H100i V2 coolant temp

Trying to make sense of this.

I recently changed Mobo and CPU to an i7 7600k and reinstalled the H100i V2. The CPU is idling around 42C though and under load hits 74. No indication of thermal throttling, which apparently should happen at 80, and nothing above that. But the coolant temp reported doesn't go up more than a few degrees above room temperature. (Room is 22, coolant is 26.5)

For due diligence, I checked on the odd spiking that I see with the 7600k and found a lot of people talking about it. I also noted that when load drops, the CPU temp drops instantly (under one sample cycle). A lot is indicating that there is active and substantial cooling going on, but the coolant temp oddity and the high CPU-reported temps in just about every software out there is creepy. Combine that with numerous reports of the CPU 's TIM being badly applied by Intel and it creates a mess of confusion. "Just delid your CPU and apply a new TIM"... Riiiiiiight. O.o

The temp reports are also uninformative in many ways, with the multi-core reports showing only one core hitting insane temps, and the mobo sensors staying down around 32-40. But the temp sensors are clearly labeled "#1 #2 #3" and such and that doesn't say much about where they are.

So I've got temps looking like a yoyo but the coolant never heats up. Any insight? Normal? 7700k is badly-TIMmed? Squirrels rampaged?

Are you positive on the 42C idle? Is that a package temp, CPU temp (avg), or something else? It should bounce around the 26.5 coolant temp, although it might run higher if you are set up with C-states disabled and/or a fixed manual voltage.

As for the 7600/7700 generally, it does seem this one is going to be limited at the CPU by voltage, like most recent Intel models. There are some detailed Kaby reviews out there and I am sure you've read several. This review is a little different and is a cooler showdown on a 7700K overclocked to 4.8/1.25 and 5.0/1.30v. Unfortunately, the graphs do no give you the coolant temp delta and not all coolers have the ability to measure that. However, what is of note are the three sets of results for the H115i and H100iv2 across their standard Performance/Balanced/Quiet profiles. As you can see, the total CPU temp delta is only 1C different from max fan speed to quiet. This strongly suggests that removing heat from the water stream is not the limiting factor on these CPUs and the heat transmission through the CPU/cold plate will be... and thus all the delidding talk. The OCCT Linpack AVX test used in this comparison is quite brutal, so hopefully your CPU temps are not this high. Also, don't get to concerned about momentary temp spikes compared to sustained averages. Most days my highest recorded CPU core temp is going be recorded while opening a program or some other strangely written command. You won't wear out your CPU that way. Try monitoring with something like AIDA that keeps a rolling line graph of individual core temperatures. It may shed some light on actual behavior.

If you want to count 3-10% load as "idle" I suppose. AIDA is out of the "want to spend this much" bracket.

So, going between things:

BIOS read: 40-42

CL4: Shows, as described, "Temp #1", #2, and #3, and calls them all "MB Temp". #1 Stays at 33 "idle", max 44; #2 stays at 34 "idle", max 46; #3 is almost like remapping CPU load to a temp between 35 and 74, fluctuating regularly between 39 and 43 as load fluctuates between 4 and 12%. One thing I just noticed though is that CL4 is reporting VCPU at 1.86 stable.

Speedfan 4.52:

... Appears to be Mostly Useless. Claims to be reading two ACPI temps and all four physical cores of the CPU, but it's listing the physical cores as 19, 20, 21, and 27 respectively. ACPI 1 and 2 are showing at 28 and 30, which are more believable than "CPU Core temps below room temp".

Intel ETU:

This is kind enough to claim that the dramatically-fluctuating temperature is the package temperature. Core temps fluctuating relatively quickly between 32 and 47 at minimal load. Interesting that the core voltage is also fluctuating between 1.2900 and 1.3990V, averaging 1.3462. The fact that people are getting a 5.0 overclock at 1.3 indicates that if these voltage readings are accurate, they're pretty high, even with the CPU's adaptive voltage mode running.

I'm going to poke some voltages live in ETU and see what happens. If you don't see me again, you know I screwed up.

I can't make sense out of the CL4 readings. Maybe someone else can. VCPU in Link is your Input Voltage. 1.86 is a fairly normal number, but I don't know enough about the tendencies on Gigabyte Z270's to predict LLC and Input voltage change frequency.

So are you at stock frequency with the full AUTO voltage? If so, that is something that is easy to address. Even if you are not going to overclock, pick an adaptive voltage and set that as a defined number. It should cut down on all the voltage overshoot common to the auto voltage on the Lake series CPUs and settle the large fluctuations down as well.

You should be able to configure XTU to show all 4 cores in line graph form. It's not as handy as AIDA, but it will do for the short time period until you don't care to look at this stuff anymore.

Thank you for the input. This is getting even more interesting.

I snagged HWiNFO64 (v5.44-3070) and it gives much more reasonable information:

Package temp of 38 (ranging from 32 - 71); individual cores fluctuating wildly from 33-51 with 0 and 3 hitting the highest in general, but nothing going about 72; the core voltage reported by XTU seems to primarily be the CPU VID, while board vcore is indicating 1.380 average (1.322 to 1.428)

On XTU I went from voltage auto/adaptive to 1.300 static and that dropped XTU's temp readings, but increased the VID readings. I can definitely say things are more complex than a decade ago. Moreso since not everything seems to know what it's doing software-wise.

When I see odd values swinging all over the board, I want to make sure things are stable before trying to change anything clock-wise.

Everything though is showing that despite the base frequency being 4.2 and turbo being 4.5, the cores are running at a consistent 4.5 100% of the time and only the cache and ring is running at 4.2.

The CPUID SDK is too old so CL4 fails to report Kaby Lake CPUs as it should, further I suspect many utilities will not report the motherboard sensors as well as they should. If you post the initial + Menu->Hardware->Sensor Debug screen shots from my SIV utility I could tell far more. CL4 did rather poorly on my Gigabyte Z270-HD3P system 'till I contacted the CPUID SDK developers, see http://forum.corsair.com/forums/showthread.php?p=891269.

Does CL 4.5.0.55 report all four of your LED DIMMs? It fails to report mine and I have only seen posts saying the same even though it' supposed to fix LED DIMM support. It would be good if you posted the CL 4.5.0.55 [Home] tab screen shot. SIV can report and/or control the LED DIMMs, see http://forum.corsair.com/forums/showthread.php?p=888742.

Interesting discoveries so far...

Re-pasted and seated the cooler when a prime95 mixed test reached 101 despite thermal throttling. -> No benefit. Still heavily-fluctuating CPU temps for both the package and cores, as well as for the mobo sensor. Coolant temp finally got up to 42 when CPU reached 95+. CPU load dropping cools it down within a poll cycle, but this is not-good regardless.

Found out that the mobo bios will not allow me to change the CPU vcore voltage from auto of 1.300. Changes in XTU also do not take effect.

Bios also will not allow manual control of the CPU multiplier. The CPU is running at 4.5 constantly, even when Turbo is disabled. Uncore/Rng is running at 4.2 at least, but the fact that I can't actually make any changes in the BIOS that actually do anything is disturbing. Even Bios shows it running constantly at 4.5.

HWiNFO seems to be the least unreliable at reading the data. It's still giving me annoying news though heat-wise.

red-ray: CL 4.3 showed "one" DIMM LED listing, 4.5 shows none.

SIV shots attached, CL4 shot skipped because it only involves the LED DIMMs that broke between 4.3 and 4.5

1. Thank you for the SIV screen shots, SIV looks to reported all the sensors correctly. Are the names that SIV uses for the temperatures the same as the names in the BIOS? I suspect they are on the Q-Fan setup screen. I asked for [sensor Debug] so I could adjust things if I needed to and for general use [status] (Menu->Hardware->Hardware Status) should be used.
    
   
2. Which power scheme do you have selected? I suspect performance and if you change to balanced the CPU speed will drop. It's also worth checking you have the latest BIOS as I needed to update the BIOS on my Z270-HD3P.
    
   
3. I suspect SIV reports the same as HWiNFO as Martin and I worked together on the Gigabyte IT8686 reporting. It's not just the LEDs I was interested to see, but also what motherboard sensors CL4 reports. I wonder if CL 4.5.0.55 ever reports LED DIMMs, Corsair seem to have messed up even more than usual with this release.
    
   
4. Does SIV report the LED DIMMs? You need to use the command SIV64X -AIOLED for it to do this, see http://forum.corsair.com/forums/showthread.php?p=888742. Assuming so please post [Link Status] and either way [sMB Bus] so I can see the DIMM LED firmware revision. I am also wondering if your board has PWM power controllers so please post [PMB Bus] so I can check.
    
   
5. I recall posts that prime95 pushes things to much for Slylake/Kaby Lake. You could try Menu->Machine->CPU Stress and see how the temps look with that.

BIOS doesn't give them very informative names. "System Temp", "CPU Temp" is what BIOS shows. F3 is the most current release BIOS with F4a being in alpha (Sorry, but BIOS is one of those things that I'm not keen on alpha versions of, what with $1,000 of hardware on the line and "At your own risk" involved in the big print).

There is no sign of a "Power Scheme" or anything similar in BIOS. and no notes in TFM that indicate why certain options claim they should be changeable but just won't bring up a menu to do so.

CL4 4.3 reported one LED DIMM and could change it back and forth between "Default" and "Steady. As to board sensors, CL4 home attached for your edification.

SIV Does. Link Status and PMB and SMB included as well because reasons.

CPU Stress on SIV is definitely less impolite. Cores shot up instantly to about 72-75 with a max at 78, package stayed 72-75 with a max of 78 as well. Coolant started to slowly climb from 26.2 and at 15 has reached a whopping 29.2. And I just noticed... Why the {censored} has the clock gone to 4.6? Interesting that VID actually drops under load though. O.o It's getting up as high as 1.417 at lower load and down to 1.270 at full load, which is an inverse of what it normally should be per the CPU's internal adaptive regulator.

So I'm still in a place where package and core temps are hitting 75 and coolant is staying stable below 30, which is the source of my worry in this thread.

Why the {censored} has the clock gone to 4.6?

Thank you for all the screen shots.

I also noticed the x46/4.60GHz and assumed you had intentionally set this overclock up in the BIOS. I wonder if Menu->Hardware->CPU Detail->CPU Power Levels or ...->CPU OC Data will provide and insight. Every other 7700K I have seen report as attached with the maximum being x45. I don't overclock my CPUs so I am not sure what to setup and/or what all the voltages should be.

It looks like there aren't any PWM power controllers and I can see you have the original revision 0x03 rather than the later 0xC0 LED DIMMs. SIV should be able to control all 4 via the [Link LEDs] panel.

My Z270-HD3P reports all the temperatures on the Smart Fan 5 screen, see attached.

There is no sign of a "Power Scheme" or anything similar in BIOS.

You need to select the Power Scheme in Windows via the power options in the control panel. From SIV do Menu->Control Panel->Power.