Last night, I finally did a little session under LN2 with the i9-13900K coupled with our ROG Maximus Z690 APEX. We chose this motherboard while waiting for the new ROG Maximus Z790 APEX to arrive. It had been a while since I had taken the time to do some tests with LN2 and I missed it. First unpretentious session to see how our i9-13900K (SP 99) behaves and the performances we can expect with it under cold conditions.
For this session, in addition to the i9-13900K and the ROG Maximus Z690 APEX (BIOS 2004), we used the Teamgroup Delta RGB 7200 MT/s CL34 memory kit, the LN2 T-Rex gold bucket from KingPiN and the KingPiN Inferno. The isolation was done mostly with charcoal gum and paper towels.
In order to monitor the temperature of the bucket, we will use the KTH K-type thermometer as well as a multimeter to monitor the voltages.
The tests under LN2:
Initially, we focused on the performance of the processor. A next session will be devoted to memory, even if we know that A-die chips don’t like the cold. Once again, thanks to the company Flandres Azote which, despite the current context, still continues the deliveries.
Here we go, we pour the first liters of nitrogen and we make sure that the i9-13900K stays cool, proof of a good contact with the bucket. Then, the temperature is lowered in steps to detect any CB or CBB. As a reminder, the Cold Bug (CB) is a concern related to cold, when the processor is too cold, the OS can freeze and force you to restart the configuration. The Cold Boot Bug is the difficulty of the card to boot when the temperature is very low.
So, I’m not able to bencher full pot. That is, if I fill the bucket to -190°C, I have a Cold Bug but I can easily reach -185°C. You just have to be careful not to exceed them at the risk of having to reboot the motherboard. As far as the CCB is concerned, no real concern since the board accepts to start at -175°C so there is no need to warm up the bucket too much in case of crash.
The results with our i9-13900K (SP 99):
I knew it, we are not on a golden. The overclockers talk about a “golden” processor when it is very efficient either in ambient or under LN2. For this first session, I managed to reach with the 8 P-Core 7000 MHz and 5500 MHz for the 16 E-Core (Atom). On the other hand, a small bug for the uncore frequency which is always displayed at 4500 MHz although it should be 5000 MHz. I will try to solve this problem with the next LN2 session.
When I disable the E-Core, I was able to gain a little in P-Core frequency since it went up to 7100 MHz. Regarding the voltage, I limited myself here to 1.60 volts (LoadLine calibration 7), which corresponds to a real voltage of 1.52 volts on the multimeter.
The scores aren’t too bad but I still hope to get close to 7200 MHz for the next session. I’ll also have to focus on the memory which was just activated in XMP here. Here are the findings I was able to make during the tests:
- My i9-13900K suffers like a majority of 13900K it seems from a Cold Bug and thus no full pot.
- The motherboard behaved very well and rebooted without difficulty from 175°C.
- I have a bug in the display of the uncore frequency.
- No condensation under the socket thanks to the KingPin Inferno but under the PCB next to the memory slots. I had some stability problems at the end of the session. I will have to isolate this part better.
- My i9-13900K is not a golden:)