Overclocking 13900K

[Emmitt Jay DeLong II]

I know this probably gets tiring but I could use references.

Just finished my new PC, specs follow. I want to try overclocking. I hit up Google and YouTube with a handful of permutations of "overclocking beginners guide" and ended up with a headache. I tried the forums here and found a few suggestions but most were intermediate to advanced levels of steps. If anyone could point me to some decent guides, I would be appreciative. This will be MY first overclocking. I have allowed other builds and MB's to AI Overclock before but I want to learn it and since I'm dense I'll need a 099 course.

CPU: I9-13900K
MB: ASUS TUF Z790 WiFi (DDR5 varient)
RAM: Triton 32GB (2x16) 6000MHz
GPU: NONE - Saving money for a 4090. Currently running off the iGPU but I'm not running games atm. New PC and my old one is still in service during build.
COOLING: iCUE H150i Elite Capellix AIO, 6 x Noctua NF-S12B Redux-1200 PWM High Performance Cooling Fan
CASE: Lian Li O11 Dynamic (Designed for water cooling I know but I was directed to it via disability services as being a good case for me to build in given its design and openess.)

[Somerandomtechyboi]

Beginner oc guides can fuck off, if you follow the tuning methodology those use it will take absolutely forever to tune your system

 

I assume you are just ocing your 13900k so i would suggest seeing what other ppl with 13900k oc settings

 

Heres a reference from an article but reddit posts or forum posts are also fine

 

Another reference from a skatterbencher article aka an actual overclocking guide that isnt a complete joke

 

Using the reference from those articles i would suggest setting static vcore to 1.35v, no llc, vccia/ecore volt = 1.25/1.3v, a multiplier of 55 on all pcores, alongside a multiplier of 40 on all e cores

 

For further tuning you can try raising multis and voltages till you hit 100c tjmax or voltage rollover (more volt negatively impacts stability, so 5.5ghz is stable at 1.3v 100c but not stable at 1.35v 108c as an example)

[RONOTHAN##]

So the first thing to point out is that the 13900K has very little headroom in it, at least at ambient cooling. It's rated for 5.5GHz all core on the P cores, and the most a lot of 13900Ks can do is 5.6GHz fully stable, at least without doing things like the E cores and Hyperthreading which by that point you should've saved your money and gotten either a 13600K or 13700K, both of which do actually have some headroom that can be noticeable. While the chip is overclockable technically, in practice it's not worthwhile to overclock it outside of competitive benchmarking and you're better off undervolting.

 

If you still want to do CPU overclocking, if for nothing else but the fun of it, here is a rough guide for it, at least the method I've always done for 13th gen. The main thing to know about overclocking is that it's a giant balancing act between temps, voltage, and clock speeds. More voltage means it's more stable at higher clocks, but it also means it's got higher temperatures. Depending on how temperature sensitive the specific CPU is, raising the voltage beyond a certain point can cause it to stop scaling and you actually get less stable with higher voltage, though admittedly on 13th gen they don't hit that point until you've already gotten into unsafe temps. The general steps for how to do this:

1. Run some sort of benchmark to find out the initial performance. A single pass of Cinebench is usually pretty good for this. 
2. Go into the BIOS and set some initial settings.
   1. First set LLC, pretty sure on ASUS boards it's in the Digi+ VRM menu, and set it to something like Level 3 LLC. This is to help get better voltage regulation at full load at the expense of some higher idle voltage.
   2. Next, manually enter ratios for the P cores, E cores and Ring, I would do 55/42/45 for them respectively as that should be the stock values.
   3. Next, go into the Advanced CPU settings and disable all the power management stuff, so max out the power limits and disable things like C-States. This is to keep the limits from getting in the way of anything you do now, you can always turn these back on later. 
   4. Finally, set the VCore to manual/override mode and set the value to 1.4V. This is just a starting point and you will change it later. 
3. Save the settings and go into Windows. Run the benchmark you did initially to make sure that it's not incredibly unstable and make sure performance is actually increasing. 
4. Run a stress test of your choice for ~5 minutes to make sure it's not super unstable. A few common picks are Linpack Xtreme, OCCT, and Prime95 Small FFTs, all with their pros and cons. 
5. During the stress test, monitor frequency and temperature to make sure that the CPU isn't overheating and dropping clocks. HWInfo64 is the preferred tool for this. You want to make sure the temperatures stay below 100C. 
6. Once the stress test completes/crashes, do one of the following:
   1. If it's stable and temps are fine, increase the P core multiplier. 
   2. If it's not stable but temps are fine, increase voltage by 0.01V. 
   3. If it's not stable and temps are at the limit, decrease the P core multiplier. 
7. Repeat from Step 3 until you feel like you've balanced temps, voltage, and clock speed correctly. 
8. Once you feel like you've balanced the P core clocks correctly, repeat from Step 3 with the E core multiplier instead. 
   1. The advanced way is to also mess with the E Core L2 Cache voltage, this can sometimes be used to help stabilize higher E core clocks, though personally I've never had it help much so I wouldn't bother. 
9. Once you feel the E cores and P cores are at their limits, repeat from step 3 with the Ring clock instead. 
10. After all 3 are balanced for the 5 minute quick test, start decreasing V Core to find the lower limit with all of them. This part isn't necessary, but it can help temps a bit.
11. Run a long stress test (I.E. 2+ hours) to make sure it's fully stable. Make sure to monitor frequency and temperatures during so, and follow the substeps in step 6 for dealing with it in case it crashes. 
12. Re-enable power management stuff if you want to. 

That should be all you need to do to get a CPU overclock dialed in. If you want to do an undervolt (usually the easier and better option), there's two methods for that, either the static voltage mode or the offset voltage mode. Offset voltage is a bit easier though can result in some odd issues depending on the board, while static mode is basically manually overclocking except you never raise the clock speeds, only lower the voltage. Here's the steps for offset voltage:

1. Go into BIOS and set voltage mode to offset, and set the offset to negative. 
2. Increase the negative offset by 10mV. 
3. Go into Windows and run a stress test of your choice for 5 minutes. 
   1. If it passes, go into BIOS and repeat step 2. 
   2. If it crashes, go into BIOS and increase the negative offset by 10mV. 
4. Go into Windows and run a stress test of your choice for 2+ hours. 
   1. If it crashes, go into BIOS and decrease the negative offset by 10mV. 

Once it passes, you're done the undervolt. 

 

[Emmitt Jay DeLong II]

4 hours ago, RONOTHAN## said:

So the first thing to point out is that the 13900K has very little headroom in it, at least at ambient cooling. It's rated for 5.5GHz all core on the P cores, and the most a lot of 13900Ks can do is 5.6GHz fully stable, at least without doing things like the E cores and Hyperthreading which by that point you should've saved your money and gotten either a 13600K or 13700K, both of which do actually have some headroom that can be noticeable. While the chip is overclockable technically, in practice it's not worthwhile to overclock it outside of competitive benchmarking and you're better off undervolting.

 

If you still want to do CPU overclocking, if for nothing else but the fun of it, here is a rough guide for it, at least the method I've always done for 13th gen. The main thing to know about overclocking is that it's a giant balancing act between temps, voltage, and clock speeds. More voltage means it's more stable at higher clocks, but it also means it's got higher temperatures. Depending on how temperature sensitive the specific CPU is, raising the voltage beyond a certain point can cause it to stop scaling and you actually get less stable with higher voltage, though admittedly on 13th gen they don't hit that point until you've already gotten into unsafe temps. The general steps for how to do this:

1. Run some sort of benchmark to find out the initial performance. A single pass of Cinebench is usually pretty good for this. 
2. Go into the BIOS and set some initial settings.
   1. First set LLC, pretty sure on ASUS boards it's in the Digi+ VRM menu, and set it to something like Level 3 LLC. This is to help get better voltage regulation at full load at the expense of some higher idle voltage.
   2. Next, manually enter ratios for the P cores, E cores and Ring, I would do 55/42/45 for them respectively as that should be the stock values.
   3. Next, go into the Advanced CPU settings and disable all the power management stuff, so max out the power limits and disable things like C-States. This is to keep the limits from getting in the way of anything you do now, you can always turn these back on later. 
   4. Finally, set the VCore to manual/override mode and set the value to 1.4V. This is just a starting point and you will change it later. 
3. Save the settings and go into Windows. Run the benchmark you did initially to make sure that it's not incredibly unstable and make sure performance is actually increasing. 
4. Run a stress test of your choice for ~5 minutes to make sure it's not super unstable. A few common picks are Linpack Xtreme, OCCT, and Prime95 Small FFTs, all with their pros and cons. 
5. During the stress test, monitor frequency and temperature to make sure that the CPU isn't overheating and dropping clocks. HWInfo64 is the preferred tool for this. You want to make sure the temperatures stay below 100C. 
6. Once the stress test completes/crashes, do one of the following:
   1. If it's stable and temps are fine, increase the P core multiplier. 
   2. If it's not stable but temps are fine, increase voltage by 0.01V. 
   3. If it's not stable and temps are at the limit, decrease the P core multiplier. 
7. Repeat from Step 3 until you feel like you've balanced temps, voltage, and clock speed correctly. 
8. Once you feel like you've balanced the P core clocks correctly, repeat from Step 3 with the E core multiplier instead. 
   1. The advanced way is to also mess with the E Core L2 Cache voltage, this can sometimes be used to help stabilize higher E core clocks, though personally I've never had it help much so I wouldn't bother. 
9. Once you feel the E cores and P cores are at their limits, repeat from step 3 with the Ring clock instead. 
10. After all 3 are balanced for the 5 minute quick test, start decreasing V Core to find the lower limit with all of them. This part isn't necessary, but it can help temps a bit.
11. Run a long stress test (I.E. 2+ hours) to make sure it's fully stable. Make sure to monitor frequency and temperatures during so, and follow the substeps in step 6 for dealing with it in case it crashes. 
12. Re-enable power management stuff if you want to. 

That should be all you need to do to get a CPU overclock dialed in. If you want to do an undervolt (usually the easier and better option), there's two methods for that, either the static voltage mode or the offset voltage mode. Offset voltage is a bit easier though can result in some odd issues depending on the board, while static mode is basically manually overclocking except you never raise the clock speeds, only lower the voltage. Here's the steps for offset voltage:

1. Go into BIOS and set voltage mode to offset, and set the offset to negative. 
2. Increase the negative offset by 10mV. 
3. Go into Windows and run a stress test of your choice for 5 minutes. 
   1. If it passes, go into BIOS and repeat step 2. 
   2. If it crashes, go into BIOS and increase the negative offset by 10mV. 
4. Go into Windows and run a stress test of your choice for 2+ hours. 
   1. If it crashes, go into BIOS and increase the negative offset by 10mV. 

Once it passes, you're done the undervolt. 

 

So, I'm not planning to compete at all, lol. I had read that the 13900K had fairly good overclock potential so I decided to try it out this time for the education and experience of it. IF I understand what you said at the top, the overclock info was either wrong or I misread it.

The steps provided are very deep and robust. I REALLY thank you for it. If I was to want to look into more info for these steps, like what the ratios on the cores are actually doing and such, where would I go?

The undervolt thing, is NOT the first time I'm reading it. What benefit does it actually provide? I plan to just play games, all be it at max settings and on a better monitor then right now, but still, just planning on gaming. How will undervolting benefit/impact my gaming performance?

Last thing, should I wait to do this until after I get the GPU. As I said, I'm currently running of the iGPU for the chip so I know there’s extra performance sitting there. I assume it would be better to.

[RONOTHAN##]

7 minutes ago, Emmitt Jay DeLong II said:

IF I understand what you said at the top, the overclock info was either wrong or I misread it.

Yeah, there just isn't a ton of headroom on those chips. They're so thermally limited that unless you're doing something like chilled water or some other sort of sub-ambient cooling there's just not enough headroom to get anything useful out of it. They have headroom with memory overclocking, and that can in some instances give pretty big performance uplifts, but just not with actual core overclocking, and memory overclocking is extremely tedious and difficult. Plus you're on a board that's not really good at memory overclocking, so you won't get too far when doing so. 

 

8 minutes ago, Emmitt Jay DeLong II said:

If I was to want to look into more info for these steps, like what the ratios on the cores are actually doing and such, where would I go?

There are tons of different places that exist that cover some of this stuff, Actually Hardcore Overclocking and Skatterbencher are the two that come to mind immediately, plus Intel does tend to release documentation on how all the settings for their CPUs work (good luck finding it though, it's scattered throughout their website and kind of difficult to find a lot of the time). What the actual ratios do though is determine the actual clock speeds. There are two factors that determine the clock speed of a particular thing, the base clock and the ratio. The base clock by default is 100MHz, so when you take a say 50x ratio, that would mean that the particular core would be running at 50x100MHz, AKA 5000MHz, AKA 5.0GHz. The other way of overclocking would be to just manually change that base clock, so say a 40x ratio with a 125MHz base clock would result in a 5GHz overclock as well, it's just that the base clock affects more than just one particular clock speed, it affect things like memory as well, so it's not recommended to do for beginners as you have to balance a ton of different variables at the same time. 

 

When I talked about ratios, I referred to three different ones: P core, E core, and Ring. 13th gen CPUs are of a hybrid architecture, so they have performance (P) cores for single threaded tasks, and they have a ton of smaller efficiency (E) cores. They both clock to different frequencies, hence the different ratios. The Ring is the interconnect between all of the cores, the memory controller, and the PCIe devices. In more IO heavy workloads like games, this can become the bottleneck, though in my experience this is also what takes the most voltage to get clocked high. Most chips can do 4.8GHz ring, with some of the best chips being able to do 5.2GHz ring up from the stock 4.5GHz. 

 

23 minutes ago, Emmitt Jay DeLong II said:

What benefit does it actually provide? I plan to just play games, all be it at max settings and on a better monitor then right now, but still, just planning on gaming. How will undervolting benefit/impact my gaming performance?

The main benefit from undervolting is lower temps and power draw. In all core workloads where you're bound to have all your cores active, the chip hits 100C, and starts throttling/riding the power limits, undervolting the chip will allow it to boost higher and get a bit more performance (usually not a ton, but measurably more). If you're just gaming, you'll see the CPU be at say 65C rather than 75C, so a noticeable drop in temps and a lower power bill, but it'll still be at the same clock speeds no matter what so no performance impact. 

 

26 minutes ago, Emmitt Jay DeLong II said:

Last thing, should I wait to do this until after I get the GPU.

I would wait. The iGPU causes some really weird behavior when trying to overclock a lot of the time, more so with memory overclocking than core overclocking, but still when doing core overclocking. Unless you need the iGPU for Quicksync or something, it's generally considered best practice to disable the iGPU when overclocking, though if you're just undervolting it's not likely to matter as the issues don't usually show up until you're at the limits of the system. 

[Emmitt Jay DeLong II]

16 hours ago, RONOTHAN## said:

Yeah, there just isn't a ton of headroom on those chips. They're so thermally limited that unless you're doing something like chilled water or some other sort of sub-ambient cooling there's just not enough headroom to get anything useful out of it. They have headroom with memory overclocking, and that can in some instances give pretty big performance uplifts, but just not with actual core overclocking, and memory overclocking is extremely tedious and difficult. Plus you're on a board that's not really good at memory overclocking, so you won't get too far when doing so. 

 

There are tons of different places that exist that cover some of this stuff, Actually Hardcore Overclocking and Skatterbencher are the two that come to mind immediately, plus Intel does tend to release documentation on how all the settings for their CPUs work (good luck finding it though, it's scattered throughout their website and kind of difficult to find a lot of the time). What the actual ratios do though is determine the actual clock speeds. There are two factors that determine the clock speed of a particular thing, the base clock and the ratio. The base clock by default is 100MHz, so when you take a say 50x ratio, that would mean that the particular core would be running at 50x100MHz, AKA 5000MHz, AKA 5.0GHz. The other way of overclocking would be to just manually change that base clock, so say a 40x ratio with a 125MHz base clock would result in a 5GHz overclock as well, it's just that the base clock affects more than just one particular clock speed, it affect things like memory as well, so it's not recommended to do for beginners as you have to balance a ton of different variables at the same time. 

 

When I talked about ratios, I referred to three different ones: P core, E core, and Ring. 13th gen CPUs are of a hybrid architecture, so they have performance (P) cores for single threaded tasks, and they have a ton of smaller efficiency (E) cores. They both clock to different frequencies, hence the different ratios. The Ring is the interconnect between all of the cores, the memory controller, and the PCIe devices. In more IO heavy workloads like games, this can become the bottleneck, though in my experience this is also what takes the most voltage to get clocked high. Most chips can do 4.8GHz ring, with some of the best chips being able to do 5.2GHz ring up from the stock 4.5GHz. 

 

The main benefit from undervolting is lower temps and power draw. In all core workloads where you're bound to have all your cores active, the chip hits 100C, and starts throttling/riding the power limits, undervolting the chip will allow it to boost higher and get a bit more performance (usually not a ton, but measurably more). If you're just gaming, you'll see the CPU be at say 65C rather than 75C, so a noticeable drop in temps and a lower power bill, but it'll still be at the same clock speeds no matter what so no performance impact. 

 

I would wait. The iGPU causes some really weird behavior when trying to overclock a lot of the time, more so with memory overclocking than core overclocking, but still when doing core overclocking. Unless you need the iGPU for Quicksync or something, it's generally considered best practice to disable the iGPU when overclocking, though if you're just undervolting it's not likely to matter as the issues don't usually show up until you're at the limits of the system. 

Once again, linustechtips forums comes through. Thank you all for the information. I had looked at one of those sites but the information I skimmed was thick and I couldn't absorb it well. What you said makes sense so perhaps I should have another look.

Since I'm both board and chip limited, I'll just apply the factory setting. I had hoped the AiO cooler could help with overclocking but I guess not.

Give the info provided, I'm a little concerned that my thermal setup may not be enough. I do have all 9 fans in the case right now and the CPU temps have been fairly steady 40 to 70. But I don't have the GPU in, of course, and since that will increase ambient temps, should I look into another case? My current setup is the AiO on to, exhausting, three on the sideas intake, and, three on the bottom as intake. If this is going to be an issue, please let me know so I can adjust the build.

Thanks again all.

[RONOTHAN##]

28 minutes ago, Emmitt Jay DeLong II said:

I had hoped the AiO cooler could help with overclocking but I guess not.

Nope, the most it does is keep it from thermal throttling. To have a shot at overclocking a 13900K, you really need to do custom water cooling, and even then you won't really get that far. Technically you could get a bit farther by disabling the E cores, they don't really help with gaming performance and you might be able to get 5.8GHz stable all core, but that's maybe a 5% bump in performance with a pretty significant drop in multicore performance, and at that point you probably should've just gone 7800X3D or 13700K. 

 

32 minutes ago, Emmitt Jay DeLong II said:

Give the info provided, I'm a little concerned that my thermal setup may not be enough. I do have all 9 fans in the case right now and the CPU temps have been fairly steady 40 to 70. But I don't have the GPU in, of course, and since that will increase ambient temps, should I look into another case? My current setup is the AiO on to, exhausting, three on the sideas intake, and, three on the bottom as intake. If this is going to be an issue, please let me know so I can adjust the build.

For a gaming load, it's fine. A GPU will increase ambient temps inside the case but b6 about 5-10C. You'll still be well within safe temperature range for the CPU, so nothing to worry about.