CPU Constant Voltage

Victorx588

Reputable
Mar 4, 2017
8
0
4,510
I'm new to overclocking so this might seem like a dumb question. I've just overclocked my i5-6600K to 4.8 GHz on 1.33V, everything is fine but the only issue i have is when the CPU is idle, the Voltages stay constant (800 MHz, still 1.31-1.33V). I have a Asus Sabertooth S, is there any option i can enable in my BIOS to lower the voltage on idle? (I've tried enabling Intel speedstep).
 
Solution
You wouldn't be wasting power leaving it at constant voltage. It doesn't make a whole lot of difference in life span either. It helps but you by the time you'd see a difference the chip would be well outdated.

Victorx588

Reputable
Mar 4, 2017
8
0
4,510
It was on auto by default, I thought putting it on enabled would help but no luck. Also my motherboard doesn't have an adaptive setting, it's just auto, enabled or disabled.
 

Victorx588

Reputable
Mar 4, 2017
8
0
4,510
I'm blind, xD it was in the CPU voltage drop down menu just missed it cause I was to excited about booting up windows and testing the overlock. One more question though, I put it on adaptive and as you said it hits the CPU with way to much voltage (1.53) is there any way to set up a max voltage limit?
 

trawetSluaP

Distinguished
Jun 5, 2013
389
0
18,860
Don't use the "Auto" setting. The motherboard will pump WAY too much voltage into the CPU.

If you highlight "Auto" where CPU Voltage is you can manually enter a figure and it will stick to that.

As for it not dropping the voltage this isn't an issue. Many people say they run their system like this (because Adaptive is rubbish) for years and have no issues, as long as the temps are fine.
 
To clarify what these guys are saying, adaptive mode will only cause overvolting on the most ridiculously demanding apps that exist for PC (like Prime95).

So still use adaptive, I use it all the time on Haswell (Haswell is WAY WAY WAY more prone to overvolting), and I've never had a single overvolting problem. Like I said above, only certain apps like Prime95 and I think Linkpack will cause any sort of chance of overvolting.

However, I also heard that skylake/Kaby Lake got a updated version of adaptive voltage, meaning that it shouldn't overvolt no matter what now. However I can't fully back this up.

EDIT: I didn't see that you are getting overvoltage. What vcore are you putting in the adaptive voltage? Is your BIOS up to date. Cause overvolting right off the bat is completely not normal.
 

The_Staplergun

Estimable
Jan 30, 2017
1,395
0
2,960
I can tell you that's wrong. I had the adaptive voltage turned on. AI Suite allows you to set a "max voltage" and I had it at 1.375. it overvolted as high as 1.42 just gaming. I don't trust any automated functions.

Using manual I have LLC set to auto and it puts it at 6 (from 1-7) and I see a zero vdroop at that.
 


That is very very strange, I've never had this problem with Haswell, and Haswell is known for this kind of issue (far more than Skylake). So this is all on adaptive mode?

I need to do some research.

Are you trying this in the BIOS too? I don't trust software for overclocking.

Update, ok for both OP and you, try using the adaptive mode's offset functionality, to compensate for the overvolting.
 

The_Staplergun

Estimable
Jan 30, 2017
1,395
0
2,960
It's kaby lake.

Adaptive set in bios.

With asus boards you can manually tweak the adaptive curve in ai suite. You can also change some stuff around when it's set to "auto" voltage in the bios.

I might have not played around with it enough as I just said forget it when I saw 1.42.
 

The_Staplergun

Estimable
Jan 30, 2017
1,395
0
2,960
Just because the clock is that high doesn't mean it's using that much wattage.

Basic laws of electricity. Voltage is pushed by amperage (data signals) through resistance (logical transistors in the die)

If there's no data signal even though the supplied voltage is high there's no calculations being done so there's no resistance from transistors and no amperage required.

I'll be sitting at 4800 MHz at 1.280 volts and have like 5 or less watts used at idle. Under load is when you get heat and wattage from calculations and data
 


Really? I did not know that. Thank you!
 

The_Staplergun

Estimable
Jan 30, 2017
1,395
0
2,960
So, to kind of highlight this and the oddities of computer processing
Here is four images doing different things and at different voltage.

This is the computer processor downclocking itself during an idle cycle to 800mhz. Slightly lower wattage, but same voltage.
Nlen7Z3.jpg

The processor bouncing around frequencies as it's needed. Here it is at 2700mhz. Notice that the wattage draw is actually 1 watt lower than 800mhz.
s4eF6m1.jpg

Here it is stepped in at 3300mhz. Slightly higher wattage, matching the 800mhz.
eTiRlmN.jpg

Here is at idle at 4800mhz with nothing running but the standard background apps. Notice its around 10 watts even with a high frequency and the volts at 1.280.
ELnxmSB.png

Here is my system at a higher voltage threshold (1.320) but under a different type of load (CPUz instead of Prime95 down below). Note the 78 Watts
YoipC0N.jpg

Here is doing large in place FFT's which generates the most heat supposedly. Notice my temp is higher than the CPUz test even being at a lower voltage, but the wattage is much higher.
GpKY1Bf.png

Here is the CPU doing blended testing under Prime95 which tests a bunch of different components without as much stress directly on the CPU (not for thermal). Note that the temps are slightly lower by a small margin, but the wattage is actually about the same as the CPUz load.
7lVA308.png


So, as odd as computing is, different loads exercise the CPU in different ways. It all depends on the size of the data being sent in.

Even at a higher voltage or a higher frequency, it doesn't directly correlate with higher temps or wattage usage. It's based on load provided and how much the CPU needs to intervene. I would gladly do a load under gaming but I have no GPU right now as I'm waiting for the 1080ti from asus to come out.
 

Victorx588

Reputable
Mar 4, 2017
8
0
4,510


Just updated my BIOS, didn't make a difference though. I tried playing around with the C states but i don't know how any of that works so i just left it alone.
 

Victorx588

Reputable
Mar 4, 2017
8
0
4,510
One more thing, I noticed that my motherboard lets me set an off set voltage when inn adaptive, I was reading up on what offset voltage is and how to use it. Apparently you can subtract the voltage it gives to your CPU, the only problem is it does it from both idle and full load, I'll try messing with it tomorrow and see if I can get any successful results.
 

The_Staplergun

Estimable
Jan 30, 2017
1,395
0
2,960
You wouldn't be wasting power leaving it at constant voltage. It doesn't make a whole lot of difference in life span either. It helps but you by the time you'd see a difference the chip would be well outdated.
 
Solution

The_Staplergun

Estimable
Jan 30, 2017
1,395
0
2,960
Not to double post but a bit more into wattage:

A watt is measured directly as Volts x Amps.

Since the volts are so tiny, the amps are what makes up the difference in all that power.

Your home electric bill is measured in kilowatt hours.

A chip running at 80 watts (which is a unrealistic load for a regular gamer) is 80 watt-hours. Multiply that by 24hours in a day and you get 1920 watt hours, which is 1.92 kilowatt hours. That's in a single 24 hours of unrealistically loading down this chip. You probably pay 10-15 cents a kilowatt hour, so that's literally 20-30 cents in a day. Over the span of a month??? At 30 cents a day for 30 straight days of retardation on this chip would be $9 usd.

Most of the time you would game 4-6 hours a day so cut that down to 1/6 or 1/4 of the cost....So $2-$3 a month very roughly. You're not gonna cost a lot.

The big concern is over volting (1.4 volts or more) which you're not doing. You will not see significantly reduced lifespan. You're good man.