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

i5 6600k OC - Need some help to stabilize

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

karnalta

New Member
Joined
Jan 20, 2009
Hi,

Yesterday I started to play with my new computer and I did a bit of oc. I haven't done this since long time so I have some interrogations ;)

My setup :

- Intel Core i5 6600k
- ASUS Z170 Pro Gaming
- 2x8Gb Corsair LPX 3200Mhz
- Custom loop WC (only my CPU)

I was first concentrate on the CPU oc, so I have set my memory to 2133Mhz, BCLK to 100, FCLK to 800Mhz. I have set LLC to level 5, vcore to 1.35v and I just played with the multiplier.

I have been able to boot Windows and run some superPi up to 5Ghz, I wasn't expecting that much with so few tuning ! But, to try to find a stable OC, I lowered to 4.7Ghz and I am now trying to get it stable at the minimal vCore.

Here comes my questions :

- If I try to lower the vCore, the core #0 of my CPU is always failing after 1-2 min under Prime95 while other cores can run for a hour (I then stopped the test), do my CPU has a weakness or is it normal that one core fail first ?

- Does Prime95 FFT Test is really representative of reality ? Cause at 1.35v the Blend test get my CPU to around 65°C, that's a good temperature I presume for the life of the CPU. But FFT hammer it so much harder, I get up to 85°C.. That's too much, but I don't really think that any real life application is going to hit that strong the CPU. Even VRay rendering is lighter.

- There are so much parameters in the BIOS that I didn't even touch, they are probably other stuff to ajust to increase my OC stability and being able to reduce a little bit the vCore ? What are the others important parameter I should test ? I feel like I have only touch the very essential setting at the moment.

- How much should I run Prime95 Blend test for a real life stability ? My computer will mainly do office stuff, a bit of CGI and have some gaming session of 1-2 hours, it's not going to be part of a rendering farm for Lucas Film :p

- Does the FCLK frequency OC help stability or so ? Cause performance gain seem so small that I am thinking why I should do it.

Thank for your help and sorry for my english mistakes ;)
 
It is normal for one core to fail first on Prime95. All CPUs have cores that are weaker than others.

You should use a different program to test system stability, as Prime95 is considered to be too extreme. ASUS RealBench is a good one, as is AIDA64. They will both stress your entire system.

FCLK should be set to 1GHz and left there unless you are going to try BCLK overclocking. 1GHz is the standard speed, and it affects your PCIe performance when set lower.

Unfortunately, vCore is vCore. There is no magic setting to have it stable at less vCore.
 
I'm not sure if on MSI level 5 LLC is the highest or the lowest. You simply have to check if under load, voltage of your CPU is going down or not. It should remain without bigger changes to keep CPU stable. If it's not then set higher CPU voltage.
On my MSI there is no LLC and I have to set 1.25V only to get 4.2GHz stable on 6600K. This CPU requires about 1.18V for 4.2GHz on ASUS motherboard.
 
I'm not sure if on MSI level 5 LLC is the highest or the lowest. You simply have to check if under load, voltage of your CPU is going down or not. It should remain without bigger changes to keep CPU stable. If it's not then set higher CPU voltage.
On my MSI there is no LLC and I have to set 1.25V only to get 4.2GHz stable on 6600K. This CPU requires about 1.18V for 4.2GHz on ASUS motherboard.

Karnalta,

I'm not too surprised you can get to 5GHz water cooled with these Skylake chips but each chip will vary getting to their peak.

This Asrock board only goes to 3 LLC so that's where I set it. Vcore set to 1.2V and AIDA64 says ~ 1.150V under load
which is what I have seen in the past on most boards. There was a time I would try to find a direct motherboard
reading point to use a DMM to check Vcore but it's getting to be too risky anymore to do that and slip off with the probes. :)

I'm sure there is 4.2 GHz left on this CPU at 1.2Vcore with standard cooling. I don't bother overclocking the RAMs.
 
Thank you for the informations.

I am going to give a try to AIDA64 tonight, look like a nice program.

On my ASUS LLC max level is 7 (where it mean minimum vDroop), I could try to set it from 5 to 7 and lower slightly my vCore. Yesterday I lowered my OC to 4.6 Ghz because of my failing core #0, and I was stable under Prime95 blend test. Does AIDA64 can test show failing core ? Or is it a global stress test ?

And what about all others voltages like : PLL, VCCIO, SystemAgent, PCH. Are they useful to reach CPU oc stability ? I guess PCH voltage is useful when you try to OC the BCLK ?

It's a bit frustrating to know that all is stable under 4.8Ghz except that one core :(
 
Regardless which core is failing, it won't help you in anything and you have to stabilize all cores. AIDA64 won't tell you which core fails but you can run different tests like cache, memory, FPU separately so you can check what is causing issues.
Overclocking bclk on unlocked CPU is pointless so really other voltages are useless but you can of course play with various settings.


This is my setting since yesterday.

42.jpg
 
Hey guys, I am a new OC as well, and I thought I would pop in to this thread since it seems to be the most current. Just got my new i5 6600k last week and have been playing around with the OC settings. Achieved a stable 4.2 Ghz on a 1.25V setting. Last night I ran an 8+ hour stress test in Aida64 running at 4.3 Ghz with 1.29V. I have liquid cooling, so CPU temps barely hit 50C. On my tests the core frequency was rock solid at 4300 Mhz during the entire test, so it seems to be stable at these settings.

I have two questions though:

1) With it being stable at 1.29V @ 4.3 Ghz, should I bother trying to lower the Voltage (say 1.285 or 1.28) and trying for stability? Or do you guys think there is any harm that I leave it at 1.29V since it seems to hold perfectly stable?

2) Looking at my Hardware Monitor program, while under load the frequencies maintain their 4300 Mhz (@ 100%), however I did notice that the maximum frequency column on the far right had some of the cores showing much higher than 4300 (one core showed 4500, another one showed 4900). However none of the cores are operating at that frequency under normal 100% loads. Is this just a common thing where a core can have a quick spike to some large frequency? Do I need to be concerned about that, or does that sound normal?

Like I said, I am new to this, so the last thing I want to be doing is killing a brand new PC. Lol.
 
Ok, so I am nearly done with my OC, 4.6Ghz is stable with 1.35 vCore but I was concerned about my temp that can hit 80+ under hard test like Prime95 FFT. But under normal usage I never exceed 60-65 °C. I guess OC is intended to "use" the computer so I should not be too much worried about extreme test. Maybe I will watch my temp in summer to see if it's still acceptable.

Hey guys, I am a new OC as well, and I thought I would pop in to this thread since it seems to be the most current. Just got my new i5 6600k last week and have been playing around with the OC settings. Achieved a stable 4.2 Ghz on a 1.25V setting. Last night I ran an 8+ hour stress test in Aida64 running at 4.3 Ghz with 1.29V. I have liquid cooling, so CPU temps barely hit 50C. On my tests the core frequency was rock solid at 4300 Mhz during the entire test, so it seems to be stable at these settings.

I have two questions though:

1) With it being stable at 1.29V @ 4.3 Ghz, should I bother trying to lower the Voltage (say 1.285 or 1.28) and trying for stability? Or do you guys think there is any harm that I leave it at 1.29V since it seems to hold perfectly stable?

2) Looking at my Hardware Monitor program, while under load the frequencies maintain their 4300 Mhz (@ 100%), however I did notice that the maximum frequency column on the far right had some of the cores showing much higher than 4300 (one core showed 4500, another one showed 4900). However none of the cores are operating at that frequency under normal 100% loads. Is this just a common thing where a core can have a quick spike to some large frequency? Do I need to be concerned about that, or does that sound normal?

Like I said, I am new to this, so the last thing I want to be doing is killing a brand new PC. Lol.

1.) If your temps never exceed 60°C I don't see why you should try to use less vCore, moreover I think 1.29 is'nt bad for the CPU life. My old i7 860 has stay 5 years with an high vCore and is still alive :p But my WC loop never cleaned in 5 years was the real carnage... Before making my new computer, even a 3DMark bench was making my CPU hit 100°C and throttle .. But he is still alive :p

2.) I am using HWMonitor and I have'nt noticed that kind of frequency spike with my OC.. Maybe it's a simple reading bug from your monitioring software ?
 
If your board has load line calibration ( LLC ) adjustment then you should be able to set lower voltage for that clock. I mean I don't have any special chip and on ASUS I'm able to set 4.2GHz @~1.18V or something like 4.5GHz @~1.25V with max clock for longer tests ( not fully stable ) 4.7GHz ~1.36V. On MSI motherboard without LLC I have to set nearly 0.1V higher CPU voltage as during full load it's dropping. My 6700K is slightly worse but still doesn't need much more for similar clocks.
I would try lower voltage or simply higher clock at the same voltage.

If you disable all power saving options then clock will remain the same regardless what load there will be. If you set manually max ratio but keep power saving options enabled then clock will go up to that max ratio which you set and will drop or keep load balance if load will be lower. So if you set x45 then under load it should make 4500MHz and nothing more. If software shows more then it's a read error.

Hardware monitor is not the best to check voltages and clocks. Or maybe it shows almost everything but you have to know how to read it as sometimes values are in different rows/tables.
For only clock and CPU voltage read you can use CPU-Z. For other things you can probably find better soft or even application added to the motherboard as probably every manufacturer is adding something.

Max safe cpu voltage seems to be something near 1.35V but if you don't need any faster CPU then I see no point to set it so high.
 
yeah the current core frequencies fluctuate depending on what I am doing. If I am running a test through Aida 64, then the current frequency stays locked at 4300 the entire time showing at 100% load. I was just more or less curious about the third column in the hardware monitor. First is current, Second is Minimum and Third is Maximum. When I first load up HW Monitor it shows 4300 as the max frequency. Then I run a test, and I notice at some point, there must have been a quick spike up to 4500 or 4800 because that is what the new maximum values are. However, when looking at the Aida results, the core frequency shows a straight line locked at 4300 during the entire 8 hour test.

Again, just not sure if it's something I really need to worry about. Like it has been said, maybe it's just a bug in the reading.
 
If your PC can't run the Pime95 program I would not call that stable. Are we getting to the point that the desk top PC can and can't handle different programs, with the example of prime95.
 
yeah the current core frequencies fluctuate depending on what I am doing. If I am running a test through Aida 64, then the current frequency stays locked at 4300 the entire time showing at 100% load. I was just more or less curious about the third column in the hardware monitor. First is current, Second is Minimum and Third is Maximum. When I first load up HW Monitor it shows 4300 as the max frequency. Then I run a test, and I notice at some point, there must have been a quick spike up to 4500 or 4800 because that is what the new maximum values are. However, when looking at the Aida results, the core frequency shows a straight line locked at 4300 during the entire 8 hour test.

Again, just not sure if it's something I really need to worry about. Like it has been said, maybe it's just a bug in the reading.

Can you put a signature in your profile so we all know your hardware?
I just finished the 6600K build in my signature and I use AIDA64 and
Orthos (Prime 95 with a niice GUI over it). Orthos runs a few hours
just fine as does AIDA64 so I'm stable.

Unless there is a problem with your motherboard or software testing
I have found AIDA64 and my other HWINFO information programs to
have stable but slightly lower voltage readings than the BIOS and
without any spikes in speeds.

http://www.hwinfo.com/download.php

http://www.softpedia.com/get/System/Benchmarks/Orthos-Prime.shtml

If you're stable after running Prime95 for a few hours and the same
with AIDA64 I wouldn't be concerned with software spikes. You'll know
if there is a problem in real world use.
 
PC was stable for over 8 hours using Aida64. I actually swapped out the MOBO since the last post, so now I get a stable 4.3 Ghz at only 1.25V instead of 1.29V on the previous board. I was just wondering if software like HW Montior can have glitches that would cause the max frequency to read a number higher than what the CPU frequency was set to. Again, all numbers were stable and rock solid at 4300 during the entire test, and temps never got above 52 on any of the cores.

How do I get my signature to display? LOL
 
Back