• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Overclocking Skylake 6700k

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

serge89

New Member
Joined
Aug 17, 2015
Location
Sydney
Hi everyone,

I am new to the forum and never was really involved in to posting anything myself before, so please sorry in advance in case I am doing anything wrong.

Back to the topic, I just built my new system jumping from x58 Bloomfield i7 960 on to z170 Skylake 6700k. My first overclocking experience with Bloomfield was relatively smooth other then I got a mediocre chip. So I was able to push it to 4.2 on water with 1.325 in bios without V-droop with temps around 60s for the chip and 70`s for cores under max load.

Now, considering Bloomfield is soo far behind in terms of cpu architecture, there is a bunch of new features and other things to consider while overclocking. Things like Dynamic V-core for instance, different LLC states, power-saving features etc etc. I know these features (at least some of them) have been around for a while already and I know these things quite well, in terms of how these features work and what kind of effect they have on a system, however this makes me a bit confused. I am not sure where to start my overclock considering the difference in platforms.

Therefore I am here and would like to get some input from you good people, who are on a newer gen platforms and know what they are doing.

Basically my main points / questions are:
1. How different the overclocking process is comparing to, say, Bloomfield vs Haswell (or is it more relevant to compare Bloomfield vs Skylake considering it has unlocked blk this time around)?
2. What are the new things to consider (again comparing different platforms)
2. Overclocking blk vs higher multiplier
3. Anything important I need to know re Uncore Frequency (for instance with Bloomfield, UF was twice the ram multiplier)
4. Should I use LCC / Vdroop-off? If yes, what states should I be considering the most.
5. Keep Speedstep enabled and set up dynamic vcore, or disabling it all together
6. Cpu turbo (on vs off)?

I believe I have covered most of my concerns here and I very hope everything will make sense (apologies if it does not, as it has been a while since my last overclock and I might have forgotten some terms etc.)

Any help is very appreciated. Thanks in advance. :rolleyes:

Also please note, I did not have a chance to have a look in to newer bios (in my case its Asus z170), so I have no idea what to expect. I was using Evga FTW3 on my x58 system.

(system specs: z170 Asus Maximus 8 Hero, i7 6700k, Thermaltake Water 3.0 Ultimate - cpu cooling, Vengeance LPX 16GB (2x4GB)x2 DDR4 3200MHz C16, Nvidia GTX Titan X, Antec HCP-1000 Platinum PSU)
 
Last edited:
Welcome to the OCF :)

Overclocking on new systems is much easier than on these older ( regardless what people say ).
- On Skylake all you have to do is to set CPU ratio ( best is to set it manually disabling optimized ASUS core control ) and adjust CPU voltage.
- Bclk is overclocking good and you can push it really high. You don't have to worry about pcie bus as it's always at 100MHz. However bclk is not helping in general performance so you don't have to waste time on that.
- Cache voltage and CPU voltage are tied so raising one you are raising the other one too. 1.4V seem reasonable max for 24/7 work and should let you OC CPU to about 4.6-4.8GHz depends from CPU itself.
- Cache clock is helping in memory performance but is not really important. Still that 1-5% performance gain and you have to raise cache voltage while overclocking CPU clock anyway so why not to raise cache clock too.
- you don't have to worry about LLC but try higher values if you want to keep voltage more stable.
- enable turbo, disable C stages, disable eist - this is optimal for overclocking but settings with enabled eist/c may also work
- I don't know if your memory is working at XMP as most kits have problems with that, If it's not then set the same main timings but check Maximus Tweak 1 or 2 option ( on the top of the memory timing window ). It usually helps with booting.

ASUS M8H is setting most voltages and other settings automatically so you don't have to worry about most of them while overclocking below 5GHz. You won't pass 5GHz on air/water anyway. What is important is CPU/cache voltage, memory voltage and memory clock. Changing only these options will let you set ~4.7GHz.
 
Thank you for your reply.

Looks quite straight forward so far.

Hopefully I will be able to hit 4.8 mark without too much of a voltage bump. Do you have any ideas re preferable max load temps on both cpu and cores? I was hearing something within 90c range should be safe, however it seams a bit too hot if you ask me.
 
I can't see clear info on Intel site but probably the same as in other series anything below 90*C is still good. Of course better if lower when you are overclocking.
 
I can't see clear info on Intel site but probably the same as in other series anything below 90*C is still good. Of course better if lower when you are overclocking.

Sorry for such a late reply. Took me a while to get my system up and running stable.

So for those who is interested, overclocking was a pain. I was running in to tons of issues in order to get my system stable. However all the issues ended up being Software related, not hardware, especially if you are running Windows 10. If any of you experiencing bsod`s while stress testing, go back to a very beginning with stock voltages and make sure to clear your cmos if you had not done it already.

Windows 10 incompatible software that was causing bsod (even on stock speeds/voltages). Some versions of CPU-Z(HWmonitor worked for me), Asus AI suite, MSI afterburner (use Asus Tweak 2), Daemon tools, possibly optional Intel Ethernet controller driver. Again, if you cant get your system stable, try removing any of these and re-test your system. Also fresh Windows install would be good.

So back to overclocking. After isolating all the soft issues, I ended up with some interesting results. 6700k is running at 4.7 1.35 adaptive voltage (llc level 4), cache oc to 4.3 with ram sitting at XMP 3000. Voltages under full load are sitting withing 1.328 - 1.344 range. C-states are set to auto along with Speedstep and Turbo enabled. Re temps, 8 hour Realbench stress test was giving me low 50c on the cpu and low 60 on cores with Realbench benchmark socre of almost 99000. Idle cpu temps are 26 on the cpu and low 20s on cores.

I hope this will be helpful.
 
Last edited:
Since you have Maximus VIII Hero and I don't want to start new thread then I just wanted to add that you can play with LEDs. ROG logo on this board can change colors and there are couple of modes. You can set red, green, blue, purple, white or anything else or just run rainbow mode. There is also mode to switch between green, yellow and red depends from CPU temperature.
Once you install soft from ASUS site ( in download/support section ) to control it and configure LED how you like then you don't have to run soft anymore. It will work even when you turn off PC ( but keep power cord plugged of course ).
Since you can change LEDs then idea to make this board less red than earlier ROG series seem not that bad. Anything greenish or blueish looks nice with grey/black parts of the mobo.
Nothing really important when you decide to buy motherboard but it's nice that ASUS added that option. I wish it could be done for audio card LEDs too as they're only red but you can turn them off.
So playing with LEDs was all I was doing on Skylake platform in last days :p

Btw. there is new BIOS 0802 which works fine so far. I've noticed that ASUS released 3 more BIOS versions after 04 but 2 of them were quickly removed. On the website only 04xx 06xx and 08xx left. I guess that other versions had some instability issues.
 
Sorry for such a late reply. Took me a while to get my system up and running stable.

So for those who is interested, overclocking was a pain. I was running in to tons of issues in order to get my system stable. However all the issues ended up being Software related, not hardware, especially if you are running Windows 10. If any of you experiencing bsod`s while stress testing, go back to a very beginning with stock voltages and make sure to clear your cmos if you had not done it already.

Windows 10 incompatible software that was causing bsod (even on stock speeds/voltages). Some versions of CPU-Z(HWmonitor worked for me), Asus AI suite, MSI afterburner (use Asus Tweak 2), Daemon tools, possibly optional Intel Ethernet controller driver. Again, if you cant get your system stable, try removing any of these and re-test your system. Also fresh Windows install would be good.

So back to overclocking. After isolating all the soft issues, I ended up with some interesting results. 6700k is running at 4.7 1.35 adaptive voltage (llc level 4), cache oc to 4.3 with ram sitting at XMP 3000. Voltages under full load are sitting withing 1.328 - 1.344 range. C-states are set to auto along with Speedstep and Turbo enabled. Re temps, 8 hour Realbench stress test was giving me low 50c on the cpu and low 60 on cores with Realbench benchmark socre of almost 99000. Idle cpu temps are 26 on the cpu and low 20s on cores.

I hope this will be helpful.

Hi there! I'm also new here, aaaand I have kinda similar issue. The thing is, that I've the same CPU, which easily ( I mean with just 1-click-kill) went up to 4.7 GHZ ( 100 * 47 ). What bothers me here is the fact that while at idle it remains at about 24 degrees, but at AIDA64 stress test it shows an average of 70 with max at sky-high 98 degrees(thought it only happened once, but max temps on all cores reached 90)! I also have a thermaltake water 3.0 ultimate which should be an overkill for this CPU, so I'm kinda surprised. Maybe there is a chance I forgot to modify something, voltages or something else ... ? This seems to be really high..


Thank you for your time and help!
 
Last edited:
Double check the waterblock is seated well, and what voltages is the board picking for you? It might be on the generous side so lowering it a bit could help.
 
Hi there! I'm also new here, aaaand I have kinda the similar issue. The thing is, that I've the same CPU, which easily ( I mean with just 1-click-kill) wen't up to 4.7 GHZ ( 100 * 47 ). What bothers me is the fact that while idle it remains at about 24 degrees, but at AIDA64 stress test is show average of 70 with max at sky-high 98 degrees(thought it only happened once, but max temps on all cores reached 90)! I also have a thermaltake water 3.0 ultimate which should be an overkill for this CPU, so I'm kinda surprised. Maybe there is a chance I forgot to modify something, voltages or something else ... ? This seems to be really high..


Thank you for your time and help!
When you say 1 click are you using some sort of overclocking software either in widows or in BIOS? If you are a lot of times these will give the Cpu way more voltage then needed.
 
Seems about right for that stress test and those clockspeeds...

I would bet, as mandrake already said, that you are auto overclocking and there is likely too much voltage for the clockspeed.
 
Seems about right for that stress test and those clockspeeds...

I would bet, as mandrake already said, that you are auto overclocking and there is likely too much voltage for the clockspeed.

When you say 1 click are you using some sort of overclocking software either in widows or in BIOS? If you are a lot of times these will give the Cpu way more voltage then needed.

Double check the waterblock is seated well, and what voltages is the board picking for you? It might be on the generous side so lowering it a bit could help.

Thank you guys for quick responses!

So, originally I went with pre-installed thermal compound, replacing it with an aftermarket one and running a non-long stress-test (just to let it warm-up a little bit for 30 minutes ) gave me next results :

Screenshot_1.png


So, while the avg temps did lower ( thought I am not sure if it will maintain this way during longer tests, unfortunately I was unable to check this yet ), the max temps seem to be still pretty high.

I'm not a professional, so should I lower the voltages ?

I did use the tool that comes with Maxumus VIII Extreme in UEFI, to modify ratio, so yes, the voltages are auto.
I would be really thankfull if you could tell me what would be a better option here :)

Thank you all again and have a good time! :)
 
Last edited:
Xedin can you create a signature like we have that shows all the components in the pc, it's helpful to know exactly what is in the system. Using an overclocking tool isn't recommended, like I said previously a lot of times it will usually give the Cpu more voltage then needed. What I suggest is doing it manually, try setting the Multiplier to 4.5, the Cpu voltage manually to 1.35 V and see if you can boot into windows. If you can stress test with Aida 64 for at least 2 hours, keeping the temps below 90c. If you can't get into window or cannot pass Aida 64 raise the Cpu V .02 and try again. You want to keep the voltage below 1.42 for a daily Oc and temps below 90c.
 
Xedin can you create a signature like we have that shows all the components in the pc, it's helpful to know exactly what is in the system. Using an overclocking tool isn't recommended, like I said previously a lot of times it will usually give the Cpu more voltage then needed. What I suggest is doing it manually, try setting the Multiplier to 4.5, the Cpu voltage manually to 1.35 V and see if you can boot into windows. If you can stress test with Aida 64 for at least 2 hours, keeping the temps below 90c. If you can't get into window or cannot pass Aida 64 raise the Cpu V .02 and try again. You want to keep the voltage below 1.42 for a daily Oc and temps below 90c.

Thank you again for the quick response!
I really wanted to remain on 4.7 ( but I guess I'll have to fall to 4.5 if there are no other options left ), and did some tests & followed your advises :)

So, I went to UEFI and set the Core CPU voltage to 1.35 -> booted, but failed at stability test, modified to 1.37 - everything is fine... but it's not :)
The thing is, that while it did become cooler ( though still had surprisingly big max temps ) and is still stable, and somehow just ... didn't care of the voltage change. In fact, it became even bigger - as you can see, with manual voltage set to 1.370, max voltage somehow reached 1.402! When with auto, it never exceeded 1.400.
Screenshot_2.png

So, here I've 2 questions :
1) With auto mode, I've often seen voltage to drop to something like 0.79 at standby, but now - will it always remain ~1.37 ?
2) How is this even possible for voltage to bypass manually set limit of 1.37 ?
I don't actually know how to screenshot UEFI ( though there should be a way, gonna look for it later, sorry for being lame :( ) , so I went with AL Suite, and here is what I have there :
Screenshot_3.png


But at this same moment, CPUID shows :
(I've actually included AIDA's one and ROG's one, since they somehow manage to show different results ) :

Screenshot_4.png

And this is Idle... So I'm reaally confused now.

I am sorry for probably not understanding something obvious, and wasting your time .. :(


Anyway.. thank you for your help and time! :)


UPD :

Seems like ROG's one just shows something strange. Switched to AUTO to check , and saw this :
Screenshot_7.png


And here is screenshot of UEFI :

Screenshot_6.png

UPD2 :
Did some additional stressing, with auto mode, just for the interest. Here are the results, voltage seems to be actual.

Screenshot_8.png

I can't get it. In manual mode -> voltages are higher, yet temp is lower ( but still > 80 ) ... This doesn't make sense to me to be fair :confused:
 
Last edited:
Xedin, you're not wasting our time. We are here because we enjoy helping people with this hobby.

To answer your questions the reason why the voltage is going higher then the set voltage is because of the Load Line Calibration. All processors have what is called voltage droop when under load. The cpu LLC was developed to help control this, by raising the voltage. The goal is to have the Cpu LLC set where it keeps the voltage from drooping under load but not exceeding the set voltage by a large margin. You likely have the Cpu LLC set to auto and I don't feel you need to change it. A .03 over shoot above the set voltage isn't something to really be concerned about.

As far as why the temps are lower. If your ambient temps remained the same and you stress tested with the same program for the same period of time. It's likely that with the voltage is on auto, during the stress test it is staying closer to the 1.4 v for the duration of the test. With it manually set to 1.37 v it's probably staying closer to the set voltage and only going up to 1.402 for short periods of time, during heavy loads.

Where I feel you should go from here, lets start by forgetting what we want out of the chip. It's a silicon lottery, you're only going to get what the setup will give. The first temperature screen shot in post #13 is at 4.5 with the voltage set manually at 1.37 correct? If so lets work from there, you still have about 10c worth of headroom in the temperature dept. Run aida 64 for at least 2+ hours watching temps again 90c max is recommended, if it spikes above that for a brief period don't sweat it. If you can pass 2+ hours at those settings raise the multiplier to 4.6 and test again. If you cannot pass 2 hours raise the Cpu V .02 again and re test.
 
Xedin, you're not wasting our time. We are here because we enjoy helping people with this hobby.

To answer your questions the reason why the voltage is going higher then the set voltage is because of the Load Line Calibration. All processors have what is called voltage droop when under load. The cpu LLC was developed to help control this, by raising the voltage. The goal is to have the Cpu LLC set where it keeps the voltage from drooping under load but not exceeding the set voltage by a large margin. You likely have the Cpu LLC set to auto and I don't feel you need to change it. A .03 over shoot above the set voltage isn't something to really be concerned about.

As far as why the temps are lower. If your ambient temps remained the same and you stress tested with the same program for the same period of time. It's likely that with the voltage is on auto, during the stress test it is staying closer to the 1.4 v for the duration of the test. With it manually set to 1.37 v it's probably staying closer to the set voltage and only going up to 1.402 for short periods of time, during heavy loads.

Where I feel you should go from here, lets start by forgetting what we want out of the chip. It's a silicon lottery, you're only going to get what the setup will give. The first temperature screen shot in post #13 is at 4.5 with the voltage set manually at 1.37 correct? If so lets work from there, you still have about 10c worth of headroom in the temperature dept. Run aida 64 for at least 2+ hours watching temps again 90c max is recommended, if it spikes above that for a brief period don't sweat it. If you can pass 2+ hours at those settings raise the multiplier to 4.6 and test again. If you cannot pass 2 hours raise the Cpu V .02 again and re test.

Hi again. Sorry for long response, had a really hard work day.

Thank you again for your help.

The first temperature screenshot is in fact still at 4.7 already, I can check results with 4.5 tonight, but all the data in the last post corresponds to 4.7 with different voltage control.

I'll run stress tests tonight, in both auto auto and manual modes, so we could be sure about average/spike temps and voltages :)
 
Back