Asrock z87 voltage issues while just gaming

[tekjunkie28]

On my motherboard with the latest bios which is 2.40 I can NOT set adaptive voltage without it hitting 1.38 to 1.4v for a few seconds at a time in my core voltage is set at say 1.26. I have all C states enabled and I've been reading for weeks trying to find a solution. Setting a negative offset will lower the spike I think but only 0.3-0.5 max and I still need to check stability. I'm at 4.5Ghz with 1.265v and I'm decently stable in Aida and x264 but not prime which is the newer version.

Even with everything set to energy saving my vid does not lower on static voltage and I can confirm with a kill a watt letter that adaptive voltage setting used about 5 watts less then override voltage setting. Also to add salt to the wound I cannot monitor Vcore on the asrock board afaik.

So should the voltage spikes be ok? I have a h100i and only hit 90c under super heavy linpack.
Should I eat the extra 5 watts and keep static voltage?
Or should I lower my OC to 4.2-4.3 and lower voltage required?

I don't really wanna do static because in my mind that 1.265 continuous will degrade the processor life and I truly don't plan on replacing it for several years unless Intel gets some competition and they release a processor that's significantly faster.

Also does the ring bus overvolt like this on adaptive? Is it truly ok to hammer these with static voltage? I'm seriously debating if its even worth this hassle to overclock but I would LOVE to go farther.

- - - Updated - - -

The CPU in question is a 4670K.

 

[trents]

Are your speedstep/c states like they should be in bios?

 

[tekjunkie28]

Yes I believe so. I have the C state at C7 and everything else released to EIST is whatnot set to enabled. The options are disabled, enabled and auto. Also I dont think there is a way to monitor Vcore on the Asrock z87 extreme6, only VID, so if the Vcore was indeed downclocking (which I highly doubt it is as there was about a 5 watt increase in manual over adaptive) then I cant measure it.

On this websites OC guide to Haswell it says that on adaptive that ASUS said they only seen the voltage increase on synthetic tests but that it completely untrue in my case. But I can confirm the 0.12v increase only happens for a few seconds as a time during heavy load. It even does this at stock auto settings going from normal 1.167 to 1.285 for a few seconds.

 

[unclewebb]

If you have your C States enabled, using a fixed voltage is not such a bad idea. When idle, individual cores should be spending 99% of their time in C7. In this C State, cores are disconnected from the voltage rail so they are actually getting zero volts. Whatever voltage monitoring software you are using and whatever it is telling you becomes completely meaningless because it is going to be wrong 99% of the time when cores are in C7. The only time your cores will be receiving a fixed voltage is when the core becomes active and enters C0 to process a task. When active, the core needs this fixed voltage. When inactive, it goes back to C7 and is once again disconnected from the fixed voltage rail so no worries that your CPU will always be getting a fixed voltage.

Just make sure that your cores are actually using the C States. Some C State bios options do not actually work when you are overclocking.

 

[tekjunkie28]

WOW I have never seen this software before, I'll look in to this and post back, if this is true then I can go to a much higher overclock as i have a AIO water cooling H100i.

OK! I tried it out. The program seems a little buggy on detecting C7 states after some messing around or maybe its my PC but it just says 0 across the board for some reason sometimes. It only seems to work correctly for a few moments after a restart.

From what it tells me is that C7 was being used around 90-95% plus and package was C2. C3 and C6 were at 0 percent.

Well I so glad that I found that program. I was having issues with it and it would stop working about 2 minutes after windows would load up. I could pull up window performance monitor and the c states would start reading 0% at the same time there was some loading activity. I reverted my OC and windows still done this so I started looking into corrupt software and programs that might cause it and I am 99% sure it was the intel driver update software. There is a service that I had been seeing lately called WILLIAMETTE in windows services. It was removed when I removed the intel driver update utility and the command prompt windows displayed something like WILLIAMETTE perfmon so I assume it was keeping the CPU from entering any C state. Also my Kill-a-watt meter is now showing a 15 watt decrease in power consumption which is shocking. I also removed my A-tuning asrock mobo program but it wasn't setup to run anyways.

 

[unclewebb]

ThrottleStop has been around for 6 years and it is popular in the laptop community but most desktop owners have probably never heard of it. TS will let a user control their CPU in ways that no other software will give you access to. It has important options that Intel XTU does not.

The C States that the Windows performance monitor reports are completely different than what ThrottleStop reports. ThrottleStop monitors high performance timers within the CPU using a method recommended by Intel so it can accurately determine what C States the cores are in. When it shows 0.0, that core is definitely not using that C State. One bad driver or misbehaving program can prevent a core or cores from using the deeper C States. There are also a lot of motherboards out there that have C State bios options that simply do not work. Manufacturers get away with this because not enough people know how to monitor what C States the CPU is in.

Some manufacturers also deliberately disable the Package C States or limit the CPU so it can only enter Package C2. My Skylake Asus board has bios options for package C3 / C6 / C7 / C8 but none of them work. If the individual cores are entering the deeper C States like C7 and the package is only showing C2, that is what is going on. The deeper package C States can be disabled by the bios and there might not be any option in there to turn them back on.

Also my Kill-a-watt meter is now showing a 15 watt decrease in power consumption which is shocking.

Less power consumption means less voltage is going to your CPU, just what you wanted. I have been twisting people's arms for years, trying to get them to give ThrottleStop a try. Most monitoring software is meaningless when the C States are being used so it is nice to have a tool that shows you exactly what your CPU is really doing. If you have any questions, post some pics so I can have a look.

 

[tekjunkie28]

Alright so I cleaned out my entire case today and moved some wires around just for looks and my PSU was A LOT more dirty then I could see. After 3 hours my case looks great but the temps didnt go down which the radiator was clean to began with bc of case fan filters.

I can confirm that the WILLIAMETTE service is the cause of my Cstates not working. Idk how long thats been going on The strange part is that even at stock the VID still dropped but the Cstates were still reporting 0%.

Are some manufacturers better as having higher Cstate options then others? This is my first Asrock board and all my previous ones were Asus mainstream boards like the P5K and P5Q.

So now that I have that sorted out how far can I OC this think voltage wise for 24/7 OC? I will reach a voltage limit before i hit a temp limit for everyday use I think. The pc normally gets turned off at night.

 

[wingman99]

I use Adaptive offset on my motherboards over the last 6 years for overclocking to save the processor and power. From what I read you can't monitor Vcore, have you tried Intel Extreme Tuning Utility? https://downloadcenter.intel.com/download/24075/Intel-Extreme-Tuning-Utility-Intel-XTU-

VID sends the the signal to the VRM to supply Vcore. With Adaptive offset of the VID you can't see what your Vcore is, you will have to do the math unless you can find software that works. VID only changes with the clock to a extent and when you change the Adaptive offset VID does not change, the Vcore offset does the changing that you can see.

So to sum it up VID is built into the CPU by Intel and it can't be changed, it adjust the voltage dynamically, then you can add offset to it.

VID (Voltage identification)

The correct supply voltage is communicated by the microprocessor to the VRM at startup via a number of bits called VID (voltage identification). In particular, the VRM initially provides a standard supply voltage to the VID logic, which is the part of the processor whose only aim is to then send the VID to the VRM. When the VRM has received the VID identifying the required supply voltage, it starts acting as a voltage regulator, providing the required constant voltage supply to the processor.

Instead of having a power supply unit generate some fixed voltage, the CPU uses a small set of digital signals, the VID lines, to instruct an on-board power converter of the desired voltage level. The switch-mode buck converter then adjusts its output accordingly. The flexibility so obtained makes it possible to use the same power supply unit for CPUs with somewhat different nominal supply voltages and to reduce power consumption during idle periods by lowering the supply voltage.[5] https://en.wikipedia.org/wiki/Voltage_regulator_module

Serial Voltage Identification "SVID": A few generations back, Intel introduced serial voltage identification (SVID) which is a protocol the CPU uses to communicate with the voltage regulator. The power control unit inside the CPU uses SVID to communicate with the PWM controller that controls the voltage regulator. This allows the CPU to pick its optimum voltage depending on current conditions (temperature, frequency, load, etc.). You can actually use a combination of SVID and LLC to get an optimal VCore instead of manually setting it. If you start your system without making any changes, your VID (which some refer to as the stock voltage) might be 1.25v, but if you lower your CPU multiplier and restart, you will find your VID has dropped automatically. The reverse happens if you increase your clock and do not set any VCore. Intel's latest CPUs are able to pick their own voltage, and this comes into play if you want to utilize "offset" / "adaptive" voltage. The good news is that if you come from Haswell, you should look forward to a CPU that has the same or better durability.

Read more: http://www.tweaktown.com/guides/748...-intel-skylake-overclocking-guide/index5.html

I have a Gigabyte motherboard and all I do to use all the power saving features is leave everything on default and just change multiplier 45 then DVID to +0.075 and I was done after stress testing.

 

[tekjunkie28]

Yea I have tried all types of programs to see Vcore and it doesn't work. Vcore is apparently only able to be monitored by a separate chip that only some mobo manufacturers have used. There is absolutely no why that I know of that will keep the Vcore (or VID) from going to 1.4 when the adaptive is set to 1.26 or so. Is the spike to 1.4 dangerous? Is already spikes up to 1.28 on stock so it must not to too bad, right or wrong?
I also thought that Vcore cannot be higher then VID?




Pro

 

[wingman99]

Yea I have tried all types of programs to see Vcore and it doesn't work. Vcore is apparently only able to be monitored by a separate chip that only some mobo manufacturers have used. There is absolutely no why that I know of that will keep the Vcore (or VID) from going to 1.4 when the adaptive is set to 1.26 or so. Is the spike to 1.4 dangerous? Is already spikes up to 1.28 on stock so it must not to too bad, right or wrong?
I also thought that Vcore cannot be higher then VID?




Pro

Well VID is not what the CPU gets completely for Vcore, there is Intel (load line) that causes Vdoop so the voltage is lower. I would not worry about stock 1.28v the CPU is built for it. voltage can go as high as 1.44v and that is ok.

Can you set Adaptive offset negative - 0.160?

Have you contacted Asrock to ask what software you can use to monitor the Vcore? if your getting VID you should get Vcore the chip is there.

 

[tekjunkie28]

Well VID is not what the CPU gets completely for Vcore, there is Intel (load line) that causes Vdoop so the voltage is lower. I would not worry about stock 1.28v the CPU is built for it. voltage can go as high as 1.44v and that is ok.

Can you set Adaptive offset negative - 0.160?

Have you contacted Asrock to ask what software you can use to monitor the Vcore? if your getting VID you should get Vcore the chip is there.

The offset doesn't do anything like what everyone says it does. If I set 1.26 adaptive then a -0.16 offset it wouldn't even boot also it it just lowers the whole VID. I have tried everything but nothing can avoid the 0.12 spike in voltage under heavy load.

I have not contacted Asrock as I figured someone would have figured it out after almost 3.5 years. Worth the haswell FIVR it makes that nearly impossible and LLC doesn't do anything to Vcore because of the FIVR.


Pro

 

[Johan45]

I'm pretty sure the issue is adaptive voltage. I recall when haswell was introduced there were many issue between it and P95. There were reports of blown CPUs etc. and it was found that when in adaptive mode the BIOS/CPU will really hammer the voltage when AVX2 instruction sets were detected. P95 was not recommended as a stress test or to use one of the older versions if you were insistent on p95.
So IMO if you take the BIOS off adaptive and use offset only with EIST,CStates it should run a bit more level and not give you the voltage spikes tht you are seeing.

 

[tekjunkie28]

so your saying use auto with offsets?? or overide with offsets? If I set adaptive with offsets not mater what it spikes with ~0.13v. If I use offsets then its useless from what im setting, lets say i use override of 1.3 with a offset of -0.040 it just gives me 1.26 and with heavy loads such as civ6 even just loading a map gives a spike up to 1.394 and it only does it for a few cores as a time and only for very little of time. I'm using both Intel XTU and hardwareinfo64 to get these readings.

Johan45 I have used your link in your sig to overclock with and another guide but neither one say anything about what is safe voltages or how much voltage the processor is actually getting on adaptive with that spike. It even says in your guide you linked that there is nothing you can do about it but goes onto say that it should only be see in synthetic tests.

 

[Johan45]

I would use auto with offset, what you need is the VID for the CPU which is your starting point. Coretemp makes that easy http://www.alcpu.com/CoreTemp/ It takes a bit of experimenting to get it just right so you'll need to monitor voltages/temps etc. Until you find stability. As for max voltage you should be fine with spikes up to 1.4v as long as your temps are under control.

 

[tekjunkie28]

I tried auto with offset. It just makes the voltage static and XTU confirms that's it's set to static(override).

Temps are excellent. The H100i keeps them in the 40s and 50s @ 1.28 volts gaming and x264 is around 62


Pro

 

[tekjunkie28]

What do you mean VID? I can tell u that on stock idle it's around 0.717 and load it goes to 1.167 according to core temp. That spikes to 1.283 inter heavy load for just a few seconds just like adaptive would if I was overclocked just with lower voltage thought. So afaik there is NO way to monitor actual Vcore.


Pro

 

[Johan45]

You'll see when you run Coretemp. The VID is the programmed voltage for it's rated speed. So assuming the stock voltage is the VID 1.167 if you need it to get to 1.3v you add an offset of +.133v

 

[tekjunkie28]

If I put in the offset it will force override mode even thought the bios still says auto. It will also set 1.25 as the voltage. There is no way to use auto and set a voltage offset. The only way is adaptive. And then to set a voltage and then once i put in an offset it just offsets that voltage i put in.

It seems asrock sticks to the intel spec extremely tightly. There is a pic that shows how that stuff works and my mobo works exactly as the pic shows.


Pro

 

[Johan45]

Can't say I'm that familiar with Asrocks BIOS but typically, using offset the BIOS inserts the VID and your offset either adds or takes away from that number. Maybe the VID for that chip is 1.25v??

 

[tekjunkie28]

Idk. If I'm gaming (mostly strategy) then the voltage (VID according to hwinfo) sticks to 1.167 but if I run prime 95 28.5? At stock it gots all the way to 1.285...




Pro