Question Things have changed. What is your (over)clocking method for your 13/14 then Intel CPU?

[Hulk]

This question is most likely aimed at users of 13-14th gen Intel processors, which are more constrained by power and thermals than the limits of the cpu in question. Things have changed with regards to overclocking. When you had one core and it was relatively easy to cool you'd just pick a voltage you felt was safe, find the max frequency and be done with it.

Now we have to deal with as many as 24 cores and with out custom loops and possibly chillers there is no way to run a 14900K at max frequencies, I think you have to create a plan to accommodate your cooling solution. Actually even if you do have the cooling to run a 14900K full out it's not the most efficient plan of action. Anyway...

As an example I have a 13600K currently. Different applications load the cpu differently as we all know. If I'm mixing down a song in Presonus Studio One the application will use most of the threads but it won't hit them that hard. I know this because temps/power remain under control. it's only pull 90W package power. So for this app I could run 5.4GHz all cores and 4.2GHz on the E's and temps and power would be fine.

But if I were to then encode some video in Handbrake with these settings I'd get a fail/lock up.

So the question is how to set up the system so that application that hit the cores lightly and can run them to high clocks with low power will be allowed to do this while limiting frequency/temps/power for applications that saturate the cores?

After a bit of tweaking I have my system set for a PL1/2 max of 125W. The P cores are limited to 5.4 globally and then cores 2 through 5 are limited to 5.1GHz. These are the ones that get really hot when heavily loaded. I have the E's set to 4GHz.

Set in this manner when running Studio One I'll use about 90W and Studio One will put the most work on the cores running 5.4GHz. But when I run a Handbrake transcode it'll it the power limit and do about 5GHz on the P's and 4GHz on the E's at 125W.

I realize I could push the power up but honestly the performance gained by adding 50W is minimal but the heat and noise are not.

Perhaps if I had a better cooler.

Just curious how others with 13 and 14th gen Intel CPU's have them set up?

 

[H433x0n]

Not to toot my own horn but I feel like I’ve got this down pretty good. I’m fairly confident the following things will work pretty well for Intel CPUs regardless of bin.

YMMV may vary:

Set an undervolt of -0.02 for P-Cores and -0.01 for E-Cores. There is usually enough guardband where this tiny undervolt will always work. If you have a higher end motherboard with more granular control -0.03/-0.02 will probably work. This is setup this way since if you try to do a global undervolt, you’ll notice e-cores crashing before p-cores and this decoupling will allow more finely tuned control. If you plan to overclock, then undervolting is off the table.

Do not rely on a global PL1/PL2 to limit power. Use IccMax to limit power. This alone will probably fix your issue. To do this you’ll want to go in and disable unlimited IccMax (this is usually set to 512A by default). IIRC, the default IccMax for your CPU is 200A. You can move this up or down to meet your preferences. By utilizing this value instead of a PL2, it will dynamically underclock more effectively and probably will let you run those higher frequencies for the lighter apps and limit current (therefore clocks) for the heavier avx type tasks.

 

[Khato]

Interesting approaches. I hadn't given any thought to doing more than constraining the i9-14900k to operate within a sane 150W PL1/PL2 that a Noctua NH-D15 can easily handle in near silence. Sounds like I should try out some alternate approaches to see what difference it makes.

 

[Hulk]

Not to toot my own horn but I feel like I’ve got this down pretty good. I’m fairly confident the following things will work pretty well for Intel CPUs regardless of bin.

YMMV may vary:

Set an undervolt of -0.02 for P-Cores and -0.01 for E-Cores. There is usually enough guardband where this tiny undervolt will always work. If you have a higher end motherboard with more granular control -0.03/-0.02 will probably work. This is setup this way since if you try to do a global undervolt, you’ll notice e-cores crashing before p-cores and this decoupling will allow more finely tuned control. If you plan to overclock, then undervolting is off the table.

Do not rely on a global PL1/PL2 to limit power. Use IccMax to limit power. This alone will probably fix your issue. To do this you’ll want to go in and disable unlimited IccMax (this is usually set to 512A by default). IIRC, the default IccMax for your CPU is 200A. You can move this up or down to meet your preferences. By utilizing this value instead of a PL2, it will dynamically underclock more effectively and probably will let you run those higher frequencies for the lighter apps and limit current (therefore clocks) for the heavier avx type tasks.

Thanks, I'm going to try this. Funny thing is I thought I was stable but I just had a lock up. I was running Topaz Video ai while compressing a video in Vegas Pro while doing some CorelDraw work. Boom! Lock up.
Need a new strategy.

 

[TheELF]

So the question is how to set up the system so that application that hit the cores lightly and can run them to high clocks with low power will be allowed to do this while limiting frequency/temps/power for applications that saturate the cores?

That's why intel has released intel extreme tuning utility like 100 years ago.
You set up your bios the way you want your system to run most of the time, either low power or high power, and then you set individual profiles for apps that you want to run differently from the default.

 

[Hulk]

That's why intel has released intel extreme tuning utility like 100 years ago.
You set up your bios the way you want your system to run most of the time, either low power or high power, and then you set individual profiles for apps that you want to run differently from the default.

Thanks for letting me know! Better 100 years late than never!

I've always avoided manufacturer "tuning utilities" because back like 100 years ago they were always bloatware garbage and I always ended up in the bios anyway. I guess things have changed in the last 100 years.

 

[TheELF]

Thanks for letting me know! Better 100 years late than never!

I've always avoided manufacturer "tuning utilities" because back like 100 years ago they were always bloatware garbage and I always ended up in the bios anyway. I guess things have changed in the last 100 years.

Yeah, unless you find a mobo that has a bios with dozens of profile slots for different apps the only way to do this is with software.

 

[Hulk]

All joking aside I do appreciate the advice.

 

[Hulk]

Not to toot my own horn but I feel like I’ve got this down pretty good. I’m fairly confident the following things will work pretty well for Intel CPUs regardless of bin.

YMMV may vary:

Set an undervolt of -0.02 for P-Cores and -0.01 for E-Cores. There is usually enough guardband where this tiny undervolt will always work. If you have a higher end motherboard with more granular control -0.03/-0.02 will probably work. This is setup this way since if you try to do a global undervolt, you’ll notice e-cores crashing before p-cores and this decoupling will allow more finely tuned control. If you plan to overclock, then undervolting is off the table.

Do not rely on a global PL1/PL2 to limit power. Use IccMax to limit power. This alone will probably fix your issue. To do this you’ll want to go in and disable unlimited IccMax (this is usually set to 512A by default). IIRC, the default IccMax for your CPU is 200A. You can move this up or down to meet your preferences. By utilizing this value instead of a PL2, it will dynamically underclock more effectively and probably will let you run those higher frequencies for the lighter apps and limit current (therefore clocks) for the heavier avx type tasks.

I've been thinking about this a bit and while I don't doubt you are correct here I'm having trouble wrapping my mind around how the motherboard/processor responds differently to a max current vs. a max power? Since they are both related through the voltage wouldn't this essentially limit the cpu in the same manner?

Let me try an example to see if I can work out why myself. Let's assume 1.288V at 5.1GHz, after Vdroop it'd probably be about 1.22 or so. 1.288 x 200 = 257.6W, which is more than the 181W max turbo Intel specifies for the 13600K. Backing into 181W would give us a voltage of 1.105V.

Anyway, I'm still not understanding why limiting current is more effective than limiting voltage and I'd like to learn why?

 

[aigomorla]

To be honest, i don't think you will gain much from overclocking like you once did in the past.
This is because of boosting profiles the CPU's now have.

Unless your the type to constantly load 100% of your cpu all the time, or need a constant ghz across all cores, its better off to let it handle boosting as needed, and eco mode when not.

This also applies to AMD CPU's as well.

 

[Hulk]

To be honest, i don't think you will gain much from overclocking like you once did in the past.
This is because of boosting profiles the CPU's now have.

Unless your the type to constantly load 100% of your cpu all the time, or need a constant ghz across all cores, its better off to let it handle boosting as needed, and eco mode when not.

This also applies to AMD CPU's as well.

From a practicality point of view you are definitely correct. It can be a time vampire.