• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

My i9-14900K Benchmarks

El_Capitan

El_Capitan

[H]ard|Gawd
2FA
Joined
Nov 11, 2012
Messages
1,479
I finally got an i9-14900K thanks to CaffeineMan

5/12/2026:
Replaced my capable overclocking i5-13600K with the i9-14900K. The motherboard is an ASUS TUF Gaming Z690-Plus WiFi D4 motherboard with the latest available BIOS (Version 4505). I kept it simple and had the CPU running at stock, and my memory running at XMP1 and Command Rate at "Real 1N". The memory is Crucial Pro DDR4 3200MT/s CL22 4x32GB for a total of 128GB.

For cooling, I'm using a DeepCool LD360mm AIO with a contact frame, and Thermal Grizzly Duronaut for the TiM. Temps so far at stock settings hasn't gone over 83C for the hottest core.

Geekbench 6:
Single Core: 3231
Multi Core: 16224
2026-05-12_16-58-29.png


CPU-Z:
Single Core: 972.9
Multi Core: 16500.5
2026-05-12_17-41-35.png


20260512_174012.jpg


5/14/2026:
Lol, for some reason I thought it was the opposite, Gear1 for DDR5, and Gear2 for DDR4. I should probably stop listening to random search results.

I upped my DRAM voltage to 1.35V, Frequency to 3600MHz, Gear1, 18-22-22-42-2T. I left uncore on auto, but seen it hit 5000MHz during load and seen it downclock near idle. I'll try to hit 4000MHz.
2026-05-14_11-27-33.png2026-05-14_11-29-29.png

I was not able to hit 3800MHz or 4000MHz. My scores were faster at BCLK 100:100 than 100:133. I also was stable at Command Rate 1T, but scores were better at 2T.
2026-05-14_14-05-32.png2026-05-14_14-07-02.png



To be continued...
 
Last edited:
First, Looks very nice! ..but why tho? 83c on a 360 AIO with DDR4? I'm getting ~10800 in cpu z for reference.
 
cjcox said:
Possibly. Maybe not the best "CPU" benchmark choice. :) Edit: Note, I'm at JEDEC speeds though.
Click to expand...
DDR4 seems to do better with Single Core/Thread, possibly due to timings.
DDR5 does better with Multi-Core/Threads, mostly due to bandwidth. Amount of RAM also factors in.

i7-14700T 64GB DDR4 3200MHz:
2026_05_13_i7_14700T_64GB_DDR4_3200MHz_Geekbench6.jpg2026_05_13_i7_14700T_64GB_DDR4_3200MHz_CPU-Z.jpg

i7-14700T 32GB DDR5 6400MHz:
2026_05_13_i7_14700T_32GB_DDR5_6400MHz_Geekbench6.png2026_05_13_i7_14700T_32GB_DDR5_6400MHz_CPU-Z.png
 
toddw said:
Why change out your 13600K? The heat, power suckage from a 13600 to a 14900K is massive. If the answer is because you're [H]ard, fair enough... just curious.
Click to expand...
I love the 13600K, and I always will. It's my Prius for commuting. However, sometimes you just need the F350 or E-350 to do some heavy lifting.

So, after having three 13600K's, only one was a capable overclocker at 1.35V's. From 5700MHz single core, scaling down to 5400MHz all-core. One could only do 5400MHz all-core, and the other only 5300MHz all-core. Overclocking ability is hit or miss with them.

With the 14900K, at stock it's 6000MHz for single core, and 5700MHz for all-core.

I do a lot of video editing, so the 14900K helps a lot.
 
Last edited:
  • Like
Reactions: toddw
like this
On my 13900KF @ 5.4GHz/4.3GHz flat with zero turbo boost and without Hyper-Threading so 24 hardware threads.
1778703514189.png

DDR4 4000MT/s CL19 2T

My only load that I care for (Quartus) runs mostly on single core (so much so that giving it 1 P-core and 15 E-cores would yield only 1s slower performance versus 8 P-cores + 8 E-cores on minute long compile) and I still wouldn't get any benefit with default turbo boosting to 5.7GHz outside benchmark-like environment (so basically nothing else running) so I prefer non-degrading quiet operation with CPU at 1.25V rather than fake single core turbo boost. HT is of course disabled because HT/SMT sucks.
 
XoR_ said:
On my 13900KF @ 5.4GHz/4.3GHz flat with zero turbo boost and without Hyper-Threading so 24 hardware threads.
View attachment 803046
DDR4 4000MT/s CL19 2T

My only load that I care for (Quartus) runs mostly on single core (so much so that giving it 1 P-core and 15 E-cores would yield only 1s slower performance versus 8 P-cores + 8 E-cores on minute long compile) and I still wouldn't get any benefit with default turbo boosting to 5.7GHz outside benchmark-like environment (so basically nothing else running) so I prefer non-degrading quiet operation with CPU at 1.25V rather than fake single core turbo boost. HT is of course disabled because HT/SMT sucks.
Click to expand...
Lol, now I got to figure out why my low Multi-Score is happening. Maybe I need to set my Command Rate at 2T and retest.
 
El_Capitan said:
Lol, now I got to figure out why my low Multi-Score is happening. Maybe I need to set my Command Rate at 2T and retest.
Click to expand...
On CPU-z you have mem controller frequency 800MHz. I have 2000MHz. Are you running in Gear 1? You should for DDR4 run in Gear1 as Gear2 is for DDR5 basically and runs part of the IMC at half the clock. My understanding is that its better to run Gear1 than 1T if either one had to be at 2.
Also I have uncore at 4700MHz vs your 4000MHz and this affects some things like L3 caches apparently.

Command rate 1T is better than mine 2T but you have only 3200CL22
I would say to try OC the memory even if at 2T command rate.

p.s. My score with P-cores bumped to 5.5GHz https://browser.geekbench.com/v6/cpu/17987935
1778734998704.png
 
XoR_ said:
On CPU-z you have mem controller frequency 800MHz. I have 2000MHz. Are you running in Gear 1? You should for DDR4 run in Gear1 as Gear2 is for DDR5 basically and runs part of the IMC at half the clock. My understanding is that its better to run Gear1 than 1T if either one had to be at 2.
Also I have uncore at 4700MHz vs your 4000MHz and this affects some things like L3 caches apparently.

Command rate 1T is better than mine 2T but you have only 3200CL22
I would say to try OC the memory even if at 2T command rate.

p.s. My score with P-cores bumped to 5.5GHz https://browser.geekbench.com/v6/cpu/17987935
View attachment 803170
Click to expand...
Lol, for some reason I thought it was the opposite, Gear1 for DDR5, and Gear2 for DDR4. I should probably stop listening to random search results.

I upped my DRAM voltage to 1.35V, Frequency to 3600MHz, Gear1, 18-22-22-42-2T. I left uncore on auto, but seen it hit 5000MHz during load and seen it downclock near idle. I'll try to hit 4000MHz.

2026-05-14_11-27-33.png

2026-05-14_11-29-29.png
 
  • Like
Reactions: XoR_
like this
Update OP. No luck at 3800MHz or 4000MHz. Looks like setting BCLK at 100:100 is faster than 100:133, and 1T timings might not be entirely stable, as I had slightly lower scores.
 
1800MHz (3600MT/s) at these timings is pretty good considering I got overpriced underperforming 4400 ram and when I got second pair they did together similar to your 3200 sticks and at considerably higher IMC voltages than what I needed for two sticks.
Or in other words I have/had much more expensive seemingly better DDR4 sticks but couldn't get better performance out of them when I had 4 of them. It might even be that if you had only 2 sticks they might go faster... or not, hard to say.

BTW. I recommend to reduce reliance on "auto" on these CPUs. They have bad track record with their auto settings XD
I mean Intel was supposed to fix the issue RPL had and they probably did to a degree but it doesn't change the fact that these are extremely factory overclocked CPUs and voltages in such case will be too high compared to what is actually needed. Especially with turbo boost if you want to use it being dependent on power and power being dependent on voltage it makes sense to optimize voltage.
 
FWIW... I am on air, and prefer air, so this is the best I'll ever be able to do on this platform, cpu wise. Note I have been running a 104.23 base clock for over a year now and have noticed zero issues, I credit Z790. 1.39v on memory.

1779021555881.png
 
toddw said:
FWIW... I am on air, and prefer air, so this is the best I'll ever be able to do on this platform, cpu wise. Note I have been running a 104.23 base clock for over a year now and have noticed zero issues, I credit Z790. 1.39v on memory.
Click to expand...
Wait... is your 13600K doing 5.7GHz with vcore at 1.224V... how is this possible? 🤯 Am I reading it correctly?
 
XoR_ said:
Wait... is your 13600K doing 5.7GHz with vcore at 1.224V... how is this possible? 🤯 Am I reading it correctly?
Click to expand...
Some of the 13600K/14600K are amazing overclockers. His is a 14600K, which I never got to try, but it may be more efficient than a 13600K. It also may just be 1 core that goes up to 5.7, the next cores 5.6, all cores 5.5 or 5.4, etc. That's what I did with my 13600K at 1.35V. Most 13600K's are hit or miss, like only 5.5GHz all core at 1.35V, or 5.3GHz all core. Special ones can hit 5.7GHz for a few cores running lower at low-ish voltage.
 
  • Like
Reactions: XoR_
like this
El_Capitan said:
Some of the 13600K/14600K are amazing overclockers. His is a 14600K, which I never got to try, but it may be more efficient than a 13600K. It also may just be 1 core that goes up to 5.7, the next cores 5.6, all cores 5.5 or 5.4, etc. That's what I did with my 13600K at 1.35V. Most 13600K's are hit or miss, like only 5.5GHz all core at 1.35V, or 5.3GHz all core. Special ones can hit 5.7GHz for a few cores running lower at low-ish voltage.
Click to expand...
You mean it as the max clock in hwinfo64 not being for all core but merely momentary reding across all cores?
To be honest I have no idea how turbo boost actually behaves in this platform other than my tests indicating I didn't really see much if any improvements doing benchmarks in typical for me workloads so I disabled it to avoid voltage spikes.

If its somehow possible to keep some TB without excessive voltages it might be worth investigating... then again I would expect stability of such configuration to be very hard to test
 
XoR_ said:
You mean it as the max clock in hwinfo64 not being for all core but merely momentary reding across all cores?
To be honest I have no idea how turbo boost actually behaves in this platform other than my tests indicating I didn't really see much if any improvements doing benchmarks in typical for me workloads so I disabled it to avoid voltage spikes.

If its somehow possible to keep some TB without excessive voltages it might be worth investigating... then again I would expect stability of such configuration to be very hard to test
Click to expand...
There's a few different methods of overclocking the LGA1700 series, it can get confusing. The simplest way is setting All Core, so no matter what, it's always at say, 5.5GHz, or always at 5.3GHz, no matter if a single core is running, or all the cores are running.

Another way of doing it is setting the top core, or top two cores to run at say, 5.5GHz, then when all the cores are running, it only runs at 5.3GHz. Depending on your LLC, you can run lower Vcore than if running all cores at 5.5GHz.

An additional way of doing it, is if only two cores are running, run it at 5.7GHz. If only three or four cores are running, run all them at 5.6GHz. If only five or six cores are running, run them at 5.4GHz. Again, depends on your LLC and Vcore.

Another way of doing it is by Thermal Boost Velocity (TBV), but I don't like doing it this way, although you can get some of the best overclocks this way with good cooling and ambient temps.

How I test for stability is just comparing benchmark scores. They're usually consistent, and a bad overclock will get you non-boot to BIOS, errors running a benchmark, or lower scores. When scores are higher, I run a bunch more benchmarks and stress test for stability.

Here are some examples of my best 13600K that could overclock:
2026-05-10_9-35-56.png2026-05-10_9-37-57.png
 
Last edited:
In games and cpu-z bench, all cores go to 5.7, with one dipping then coming back every now and then. ...but most productity, NO. For example, Cinebench 2026 all cores are 4.7 - 5.3. Tho I imagine if I had 360 AIO and set the voltage 1.3X the cores would maintain 5.5 - 5.7.

This is mainly a gaming rig, I found this voltage is the lowest while maintaining all cores 5.7 while gaming.
 
My attempts with anything turbo resulted in much higher vcore than I have expected to see and clocks dropping like crazy with load so I gave up on it quickly.
Looking at discussions about RPL overclocking a lot of people use offsets and get good results. They seem to control some strange settings I am not willing to touch though.

I shifted my OC to 5.6GHz all P-core and 5.4GHz for E-cores at 1.328V with slight vdroop all core. With increased power limit vdroop is higher and Cinebench R23 would crash but with 255W limit it seems fine. I do get some power throttling using 21+ threads in Cinebench R23 but that is not the load I ever expect to see in normal usage (these CPUs have already too many cores 🫨) and some other programs including newest Cinebench 2024 seems fine almost running 24 threads only shaving 100MHz off E-cores and rarely. I could probably increase power limit a bit and it would be fine but for now I am testing stability with current settings and so far so good.

3140 in geekbench ST so there is some improvement from 3035 @ 5.4GHz. Fan is also louder at high load but its still impossible to hear computer in normal usage. Fast compile in Quartus causes CPU power consumption to peak at 130W 😴
OC from 5.4GHz to 5.6GHz shaved less than 2 seconds in compile time. Its a game changer 😃
 
You must log in or register to reply here.
Back
Top