Rocket Lake; Core Temp Mins 0

ezflma

Member
Hi, I have an i7-11700K (running stock) on a Gigabyte Z490 Vision D motherboard with the latest BIOS (F20d). I've noticed that v7.03-4450 during Sensor Status will occasionally / eventually report the individual Core minimum temps to 0c, which also affects the Core Distance to TJMax to then read 100c.

Resetting the timers and values will trigger the max temps to appear again. Using the same board and BIOS version, but with an i7-10700K, did not exhibit this behavior. I also tested the Rocket Lake chip on 6.43-4385 and had the same behavior.
 

Attachments

  • 0c-core-min-11700k.jpg
    0c-core-min-11700k.jpg
    65.2 KB · Views: 28
This might be an occasional glitch in sensor readout, how hard to say exactly until I can see the raw data captured in the HWiNFO Debug File.
 
It's strange, but that's really what the CPU reports. It's most likely some bug, perhaps some BIOS update will fix this.
 
Thanks for taking the time to respond; there’s a “new” bios posted on tweaktown but they haven’t posted to Gigabyte’s site yet. When it does, I’ll flash the board, retest and respond back. Thanks again.
 
I did upgrade to the new F20 (non-beta) BIOS that Gigabyte posted last night. Continue to see 0c. I'm on microcode 0x34, which is older, but I'm at the mercy of Gigabyte's BIOS releases. Apparently 0x39 and 0x3c are newer, but Gigabyte hasn't released these microcode BIOS updates for their Z490 boards.
 
For what it's worth, I returned the i7-11700K for an i9-10850K -- no issues. Either it was a BIOS issue, or given the issues I've had with that RKL chip, perhaps the CPU's thermal sensors were faulty.
 
could this 0 reporting have anything to do with C states cutting power to the core which then report 0 degrees ? I think so.
 
So this question; what prevents you from adding a routing that ignores only the 0 values and keeps the positive and negative (LN2-users ) values?
 
And what if the value is really 0? I think this should be fixed by Intel as the CPU returns invalid data.
 
a value of really 0 does not really occur, and even if it would the next best number reported would still be only -1 or +1.
 
You can define your own invalid value range in sensor settings, so you can set core temperatures to Invalid if: < 1
 
You can define your own invalid value range in sensor settings, so you can set core temperatures to Invalid if: < 1
Morning, everyone.. I’m glad to see that there’s been additional convo and exploring this since my first post. Wanted to share a couple things:

I returned the i7-11700K and up/downgraded to an i9-10850K using the same Gigabyte board and never had the 0c issue.

Recently, I upgraded to a MSI Z690 board and i9-12900K (retail), and the 0c issue shows up on this chip as well, but only on the P-Cores. E-Cores were fine. While annoying, glad to see it’s harmless.

Did disabling C-States resolve the issue? I know I can try it; I have an open ticket with Intel. I have no desire to return or RMA the chip… just wanted to bring it to their attention (again?).

Good idea about setting an invalid sensor range. Love it. Thanks.
 
Good info @ezflma, this points to MSI most likely being the cause for this 0 reporting in my opinion going by what i hear. This could be caused by the microcode version being used in the mobo bios.
I know that for my MSI Z590 (tomahawk-wifi) edition the microcode that is in their latest bios V23 is at version 44, while the latest microcode from intel is already 4 versions further at microcode 50.
This could make a difference and maybe Intel has solved this 0 reporting in newer mc versions then the 44 which comes with this MSI board.
Could others post their versions of microcode used so maybe we can get an idea whether it might be related to the actual microcode being used or if it would maybe be something other that MSI does.
Could maybe even have to do with the intel management version being used by some Motherboard makers including MSI. My motherboard is using Intel Management Engine 15.0.23.1777 (the bios part, not the windows driver which has a complete different version number and is of no real importance to the problem). It could also be that MSI does it's own polling of the core-temps in such a way that too many requests from different monitors causes the sensor to not respond to every call from monitoring programs...
 
@Martin and @mnemonic -- revisiting this. I used your suggestion about C-States. Sure as heck, after DISABLING C-States in the MSI bios, voila ... no 0c readings for the 1.5hr of mixed use (idle, gaming, etc.). I went back and changed C-States to "AUTO" [default setting] as well as C1E Support to DISABLED [default setting], and a 0c popped up after about 2-3 mins of the system booting up and loading what it needed to. Screen captures attached. Again, only the P-Cores are affected.

I think you hit it on the head. I updated my Intel service inquiry. Not sure if it's MSI's issue or Intel microcode; I suspect latter. If I still had my Gigabyte Z490 and i7-11700K, I'd do the same test.

If Intel responds back acknowledging the issue, I'll respond back.
 

Attachments

  • cstates-auto-0c.jpg
    cstates-auto-0c.jpg
    65.3 KB · Views: 5
  • cstates-disabled-no issue.jpg
    cstates-disabled-no issue.jpg
    77 KB · Views: 5
This is clearly an Intel issue, all Rocket Lake and Alder Lake CPUs seem to be affect. I notified Intel about this several months ago, they are very aware, but didn't seem to fix it yet.
 
Back
Top