One unit shows poor thermal performance at 12900KS.
One unit shows poor thermal performance at 12900KS.
Yes the chip does throw out a lot of heat, but under stress testing with CineBench R23 P-Core 5 hits 100C while the other cores are mid 80s to low 90s. The chip is limited to 4.7GHz in CineBench R23 because of current limits. I've tried varying the current down from 500A to 400A and 300A, each step down limits both the P- and E-cores further, so I just keep it at 500A. I've tried downgrading P-Core 5's multiplier from 52x (5.2GHz) down to 42x (4.2GHz). (The 12900KS has two P-cores at 55x and 6 at 52x. E-cores are at 40x.) This downgrade barely works, it bounces between 98C & 100C. I've started an RMA but I want to be sure. I've been using XTU to change CPU multipliers and monitor under load. The stress test built into XTU makes the cores limit to about 4.9GHz, and while P-core 5 is the worst it's only by about 2C-3C, and it only gets to about 91C. Cyberpunk allows the CPU to go to 5.2GHz. My cooler is a AIO Thermalright Frozen Warframe 240, I have lots of experience with PC's but this is my first time watercooling. The 12900KS is notorious for heat and a 360 or 420 won't fully tame the beast AFAIK. Thought's good people? I've attached Intel XTU monitoring with P-Core 5 set to 52x and then 42x.
Are you using a contact frame? Just to confirm, haven't you removed the lid yet? Likely it's a thermal paste problem. If you ran it without the lid, an uneven amount of paste would be more noticeable, but with the IHS on, heat distributes evenly across the die. Still unsure—maybe the paste inside the IHS isn't applied uniformly. You might want to reapply it. Also, if you're getting a 12900KS, investing in a good 360 mm AIO would be wise.
I'm still figuring this out. I'll look again in about 24 hours to see the paste. My setup only works with a 240mm, it responds at 5.2GHz when background tasks are minimal, which is decent, but CineBench R23 really highlights some issues.
The issue was resolved by removing the block, which allowed good coverage except for a channel in the middle. That channel had a thin layer of paste, but I added more. P-Core 5 now operates just 2°C above the other cores, reaching around 94°C at peak during testing with CineBench R23. The stress test on Intel XTU pushed P-Core 5 to 90°C, so reapplying it and reinstalling fixed the problem.
The cores seem to be handling the same workload. It’s possible Windows is shifting all the stress onto a single core. Check if you’re using the newest chipset drivers and UEFI settings.
With the newest Intel Chipset drivers and UEFI installed, your evaluation makes sense (unless there are unusual mounting or bracket problems).