My Experience Undervolting 13900k with Strix Z790-E

[weidemanTV]

I wanted to do a post about my experience undervolting my 13900k with the ASUS Strix Z790-E MB. I could not find much help online and had to piece together much different information about this, so hopefully this helps. I will be as detailed as possible and will include my entire build list. I am by no means a professional, so use this as a guideline to help you, not perfect advice. To do the actually undervolting, people were using Intel XTU to do it real time. I was having problems with this, so this is just how I did it in the ASUS Strix ROG BIOS.  I will be putting the BIOS settings needed in all caps and red lettering below so that you can easily find them. I really hope that this helps someone. 

I use CinebenchR23 to do my testing. 

My PC Build:

Case- Lian LI O11 Dynamic Evo

GPU- Gigabyte Oc Gaming 4090

CPU- I9-13900k

Cooler- DeepCool LS720 360MM AIO 

MB- ASUS Strix Z790-E

PSU- Corsair Shift series RM1200x Gold

Ram- Team Group Delta 32GB DDR5 7000

Fans- I Have 6 Lian Li SL120 Intake Fans, 3 on the side, 3 on the bottom (these are awesome fans btw they eliminate a lot of cable management). For exhaust I have the 3 fans on the Deepcool720 on top and one more Lian Li Uni 120 fan in the back. 

With that being said, I will get into the statistics. I use HWMonitor. I have heard people say use something different, but I like its ease of use and its easy for me to read. I play a lot of Escape From Tarkov. Before I started undervolting my pc would idle around 40-50 c. When I was gaming it would be anywhere between 60-70c with 75c spikes. I wanted to get these numbers down a little bit, especially after doing my first benchmark.

When I first did the CineBench 10-minute test, my CPU temps would get up to 100c and then start throttling. I was getting mid 300s on package wattage like 330-350. Everything I read said that these CPUs run hot and its fine, and really without undervolting the only way for me personally to get these down was to get custom loop cooling, which I do not want to do at this time. 

I was able to at least get the max temps down all the way to 88c just by going into my bios under AI TWEAKER, and changing the ASUS MULTICORE ENHANCEMENT setting to DISABLED - ENFORCE ALL LIMITS. This was already a huge step just by the one thing, but I wanted to bring it down even lower. This is where the undervolting comes into play. On a side note, the 13900k had MAX SHORT DURATION and LONG DURATION  PACKET POWER limit stock at 253 W. I changed the LONG DURATION max to 125 W. This may not be necessary, but it was something that I picked out of one of the videos I was watching that said it could help with temps. This can be done under AI TWEAKER - INTERNAL CPU POWER MANAGEMENT. 

Now for the undervolting. The ONLY setting that I messed with for this is the GLOBAL CORE SVID VOLTAGE. It will need to be put on adaptive mode. The OFFSET MODE SIGN should be negative (-). Then I started tweaking the offset voltage. I would input a value, which I started at .08, but many recommend starting with .05, then I would start my computer, run a CINEBENCH test with the minimum test duration off, and log my scores, and temps. Like I said I started with -.08. Here were my scores:

Stock Voltage :

• CINEBENCH score of 37884
• Max VID Voltage was 1.442V
• Max Package temps were 97c
• Max package power was 313W

-.08 Voltage Offset:

• CINEBENCH score of 37879 
• Max VID Voltage was 1.353V
• Max Package temps were 84c
• Max Package Power was 253.67W 

I wanted to see how much more I could undervolt while being stable, and I read people using -.10 with no issue. So I pushed it to -.12. My computer started crashing before startup showing me blue screen and I could not get it going, so I restarted it, and set it back to -.11. This is where I found some stability. 

-.11 Voltage Offset:

• CINEBENCH score of 35780
• MAX VID Voltage was 1.280V
• Max Package temps were 83c
• Max Package Power was 251.7W

I did a 30-minute stability test and it did not crash. My games run good and I am happy where I am at.

I backed it up to -.10. Here are the results:

• CINEBENCH score of 37429
• MAX VID Voltage was 1.352V
• Max Package temps were 83c
• Max Package Power was 253.57 

CONCLUSION: -.10 seems to be the sweet spot for this setup with my cooling situation. I don't fully understand the CINEBENCH score, but from what I do understand is that it is your performance score, so the higher the better. I sacrificed a tiny bit of performance according to the score if any performance at all, but I was able to get my temps down a lot as you can see. When I am idling, they stay between 30-40 c with little spikes up when something is loading, or CPU is in use. When gaming I am getting 55-65c and when stress testing it, well you can see the difference. Thank you for taking the time to read this and I hope it is informative and can help someone else with a similar setup. 

Stay COOL yall

[Silent_Scone]

Thanks for the detailed info, always welcome.

Did you try simply adjusting SVID Behaviour to Best Case?

[weidemanTV]

I did not do that. Have you done it?

[Silent_Scone]

You can try it. Doing so manipulates the VID the CPU requests according to load, rather than applying a global offset.

[weidemanTV]

Thank you I am playing with all that trying to find best fit. Honestly I want to delid and direct dye custom loop this thing. But I keep hearing about ek’s problem but I like how user friendly their stuff is. I have looked into supercool and Iceman but don’t have enough info on those to convince me yet. I talked to ek and they said they fixed their stuff but I’m going to wait to hear from other peoples results. 

[weidemanTV]

I have checked it out and I do not have a best case mode. There is adaptive auto and manual. 

[Silent_Scone]

I would recommend the Thermal Grizzly contact frame at least, but the EK Direct Die has had some engineering issues, yes. The replacement frames seem to be much better having seen some user feedback. Delidding is needed if wanting to extract the most out of these chips, majority are very thermally constricted.

[weidemanTV]

so dude ya im trying to get my temps as coool as possible. I reallly want to get the ekwb kit I built on their customization part of their site. I am going to include GPU and CPU. and I really want to de lid and direct die too. My AIO cooler does a good job for what it is and for much power everything is drawing. I have been trying to find updated reviews on this direct die kit but I can't find anything newer than a few months ago when they were having issues. Another thing I am worried about is that I will spend the 1300 bucks and get basically the same temps. The setup would be a two-radiator setup. I got the itch though I want to "tinker" with it. Another thing I was thinking was getting two pumps and two reservoirs and having basically one system for the GPU and one for CPU. But I dont think my case would support that. That type of setup makes me feel in over my head. I like having a map so to speak with instructions like using the EKWB configuration. 

[Silent_Scone]

1. My understanding is EKWB were replacing the first batch of frames under warranty. The replacements should be machined correctly. I’m unaware of issues on later batches

2. Dual loop may not be beneficial. I personally don’t recommend this configuration as much anymore due to lack of multi gpu systems.

If not doing so already, you may wish to consider looking at conserving power by restricting your 4090’s TDP. The ideal efficiency window is around 70-80% Power Target.

[weidemanTV]

cool thank you. I went ahead and decided to just get one tank one pump, 2 x360 rads and a gpu waterblock and cpu direct die block and everything else i will need to run this loop. I will be delidding. i plan on  I'm excited to mess with this. thank you for advice and help! i will def make a post when all said and done. its all ordered. EKWB.