Hello everybody
First of all: Sorry for bad english! I Google translated it from german (https://www.hardwareluxx.de/communi...pannungs-offsets-eigenmaechtig-1189784.html). Here can also be found mentioned attachements.
Then important: The following remarks relate not only to the overclocked state but also to the factory settings, both CPU and RAM without XMP. A problem with overclocking is therefore excluded. The attached pictures (1-4) have just been created, while I write this article without OC on a freshly installed Windows. At the moment therefore not really many drivers are installed. But I have already tried different driver versions. With / without Windows Update, from Intel and also the sensor driver from ASUS. But changes everything little to nothing at all.
Have recently got an i7-6850K on an ASUS X99-S with 4x 8GB Corsair Vengeance LPX memory with 3200 MHz XMP. In my opinion, I overclocked the CPU quite successfully to 4.4 GHz - with air cooling and compared to other people pretty low voltages (especially SA, Cache and VCCIO). At some point I noticed that the voltage offsets displayed in HWMonitor change in part. Sporadically (or rather regularly) for a very short time, they jump from 0 or the set values to significantly higher - if not more dangerous / deadly high values. Rarely, they occasionally fall off at fixed voltages or manual offsets.
No matter what I do (increase tensions, minimize, change the most unlikely attitudes), nothing will be better. I've been wandering around for weeks now with the most impossible settings the mainboard offers. It has little or no influence on these changing values.
In any case, I have a huge problem with this board anyway to determine voltages and temperatures like I have never had. Each tool displays different things, so different amounts, different (wrong) named, in different places in different categories. And then different values are also implausible (I hope it's hard, there is nowhere on the motherboard a bit beyond 100 ° - these two values jump also tremendously and by feeling I could not find anything hot on the board anywhere). I also hope that the offsets I mentioned are also implausible or wrong. However, I would like to get rid of that - for reassurance and anyway. But unfortunately I could not find another tool that shows me something like that (HwInfo, Aida, Open Hardware Monitor, ...) to compare that.
It already begins with the reading and determine of basic voltages, which I have configured in the bios (Figure 5). What is displayed in all tools under the sensor voltages as "Vcore" or "CPU Vcore" is anyway complete nonsense or another tension. VIN4 in the HWMonitor, I am also not sure if that is really the core voltage. At least it does not match the values of the Intel Xtreme Tuning Utility, but with the core voltage in AIDA. In the Intel XTU it shows me a core voltage, which agrees with the VID in AIDA and HWMonitor. The individual core voltages in the XTU deviate slightly from this (does this CPU have differentiating core voltages? The XTU can not display the VID (s) !?). The VID actually comes from the CPU itself. I change in the Bios but Vcore, so changes in the tools, the VID, so and u. a. Meanwhile, I assume that the "VID" rather corresponds to the core voltage, while "VIN4" would rather be the VID. In addition, set in the BIOS fixed voltages z. B. in HWMonitor always displayed with an offset higher, so the "VID" is correct, I think and then the "IA" is higher by an offset and the "VIN4" (HWMonitor) or "CPU core" (Aida) is also different again , Also, "VIN4" or "CPU Core" drops to 0 or almost 0, depending on the configuration of C-States and Windows power settings. With fixed "cache" voltage (Bios) it behaves with HWMonitor same as with IA, which indicated "LLC / ring" is I think about 0.03 higher, than the adjusted.And then this mysterious "GT offset" it may not be synonymous with a CPU without GFX and I could not influence the way in any way directly.
In addition, I have about 3-4 different CPU temperatures. For example, "Package" in HWMonitor, "CPU" (motherboard sensor?) And "CPU Package" (much higher) in Aida. I've also seen a VcoreRef or something, possibly in HwInfo, in HWMonitor I think that would be "VIN5".
Then to a few unclear settings in the bios, where I found on the internet nowhere a real info about it:
- "PLL Reference Offset Voltage" (main page Ai Tweaker): I have nothing to do with the PLL voltage. Increasing this will increase the Vcore, Vcache, and Vsa, much as if I were entering an offset at each voltage. In addition, it changes so I think the VIN4 in HWMonitor, possibly it is synonymous but VID / IA or everything together and ev. Also the other voltages CPU voltages (at the same set in bios fixed voltages).
- VCCU Voltage Offset ("Tweaker's Paradise"): On the Internet in various forums, there is a mess or mean it with Cache Voltage, Ring Voltage or Uncore Voltage I always believe the same thing. However, I have a cache voltage on the main page (Aida: "CPU Cache", HWMonitor "LLC / Ring") and then in a submenu called VCCU Voltage Offset (Uncore Input Voltage à la CPU Vinput / VCCIN?). However, I notice no changes, neither in behavior nor in readings. In the Intel XTU but there's this setting synonymous.
- PLL Voltage Offset ("Tweaker's Paradise"): I have no settings, no display in the bios and not in any tools where I have a PLL voltage or a change is displayed, as I know it from other threads or my previous CPUs , The impact is minimal.
As I said, I've tried pretty much everything. Different Speeds (OC / Stock), Voltages, LLC, various VRM Settings (Current Limit etc.), C-States on, C-States Off, Fixed Voltages, Offset Voltages, Adaptive, Balanced Energy Plan, and Highest Power "etc. Various drivers. BIOS is the newest. Now reinstall Windows and before that and on different discs. Funny thing is that the offsets in HWMonitor are always the same values and as seen in the pictures sometimes identical in different points identical.
I would have to test it again now with 2 DIMMS in dual-channel. So I could not determine the behavior right away, but had no desire to play around or wait long. Funny thing is that the behavior is independent of the energy-saving options primarily in idle or u. U. occurs in the partial load range or in Game. Under full load with IBT, Prime, Furmark, I seldom have ever seen it. I also tested DIMMS individually and in pairs in different channels.
Still the missing Specs: 2x GTX 1080 and a Corsair AX850 Gold power supply. Otherwise, there is currently a little bit in there or in there. I assume that this is OK, unfortunately have no way to test another. Ev. Not all plugs are so great anymore. Have the day before yesterday or so times exchanged the CPU power cable. That has stabilized the 12V voltage. But because of the slightly better or worse or (in) stable has not. And a bit earlier the GPU cable. The fact is, the power supply makes things easy with FurMark and Intel Burn Test in parallel (CPU and GPU in the SLI OC!) And bores it with up to 650W (socket) or so. With older hardware OC, the part even went beyond 900W easily from the socket, which then came off losses near the 850W output power.
P.S. Take a look at the core voltages on picture 5. Are such big differences normal? Always between 1/2, 3/4 and 5/6 or 1/3/5 and 2/4/6. For AVX loads, the differences are up to 10 °!
Thank you in advance!
First of all: Sorry for bad english! I Google translated it from german (https://www.hardwareluxx.de/communi...pannungs-offsets-eigenmaechtig-1189784.html). Here can also be found mentioned attachements.
Then important: The following remarks relate not only to the overclocked state but also to the factory settings, both CPU and RAM without XMP. A problem with overclocking is therefore excluded. The attached pictures (1-4) have just been created, while I write this article without OC on a freshly installed Windows. At the moment therefore not really many drivers are installed. But I have already tried different driver versions. With / without Windows Update, from Intel and also the sensor driver from ASUS. But changes everything little to nothing at all.
Have recently got an i7-6850K on an ASUS X99-S with 4x 8GB Corsair Vengeance LPX memory with 3200 MHz XMP. In my opinion, I overclocked the CPU quite successfully to 4.4 GHz - with air cooling and compared to other people pretty low voltages (especially SA, Cache and VCCIO). At some point I noticed that the voltage offsets displayed in HWMonitor change in part. Sporadically (or rather regularly) for a very short time, they jump from 0 or the set values to significantly higher - if not more dangerous / deadly high values. Rarely, they occasionally fall off at fixed voltages or manual offsets.
No matter what I do (increase tensions, minimize, change the most unlikely attitudes), nothing will be better. I've been wandering around for weeks now with the most impossible settings the mainboard offers. It has little or no influence on these changing values.
In any case, I have a huge problem with this board anyway to determine voltages and temperatures like I have never had. Each tool displays different things, so different amounts, different (wrong) named, in different places in different categories. And then different values are also implausible (I hope it's hard, there is nowhere on the motherboard a bit beyond 100 ° - these two values jump also tremendously and by feeling I could not find anything hot on the board anywhere). I also hope that the offsets I mentioned are also implausible or wrong. However, I would like to get rid of that - for reassurance and anyway. But unfortunately I could not find another tool that shows me something like that (HwInfo, Aida, Open Hardware Monitor, ...) to compare that.
It already begins with the reading and determine of basic voltages, which I have configured in the bios (Figure 5). What is displayed in all tools under the sensor voltages as "Vcore" or "CPU Vcore" is anyway complete nonsense or another tension. VIN4 in the HWMonitor, I am also not sure if that is really the core voltage. At least it does not match the values of the Intel Xtreme Tuning Utility, but with the core voltage in AIDA. In the Intel XTU it shows me a core voltage, which agrees with the VID in AIDA and HWMonitor. The individual core voltages in the XTU deviate slightly from this (does this CPU have differentiating core voltages? The XTU can not display the VID (s) !?). The VID actually comes from the CPU itself. I change in the Bios but Vcore, so changes in the tools, the VID, so and u. a. Meanwhile, I assume that the "VID" rather corresponds to the core voltage, while "VIN4" would rather be the VID. In addition, set in the BIOS fixed voltages z. B. in HWMonitor always displayed with an offset higher, so the "VID" is correct, I think and then the "IA" is higher by an offset and the "VIN4" (HWMonitor) or "CPU core" (Aida) is also different again , Also, "VIN4" or "CPU Core" drops to 0 or almost 0, depending on the configuration of C-States and Windows power settings. With fixed "cache" voltage (Bios) it behaves with HWMonitor same as with IA, which indicated "LLC / ring" is I think about 0.03 higher, than the adjusted.And then this mysterious "GT offset" it may not be synonymous with a CPU without GFX and I could not influence the way in any way directly.
In addition, I have about 3-4 different CPU temperatures. For example, "Package" in HWMonitor, "CPU" (motherboard sensor?) And "CPU Package" (much higher) in Aida. I've also seen a VcoreRef or something, possibly in HwInfo, in HWMonitor I think that would be "VIN5".
Then to a few unclear settings in the bios, where I found on the internet nowhere a real info about it:
- "PLL Reference Offset Voltage" (main page Ai Tweaker): I have nothing to do with the PLL voltage. Increasing this will increase the Vcore, Vcache, and Vsa, much as if I were entering an offset at each voltage. In addition, it changes so I think the VIN4 in HWMonitor, possibly it is synonymous but VID / IA or everything together and ev. Also the other voltages CPU voltages (at the same set in bios fixed voltages).
- VCCU Voltage Offset ("Tweaker's Paradise"): On the Internet in various forums, there is a mess or mean it with Cache Voltage, Ring Voltage or Uncore Voltage I always believe the same thing. However, I have a cache voltage on the main page (Aida: "CPU Cache", HWMonitor "LLC / Ring") and then in a submenu called VCCU Voltage Offset (Uncore Input Voltage à la CPU Vinput / VCCIN?). However, I notice no changes, neither in behavior nor in readings. In the Intel XTU but there's this setting synonymous.
- PLL Voltage Offset ("Tweaker's Paradise"): I have no settings, no display in the bios and not in any tools where I have a PLL voltage or a change is displayed, as I know it from other threads or my previous CPUs , The impact is minimal.
As I said, I've tried pretty much everything. Different Speeds (OC / Stock), Voltages, LLC, various VRM Settings (Current Limit etc.), C-States on, C-States Off, Fixed Voltages, Offset Voltages, Adaptive, Balanced Energy Plan, and Highest Power "etc. Various drivers. BIOS is the newest. Now reinstall Windows and before that and on different discs. Funny thing is that the offsets in HWMonitor are always the same values and as seen in the pictures sometimes identical in different points identical.
I would have to test it again now with 2 DIMMS in dual-channel. So I could not determine the behavior right away, but had no desire to play around or wait long. Funny thing is that the behavior is independent of the energy-saving options primarily in idle or u. U. occurs in the partial load range or in Game. Under full load with IBT, Prime, Furmark, I seldom have ever seen it. I also tested DIMMS individually and in pairs in different channels.
Still the missing Specs: 2x GTX 1080 and a Corsair AX850 Gold power supply. Otherwise, there is currently a little bit in there or in there. I assume that this is OK, unfortunately have no way to test another. Ev. Not all plugs are so great anymore. Have the day before yesterday or so times exchanged the CPU power cable. That has stabilized the 12V voltage. But because of the slightly better or worse or (in) stable has not. And a bit earlier the GPU cable. The fact is, the power supply makes things easy with FurMark and Intel Burn Test in parallel (CPU and GPU in the SLI OC!) And bores it with up to 650W (socket) or so. With older hardware OC, the part even went beyond 900W easily from the socket, which then came off losses near the 850W output power.
P.S. Take a look at the core voltages on picture 5. Are such big differences normal? Always between 1/2, 3/4 and 5/6 or 1/3/5 and 2/4/6. For AVX loads, the differences are up to 10 °!
Thank you in advance!