• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

help overclock i9 14900k

Joined
Jan 5, 2021
Messages
76 (0.05/day)
hello everyone i have

mobo asus z690 hero bios 2802

i9 14900k sp98 p106 e84 mc80

cooler nh-d15

i see a lot of people with sp like mine overclock the cpu to 5.7 with 1.16-1.8v

mine need 1.21-1.22 llc6 and bios prediction show 1.19 and still the cpu not stable

does i have bad cpu? does it motherbord problem? or mybe the cooler?

thanks for the help
 
I can tell you right now your not getting to 5.7 Ghz all- core (P) with a air cooler. The CPU will be pulling over 320 watts.

I would just leave It alone and let the turbo do its thing.
 
your cooler is nowhere near good enough.
if you want to overclock a 14900k (which is only a thing for tasks like games) you need a high end custom loop.
rather undervolt with an adaptive offset to get higher all core boost clocks.
 
You need a 360mm AIO minimum for that CPU before you even think about overclocking.
 
my cpu does 5.5p 44e on 1.10v load llc3 stable all tasks

i saw with the same cpu sp98 people do 5.7p 44e on 1.17-1.18v pulls on cinbench 278waat

when i try 1.18v the cpu not stable in r23 i need more volt

if i will get beter cooler can my cpu be stable in low voltage?
 
my cpu does 5.5p 44e on 1.10v load llc3 stable all tasks

i saw with the same cpu sp98 people do 5.7p 44e on 1.17-1.18v pulls on cinbench 278waat

when i try 1.18v the cpu not stable in r23 i need more volt

if i will get beter cooler can my cpu be stable in low voltage?

Yes, but not really - if you're at the point where you're unstable at temps below tjmax, then you're unstable.
 
CPU's never seem to be fast enough... :D
 
if i will get beter cooler can my cpu be stable in low voltage?
It's possible, a CPU running at 90C might behave differently than one running at 80C. But don't expect much, the point of the extra cooling is to avoid throttling, that's more important. Also better cooling means you can use higher voltages until temperatures become a problem.
 
Last edited:
Overclocking is like a bad compulsion. Not every chip is the same. You can try to push it but only if you have the right tools and parts to keep it cool and under control.

Then you get a result you wanted and realize there is nothing is left to do.
 
difference at 4k between 5.7 and 5.5 is actually nothing.

majority of the gaming speed comes from ram tuning.
 
difference at 4k between 5.7 and 5.5 is actually nothing.

majority of the gaming speed comes from ram tuning.
I wish there was more test data available for this (memory scaling and tuning differences). Guru3D's test is the only one I've seen and they show no difference at 4k and negligible difference at 1440p. Benchmarks, sure...but in games?...I haven't seen any proof of this outside claims from people who charge customers for tuning to get "more FPS" in games like CoD, CS:GO, and Apex. I'm not saying you're wrong, but I am saying I would like to see real tangible evidence of this from credible sources.
 
I wish there was more test data available for this (memory scaling and tuning differences). Guru3D's test is the only one I've seen and they show no difference at 4k and negligible difference at 1440p. Benchmarks, sure...but in games?...I haven't seen any proof of this outside claims from people who charge customers for tuning to get "more FPS" in games like CoD, CS:GO, and Apex. I'm not saying you're wrong, but I am saying I would like to see real tangible evidence of this from credible sources.
Apex has a 300 FPS engine limit which can easily be reached with a 11800H and a RTX 3080M at competitive settings (1080p). So tuning both CPU and GPU wouldn't help much at all with this game on higher-end hardware like what OP has.

@tomer20vt You would need an AIO or custom water-cooling to get better results. If the CPU sees that it isn't near its thermal (not TDP, but really heat) limit, it will continue to push higher.

I have tested this with a i7-12700K and i5-13500 with a stock Intel cooler (Laminar RM1), a NH-D12L, a Thermalright FC140 and an EVGA CLC 280mm AIO. The CPUs were able to get better, more consistent clockspeed with the AIO (I saw around 65C max) compared to the three air coolers. The FC140 did keep high clocks for long but at a max of 88C, which is still not near the thermal limit of 100C but proves that if the CPU is cooler = more clockspeed. All tested with Cinebench R23 multi-core at the time.
 
%1 lows matter more and that's were RAM speed / timings come into equation. Extra 200mhz for the CPU isn't going to do anything if it's not at 1080 low playing CS2
 
I wish there was more test data available for this (memory scaling and tuning differences). Guru3D's test is the only one I've seen and they show no difference at 4k and negligible difference at 1440p. Benchmarks, sure...but in games?...I haven't seen any proof of this outside claims from people who charge customers for tuning to get "more FPS" in games like CoD, CS:GO, and Apex. I'm not saying you're wrong, but I am saying I would like to see real tangible evidence of this from credible sources.
There are sources out there, it's difficult testing to do so time-bound tech writers working for larger outlets tend not to want to dive into it. Also that test was done on a 3090 - which was hard bottlenecked at 4k.

So for instance, in my case, this is the exact same system, same clock speed CPU, one with DDR5 76000 vs DDR5 6400:
1700170897490.png


Same Rig at 7600 XMP (no tuning yet):
1700171369602.png


This is everything identical except ram -

so you can see the cpu minimum FPS went from 208 to 224 (usually about 230ish - I have stuff open now, not a clean boot).


HWUB did some stuff:
1700171549205.png


1700171832381.png

Here you can see ram tuning on intel 13th gen -- look at the DDR5 6000 tuned. You're gaining on average about 5-10% on just tuning timings. You're not going to get that from a 5500mhz to 5700mhz a (3.6% bump in clockspeed) - especially when the clockspeed is not a limiting factor for the lows..

Combine this with some windows tweaks (disabling vbs, priority tweaks, ultimate performance mode etc.) and you will net another few % and you have a rig that's running 10-15% faster with no clockspeed bump. But if you spent your time chasing the last 3.6% clocks you would get a rig that's like 1-2% faster and consumes 800W and heats your room.
 
Last edited:
There are sources out there, it's difficult testing to do so time-bound tech writers working for larger outlets tend not to want to dive into it. Also that test was done on a 3090 - which was hard bottlenecked at 4k.

So for instance, in my case, this is the exact same system, same clock speed CPU, one with DDR5 76000 vs DDR5 6400:


Same Rig at 7600 XMP (no tuning yet):


This is everything identical except ram -

so you can see the cpu minimum FPS went from 208 to 224 (usually about 230ish - I have stuff open now, not a clean boot).


HWUB did some stuff:

Here you can see ram tuning on intel 13th gen -- look at the DDR5 6000 tuned. You're gaining on average about 5-10% on just tuning timings. You're not going to get that from a 5500mhz to 5700mhz a (3.6% bump in clockspeed) - especially when the clockspeed is not a limiting factor for the lows..

Combine this with some windows tweaks (disabling vbs, priority tweaks, ultimate performance mode etc.) and you will net another few % and you have a rig that's running 10-15% faster with no clockspeed bump. But if you spent your time chasing the last 3.6% clocks you would get a rig that's like 1-2% faster and consumes 800W and heats your room.
Thanks for that info!

Yeah, I was not disagreeing or trying to put up an argument for the clock speed vs memory situation, I was just saying I haven't seen much data for 1440p/4k testing with memory tuning.

That said, I saw the 3090 in guru3d's test and agree, the HUB tests you posted were 1080p, and your tests were the most helpful but still 1440p. I would definitely say that's not "negligible" as I had previously seen and stated above.

I have some DDR5-6000 that is B-die and doesn't tune...long story, but I have been considering trying to find some A or M-die 7200 RAM that is more tunable...I just have not found any real evidence to make it worth spending a couple hundred bucks and then trying to sell mine if I game at 1440p and my gains will be as I had previously thought: "negligible". I will also say that from the research I've done, and watching Buildzoid's videos, it certainly seems a risky endeavor and I don't think I really want to risk my Windows install...so I'd have to come up with another HD with a sacrificial Windows install that I can mess with the tuning on, then switch back over once I have things stable (so I don't risk corrupting my OS or files)...which I don't feel like doing, so I'd likely be on XMP or something very close to that, and that again makes me think that 7200-7600 memory at XMP is not going to be 5-10% faster than mine. Sorry for the book of a response, I just wanted to explain why I was so curious. Definitely not going deeper into this as it would detract from OP and I was just trying to settle the related question of what they'd gain from memory tuning.

Apex has a 300 FPS engine limit...
Sorry, Apex was probably a bad example. I just see ads for those people here and there on social media for whatever e-sports-oriented shooters people are into. I like FPS games, but generally not those where people are looking for hundreds of FPS.
 
Thanks for that info!

Yeah, I was not disagreeing or trying to put up an argument for the clock speed vs memory situation, I was just saying I haven't seen much data for 1440p/4k testing with memory tuning.

That said, I saw the 3090 in guru3d's test and agree, the HUB tests you posted were 1080p, and your tests were the most helpful but still 1440p. I would definitely say that's not "negligible" as I had previously seen and stated above.

I have some DDR5-6000 that is B-die and doesn't tune...long story, but I have been considering trying to find some A or M-die 7200 RAM that is more tunable...I just have not found any real evidence to make it worth spending a couple hundred bucks and then trying to sell mine if I game at 1440p and my gains will be as I had previously thought: "negligible". I will also say that from the research I've done, and watching Buildzoid's videos, it certainly seems a risky endeavor and I don't think I really want to risk my Windows install...so I'd have to come up with another HD with a sacrificial Windows install that I can mess with the tuning on, then switch back over once I have things stable (so I don't risk corrupting my OS or files)...which I don't feel like doing, so I'd likely be on XMP or something very close to that, and that again makes me think that 7200-7600 memory at XMP is not going to be 5-10% faster than mine. Sorry for the book of a response, I just wanted to explain why I was so curious.
I actually had this same dilemma, and please -- by all means disagree - this is a nerd forum after all.

I had 64GB of 6200/6400 memory and was super curious to see what would happen with 7600 -- so i grabbed some 48GB sticks really cheap Oloy kit (that I will get around to tuning or so i tell myself).

I can say that cyberpunk (the game I'm currently playing) definitely does feels a smoother in 4k but it's because of my settings --i'm locked at 120FPS with DLSS no RT and only 80-90% usage on the 4090, so running straight CPU bottleneck the entire time (100% load on all P cores). They all do feel slightly better varying degrees. It's really because of the latency i think -- I can feel the difference between 65 and 55 latency and see it in the numbers, but the speed seems somewhat irrelevant.

When you bottleneck the GPU fully at 4k then no - ram's not gonna make a difference. If you're playing competitive or high FPS you can still be latency bound even at 4k,.

Is it night and day difference? definitely no - but it is more difference than a 200Mhz cpu bump - absolutely.
 
hello everyone i have

mobo asus z690 hero bios 2802

i9 14900k sp98 p106 e84 mc80

cooler nh-d15

i see a lot of people with sp like mine overclock the cpu to 5.7 with 1.16-1.8v

mine need 1.21-1.22 llc6 and bios prediction show 1.19 and still the cpu not stable

does i have bad cpu? does it motherbord problem? or mybe the cooler?

thanks for the help
NH-D15 is a great cooler, don't misunderstand, but it's not going to get you too far with a 14900k. The main issue with your CPU is the thermal density, you'll hit thermal limitations long before voltage.. unless, you disable hyperthreading - which is what I do (you can check my specs).

If all you do is game, you can really benefit from three things when it comes to CPU-related overclocking.

1. Core frequency - of course, pushing it higher is often better. Most games only care about single threaded performance on modern CPU's. Once you have enough multithreaded performance, single is often what holds you back the most. For the games I personally play, frequency is all that matters - hyperthreading does absolutely nothing whatsoever to boost my framerate. My testing in Cyberpunk 2077 showed an 8% boost with HT enabled, and is one of the few titles that actually scale with HT. Other games show little to no difference.

2. RAM tuning. I recognize your screen name because I helped you previously tune your memory in another thread. It's an easy +8% if you do it correctly.

3. Process lasso. I don't see very many people discuss this, but it's a wonderful application that I recommend using. It lets you bind applications, permanently, to your selected cores. I leave all of my necessary bloatware pinned onto the efficiency cores. While the boost in framerate is pretty miniscule (2% at most) - the frametimes, in my experience, are better in CPU bound scenarios.

Overclocking is a lot different than what it once was. You can't really set a big OC without utilizing a power limit these days, unless you're direct-die & custom looping. I've tested and tuned my system extensively and can tell you the following:

In the harshest CPU bound environment I tested (Cyberpunk 2077, 720p, no frame rate cap) - the CPU draws 150 watts with spikes up to 165 watts when set to 1.44V (1.41 under load) - Hyperthreading disabled. Because the game utilizes efficiency cores, I would expect around 180 watts of power draw at similar voltages, since you have 8 more e-cores.

My power limit is set to 175W.
I would recommend using 200 to 220 watts for yourself.
You have to change short duration and long duration to the same amount (that's what PL1=PL2 means) - it caps your CPU at the set wattage. So 200 or 220 for both.

The best way to overclock currently, assuming you don't play things such as simulators that commonly push hyperthreading fairly hard, is to disable hyperthreading altogether. It lowers the thermal density greatly. As a disclaimer, if you're doing video editing or number crunching or any other type of work that actually loads the CPU fairly heavily, just leave HT enabled. Otherwise, you'll benefit greatly from disabling it from a thermal and voltage standpoint.

I recommend skimming through this video by Buildzoid.

Addressing your voltage of 1.22 @ LLC6 - this is normal. If you're on an ASUS BIOS, in my experience, you often need around 0.0200 mV higher than what's stated for the prediction - with that amount increasing the nearer you get to the frequency wall. I'm running LLC6 as well on my ASUS board, the stated vdroop is fairly minimal - but it's likely much lower, since there's a voltage drop across the CPU socket (which is why I'm fine with running 1.45v daily)

So my recommendations for core tuning are the following:

Leave the cache alone, and if you do tune it, don't go above 48x / 1.300v (it's called CPU L2 in hwinfo64) - realistically you should need only 1.200v for 46x. Don't test your luck with pushing cache voltages. It will break your chip and gives minimal gains.

Disable hyperthreading and try 1.35v LLC6 at 5600 MHz (realistically you should be able to push 5700+ on a nicely binned CPU).

Once you've finished stability testing, set a short duration & long duration power limit of 200 or 220 watts.

Don't go above 1.38v / 100 C because I don't want to be responsible for your CPU degrading. I don't know if this is your first time overclocking, but if it is, the 14900k is one hell of a chip to tame and not beginner friendly whatsoever.
 

Attachments

  • lasso.png
    lasso.png
    81.9 KB · Views: 159
thanks for everyone for the help

i want to ask another question

if i rise core pll can higher value can lower the vcore?

my stock core pll is 0.900 how much can i rise for safe?
 
thanks for everyone for the help

i want to ask another question

if i rise core pll can higher value can lower the vcore?

my stock core pll is 0.900 how much can i rise for safe?

In my experience, the only thing that you *MAY* have to tune is CPU Input voltage, which by default is 1.800
Pushing that to 1.8800 can help with stability at the far end of overclocking (1.45v or higher).

At the day's end, what matters is how much power the CPU is receiving. There's no silver bullet to getting a higher clockspeed. Even if you manage to get a lower vcore, the transistors are still going to necessitate a certain amount of current to run through them for switching on and off.
 
%1 lows matter more and that's were RAM speed / timings come into equation. Extra 200mhz for the CPU isn't going to do anything if it's not at 1080 low playing CS2
or 720p, remember, it's "cybersport settings":D
 
my cpu does 5.5p 44e on 1.10v load llc3 stable all tasks

i saw with the same cpu sp98 people do 5.7p 44e on 1.17-1.18v pulls on cinbench 278waat

when i try 1.18v the cpu not stable in r23 i need more volt

if i will get beter cooler can my cpu be stable in low voltage?

No, and I strongly suggest you give up before you damage your hardware. The 13900KS/14900K is already the fastest CPU around.
 
Back
Top