• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Bad 6700k?

J

JayPhizzt

n00b
Joined
Jun 17, 2013
Messages
6
I'm trying to OC my 6700k, but I must have gotten a bad chip or something because I'm getting really poor results.
Most people seem to easily be able to get up to around 4,5-4,6GHz, but not me.
At the moment I'm running at 4,3GHz with the vcore at 1,325V and with LLC set to "high". This is about the most stable I seem to be able to get it.
I have managed to get it to 4,5GHz with the vcore at 1,375V, but then I get errors in Prime95 after just a short while. And if I set the vcore lower at 4,5GHz windows just crashes/freezes as soon as I start Prime95.

Have I just got a bad chip, or could it be my motherboard limiting my OC? Or could there potentially be any setting(s) on my motherboard limiting my OC? I have a Gigabyte GA-Z170N-Gaming 5.
 
You are most likely damaging the CPU with too high voltage and putting core temps to maximum. This will reduce your max overclock.
What you set Vcore to and what it actually runs at can be wildly different.
The highest setting of LLC will boost the load voltage substantially and can make the idle voltage too low, forcing you to increase the basic voltage settings to get stability back, making it run at an even higher voltage under load!!
I damaged my old 2500K by accidentally leaving LLC on max when it appeared as a new feature after a bios update.

Turn LLC off (set it to auto) and see what you can get. See if you can improve on it with a lesser LLC tweak once you have a baseline.
Hopefully you havent wrecked the CPU.

You dont even appear to be monitoring CPU core temps!
The guiding line for max voltage is the max core temperature.
I dont let mine get near 80C.
Very high temps and very high voltage are what will kill your CPU.

Note that the core voltage reported by general monitoring tools is not usually correct.
You may need to use the clocking tools that came with your motherboard to guarantee seeing the correct voltage.
Anything that says VID is "not" the real voltage. That is the the bios level voltage setting used when you leave the voltage on automatic, not what you are using when manually setting the voltage.
(note that VID is the voltage set, not what is applied. You again need to use a voltage monitoring tool that reads the correct voltage when relying on VID)

I'm not going to write you a guide on overclocking a 6700K, there are tons out there.
Read a few.
The above are basics and a few warnings.

ps
max voltage on air is 1.4V, max on water is 1.45V. This is to prevent damage.
That isnt what you set in the bios, that is the actual voltage in use as read using the motherboards own overclocking software.
 
Last edited:
IMO each cpu can be a bit different depending on mobo too. Above poster has a good example of LLC doing bad things. I've had the opposite where not putting enough LLC in I'd have to raise up my resting vcore so high and then it was still unstable at load. whereas putting in lowest stable vcore and applying LLC helped me gain stability. But I do think it's very dependent on the board, and how the cpu reacts to voltage.
 
I didnt explain LLC much, I prefer that the Op go read a few guides to pick up other essential info for his understanding.

I'll cover what I said here because I didnt make it clear what I meant.
The highest setting of LLC will boost the load voltage substantially and can make the idle voltage too low, forcing you to increase the basic voltage settings to get stability back, making it run at an even higher voltage under load!!
Click to expand...
When I said high LLC can make the idle voltage too low, that was in the context of the overall clocking experience.
LLC mostly just changes the load voltage to apply a compensation for vdroop under load. It can apply less or more than vdroop depending on the level used.
It can also affect idle voltage slightly at higher settings, usually by raising them a small amount.
This appears to oppose what I said so is worth going over.

You boost the voltage so that a higher overclock can remain stable.
This higher overclock might also need a higher idle voltage depending on how you apply your overclocking.
ie
if you use all the power and clock states then clock speed and voltage increase / drop with load. This allows idle to use low clocks and can make use of a lower idle voltage.
You can also set a rigid base voltage level and disable low power states so that the clock speed is always maximum, even at idle. This may need a voltage boost for idle as well as load if you raise the clock speed.
 
Nenu said:
You are most likely damaging the CPU with too high voltage and putting core temps to maximum. This will reduce your max overclock.
What you set Vcore to and what it actually runs at can be wildly different.
The highest setting of LLC will boost the load voltage substantially and can make the idle voltage too low, forcing you to increase the basic voltage settings to get stability back, making it run at an even higher voltage under load!!
I damaged my old 2500K by accidentally leaving LLC on max when it appeared as a new feature after a bios update.

Turn LLC off (set it to auto) and see what you can get. See if you can improve on it with a lesser LLC tweak once you have a baseline.
Hopefully you havent wrecked the CPU.

You dont even appear to be monitoring CPU core temps!
The guiding line for max voltage is the max core temperature.
I dont let mine get near 80C.
Very high temps and very high voltage are what will kill your CPU.

Note that the core voltage reported by general monitoring tools is not usually correct.
You may need to use the clocking tools that came with your motherboard to guarantee seeing the correct voltage.
Anything that says VID is "not" the real voltage. That is the the bios level voltage setting used when you leave the voltage on automatic, not what you are using when manually setting the voltage.
(note that VID is the voltage set, not what is applied. You again need to use a voltage monitoring tool that reads the correct voltage when relying on VID)

I'm not going to write you a guide on overclocking a 6700K, there are tons out there.
Read a few.
The above are basics and a few warnings.

ps
max voltage on air is 1.4V, max on water is 1.45V. This is to prevent damage.
That isnt what you set in the bios, that is the actual voltage in use as read using the motherboards own overclocking software.
Click to expand...

Nenu said:
I didnt explain LLC much, I prefer that the Op go read a few guides to pick up other essential info for his understanding.

I'll cover what I said here because I didnt make it clear what I meant.

When I said high LLC can make the idle voltage too low, that was in the context of the overall clocking experience.
LLC mostly just changes the load voltage to apply a compensation for vdroop under load. It can apply less or more than vdroop depending on the level used.
It can also affect idle voltage slightly at higher settings, usually by raising them a small amount.
This appears to oppose what I said so is worth going over.

You boost the voltage so that a higher overclock can remain stable.
This higher overclock might also need a higher idle voltage depending on how you apply your overclocking.
ie
if you use all the power and clock states then clock speed and voltage increase / drop with load. This allows idle to use low clocks and can make use of a lower idle voltage.
You can also set a rigid base voltage level and disable low power states so that the clock speed is always maximum, even at idle. This may need a voltage boost for idle as well as load if you raise the clock speed.
Click to expand...


The very reason I enabled LLC is because of vdroop. With LLC set to "auto" or "standard" windows crashed/froze as soon as I started Prime95. The vcore dropped to around 1,236V, so it's not suprising. And of course I'm monitoring temps, I'm using HWmonitor. At 4,5GHz and 1,375V it did run a bit hot(84C). Of course it only reached that temperature briefly and it was mostly at around 70C. I also try to keep it lower than 80C. During normal usage it never goes above 80C though.
 
Which version of Prime95?
There are versions which run instructions that you will never need (I think AVX instructions).
These bump power use and core temps up in a big way.

I have a 2015 version (v 28.5.1.0) that does this, I keep it just for kicks but dont use it.
Because I test my 6700K with a very old 2009 version (v 25.11.1.0) I havent tested the latest releases so cant confirm if this has been fixed.
I can only recommend you try and find the version I use until someone else can confirm the current status.
(or you could see if a new version helps)
 
Nenu said:
Which version of Prime95?
There are versions which run instructions that you will never need (I think AVX instructions).
These bump power use and core temps up in a big way.

I have a 2015 version (v 28.5.1.0) that does this, I keep it just for kicks but dont use it.
Because I test my 6700K with a very old 2009 version (v 25.11.1.0) I havent tested the latest releases so cant confirm if this has been fixed.
I can only recommend you try and find the version I use until someone else can confirm the current status.
(or you could see if a new version helps)
Click to expand...

Testing with the old version just means you aren't stressing the chip. Which pretty much defeats the purpose of a stress test once you decide "Well those AVX instructions work the CPU too hard, better use an old version" :p
 
Ocellaris said:
Testing with the old version just means you aren't stressing the chip. Which pretty much defeats the purpose of a stress test once you decide "Well those AVX instructions work the CPU too hard, better use an old version" :p
Click to expand...

Not true.
I havent had a single crash and I game hard.
If there was a problem I would clearly be using a different test method.
 
Nenu said:
Not true.
I havent had a single crash and I game hard.
If there was a problem I would clearly be using a different test method.
Click to expand...

Everyone comes with their own definition of stable. I'd rather find the edge case maximum stress and shoot fork that rather than make the system stable enough that it doesn't crash (yet). People BSing about how they run tests certain ways to check for stability is often why it's hard to compare numbers. 4.6GHz isn't the same between the systems if one can't pass tests that others can.
 
Just adding my own thoughts here as i'm going through the same issues with my asus z170-a + h100i v2 cooler + 6700k..

On the side of stress testing, i'm using either 27.9 or 28.9, but in the local.txt i made sure to include CpuSupportsFMA3=0 per various tips online about a skylake bug that can cause a freeze (unless the bios is now updated).. at any rate.. i use prime95 with custom and 8k for the setting for 1 hour as one test.. or Prime95 blend.. usually if i'm really unstable, P95 will crash or report errors within 10-25 minutes on a blend test. I also use the rog realbench with does x264 encoding cycles.. I think i've seen it recommended to run this one for overnight periods, but i run into errors usually within 15 minutes on obvious issues.. the other is the x264 custom loop 16 thread test you can find online.. version 2.0.6 currently.. generally also ran overnight.. then IBT for 50 cycles (sometimes i fail at 2.5 hours). Linpack is the other but i cant run it, even at 4.4ghz and around 1.35 load, it heats the cpu to 98C in seconds (normal?).

My typical testing temps are 82-85C with Prime blend and around 75-78 with realbench.. happy with realworld results that are 75c or less (right now at 1.32v load + 4.4 i'm 74C with an acronis going on, max).

Back to the overclock, i have also had issues going above 4.4ghz (seems to like 1.29 to 1.32 under load here).

The ticket for me so far, has been to do the auto overclock asus tool in windows.. it actually uses auto for volts (I know bad right).. but it ended up under load with 1.32 after leaving it auto and -0.040 in the voltage setting in the asus tool.. i'm guessing i can go adaptive 1.32 with additional voltage (zero?) at some point.

When i tried for 4.5 or higher, no matter what LLC + voltage combo i tried, it would cause p95 to crash after 3-15 minutes or error out. I went as high as 1.40 with llc at 4 or 5 out of 7, to keep the load close to that 1.40 (which was my understanding of the need for the llc value to be higher).. and 4.5 would not work for me (stock ram 2133 at 1.4 volts too, just to be safe).

So for now i gave up on anything higher than 4.4, which is unfortunate given others success.. or maybe try for another chip, unsure.

OP: Which cooler are you using ? edit: nm i see now
 
Last edited:
See Prime 95 Stress Test i7 6700k

"Do NOT run any versions of Prime95 later than 26.6. Here's why:

Core i 2nd through 6th Generation CPU's have AVX (Advanced Vector Extension) instruction sets. Recent versions of Prime95, such as 28.9, run AVX code on the Floating Point Unit (FPU) math coprocessor, which produces unrealistically high temperatures. The FPU test in the utility AIDA64 shows similar results.

Prime95 v26.6 produces temperatures on 3rd through 6th Generation processors more consistent with 2nd Generation, which also have AVX instructions, but do not suffer from thermal extremes due to having a soldered Integrated Heat Spreader and a significantly larger Die.

Please download Prime95 version 26.6 - http://windows-downloads-center.blogspot.com/2011/04/pr...

Run only Small FFT’s for 10 minutes.

Use Real Temp to measure your Core temperatures, as it was developed specifically for Intel processors: Real Temp -http://www.techpowerup.com/downloads/2089/real-temp-3-7...

Your Core temperatures will test 10 to 20C lower with v26.6 than with v28.9."
 
You must log in or register to reply here.
Back
Top