idle voltage: 1.312v to 1.328v
Maximum voltage: n/a
Result: freezes before/during IBT test

BIOS vcore: 1.31v
LLC level: two
Startup/idle voltage: n/a
Maximum voltage: n/a
Result: freezes before IBT test

BIOS vcore: 1.31v
LLC level: four
Startup/idle voltage: 1.312v
Load voltage: this one drops, to 1.296v
Result: freezes during IBT test

BIOS vcore: 1.32v
LLC level: four
Startup/idle voltage: 1.312v to 1.328v
Load voltage: this one drops, to 1.296
Result: freezes during IBT test

BIOS vcore: 1.33v
LLC level: four
Startup/idle voltage: 1.312v to 1.328v
Load voltage: stays at 1.312v
Result: freezes during IBT test

BIOS vcore: 1.34v
LLC level: four
Startup/idle voltage: 1.328v to 1.344v
Load voltage: stays at 1.328v to 1.344v
Result: freezes during IBT test

BIOS vcore: 1.345v
LLC level: four
Startup/idle voltage: 1.344v
Maximum voltage: 1.328v
Result: freezes during IBT test

BIOS vcore: 1.345v
LLC level: three
Startup/idle voltage: 1.36v
Maximum voltage: 1.376v to 1.392v
Result: passed IBT, but voltage is too high

BIOS vcore: 1.34v
LLC level: two
Startup/idle voltage: 1.36v to 1.392v
Maximum voltage: 1.376v to 1.424v
Result: passed first IBT run, stopped it due to voltage

BIOS vcore: 1.32v
LLC level: two
Startup/idle voltage: 1.36v to 1.376v
Maximum voltage: 1.36v to 1.392v
Result: passed IBT, but voltage is too high

BIOS vcore: 1.31v
LLC level: two
Startup/idle voltage: 1.328v to 1.344v
Maximum voltage: n/a
Result: freezes during IBT test

BIOS vcore: 1.315v
LLC level: two
Startup/idle voltage: 1.328v to 1.344v
Maximum voltage: 1.344v to 1.392v
Result: passed IBT, but voltage is too high

BIOS vcore: 1.3v
LLC level: one
Startup/idle voltage: 1.344v to 1.36v
Maximum voltage: 1.328v to 1.376v
Result: freezes during IBT test

BIOS vcore: 1.3v
LLC level: two
Startup/idle voltage: 1.312v to 1.344v
Maximum voltage: 1.328 to 1.376v
Result: freezes during IBT test

BIOS vcore: 1.33v
LLC level: four
Startup/idle voltage: 1.328v
Maximum voltage: 1.328v
Result: freezes during IBT test

BIOS vcore: 1.34v
LLC level: four
Startup/idle voltage: 1.328v to 1.344v
Maximum voltage: 1.312v
Result: freezes during IBT test

BIOS vcore: 1.29v
LLC level: one
Startup/idle voltage: 1.296v to 1.312v
Maximum voltage: 1.328v to 1.344v
Result: no IBT test. eventually froze up while playing Battlefield 1

BIOS vcore: 1.3v
LLC level: one
Startup/idle voltage: 1.312v to 1.344v
Maximum voltage: 1.36v
Result: freezes shortly after closing Battlefield 1 after playing a round

BIOS vcore: 1.33v
LLC level: four
Startup/idle voltage: 1.312 to 1.328v
Maximum voltage: remains at 1.328v during light use
Result: no IBT test. eventually froze while typing this post

After the last freeze I changed my CPU/cache/min multiplier to 42. All good so far but haven't run any tests or played Battlefield 1. In the week or so that I've had this CPU I have yet to run into any crashes at 4.2GHz even in 25+ run IBT tests. I really don't get why I can't get 4.4GHz running reliably under load at under 1.35v

I will probably try disabling XMP and using lower cache and min multipliers in my next round of testing, not sure what else would be causing these problems.

 

Why are you decreasing the LLC Level, after it fails to pass stability at a certain voltage and LLC Level ? That makes no sense whatsoever, because if for example it fails stability testing at 1.28 volts at LLC Level 1, then decreasing to LLC Level 2, will only result in it failing again. If it fails at LLC Level 1, that means what you have your vcore volts set to, is not enough voltage for your cpu at 4.4ghz.

All the tests you have done, have been in a sort of backward order and were only going to fail, no matter what.

At your first result of 1.28 volts at LLC Level 1 failing, then your next test should have been 1.29 volts at LLC Level 1. Instead you went to LLC Level 2 at the same 1.28 volts, that would never work as all you are doing by lowering the LLC Level is increasing the amount of vdroop and providing less voltage stability to your cpu.

From your results, the highest you have tested 4.4ghz both core and cache, is with a BIos Vcore of 1.3 volts at LLC Level 1 and it failed. So your next test should be a Bios Vcore of 1.31 volts at LLC Level 1 and try again. If that fails then increase to 1.32 volts at LLC Level 1 and test again. If that fails then increase to 1.33 volts at LLC Level 1 and test again. And so on and so on until it is stable. Do not drop the LLC Level at all, keep it at Level 1, just keep increasing your Bios Vcore volts as above. Once you achieve 4.4ghz stability, note what the Vcore volts is set to in your Bios and if it is higher than 1.35 volts, then unfortunately you do not have a good overclocking 6700K and you will struggle to overclock much higher without having a really high Vcore voltage.

Always keep an eye on your temperatures, at all times during stability testing and make sure it doesn't go above 80-85 degrees. You could let it stretch out to 90 degrees max during stress testing, such as Prime95, but only if its like hovering around 85 degrees and peaks to 90 degrees and then goes back down to 85-86 degrees. If it goes over 90 degrees and stays there, then I would stop the tests, as you have reached the limit of your cooler.

 

I raised the LLC level because level one resulted in voltages over 1.35v under load when set to 1.28v to 1.3v in the BIOS. At that point isn't anything higher in the BIOS guaranteed to result in the same >1.35v load voltage? Or are those load voltages OK as long as I have it set to under 1.35v in the BIOS?

 

What program are you using to monitor the load voltage ?

The Vcore voltage in the bios is what you want to be keeping under 1.35 volts for your 4.4ghz overclock.

 

Quote:

Originally Posted by tknight

What program are you using to monitor the load voltage ?

The Vcore voltage in the bios is what you want to be keeping under 1.35 volts for your 4.4ghz overclock.

I have been using both HWMonitor and CPU-Z at the same time, launching them right when I get into Windows moments before trying to run IBT. They seem to be reporting pretty much identical voltages at any given time, so I feel like they are accurate.

 

Huh. I just installed HWinfo64 and ASRock's tuning utility (for monitoring purposes only) and they both report lower voltages than the (matching) HWMonitor/CPU-Z voltages. I figured those two were accurate since they matched each other, but maybe both are wrong...same developer so wouldn't be shocked despite both being on the current version.

If these new, lower readings are truly accurate then everything starts to make sense since I was passing IBT when the latter 2x applications reported over 1.35v under load despite being set much lower in the BIOS. Perhaps it was truly under my vcore limit resulting in a stable OC at safe 24/7 voltages.

I think I am going to try 4.4GHz again with level one LLC using HWinfo64/ASRock to monitor voltages tonight.

 

Yes different software monitoring programs can give different readings, and aren't always accurate.
Since you have an Asrock board, use Asrock Formula Drive to monitor your board as that will be reporting the same figures, as the hardware monitor in your bios.
Also apart from focusing on just your vcore voltage alone, you need to be checking what temperatures you are getting when testing for stability. The whole point of trying to achieve the lowest stable vcore volts, is to keep your temperatures as low as possible.
It doesn't matter if you see more than 1.35 volts in CPUZ while under load, as long as its stable and your temperatures are in check, then thats what is most important.

 

Quote:

Originally Posted by tknight

Yes different software monitoring programs can give different readings, and aren't always accurate.
Since you have an Asrock board, use Asrock Formula Drive to monitor your board as that will be reporting the same figures, as the hardware monitor in your bios.
Also apart from focusing on just your vcore voltage alone, you need to be checking what temperatures you are getting when testing for stability. The whole point of trying to achieve the lowest stable vcore volts, is to keep your temperatures as low as possible.
It doesn't matter if you see more than 1.35 volts in CPUZ while under load, as long as its stable and your temperatures are in check, then thats what is most important.

During all of my testing I never hit 80C. The vast majority of the tests resulted in max temps of 69-74C. I think I will trust ASRock (with HWinfo64 as my secondary) for voltage monitoring on my upcoming tests since it's their motherboard.

In regards to that last part, I will be running this computer 24/7 for years to come so I'd like to keep voltage at a safe level. If CPU-Z is accurate, then over 1.35v worries me. If CPU-Z is fibbing, then I feel way more confident about getting a stable OC under that voltage. Again, not trying to break any records but based on what I've read 4.4GHz at under 1.35v should be a walk in the park.

 

You should be doing 4.4ghz at a lot less voltage than 1.35 volts, infact to be honest with you it should be a lot less than 1.30 volts too. The fact that you failed stability at 1.30 volts indicates that unfortunately you have a below average 6700K.

You can run a 6700K at more than 1.35 volts for 24/7 use, as Intel's recommended limit is 1.45 volts.

As you can see from the following chart, an average 6700K only needs 1.23 volts to achieve 4.4ghz and thats while running Prime95 1344K test. This was also with the ram running at 3200mhz C16.

 

Quote:

Originally Posted by tknight

You should be doing 4.4ghz at a lot less voltage than 1.35 volts, infact to be honest with you it should be a lot less than 1.30 volts too. The fact that you failed stability at 1.30 volts indicates that unfortunately you have a below average 6700K.

You can run a 6700K at more than 1.35 volts for 24/7 use, as Intel's recommended limit is 1.45 volts.

As you can see from the following chart, an average 6700K only needs 1.23 volts to achieve 4.4ghz and thats while running Prime95 1344K test. This was also with the ram running at 3200mhz C16.

no offense and i'll admit you seem to give good advice but that chart is based off of one sample.

looking at the skylake chart here on OCN it seems that ~1.3 or above for 4.5* isn't uncommon

*yeah i know 4.4 is what is pointed out but there is just one sample there ( at 1.36v

) but i am looking at a .07 vcore increase for 100mhz? (4.4@1.23 vs 4.5@~1.3+)

i'm just saying based on a larger sample size i don't know if TMallory has a less than average chip is all.

E;
grammar.

 

Quote:

Originally Posted by looniam

no offense and i'll admit you seem to give good advice but that chart is based off of one sample.

looking at the skylake chart here on OCN it seems that ~1.3 or above for 4.5* isn't uncommon

*yeah i know 4.4 is what is pointed out but there is just one sample there ( at 1.36v

) but i am looking at a .07 vcore increase for 100mhz? (4.4@1.23 vs 4.5@~1.3+)

i'm just saying based on a larger sample size i don't know if TMallory has a less than average chip is all.

E;
grammar.

No offence taken at all. Yes that chart is from one sample, but the person who made the chart has tested a large number of 6700k's and knows what is considered as an average 6700K.

1.3 volts and a bit above, is 4.5ghz territory for an average 6700K. 4.4ghz on average needs about 1.23-1.25 volts. The skylake chart here on ocn hardly has a couple of 4.4ghz entries, so not much to compare there. But if you look at the majority of the 4.5ghz entries, they range from 1.28- 1.32 volts. Therefore a 6700k needing more than 1.3 volts for 4.4ghz is below average.

If you look at the following spreadsheet of 6700k overclocking results from hwbot, it tells a completely different story as to what is average voltages for a 6700K. You will see that 1.35-1.40 volts is mainly 4.7ghz to 4.8ghz clock speed territory.

http://forum.hwbot.org/showpost.php?p=400330&postcount=1

 

ah, i see a bit better, thanks for the HWBot link. i don't go to that forum (much) and should (more).

 

Been using RealBench for brief stress tests instead of IBT to get an initial idea regarding stability and had some better results. I boot into Windows, confirm the idle voltages, and run a quick RealBench test before rebooting back into BIOS. Now using ASRock's utility and CPU-Z to monitor voltages - HWinfo64 turned out to show 1.26v no matter what I had the voltage set to in the BIOS so I stopped using it. That being said, ASRock and CPU-Z matched up consistently. Still using the balanced Windows 10 power profile (CPU speed drops to 800MHz at idle) and 1.2v for VCCSA and VCCIO. I still have XMP enabled to achieve 3200hz at 1.35v. ASRock reported slightly higher voltages for memory, VCCSA, and VCCIO under load. CPU, cache, and min are all at 44x. My fan settings are fairly low and my cooler is installed with only one of two fans, so these temps can be dropped if needed.

BIOS voltage: 1.32v
LLC level: one
ASRock idle voltage: 1.33 to 1.335v
CPU-Z idle voltage: 1.328v
RealBench: ran on default settings (4GB RAM) for five minutes and succeeded. Both monitoring sources report up to 1.392v under load. Max temp on the four cores range from 75-82C.

BIOS voltage: 1.31v
LLC level: one
ASRock idle voltage: 1.328 to 1.33v
CPU-Z idle voltage: 1.328v
RealBench: default test again for five minutes and succeeded. Load voltages up to 1.376v from both sources. Max temp on the four cores range from 75-82C.

BIOS voltage: 1.3v
LLC level: one
ASRock idle voltage: 1.312v
CPU-Z idle voltage: 1.312v
RealBench: default test again for five minutes and succeeded. Load voltages up to 1.376v from both sources. Max temp on the four cores range from 74-80C.

BIOS voltage: 1.29v
LLC level: one
ASRock idle voltage: 1.3 to 1.312v
CPU-Z idle voltage: 1.296 to 1.328v
RealBench: default test again for five minutes and succeeded. Load voltages up to 1.36v from both sources. Max temp on the four cores range from 74-78C.

BIOS voltage: 1.28v
LLC level: one
ASRock idle voltage:1.296 to 1.3v
CPU-Z idle voltage: 1.296v
RealBench: default test again for five minutes and succeeded. Load voltages up to 1.344v from both sources. Max temp on the four cores range from 74-78C.

At this point I let RealBench complete the 15 minute test. At some point during the last 10 minutes, my video cut out for a couple of seconds and returned - this seemed to have no effect on the benchmark and it completed the 15 minute run with no errors. I retried the test but on the higher 8GB setting. I get the same load voltages but when I returned from the kitchen it had crashed at some point - it was stuck at the "no video" message on my display much like the previous test but it was permanent this time. While at this screen, I checked LogMeIn from my phone and my computer was offline. This could have happened towards the end of the test or right after the test when idle.

I then set my BIOS voltage to 1.285v on the next reboot. Same idle and load voltages as the 1.28v results above. The video cuts out for a second again but returns and I pass the 15 minute RealBench test. I had to leave at this point so I let my computer sit idle and confirmed that it was online in LogMeIn before heading out the door. Checked again in 10 minutes or so and it was offline in LogMeIn again. I had my roommate shut it down until I return tomorrow to do more testing.

These crashes are different than the previous ones during my IBT tests where the computer just froze. This time, it appears to be during idle (I do have a bunch of applications that sit in the notification area and occasionally hit the CPU) My test period is obviously pretty short (20mins max on each config) so perhaps the same thing would happen even at load if I was pushing it for longer than the RealBench runs. Going to expand my test time and do idle, RealBench, gaming, media playback, web browsing, etc. but curious whether I should change anything else due to the seemingly idle crashes (ie. raising LLC to get a higher idle voltage with similar or hopefully lower load voltages).

 

I did some more testing and I appear to be stable with the following settings after trying a bunch of different configs: 4.5GHz with 40x on cache and min. XMP on for 3200MHz. BCLK is set to 100 and FCLK is set to 1GHz. I am set to fixed 1.32v in the BIOS with level two LLC and at idle multiple monitoring applications reports 1.328v. Under load it goes up to 1.376v when running RealBench (15 minute 8GB test) and 1.36v to 1.376v while playing Battlefield (one full round). No issues so far. The video hasn't cut out like it was a couple of days ago but I ran into the original freezing type crashes during today's testing. I also disabled my video card OC but still had freezes with other configs so much be unrelated. Still on the balanced Windows 10 power profile but disabled PCIe power saving.

Max temps on each core during all of this was 75-80C. Going to let it sit idle for a bit and see if it freezes. If I continue to experience zero problems, should I try to get those voltages lower? 1.32v (my BIOS setting) is perfectly fine but hitting 1.376v under load seems like a bit much. I don't think I can get the BIOS setting much lower while remaining stable but not sure if I should be going by the BIOS voltage or what these apps are reporting under load...ASRock, CPU-Z, and HWMonitor all showed the same reading.

EDIT: the BIOS voltage was actually 1.33v - going to try level three LLC to see if it reigns in the max voltages a bit. If not, I can either try level four or go back to level two with 1.325v (and below if that works).