I posted previously about this issue and have managed to work around to an acceptable degree but it's less than ideal so I was hoping for some acknowledgement from Asrock that they're aware of the issue and planning a fix.

For reference, my specs are R7 1700 @ 3.9 ~1.35V p-phase OC w/ Noctua NH-D15

Above is a picture of my HWinfo (rearranged for easier viewing- the 'CPU' temperature is normally in one of the motherboard categories).  The problem is that the temperature labeled just 'CPU' lags way behind 'CPU (Tctl/Tdie)', and I have a 1700 so I know it has nothing to do with the offset on the X series chips.

I'm uploading a short video of what happens to the temperatures and will post it when it's done but the tl;dw is this- the 'CPU' reading is what the motherboard uses to control fans (according the fan curve) when set to use CPU temperature in BIOS.  Unfortunately, the 'CPU' temperature sensor (whatever it is, I said previous it was a socket temperature sensor but that was pure speculation) lags way, WAY behind the Tctl/Tdie- so much so that I need to set my fan curve so that both CPU fans are going at 75+% at 50C and 90% at 60C because at that point, my Tctl/Tdie readings are already at ~75C.  This is an even bigger issue since I need to set all my chassis fans on a similarly aggressive curve which makes my 'silent' build not very silent.

tl;dr Let us use Tctl/Tdie to control fans in fan curve, not this mystery 'CPU' temp sensor which is slow to respond and always a few degrees low even when it catches up

And that's one of the reasons it's a poor choice to use as the temperature that fan curves respond to and the reason that every other X370 I know of uses the standard Tctl/Tdie instead.

What's 'expected' is that a company that's been in the business as long as Asrock would fix things like this before release, and failing that, at least acknowledge it nearly 3 months later as something they're looking into.

The motherboard sensor is even *more* useless for fan control, as in my months using this boards I've seen a total range of 39C to 41C.

For reference, I have a Noctua NH-D15 in a Fractal R5, CPU1 and CPU Optional are the 2 fans on the D15, Chassis1 is the front intake fan, Chassis2 is the rear exhaust, and Chassis3 is a side exhaust fan to help keep my graphics card cool (one of my work arounds for the slow sensor- my GPU would heat up because it was going at 100% and my CPU was going at 25%, causing my intake/exhaust fans to spin too slowly for optimal airflow to the GPU.  The side exhaust helps makes sure the GPU is always getting air and keeps rpms lower on my super loud RX 480 Sapphire Nitro + fans).

1700 is 65W cpu overclocked to 155W  with your settings

my vsoc 0.95V

For example, take a gas stove.  Once lit, the flame is immediately as hot as it's going to get.  You control the temperature by how much gas is supplied.  This is essentially equivalent to the Tdie temperature.

Now, put a pan on top of the flame.  The pan isn't immediately hot.  It takes time for the heat to transfer from the flame to the pan.  This is equivalent to the CPU lid and thermal paste.

Once you put water or bacon (mmm...delicious bacon) inside the pan, it yet again takes time for heat to transfer from the flame, through the pan, to the contents.  It's the same path with the CPU: Tdie (or the Zen cores) through the lid through the thermal paste and finally to the heatsink.

It doesn't matter if the fan sensors were tied to the Tdie - the heat simply would not transfer quickly enough to the heatsink where it can be dissipated.  Instead, your fans would blasting away, doing nothing, because no heat has transferred and yet the Tdie would immediately jump 70C under full load and will stay there until the heatsink catches up.  The below image is proof of concept.

http://imgur.com/crG7ZYR" rel="nofollow - http://imgur.com/crG7ZYR

With 16 threads at 100%, it takes 1 minute and 23 seconds of sustained heat transfer to increase the socket temp 12C. It takes a total of 3 minutes and 45 seconds to reach maximum heat transfer.  In fact, the heatsink and fans don't effectively reduce the CPU or Tdie temp until after the app has been running for 8 minutes.

I have my CPU fan set to go to 80% at 45C (100% at 70C).  All my testing shows this to be about the sweet spot for my CPU heatsink and fan.  It's set this way primarily for noise control.  As you can see, it maintains a very good temp under full load.  My CPU doesn't usually go above 45C while gaming.  All my case fans go to 100% at 55C.

The issue isn't what sensor ASRock ties the fans to.  The issue is how quickly can you transfer heat from your CPU to your heatsink.

Think of it this way- you're driving a car whose speedometer lags behind by 5+ seconds and shows you going 15 mph slower than you actually are.  You're trying to avoid speeding but you're not actually sure how fast you're going because your speedometer is inaccurate AND slow- maybe you accelerated a bit past the limit but you're not sure because your speedometer hasn't caught up yet.

The only solution then is to drive extra slowly and start slowing down extra early (since I only know how fast I was going 5 or more seconds ago, and not exactly even then).

"I have my CPU fan set to go to 80% at 45C (100% at 70C).  All my testing shows this to be about the sweet spot for my CPU heatsink and fan.  It's set this way primarily for noise control.  As you can see, it maintains a very good temp under full load.  My CPU doesn't usually go above 45C while gaming.  All my case fans go to 100% at 55C."

I understand how CPU cooling works perfectly fine, and my system overclocked not 'out of spec'.  This is 4th time people incorrectly explained things to me that I already knew rather than addressing the actual problem.  The problem is that Asrock is using a socket sensor (if it is a socket sensor) instead of Tdie.  It's not solvable on my end- it's possible to work around it but it's definitely a problem.

There is a user that asked the support team, and they said it's normal. So I'm leaning towards they won't fix the issue.