Arctic Freezer II 360 problems

Associate
Joined
14 Dec 2016
Posts
14
Location
London
Hi, I have an Artic Freezer II mounted on a 13900FK + Gigabyte z690 Gaming X and the CPU is throttling with intensive use without any OC. P-Cores goes to 90/98C and performance drops to 5Ghz, whereas E-core stays a 4.3Ghz at 88/90C. In gaming it does not pass from 65C. Is there anyone with same situation than me? If so, will a custom loop fix the issue or it will be the same situation?

Thanks!
 
Associate
OP
Joined
14 Dec 2016
Posts
14
Location
London
Hello, thanks for the answer. Everything is in auto, enhance multi core is disabled, profile is the default one and memory are selected in XPM profile ( ddr4 3200 c14 -> Samsung B-Die).

By intensive use I mean mainly some forensics tools (regex, entropy calculations and Yara on ~120GB file size from memory dumps), compiling some toolings, containers shenanigans, etc. On windows yes, I have tested with Prime95 only to verify what I was getting in Linux.
 
Associate
Joined
7 Apr 2004
Posts
1,180
Location
Belfast
I use an Asus board and also use the same cooler. Try setting the llc to l4, dc_ll to 1.02 and ac_ll to 0.2. At stock min max at the low 80's. Today i changed my ac_ll to 0.19 and cb23 now maxes at 77.
Gigabyte settings might be slightly different, im not sure but try these to see how it goes.
 

mof

mof

Associate
Joined
19 Jul 2011
Posts
284
There was a problem with this cooler. You can check if yours was a affected on the artic site by typing in your serial number.
Could it be to do with that?
 
Associate
OP
Joined
14 Dec 2016
Posts
14
Location
London
Thank you for the answers! I will rather not touch bios and leave it by default, at least for now. Cooler is not affected by the chemical reaction. I have been testing multiple configuration and got some improvements and conclusion, writing here in case someone else have the same issues as me:

System:
  • 13900KF
  • Gigabyte X690 Gaming X
  • TG 4x8GB Samsung B-Die 3200 mhz CL 14 with XMP profile
  • Seasonic Prime Platinum 1000W
  • Pali Nvidia RTX 3080 TI Gaming Pro
  • Lian Li 011 Dynamic

Configuration:
I worked with two configurations. One installing the radiator in the top, another in the horizontal front part of the case:

Configuration 1 (original one):
  • In: 2 Noctua 12 cm NF-P12 in the bottom
  • In: 3 Artic fans 12 cm (original from the cooler) in the front horizontal part
  • Out: Artic Freezer installed in the top of the case with 3 Noctua NF-A12
Configuration 2 (new set up):
  • In: 2 Noctua 12 cm NF-P12 in the bottom
  • In: Artic Freezer installed in the front horizontal part with 3 Noctua NF-A12 (tubes in the bottom of the case, not in the top to avoid air to the pump!)
  • Out: 3 Artic fans 12 cm (original from the cooler) in the top part of the case

Tests:
  • Throttling test via Cinebench R23 + 4 hours prime 95 + 30 sec break + cinebench again: with conf1 1 I got a cinebench score drop from 39.7K to 37.8K (~2K points), whereas with conf2 went from 39.7K to 39.6K (~0.1K)
  • Prime 95 4 hours with In-place large FFT: with conf1 I had multiple P-Cores touching 100C before throttling, average speed in P-Cores were between 5.0/5.1 Ghz and E-Cores 4.0/4.1 Ghz. With conf2 P-Cores where in the range of 5.4/5.5Ghz and E-Cores 4.2/4.3 Ghz. With conf1 P-Cores were touching 95/100C and E-Cores 90/93C, with conf2 P-Cores were between 90/95C and E-Cores 82/87C
  • High temperature (only tested with conf2): system took around 5 minutes to arrive close by to the temperatures described before. It started with E-Cores at 70/75C and P-Cores at 80/85C having one P-Core touching 5.8Ghz
  • Concerning: the temperature in the room was between 14 and 15C. I turned on the radiator and temp in the room went to 22C observing a CPU temperature increased of ~3C approx. For hot weather with temperatures +30C you might be at the risk of throttling if all cores are in use
  • Aida and Core Temp were showing an initial CPU consume of 350W and 335W, after ~1 hour approx it dropped to 253W (I do not have enhanced core activated, so surprise seeing more than 253W)
  • Not tested: dual fan in the radiator (6 fans, 3 front and 3 back) might improve temperatures

Conclusion / TLDR:
  1. Ensure the fans of the radiator takes the air from outside the case, and not from inside (avoid hot air from graphic card)
  2. You can not overclock with this cooler if multiple E-Cores are in use, at least in the 13900
  3. Hot weather will be an issue
  4. Custom made watercooling should (not tested) improve vs Artic as far as you are getting multiple radiators as initial temperatures were around 80C
 
Associate
Joined
5 Mar 2017
Posts
2,259
Location
Cambridge
Even custom loop, if left on auto, and the motherboard uses too much voltage on auto, will be thermal throttling.
Previous experience using ASUS, they always used too much voltage, but that was AMD for years and some older Intel before Ryzen. I was able to use up to -0.15 offset for a 3900x and performance was the same but temperature was way better.
 
Associate
OP
Joined
14 Dec 2016
Posts
14
Location
London
100%, I have played with the auto options to understand why they allowed the cpu to go to 350W and it was because the enhanced multi core was auto. Moved to "Disabled" and the CPU does not pass from ~250W having all the P-cores 5.5Ghz and the E-cores at 4.3Ghz; and the one of the P-Cores still touching the 5.8Ghz if not all cores in use. And for temperatures, I see a decrease on ~3/5C average.
 

Deleted member 258511

D

Deleted member 258511

The problem is that you are leaving it stock. Even with simple LLC tuning you will bring the temps in line.
 
Back
Top Bottom