I would like to get a basic OC on my I5-4690K

[pqwoerituytruei]

Why should blck be used if multiplier*bclk >= (multiplier+1)*100

for now i am going to leave it at 4.3, i think 4.4 requires around 1.275 vcore, not sure exactly cause i think the vrm was overheating when i tested it last time

since i want to use adaptive vcore id like to not use that much vcore to keep to below 1.3

how do you stress test a per core OC
if i use 44, 43, 43, 43, do i just stress 1 core?

 

[PolRoger]

4.3 @1.225 survived around 1hr 20 min, assuming stable enough
4.3 @1.225 w/ x39 cache at 1.15 failed at around 20 min
4.3 @1.225 w/ x39 cache at 1.155 failed at around 1hr 5 min
4.3 @1.225 w/ x39 cache at 1.16 TODO

Does LLC affect input or core voltage?

Core.

I'm not sure if your answer is correct Joe?...

For the Haswell/DC arch/generation cpu... I believe when LLC is applied that it affects CPU Input Voltage which goes to the chip's on board FIVR. The FIVR in turn steps down the voltage within the chip to what is requested by BIOS vcore.

For other Intel Arch... SB, IB, Skylake, Kaby Lake, and Coffee Lake(?)... Intel didn't use FIVR so motherboard LLC is applied directly to the chip via cpu vcore.

 

[pqwoerituytruei]

The way i have been doing my input voltage is based on what the sensor reads at stock, 1.792
If i use a fixed input of 1.76, i get about 1.8
so what i am using for a full time OC is a offset of -0.030V which gives me about 1.8
the way i have been looking at it is the IC that turns the input into vcore has a vdrop or forward voltage, so if forward voltage is 0.4 and i want a vcore of 1.28 i need at least 1.68 input, the smaller the difference the less heat is generated by the IC/regulator but if i cut it to close i will not get enough voltage

 

[PolRoger]

The way i have been doing my input voltage is based on what the sensor reads at stock, 1.792
If i use a fixed input of 1.76, i get about 1.8
so what i am using for a full time OC is a offset of -0.030V which gives me about 1.8
the way i have been looking at it is the IC that turns the input into vcore has a vdrop or forward voltage, so if forward voltage is 0.4 and i want a vcore of 1.28 i need at least 1.68 input, the smaller the difference the less heat is generated by the IC/regulator but if i cut it to close i will not get enough voltage

Vdrop no... forward (or overvoltage) yes.

Vdrop is the amount of vcore voltage a motherboard drops or is lost from a given specified input. Set 1.3v in BIOS and get ~1.28v after booting into Windows = .020v vdrop .

Vdroop is the amount of voltage loss/droop going from idle to load. So take that 1.28v vcore at idle in Windows and under load it droops down to 1.23v in Windows =.050v vdroop.

Your CPU Input voltage of 1.76v is either through BIOS LLC application or through motherboard/BIOS design function overvolting to ~1.8v via sensor reading. From my experience the on board chip FIVR does its own thing and LLC applied to input voltage doesn't get passed along in the voltage down stepping or conversion. I have observed with Haswell through sensor readings that the actual vcore going to the individual cores is usually higher of some what overvolted than what is specified via BIOS input. I generally try to overclock my Haswell chips with lower than nominal 1.8v input voltage with the idea that the lower voltage may helps with load temps.

 

[wingman99]

The motherboard VRM (voltage regulator module) sends the input voltage to the processor FIVR (fully integrated voltage regulator). FIVE in the processor supply's the core voltage. The FIVR contains a Load Line that is AUTO enabled to Intel's specification for voltage versus current.

Load line calibration (LLC) is for Load line.
With defualt settings there is load line because the operation of transistors is nonlinear current vs the voltage.

Load line (electronics)
In graphical analysis of nonlinear electronic circuits, a load line is a line drawn on the characteristic curve, a graph of the current vs the voltage in a nonlinear device like a diode or transistor.
It represents the constraint put on the voltage and current in the nonlinear device by the external circuit. https://en.wikipedia.org/wiki/Load_line_(electronics)

 

[EarthDog]

I'm not sure if your answer is correct Joe?...

For the Haswell/DC arch/generation cpu... I believe when LLC is applied that it affects CPU Input Voltage which goes to the chip's on board FIVR. The FIVR in turn steps down the voltage within the chip to what is requested by BIOS vcore.

For other Intel Arch... SB, IB, Skylake, Kaby Lake, and Coffee Lake(?)... Intel didn't use FIVR so motherboard LLC is applied directly to the chip via cpu vcore.

It has been a while since I was on that platform, however, I vaguely recall it hitting the input voltage now that you mention it. Good catch!

 

[wingman99]

It has been a while since I was on that platform, however, I vaguely recall it hitting the input voltage now that you mention it. Good catch!

Load line would not work with core utilization current vs voltage if it was regulated before the FIVR that regulate the core voltage and current. Load line from what your saying would be like Skylake X VCCIN lowering when there is increased current from the cores being utilized from idle. It does not work that way.

There is the PSU voltage regulator, then motherboard voltage regulator, finally the processor voltage regulator. Each one regulates voltage according to it's own specification, current demand and voltage level.

 

[PolRoger]

The motherboard VRM (voltage regulator module) sends the input voltage to the processor FIVR (fully integrated voltage regulator). FIVR in the processor supply's the core voltage. FIVR contains a Load Line that is AUTO enabled to Intel's specification for voltage versus current.

Load line calibration (LLC) is for Load line.
With defualt settings there is load line because the operation of transistors is nonlinear current vs the voltage.

Load line (electronics)
In graphical analysis of nonlinear electronic circuits, a load line is a line drawn on the characteristic curve, a graph of the current vs the voltage in a nonlinear device like a diode or transistor.
It represents the constraint put on the voltage and current in the nonlinear device by the external circuit. https://en.wikipedia.org/wiki/Load_line_(electronics)

That is interesting about the Haswell (arch) FIVR having an AUTO enabled Load line...

Load line would not work with core utilization current vs voltage if it was regulated before the FIVR that regulate the core voltage and current. Load line from what your saying would be like Skylake X VCCIN lowering when there is increased current from the cores being utilized from idle. It does not work that way.

There is the PSU voltage regulator, then motherboard voltage regulator, finally the processor voltage regulator. Each one regulates voltage according to it's own specification, current demand and voltage level.

From my experience changing LLC settings in BIOS on Z87/Z97 motherboards applies the the various LLC curve settings from the motherboard's VRM output to the CPU Input voltage.

With my ASUS Z87 Deluxe and CPU Input Voltage set in BIOS to 1.780v:

LLC Level 1 gives a CPU Input voltage reading of ~1.744v: Idle and ~1.696v: Load
LLC Level 5 gives a CPU Input voltage reading of ~1.776v: Idle and ~1.760v: Load
LLC Level 8 gives a CPU Input voltage reading of ~1.792v: Idle and ~1.808v: Load

Where as my FIVR supplied VCore readings remains more or less static with my current BIOS setting of 1.280v.

 

[EarthDog]

I wonder if it was just a board partner that did it or???????

I've been looking around for details so I have clarity, but nothing official.

 

[PolRoger]

I only have my ASUS Z87 Deluxe up and running currently but I also have two other Z97 boards on the storage/parts shelf...(Gigabyte/ASRock) and I believe both those board behave in a similar manner.

 

[EarthDog]

I have nothing here I can test.

But I have looked it up and saw (anecdotes) of what it does. Some say it was vcore and not input, others say the opposite, some say both...

 

[PolRoger]

I have SB, SB-E, IB, IB-E, Haswell, Skylake and Kaby Lake setups and I believe only the Haswell Arch has FIVR? (Haswell, Haswell-E, Broadwell and Broadwell-E) With all the other generation cpu there is no FIVR so motherboard VRM supplied LLC settings go directly CPU Vcore.

 

[wingman99]

That is interesting about the Haswell (arch) FIVR having an AUTO enabled Load line...



From my experience changing LLC settings in BIOS on Z87/Z97 motherboards applies the the various LLC curve settings from the motherboard's VRM output to the CPU Input voltage.

With my ASUS Z87 Deluxe and CPU Input Voltage set in BIOS to 1.780v:

LLC Level 1 gives a CPU Input voltage reading of ~1.744v: Idle and ~1.696v: Load
LLC Level 5 gives a CPU Input voltage reading of ~1.776v: Idle and ~1.760v: Load
LLC Level 8 gives a CPU Input voltage reading of ~1.792v: Idle and ~1.808v: Load

Where as my FIVR supplied VCore readings remains more or less static with my current BIOS setting of 1.280v.

Okay it looks like load line is from the motherboard input VCCIN.

 

[Organik]

update
So i decided to give use the "optimized 4.5GHz" setting in my bois
i looked over the settings, dropped the vcore from 1.32 to 1.3 and turned on LLC; that seems at least somewhat stable, it boots and can run a minute of stress testing without crashing at least
should i also bump the cache multiplier?
Also how do i know how much voltage i need for CPU input, the "optimized" option is using 1.9; stock reads 1.82

Don't go too high on the vcore on that chip. 1.33v or 1.34v should neild 4.4 to 4.6Ghz, that is all I will say. I don't know too much about how this chip OCes and what voltages are needed.

 

[Tech Tweaker]

Sounds like you might have a stubborn CPU, or a less than stellar sample from the 4690K lineup. Much like my previous 4690K.

1. You're going to need more than 1.8v on the CPU Input voltage to keep the system stable, pretty much guaranteed. Probably 1.85V should do it. It tends to default to 1.9V for whatever they called the in-BIOS auto-overclocking (on my board anyway).
2. If this is anything like my 4690K, I'd try 1.25V for the Core Voltage and see if that's stable (assuming you're still trying to get it stable that is...) (it's what mine took to be stable at 4.3GHz).
3. LLC is a good thing to have on, it tends to make things easier and make the Core Voltage more stable and predictable.
4. Basically ignore what Organik said about using Offset voltage, it tends to apply more voltage than necessary in my experience in order to stabilize an overclock (and thus cause an unnecessary increase in heat output). It's the easier/lazier way of overclocking, but it's not necessarily more efficient.
5. Actually, ignore what Organik said about adjusting the Base Clock (BCLK/BCLCK) as well. Most people can't get anywhere near 120MHz on this platform anyway, and the system tends to get unstable if you go much past say 104-105MHz on the LGA1150 platform (and LGA1155, for that matter). I wouldn't even bother adjusting that up at this point, as you're just going for a more basic overclock and that'll just throw more variables in and likely cause confusion. It's more of a thing you use to fine-tune an overclock to get that last few Megahertz if one is going for a high overclock while benchmarking and/or trying to break records anyway on more modern platforms from what I know.
6. On the bright side, from what I can see from one of your pictures it looks like you have an L4 Malaysia chip which means this should be a better overclocking candidate than the X5 Vietnam chip I have (X5's don't tend to perform well and/or need more voltage from my research, L3 and L4 CPU's from Malaysia and Costa Rica tend to do better).

My settings from my 4690K:
Batch number: X527A991

stock (100MHz x 35): 1.15v (voltage set to auto/default) (55-56°C max load temp)
4.2GHz (100MHz x 42): 1.179v (1.18v) (62°C max load temp)-ASRock Z97 Extreme6
4.3GHz (100MHz x 43): 1.247v (1.25v) (68°C max load temp)-ASRock Z97 Extreme6
4.4GHz (100MHz x 44: 1.29v) (1.29v) (77°C max load temp in Real Temp, 80°C in HWMonitor)-Asus Z97-A USB 3.1 -four hours Prime95 stable

 

[pqwoerituytruei]

I ordered my chip at newegg, on the release weekend (July 4th)
There was a 50$ combo with this board putting it at $300, i knew if i did not go this this platform i would have to replace my ram when i upgraded
I believe my current config is stable
43 Core;39 Cache; 1.225 vcore; 1.16 vcache; 100.0 Blck seems stable

 

[wingman99]

Sounds like you might have a stubborn CPU, or a less than stellar sample from the 4690K lineup. Much like my previous 4690K.

1. You're going to need more than 1.8v on the CPU Input voltage to keep the system stable, pretty much guaranteed. Probably 1.85V should do it. It defaults to 1.9V for whatever they called the in-BIOS auto-overclocking.
2. If this is anything like my 4690K, I'd try 1.25V for the Core Voltage and see if that's stable (assuming you're still trying to get it stable that is...) (it's what mine took to be stable at 4.3GHz).
3. LLC is a good thing to have on, it tends to make things easier and make the Core Voltage more stable and predictable.
4. Basically ignore what Organik said about using Offset voltage, it tends to apply more voltage than necessary in my experience in order to stabilize an overclock (and thus cause an unnecessary increase in heat output). It's the easier/lazier way of overclocking, but it's not necessarily more efficient.
5. Actually, ignore what Organik said about adjusting the Base Clock (BCLK/BCLCK) as well. Most people can't get anywhere near 120MHz on this platform anyway, and the system tends to get unstable if you go much past say 104-105MHz on the LGA1150 platform (and LGA1155, for that matter). I wouldn't even bother adjusting that up at this point, as you're just going for a more basic overclock and that'll just throw more variables in and likely cause confusion.
6. On the bright side, from what I can see from one of your pictures it looks like you have an L4 Malaysia chip which means this should be a better overclocking candidate than the X5 Vietnam chip I have (X5's don't tend to perform well and/or need more voltage from my research, L3 and L4 CPU's from Malaysia and Costa Rica tend to do better).

My settings from my 4690K:
Batch number: X527A991

stock (100MHz x 35): 1.15v (voltage set to auto/default) (55-56°C max load temp)
4.2GHz (100MHz x 42): 1.179v (1.18v) (62°C max load temp)-ASRock Z97 Extreme6
4.3GHz (100MHz x 43): 1.247v (1.25v) (68°C max load temp)-ASRock Z97 Extreme6
4.4GHz (100MHz x 44: 1.29v) (1.29v) (77°C max load temp in Real Temp, 80°C in HWMonitor)-Asus Z97-A USB 3.1 -four hours Prime95 stable

When you say LLC on is that enabled or disabled setting?

 

[EarthDog]

On would equal enabled. Off would equal disabled.

 

[Tech Tweaker]

I ordered my chip at newegg, on the release weekend (July 4th)
There was a 50$ combo with this board putting it at $300, i knew if i did not go this this platform i would have to replace my ram when i upgraded
I believe my current config is stable
43 Core;39 Cache; 1.225 vcore; 1.16 vcache; 100.0 Blck seems stable

Oh, sounds like you do have a better one than mine then.

On would equal enabled. Off would equal disabled.

Pretty much that.

On=ON=Enabled=Engaged=I=1 (if we get into binary coding) Off=OFF=Disabled=Disengaged=O=0 (if we get into binary coding)

I would have thought that the word "on" would have been clear enough, but I didn't think I'd have to be so technical as to say "Enabled" for most reading it to understand what I meant.

 

[wingman99]

Oh, sounds like you do have a better one than mine then.



Pretty much that.

On=ON=Enabled=Engaged=I=1 (if we get into binary coding) Off=OFF=Disabled=Disengaged=O=0 (if we get into binary coding)

I would have thought that the word "on" would have been clear enough, but I didn't think I'd have to be so technical as to say "Enabled" for most reading it to understand what I meant.

I equate enabled load line to mean fallowing Intel's specifications nonlinear current vs the voltage. Then disable load line to mean linear current vs the voltage. From what you were describing as stable voltage load line would be disabled, so the current vs the voltage would be linear.