i5-6600K, ASUS Z170-A Overclocking

[Itsmikerofl]

Dear overclockers,

This is my first post on the forum. I've lurked until now, and I'm looking for some experienced input on my i5-6600K overclock.

**Specs**

CPU: i5-6600K
Cooler: Corsair H80i GT (120mm)
Mobo: ASUS Z170-A
RAM: 2x4GB Corsair Dominator Platinum 3000MHz C15
GPU: EVGA GTX 980 Ti SC
PSU: EVGA 750B2

**Preface**

I've overclockered a small bit before. Previously, I've only messed with multiplier OC'ing on an i5-2500K. Nothing too advanced, or over anyone's head. After the release of the Skylake CPUs, and hearing how much of a champ they OC like, I decided to upgrade my four year old platform.

Out of the box, the 6600K was running at 4.4GHz at stock voltage. That's pretty darn impressive for a 3.5GHz i5 if you ask me.

Naturally, as any other human, I decided to push the limits. 4.6GHz at stock voltage stressed fine, using both ASUS RealBench for 8 hours, and AIDA64; temps under 80°C.

**The Dilemma**

I had to start upping the Voltage as I approached 4.7GHz. This seemed like the point of discomfort for my 6600K, as simply raising the Multiplier in the BIOS did not get me any results.

This is when I decided to read more into OC'ing, and found out that lowering the Multiplier while upping the BCLK created more stable conditions. It's worth mentioning that I'm not too sure what a good Multi:BCLK ratio would be for my chip.

I finally got 4.7GHz to boot at a raised manual voltage of 1.42V, while the Multiplier was set to 35 and the BCLK was set to 135. Passed both stress tests fine, and temps under 80°C.

In order to even boot at 4.8GHz, I had to go into the BIOS and raise the FCLK from default of 800MHz to 1GHz. Since I haven't done much manual overclocking before, is anyone able to explain what the FCLK is, and what raising it effects?

I have discovered that the option for Load Line Calibration can also effect stability of an OC in positive ways, primarily to counteract what's referred to as "Vdroop". Is it suggested to raise the LLC level to gain more stability?

6600K running on 4.8GHz is interesting; AIDA64 tests stressing the CPU, FPU (also one more object I'm not too informed about), and RAM will run stable for hours without a crash or a Vdroop. It gets confusing when I use AIDA64 to also test CPU Cache; which results in a "Hardware Error" after no more of a minute of testing. This happens whether I test only CPU Cache, or test it along with the other, (working), tests. I have attempted to raise the VCore all the way up to the Intel specified degradation point of 1.52V, with no change.

Is this because my CPU Cache is too high/low? If so, how should I go about adjusting it?

**TL;DR**

i5-6600K stable at 4.7GHz, 1.42V. AIDA64 returns Hardware Error while stressing CPU Cache @ 4.8GHz, any voltage up to 1.52V. Tips?

I realize I just took a really huge info-dump on my first post, but I figured too much information is better than too little.

 

[turbobooster196]

you whant to try 4.8ghz with a voltage of 1.52v?????? dont do it on a h80I just do,nt do it at all.
And if you get crasches running aida cpu cache then give the cache a litttle more voltage.
but on that cooler i would not even dare to give it that much voltage, regardless what intel says.
try stay below 1.4v is better for youre cpu.

 

[Itsmikerofl]

you whant to try 4.8ghz with a voltage of 1.52v?????? dont do it on a h80I just do,nt do it at all.
And if you get crasches running aida cpu cache then give the cache a litttle more voltage.
but on that cooler i would not even dare to give it that much voltage, regardless what intel says.
try stay below 1.4v is better for youre cpu.

Hey turbobooster!

I only tried running 4.8GHz at 1.52V for an incredibly short time, just to see if a higher voltage would effect the cache failure.

When I run at 4.7GHz, it only needs 1.42V, and temps stay under 70°C.

I have been under the impression that raising the voltage is bad because it also raises temps. If the H80i GT keeps the temps under control, does the voltage really do that much damage?

Thanks!

 

[turbobooster196]

temps is of course more inportment, if you can cool it, you can clock it, but the life span of your cpu will go down with high voltage, if its stable at 4.7ghz its fine, maybe go back to 4.6ghz you whont no the difference, but you will need less vcore, but the say 1.42v is safe

 

[Itsmikerofl]

temps is of course more inportment, if you can cool it, you can clock it, but the life span of your cpu will go down with high voltage, if its stable at 4.7ghz its fine, maybe go back to 4.6ghz you whont no the difference, but you will need less vcore, but the say 1.42v is safe

Okay. That makes sense.

The issue while on 4.8GHz is that I can boot fine starting at 1.440V, and in order to pass a CPU+FPU stress test, 1.456V is needed.

The CPU+FPU stress can run for hours without reaching 80°C, but as soon as I start a cache stress test, it returns a hardware error.

The only way to fix this is by increasing voltage? FCLK, VCCIO, SA, etc. won't help?

 

[EarthDog]

Welcome!!!! and whoa...........

Out of the box, the 6600K was running at 4.4GHz at stock voltage. That's pretty darn impressive for a 3.5GHz i5 if you ask me.

Define "stock". If you left vCore on auto, it AUTOmatically will raise it. your stock vCore you can get from setting bios defaults, booting to windows and putting a load on the CPU. Whatever that value is, will be your stock voltage. Typically you are around 1.15V (give or take). Are you telling me that at 4.7 GHz you are around that voltage?!!!!

This is when I decided to read more into OC'ing, and found out that lowering the Multiplier while upping the BCLK created more stable conditions. It's worth mentioning that I'm not too sure what a good Multi:BCLK ratio would be for my chip.

Where did you read that? Taking two things and raising them out of spec is more stable than one? I don't follow that logic, but admittedly, there is more to it than just that. I'd like to see that link......

I have discovered that the option for Load Line Calibration can also effect stability of an OC in positive ways, primarily to counteract what's referred to as "Vdroop". Is it suggested to raise the LLC level to gain more stability?

LLC prevents vdroop, correct. But if you are stable at 1.4V it doesn't matter if you use LLC to compensate for vDroop or raise vCore (which leaves you higher idle voltage - not a huge deal though). So, it really doesn't matter as it is a '6 of one, half dozen of the other' situation.

I would:

1. Leave cache alone (set it manually at its stock speed) as it really doesn't yield any gains
2. Stay at 1.42V and below if you want to keep that chip for its useful life. 1.5V will degrade that chip.

 

[PolRoger]

VCCIO and VCCSA help provide voltage to the IMC of chip and are used to help stabilize memory clocks. With Skylake the cache voltage is tied/locked to vcore and not all chips can run 1/1 with cpu multi especially at higher clocks.

You could try dropping the cache multi by 200/300Mhz to see if it can help with stability at higher clocks as well as perhaps lowering overall voltage requirements. I have a Haswell running at 45x multi/42x cache and a Skylake running at 47x multi/45x cache. Often it is recommended to test a chip's overclock margin first with low/default cache (~40x) and then later raise cache to see how it impacts overall stability.

 

[Itsmikerofl]

Welcome!!!! and whoa...........


Define "stock". If you left vCore on auto, it AUTOmatically will raise it. your stock vCore you can get from setting bios defaults, booting to windows and putting a load on the CPU. Whatever that value is, will be your stock voltage. Typically you are around 1.15V (give or take). Are you telling me that at 4.7 GHz you are around that voltage?!!!!

Hello there EarthDog!

When I say "stock", I mean at motherboard/BIOS defaults. 1.2V was the default voltage, and the 6600K clock being set to "auto" caused the rise to 4.4GHz. So yes, my 6600K is completely stable at 4.4GHz with 1.2V.

Where did you read that? Taking two things and raising them out of spec is more stable than one? I don't follow that logic, but admittedly, there is more to it than just that. I'd like to see that link......

The premise of what I said, is instead of having the BCLK at the stock speed of 100, and the stock multiplier where it was, I *lowered* the multi; while *raising* the BCLK. This kept the ratio the same.

LLC prevents vdroop, correct. But if you are stable at 1.4V it doesn't matter if you use LLC to compensate for vDroop or raise vCore (which leaves you higher idle voltage - not a huge deal though). So, it really doesn't matter as it is a '6 of one, half dozen of the other' situation.

Will it stress the CPU more as I raise LLC levels? Or does it alleviate stress? From my understanding, it makes the core voltage more consistent under load and not under load.

I would:

1. Leave cache alone (set it manually at its stock speed) as it really doesn't yield any gains
2. Stay at 1.42V and below if you want to keep that chip for its useful life. 1.5V will degrade that chip.

I've heard that adjusting the CPU cache does not yield performance gain. Does it not effect stability?

VCCIO and VCCSA help provide voltage to the IMC of chip and are used to help stabilize memory clocks.

Hi PolRoger! Thanks for helping out.

So, what I gather from this, is that VCCIO and VCCSA are for RAM stability, not affecting my CPU cache instability.

With Skylake the cache voltage is tied/locked to vcore and not all chips can run 1/1 with cpu multi especially at higher clocks.

I see. That's the wall in trying to push.

You could try dropping the cache multi by 200/300Mhz to see if it can help with stability at higher clocks as well as perhaps lowering overall voltage requirements. I have a Haswell running at 45x multi/42x cache and a Skylake running at 47x multi/45x cache. Often it is recommended to test a chip's overclock margin first with low/default cache (~40x) and then later raise cache to see how it impacts overall stability.

As I adjust the BCLK higher, and the CPU multiplier lower; the CPU Cache is also effected. It is usually automatically at a setting that is 300-500MHz below the CPU clock.

That being said, should I lower it an additional 200/300MHz as you suggested?

 

[turbobooster196]

Welcome!!!! and whoa...........


Define "stock". If you left vCore on auto, it AUTOmatically will raise it. your stock vCore you can get from setting bios defaults, booting to windows and putting a load on the CPU. Whatever that value is, will be your stock voltage. Typically you are around 1.15V (give or take). Are you telling me that at 4.7 GHz you are around that voltage?!!!!

Where did you read that? Taking two things and raising them out of spec is more stable than one? I don't follow that logic, but admittedly, there is more to it than just that. I'd like to see that link......
LLC prevents vdroop, correct. But if you are stable at 1.4V it doesn't matter if you use LLC to compensate for vDroop or raise vCore (which leaves you higher idle voltage - not a huge deal though). So, it really doesn't matter as it is a '6 of one, half dozen of the other' situation.

I would:

1. Leave cache alone (set it manually at its stock speed) as it really doesn't yield any gains
2. Stay at 1.42V and below if you want to keep that chip for its useful life. 1.5V will degrade that chip.

not quit true earthdog, depends on what bord you have, mine if i leave the voltage on auto and set everithing to stock and put load on the cpu i hafe a vcore of 1.3v
The default voltage is not programmed into the bios.
the Vid of my cpu is 1.155v

 

[EarthDog]

That type of reading is exactly what we saw upon release when things were not reading correctly. You may want to validate that reading with another piece of software (or better yet, a multimeter). 1.3V would be the highest TRUE VID I have ever seen on a Skylake CPU.

1.15V sounds right and is exactly what I mentioned.

 

[PolRoger]

6600K running on 4.8GHz is interesting; AIDA64 tests stressing the CPU, FPU (also one more object I'm not too informed about), and RAM will run stable for hours without a crash or a Vdroop. It gets confusing when I use AIDA64 to also test CPU Cache; which results in a "Hardware Error" after no more of a minute of testing. This happens whether I test only CPU Cache, or test it along with the other, (working), tests. I have attempted to raise the VCore all the way up to the Intel specified degradation point of 1.52V, with no change.

Is this because my CPU Cache is too high/low? If so, how should I go about adjusting it?

**TL;DR**

i5-6600K stable at 4.7GHz, 1.42V. AIDA64 returns Hardware Error while stressing CPU Cache @ 4.8GHz, any voltage up to 1.52V. Tips?

I didn't understand... From your post I thought you were actually trying to stress the cache at 48x along with a 48x multi...

I've heard that adjusting the CPU cache does not yield performance gain. Does it not effect stability?


I see. That's the wall in trying to push.


As I adjust the BCLK higher, and the CPU multiplier lower; the CPU Cache is also effected. It is usually automatically at a setting that is 300-500MHz below the CPU clock.

That being said, should I lower it an additional 200/300MHz as you suggested?

I would initially take the cache speed factor out of the equation for preliminary testing and use ~39x/40x to test your what your chip requires @ ~4.5/4.6/4.7GHz etc. and then test again to see if a given vcore requirement for a particular overclock will also provide enough cache voltage to run @1/1? Your chip may be able to run at 45/45 with ~1.25v or even at 46/46 with ~1.30v but perhaps not at 47/47 with 1.35v (maybe only with 46/45 cache??) or for that matter even higher at 48/48 with 1.4v(+).

 

[turbobooster196]

Earthdog what i mend was, if i leave everithing on stock, and put the cpu on load the cpu vcore goes to 1.3v what you see on the top of the picture.
But the voltage-VID is 1.115





But i have also a small quastion for you, the temp that you see in IXU that 55, is that good, ore can i go higher, if i can what would be the max that would be oke to show there

 

[turbobooster196]

And are all my temps look oke to you earthdog?

 

[Itsmikerofl]

I didn't understand... From your post I thought you were actually trying to stress the cache at 48x along with a 48x multi...




I would initially take the cache speed factor out of the equation for preliminary testing and use ~39x/40x to test your what your chip requires @ ~4.5/4.6/4.7GHz etc. and then test again to see if a given vcore requirement for a particular overclock will also provide enough cache voltage to run @1/1? Your chip may be able to run at 45/45 with ~1.25v or even at 46/46 with ~1.30v but perhaps not at 47/47 with 1.35v (maybe only with 46/45 cache??) or for that matter even higher at 48/48 with 1.4v(+).

My apologies for the late reply, as well as the misleading information.

You're correct; I am not trying to keep a 1:1 cache clock ratio.

I upgraded my motherboard in the meantime, from the ASUS Z170-A to the ASUS Z170-Premium, and I'm testing my overclock now. If the same problem happens, I'll try this advice.

Thanks in advance.

 

[wingman99]

Post back if the ASUS Z170-Premium helps?

 

[turbobooster196]

good luck, let see the results

 

[OptyTrooper]

Anything beyond 4.7-4.8 GHz is about the limit for Skylake on air @ least with the non-K chips.

 

[Itsmikerofl]

Hey all! So sorry it's taken months to update this old thread... But I've made some progress.

About a week after I last posted, I decided to use some of the tax refund I got to put towards my first custom water loop.

I integrated my CPU, 980 Ti, and Intel 750 SSD into the loop, using one 60mm thick 120mm radiator, as well as a 45mm thick 280mm radiator. This makes effectively 400mm of area for cooling.

Things got complicated, (due to the water block install), and I ended up upgrading my CPU to the i7-6700K. On air, the highest clock I could push was 4.7GHz at 1.356V. Alright compared to my i5-6600K. Temps were getting up to 75°C under load.

When I integrated it in my water loop, the temperatures at 4.7GHz/1.356V dropped immediately to 61°C under load.

I also sold my GTX 980 Ti and replaced it with a GTX 1080. This should help temps while gaming.

Now, to follow-up:

Post back if the ASUS Z170-Premium helps?

The Z170-Premium was able to give me the same clock as the Z170-A, only at 0.081V less. Good, but not good enough.

This is when I decided to try a custom loop.

good luck, let see the results

Bear with me... We're almost there!

Anything beyond 4.7-4.8 GHz is about the limit for Skylake on air @ least with the non-K chips.

This is what I've found. On a good air cooler, 4.7GHz is the limit for Skylake, and temps hit the ceiling.

On water, 4.8/4.9GHz is bootable, but very difficult to stabilize.

Now... The moment you've all been waiting for...

With my 400mm effective liquid cooling loop, I'm able to boot the system at 4.8GHz only using 1.42V. Temps do not go above 76°C while stressing with AIDA64.

When I try 4.9GHz, I can boot, but the voltage needs to be higher in order to achieve stability.

The stress test would go for anywhere from 10 minutes to an hour, at around 75°C, and it would hit a random point that temps would go above 80°C and crash.

So there you have it! Let's hope others can learn from this. I've really learned a lot over this, and I'm really glad I did the loop.

Anyone with any questions, feel free to post here!

 

[Kenrou]

On a good air cooler, 4.7GHz is the limit for Skylake, and temps hit the ceiling.

On water, 4.8/4.9GHz is bootable, but very difficult to stabilize.

mmmmmmmmmmmmmmm nop sorry, been running 4.7ghz/4.7 1.34v ~75c - 4.8ghz/4.6 1.4v ~80c for months now on a NH-D15, managed to stabilize 4.9ghz/4.7 at 1.43v ~90c.

Really depends on how lucky you are with the chip, at least for 4.7ghz a 212 EVO would be plenty for mine in particular


EDIT : shouldn't 6600/6600k run much cooler for lack of HT ? i turned mine off for experimenting and ended up with 4.9ghz/4.7 1.4v ~75c with my fans on BIOS silent profile

 

[wingman99]

Hey all! So sorry it's taken months to update this old thread... But I've made some progress.

About a week after I last posted, I decided to use some of the tax refund I got to put towards my first custom water loop.

I integrated my CPU, 980 Ti, and Intel 750 SSD into the loop, using one 60mm thick 120mm radiator, as well as a 45mm thick 280mm radiator. This makes effectively 400mm of area for cooling.

Things got complicated, (due to the water block install), and I ended up upgrading my CPU to the i7-6700K. On air, the highest clock I could push was 4.7GHz at 1.356V. Alright compared to my i5-6600K. Temps were getting up to 75°C under load.

When I integrated it in my water loop, the temperatures at 4.7GHz/1.356V dropped immediately to 61°C under load.

I also sold my GTX 980 Ti and replaced it with a GTX 1080. This should help temps while gaming.

Now, to follow-up:


The Z170-Premium was able to give me the same clock as the Z170-A, only at 0.081V less. Good, but not good enough.

This is when I decided to try a custom loop.


Bear with me... We're almost there!


This is what I've found. On a good air cooler, 4.7GHz is the limit for Skylake, and temps hit the ceiling.

On water, 4.8/4.9GHz is bootable, but very difficult to stabilize.

Now... The moment you've all been waiting for...

With my 400mm effective liquid cooling loop, I'm able to boot the system at 4.8GHz only using 1.42V. Temps do not go above 76°C while stressing with AIDA64.

When I try 4.9GHz, I can boot, but the voltage needs to be higher in order to achieve stability.

The stress test would go for anywhere from 10 minutes to an hour, at around 75°C, and it would hit a random point that temps would go above 80°C and crash.

So there you have it! Let's hope others can learn from this. I've really learned a lot over this, and I'm really glad I did the loop.

Anyone with any questions, feel free to post here!

How much cooler was your temps when you switched to Z170-Premium running with a lower voltage at the same clocks also the same cooler.

mmmmmmmmmmmmmmm nop sorry, been running 4.7ghz/4.7 1.34v ~75c - 4.8ghz/4.6 1.4v ~80c for months now on a NH-D15, managed to stabilize 4.9ghz/4.7 at 1.43v ~90c.

Really depends on how lucky you are with the chip, at least for 4.7ghz a 212 EVO would be plenty for mine in particular


EDIT : shouldn't 6600/6600k run much cooler for lack of HT ? i turned mine off for experimenting and ended up with 4.9ghz/4.7 1.4v ~75c with my fans on BIOS silent profile

Yes it does help with heat to turn off HT, less calculations per second. My Hyper 212 does just fine 84c for Prime95 at 4.5GHz also