[SOLVED] 9700k CPU temp instantly jumps to 100 C when under 100% load

May 28, 2021
21
0
10
Cooler: H115i RGB pro XT
Ram: 32GB 3000Mhz CL15
Motherboard: ASUS Maximus Hero X Wifi
GPU:6800 XT
9700k is Brand new. (8700k died last week)


Temps:
View: https://imgur.com/a/dcbjUI6


Everything is at stock settings except XMP profile for RAM is selected and ASUS Multicore enhancement is off.

The reason I wanted to turn it off is because I don't want ASUS to overclock my CPU. I want to use it at Intel's specs. It turns out Intel thinks (?) it's OK to increase the voltage to 1.45v for a split second when I start the PRIME95 Small Ffts test. Immediately, temps hit 100C and voltage goes down to 1.04V and it downclocks to 3.6Ghz.

Idles 38C at 1.04v
Under stress test (it downclocks to 3.6Ghz immediately during the test): 58C at 1.04v

Is this normal? I don't want anyone to overclock my CPU. I just want to run it at a safe temp. I am not sure if my cooler pump is working properly.
 
Solution
There's a weird 1.435 VID reading there which is too high and would cause it to hit 100c, especially on prime. Maybe there's one of these filthy auto overclock features enabled, did you mess with your bios? Some motherboards even have a physical switch for overclock like oc genie in some music boards, anyways, just to make sure use hwmonitor 64 instead if you can to double check the voltages, ranges, if it's really going as crazy high as 1.435 then you gotta fix that, Intel turbo boost by itself would never go that high.
There's a weird 1.435 VID reading there which is too high and would cause it to hit 100c, especially on prime. Maybe there's one of these filthy auto overclock features enabled, did you mess with your bios? Some motherboards even have a physical switch for overclock like oc genie in some music boards, anyways, just to make sure use hwmonitor 64 instead if you can to double check the voltages, ranges, if it's really going as crazy high as 1.435 then you gotta fix that, Intel turbo boost by itself would never go that high.
 
Solution
May 28, 2021
21
0
10
There's a weird 1.435 VID reading there which is too high and would cause it to hit 100c, especially on prime. Maybe there's one of these filthy auto overclock features enabled, did you mess with your bios? Some motherboards even have a physical switch for overclock like oc genie in some music boards, anyways, just to make sure use hwmonitor 64 instead if you can to double check the voltages, ranges, if it's really going as crazy high as 1.435 then you gotta fix that, Intel turbo boost by itself would never go that high.

I messed a lot with the BIOS. But after my 8700k died last week I am done with overclocking. I loaded the optimized default in the BIOS thinking it set everything to stock settings. Then I made the 2 changes I mentioned above. I am not aware of any buttons on the motherboard.

Can XMP profile change anything other than RAM settings?

Is 100C normal for 1.45v? If not, can there be an issue with my cooler pump?
 
you didn't list your pump or fan speeds, assume you've adjusted the fan curve!?
Radiator fans connected to the CPU or CPU/OPT header?
Pump connected to either SYSFAN or CPU/OPT header @ close to 100%?

it downclocks due to the high temp, maybe your fans aren't reacting immediately to CPU activity (leads me to suspect they're monitoring SYS temps?)
 
May 28, 2021
21
0
10
you didn't list your pump or fan speeds, assume you've adjusted the fan curve!?
Radiator fans connected to the CPU or CPU/OPT header?
Pump connected to either SYSFAN or CPU/OPT header @ close to 100%?

I run the fans at 100% just to make sure it's not the issue. Pump speed is set at extreme in the iCue software. Fans connected to the motherboard and controlled by ASUS fan software. Pump is connected to AIO_PUMP header on the motherboard.
 
Is the 9700K properly identified as the installed CPU when powered on?

Was BIOS ever hard reset (clear CMOS) after the CPU swap, and, what is the current BIOS version?

What clock speed is shown in HWMonitor during CPU-Z/bench/stress CPU? (4.6 GHz should be maintained for 60-90 seconds for Intel spec, and, to go longer without dropping clock speeds, turbo duration limits and power limits might need to be removed...; if 100C is being hit, I'd suspect the full 4.9 GHz is more likely. )
 
When you set the BIOS to default settings, your CPU will end up getting way more voltage than what it needs to be stable. With Intel's 8 and 10 core CPUs, all this does is it creates unnecessary heat.

Your CPU is probably getting 100 mV to 125 mV more voltage than what it needs. Here is an example of a 10th Gen CPU at 4800 MHz while running Cinebench.

70gfbmj.png


When checking your voltage, do not look at the VID numbers. Look at Vcore. That is the voltage going to your CPU. If your voltage is too high, reduce it. Go into the BIOS and set a negative voltage offset. Here is an example of -75 mV.

n2S74Db.png


I am using the same cooler as you are using and temperatures are not a problem at all. Did you replace the thermal paste when you installed the cooler? What did you use and how did you apply it?

Here are the Intel specs from the datasheet for an 8 core CPU with a 95W TDP rating.

ZLNa81Y.png


PL1 = 95W
PL2 = 95W + 25% = 118.75W

Your screenshot shows that power consumption went over 205W. That is well beyond what Intel originally intended. If you want your CPU to run at Intel spec, go into the BIOS and lower the PL2 turbo power limit. This will significantly reduce full load performance. That is why most motherboards bend the rules. For the 9900K, some motherboards set PL2 to 210W. Not sure what yours is set to but it might be somewhere around there. You can use HWiNFO to confirm this.

Another problem is that for most motherboards, the BIOS does not give you access to the Clamp options within the power limit register. These are necessary to truly clamp power consumption to the default values. Here is an example with PL2 set and clamped to 120W. Even during Prime95 Small FFTs, the CPU throttles and slows down as much as necessary to keep power consumption at this level.

cTxPlb9.png


I know Prime95 Small FFTs used to be a popular stress test. For Intel's 8 and 10 core CPUs, it is an unrealistic test. No other test generates that much power consumption so it is definitely overkill. If you set your CPU voltage and power limits appropriately, you will be able to run any test without your CPU having a nuclear meltdown. Do not rely on the default BIOS settings to do this. They all like to cheat.
 
  • Like
Reactions: CompuTronix
May 28, 2021
21
0
10
Was BIOS ever hard reset (clear CMOS) after the CPU swap, and, what is the current BIOS version?
Did you replace the thermal paste when you installed the cooler?

Thanks for the detailed posts. I really appreciate it. I just tested some of the things you guys said. Can you comment on my results?

CPU is identified as 9700k in bios. Current BIOS version is 2503. I just cleared the CMOS and started testing from scratch.

I removed the CPU cooler and replaced the paste. I apply the paste right in the middle of the CPU according to the directions of the most popular tech channels. It spreads the compound evenly to the CPU surface.
I use MX-4. Fans and cooler pump set to max during all tests.

Below voltage values are Vcore values I see in CPU-Z.

AFTER CMOS RESET and Re-pasting Cpu;

Test 1- STOCK: Idles 31C - 0.1v to 1.2V random
small ffts: 93C 1.37v - downclocks instantly to 3.5Ghz 1.02v 55C

Test 2- XMP ONLY: Idles 31C - 0.7v to 1.2v random
small ffts 1.31v 4.6 GHz 92C max

Test 3- XMP and ASUS MCE disabled: Idles 31C 0.8 - 1.2v random
small ffts 1.45v - instant 99C - downclocks to 3.5Ghz

Test 4- 4.8Ghz "Overclock": AVX:0 - Sync all cores 48 - MCE disabled - Vcore set to 1.29v
Idles at 35C. 1.28v
small ffts - 1.28V 85C 4.8Ghz.
CinebenchR20 score 3791 - 1.28v 67C 4.8Ghz

Test 5- 5Ghz Overclock: AVX:0 - Sync all cores 50 - MCE disabled - Vcore set to 1.33v
Idles at 36C 1.328v
small ffts - Fatal error. Unstable. Probably I should increase voltage.
CinebenchR20 score 3948- 1.328v 71C 5.0Ghz


Looks like to set the voltages manually is the best way to keep the temps below 80C without loosing significant performance. This whole thing with me tempering with BIOS and everything started 2 months ago when I noticed CPU (my old 8700k) temp jumps to 90s during gaming. I just want to keep it below 80. I don't want to loose another CPU to this. My 8700k died last week. I want to make sure this Cooler is working properly, which was replaced by Corsair 2 months ago.

If the PRIME95 small ffts temps at 99C is expected from 9700k than I am happy to keep this sytem as is and hope for the best while gaming.