Gigabyte Offset Voltage
Gigabyte Offset Voltage
I just finished an overclock of an i5 6600k, 4.6 GHz, Vcore 1.290V (BIOS) / 1.284 (CPU-Z). I'll list my settings I used DURING OC'ing below. I passed OCCT-LINPACK and OCCT CPU-Large, but when I ran OCCT CPU-Small my temps climbed above 80-degrees. I ended up needing to use the "Relax OC" under the advanced memory settings which brought the temps down 10-degrees. Not sure what "Relax OC" does because BIOS is still showing the same DRAM voltages and timings from the XMP Profile 1. This allowed me to complete all tests for 3 hours and keep temps below 80.
So now it's time to use an offset or adaptive voltage. [Note - this is a Gigabyte MOBO]. I set the CPU Voltage to "NORMAL" (in place of the 1.290V manual voltage) then set a DVID of +0.000V. Everything seemed to run just as it did when I have the manual voltage set. Check.
Questions
A) Next I enabled Turbo, EIST, and all C-states but voltage stays at 1.284V (CPU-Z) and Freq 4600GHz regardless if the system is idle or under load. Why doesn't the voltage or freq drop at idle?
B) I tried different values for DVID but all minus values caused a crash (even -0.005v) and the plus DVID just put me above the 1.284 I needed to remain stable. Does anybody know how to make offset or adaptive voltage work on the Gigabyte boards?
Setting used during OC'ing:
----For the CPU:
[1]-CPU VCore at 1.290 V,
[2]-CPU Core Ratio at 46,
[3]-Uncore Ratio at 45,
[4]-FCLK Frequency For Early Power 1000 MHz
[5]-CPU VCore Loadline Calibration (LLC) on High
[6]-Internal Graphics Disabled
-----For the memory:
[7]-Extreme Memory Profile (X.M.P.) on "Profile 1"
[8]-Memory Enhancement Settings to "Relax OC"
-----Advanced CPU Settings [Energy Savers]
[10] Intel Turbo Boost Technology DISABLED
[11] CPU EIST DISABLED
[12] CPU Enhanced Halt (C1E) DISABLED
[13] C3 State Support DISABLED
[14] C6/C7 State Support DISABLED
[12] C8 State Support DISABLED
[13] Package C-State Limit AUTO
PC Specs:
MOBO: Gigabyte Z170 Gaming 3 (F6 BIOS)
CPU: i5 6600k
CPU Cooler: EVO 212
RAM: 16GB Corsair Vengeance DDR4 3000MHz
GPU: Gigabyte GTX960OC 2GB
Noted a curious change today. I checked if my voltage was decreasing in the "stock" BIOS profile and confirmed it did at idle. Then I adjusted the CPU voltage to 1.290V in BIOS, saved the changes, and exited. When Windows started, CPU-Z showed no drop in voltage. After restarting into BIOS and reloading the stock profile, the voltage returned to AUTO. Opening CPU-Z again confirmed it wasn’t dropping at idle either.
When I shut down and restarted, the issue reappeared. It seems the settings didn’t persist properly after saving and exiting. To test further, I tried setting a small negative value for DVID (-0.020V) and observed the effect.
Current configurations:
----CPU:
[1]-CPU VCore at NORMAL,
[1a] DVID at +0.000V
[2]-CPU Core Ratio at 46,
[3]-Uncore Ratio at 45,
[4]-FCLK Frequency For Early Power 1000 MHz
[5]-CPU VCore Loadline Calibration (LLC) on High
[6]-Internal Graphics Disabled
----Memory:
[7]-Extreme Memory Profile (X.M.P.) on "Profile 1"
[8]-Memory Enhancement Settings to "Relax OC"
----Advanced CPU Settings [Energy Savers]
[10] Intel Turbo Boost Technology DISABLED
[11] CPU EIST at AUTO
[12] CPU Enhanced Halt (C1E) at AUTO
[13] C3 State Support DISABLED
[14] C6/C7 State Support DISABLED
[12] C8 State Support DISABLED
[13] Package C-State Limit AUTO
PC Specifications:
MOBO: Gigabyte Z170 Gaming 3 (F6 BIOS)
CPU: i5 6600k
CPU Cooler: EVO 212
RAM: 16GB Corsair Vengeance DDR4 3000MHz
GPU: Gigabyte GTX960OC 2GB
burnhamjs :
I just finished an overclock of an i5 6600k, 4.6 GHz, Vcore 1.290V (BIOS) / 1.284 (CPU-Z). I'll list my settings I used DURING OC'ing below. I passed OCCT-LINPACK and OCCT CPU-Large, but when I ran OCCT CPU-Small my temps climbed above 80-degrees. I ended up needing to use the "Relax OC" under the advanced memory settings which brought the temps down 10-degrees. Not sure what "Relax OC" does because BIOS is still showing the same DRAM voltages and timings from the XMP Profile 1. This allowed me to complete all tests for 3 hours and keep temps below 80.
So now it's time to use an offset or adaptive voltage. [Note - this is a Gigabyte MOBO]. I set the CPU Voltage to "NORMAL" (in place of the 1.290V manual voltage) then set a DVID of +0.000V. Everything seemed to run just as it did when I have the manual voltage set. Check.
Questions
A) Next I enabled Turbo, EIST, and all C-states but voltage stays at 1.284V (CPU-Z) and Freq 4600GHz regardless if the system is idle or under load. Why doesn't the voltage or freq drop at idle?
B) I tried different values for DVID but all minus values caused a crash (even -0.005v) and the plus DVID just put me above the 1.284 I needed to remain stable. Does anybody know how to make offset or adaptive voltage work on the Gigabyte boards?
Setting used during OC'ing:
----For the CPU:
[1]-CPU VCore at 1.290 V,
[2]-CPU Core Ratio at 46,
[3]-Uncore Ratio at 45,
[4]-FCLK Frequency For Early Power 1000 MHz
[5]-CPU VCore Loadline Calibration (LLC) on High
[6]-Internal Graphics Disabled
-----For the memory:
[7]-Extreme Memory Profile (X.M.P.) on "Profile 1"
[8]-Memory Enhancement Settings to "Relax OC"
-----Advanced CPU Settings [Energy Savers]
[10] Intel Turbo Boost Technology DISABLED
[11] CPU EIST DISABLED
[12] CPU Enhanced Halt (C1E) DISABLED
[13] C3 State Support DISABLED
[14] C6/C7 State Support DISABLED
[12] C8 State Support DISABLED
[13] Package C-State Limit AUTO
PC Specs:
MOBO: Gigabyte Z170 Gaming 3 (F6 BIOS)
CPU: i5 6600k
CPU Cooler: EVO 212
RAM: 16GB Corsair Vengeance DDR4 3000MHz
GPU: Gigabyte GTX960OC 2GB
"Relax OC" does not modify the memory timings that YOU set in the BIOS but rather values (settings) of the IMC (Integrated Memory Controller) parameters (like the write recovery time tWR, the refresh cycle time tRFC, etc.) located inside ("integrated") the CPU to "ease" (or simply make possibe) its management of the flow of data going to and from the computer's main (RAM) memory under YOUR own BIOS timings (by increasing latencies, etc.).
Interesting note today. I decided to check if my voltage was decreasing in the "stock" BIOS profile, so I loaded it and confirmed that both my voltage and CPU frequency indeed dropped at idle.
Then I switched into BIOS, adjusted the CPU voltage from AUTO to 1.290V and saved before exiting. When Windows started, I used CPU-Z and saw that the voltage wasn’t dropping at idle.
After reloading the "stock" profile and restarting, the voltage stopped dropping again.
So, I powered down and restarted—Windows loaded and now my voltage is falling once more at idle. It seems the changes didn’t stick properly after saving and exiting.
I then tried booting back into BIOS, applying a 4.6GHz profile, changing the CPU voltage from 1.290V to NORMAL (keeping DVID at +0.000V), saved, and exited. Now the voltage drops at idle. Everything appears to function as expected, though I performed a short OCCT-SMALL test and observed minor spikes in voltage.
I might experiment with a small negative value for DVID (perhaps -0.020v) to see its effect.
Current Settings:
----For the CPU:
[1]-CPU VCore at NORMAL,
[1a] DVID at +0.000V
[2]-CPU Core Ratio at 46,
[3]-Uncore Ratio at 45,
[4]-FCLK Frequency For Early Power 1000 MHz
[5]-CPU VCore Loadline Calibration (LLC) on High
[6]-Internal Graphics Disabled
----For the memory:
[7]-Extreme Memory Profile (X.M.P.) on "Profile 1"
[8]-Memory Enhancement Settings to "Relax OC"
----Advanced CPU Settings [Energy Savers]
[10] Intel Turbo Boost Technology DISABLED
[11] CPU EIST at AUTO
[12] CPU Enhanced Halt (C1E) at AUTO
[13] C3 State Support DISABLED
[14] C6/C7 State Support DISABLED
[15] C8 State Support DISABLED
[16] Package C-State Limit AUTO
When you type <<and now my voltage is dropping again at idle>> it really points out something unusual. My apologies—I could have phrased it more clearly. I should have mentioned that my voltage is naturally dropping at idle and then increasing when under load. The adaptive feature is handling it well. Now I’m just experimenting with the offset, maybe trying a -0.010 adjustment.
burnhamjs :
philipew :
When you write <<and now my voltage is dropping again at idle>> you effectively highlight an anomaly.
My apologies - I could have written that better. I should have said my voltage is appropriately dropping at idle [and ramps back up under load]. Adaptive voltage is working. Now I just want to play with the offset, trying a -0.010 offset perhaps.
philipew :
One thing I don't quite understand is your mention of <<the settings didn't apply correctly>> ...er... didn't apply WHAT correctly? Should this perhaps be read instead as "the settings
weren't applied
correctly" ? Not being a native English speaker, I don't want to miss anything here.
I had appropriately set the voltage from 1.290V to NORMAL {adaptive}, however, when I “saved & exited” it stayed as a manual voltage (1.290V) instead of the NORMAL {adaptive} I had applied. When I shut down then started it was correctly using the NORMAL voltage setting. Just odd that I had to go a shut down vice a restart to get the settings to apply.
Let light shine out of darkness.
No worries. I am sorry but for "I had to go a shut down vice a restart to get the settings to apply.", I put "vice" in Google and got in return: <<word-forming element meaning "instead of, in place of,">>. This (I am told) may be a common expression in the area where you live (?). Now I get it... So I think you actually meant: "I had to go
and do
a shut down
instead of
a restart to get the settings to
be
appl
ied
." That's no problem, good to know even.
By the way, concerning "NORMAL {adaptive}", "Normal" is a setting which usually sets the stock value, whatever it is (around 1.20 V for VCore initially). "NORMAL {adaptive}" seems to imply that "Normal" in this case corresponds to some sort of "adaptive" setting. From the other thread on DVID I gathered that VCore effectively had to be set to "Normal" for this to work in some sort of "adaptive" way (it looks more like an offset though).
During periods of high CPU demand, the VRM (Voltage Regulator Module) circuit works hard to supply the current "requested" by the processor. However, as soon as that load is gone, the VRM circuit must act quickly in order to reduce the current supply to the level needed to match the new demand. Because it's impossible for the VRM circuit to respond instantaneously, the larger the load change the greater the maximum potential peak overshoot voltage. Controlling the magnitude of these peak values is critical for maintaining system stability.
By positioning the processor's no-load (idle) voltage level higher during periods of light loading (e.g. 1.20 V is higher than 0.70 V), it is possible to sustain a larger negative voltage dip without crossing the processor's
minimum
specified voltage limit (e.g. 0.70 V). In addition, "drooping" the load voltage (by the VDroop amount) as a function of supply current, allows the VRM to effectively limit the
maximum
positive peak overshoot voltage (experienced during a heavy to light load transient) to a value below the maximum allowable max. CPU voltage. This resulting control system ensures the processor supply voltage, regardless of CPU load, never violates a specified limit (see further details here: http://www.anandtech.com/show/2404/5).
So
the CPU VID
( e.g. 1.20 V base) setting establishes the
absolute maximum
allowable processor supply
voltage
experienced during transient conditions and
is not the target idle voltage
. The DVID (Dynamic Voltage Identification, or Dynamic Voltage ID - http://www.overclock.net/t/665362/vid-vo...-explained) is said to be "dynamic" because it can change according to the CPU activity level (either loaded or idle) dynamically (i.e. automatically and in real time). Interestingly, I found from the following site:
http://www.hardocp.com/article/2015/10/1...d_review/3
...the following explanation which may help understanding (it is only an extract, please go to the IP address for the complete article) which mentions "
the dynamic vCore (DVID) setting that will light up below it
" (see below):
------------------------------------------------------------------------------------------------------------------------------------------------------------
... For example, I dislike how GIGABYTE implements voltage control for the CPU in the BIOS and that is putting it mildly. There isn’t a clearly defined specific adaptive or fixed / override voltage mode as such. These features are supported, but how these features are used isn’t exactly obvious, or it won’t be obvious for many people. You can’t hit enter and see all the values and choices present either.
....................................
If you set the voltage to "normal" then it puts the voltage control into offset mode. The voltage is appears to be fixed at 1.2v as a base setting. You can then add voltage via
the dynamic vCore (DVID) setting that will light up below it
. Any voltage offset you add or subtract seems to be based off the 1.2v setting. However the name implies the voltage mode to be
a dynamic and offset combination
. In truth, it is
a dynamic value with a max of 1.2v combined with whatever offset voltage is added or subtracted from that 1.2v value
.
The methods GIGABYTE has chosen for control do work ultimately, but it isn’t the most user friendly way they could have gone about it. Frankly, virtually everyone in the industry handles CPU vCore settings in a better way.
------------------------------------------------------------------------------------------------------------------------------------------------------------
I still need to physically test this on my rig
but from what I read and understand so far, the idea is to 'determine' the max VCore necessary for stability (e.g. 1.355 V) at the desired frequency (say 4.6 GHz for example). Then 'discover' the base setting, i.e. the "Normal" or default “stock” CPU VCore voltage automatically supplied by the motherboard's BIOS (let's say for example that it is 1.20 V) at that frequency. Then subtract the default VCore from the max VCore needed for stability (say 1.355 V – 1.200 V = 0.155 V). This resulting number is inserted as the DVID setting (say + 0.155 V). The end-value DVID (say 1.200 V + 0.155 V = 1.355 V in this example) is therefore a dynamic value made of: 1) a max. of 1.20 V combined with 2) whatever offset voltage is added (or subtracted) from that 1.20 V base setting value.
Mind you, I don't recall to ever have noticed any mention of this "DVID" setting in any of the (super thin and incomplete) documentation, either on paper or electronic media, but I can already confirm from looking at a screenshot from this review of the "GA-Z170XP-SLI" mobo's BIOS that it is there (on the second line, under CPU Vcore at the top): http://www.tweaktown.com/image.php?image...w_full.png. ...So I can't wait to be back home and try it ;-).
Everything counts here because although power consumption and temperature increase linearly with clock frequency, it increases
exponentially
with voltage !!! - And it's the overall temp that is of
great
interest to me, much more than the puny saving of a few bucks on the electricity used by my CPU.
I just hope that this thread's (http://forum.giga-byte.co.uk/index.php?topic=6849.0) specification of all C-states having to be
enabled
is not necessary (I prefer them disabled). It mentions: <<DVID = Gigabyte Dynamic Vcore. Vcore must be set to normal or DVID will not function.
All C states and EIST etc must be enabled
.>>. I must check this out.
Good on you for "unearthing" this not-so-well-known method, and further suggesting it as a thread on this forum for all Gigabyte users to benefit from using it... If I can effectively use this the way I describe it above (and of course I still have to test it all physically on my rig), I will consider this method
a pure gem
. Now let's put it into practice, and test... ;-).
PS: Thank You "grimsin" also for pointing out the "Offset" in your post... which eventually put me on this path...
I’m still trying to grasp your point about <<the settings didn't apply correctly>>... it seems the issue was whether the settings were applied properly. As a non-native speaker, I want to ensure I understand everything clearly. Yes, another way to phrase it would be: "the settings weren't applied correctly." Not being fluent in English makes me careful not to miss anything here.
Adaptive voltage testing proceeded smoothly. I adjusted DVID to -0.010V, which increased my maximum Vcore to 1.296V during the stress test. Finished three hours of OCCT-Large testing.
Stable 4.6G configurations
----For the CPU:
[1]-CPU VCore at NORMAL,
[1a] DVID at -0.010V
[2]-CPU Core Ratio at 46,
[3]-Uncore Ratio at 45,
[4]-FCLK Frequency For Early Power 1000 MHz
[5]-CPU VCore Loadline Calibration (LLC) on High
[6]-Internal Graphics Disabled
----For the memory:
[7]-Extreme Memory Profile (X.M.P.) on "Profile 1"
[8]-Memory Enhancement Settings to "Relax OC"
----Advanced CPU Settings [Energy Savers]
[10] Intel Turbo Boost Technology DISABLED
[11] CPU EIST at AUTO
[12] CPU Enhanced Halt (C1E) at AUTO
[13] C3 State Support DISABLED
[14] C6/C7 State Support DISABLED
[12] C8 State Support DISABLED
[13] Package C-State Limit AUTO
burnhamjs :
Adaptive voltage testing went well. I set DVID to -0.010V which brought my max Vcore to 1.296V during stress testing. Completed 3 hrs of OCCT-Large {see below}.
Stable 4.6G settings
----For the CPU:
[1]-CPU VCore at NORMAL,
[1a] DVID at -0.010V
[2]-CPU Core Ratio at 46,
[3]-Uncore Ratio at 45,
[4]-FCLK Frequency For Early Power 1000 MHz
[5]-CPU VCore Loadline Calibration (LLC) on High
[6]-Internal Graphics Disabled
-----For the memory:
[7]-Extreme Memory Profile (X.M.P.) on "Profile 1"
[8]-Memory Enhancement Settings to "Relax OC"
-----Advanced CPU Settings [Energy Savers]
[10] Intel Turbo Boost Technology DISABLED
[11] CPU EIST at AUTO
[12] CPU Enhanced Halt (C1E) at AUTO
[13] C3 State Support DISABLED
[14] C6/C7 State Support DISABLED
[12] C8 State Support DISABLED
[13] Package C-State Limit AUTO
Thank you for publishing your results, complete with clear screenshots, etc.
You are right, this looks good, but you are still a little too high at 1.296 V when 12 mV less (1.284 V) is sufficient for you to be stable, as you wrote previously. And since an offset of -10 drops your load VCore from 1.308 V to 1.296 V (-12 mV), I suppose that an offset of
-20
would drop your load VCore from 1.296 V to
1.284 V
(-12 mV) which is your
ideal
operating value, and also get your idle VCore even closer to 0.7 V (from 0.732 V) as a bonus. I recall that you already hinted (in a previous post) at trying this -20 offset.
The VID value and the other influencing parameters are very particular to each setup, even to each particular 6600K CPU unit. So I must really try it on my own rig and make observations as they appear, in sequence (not just when seeing an end-result) before being able to make more sense of how this works. I will probably be able to do it this week-end already, time permitting.