Question i7-12700KF -- Insane Temp Instability ?

[ForHonor13]

Hello Denizens of the OC Deep!
I seek your wisdom of the ages.

Random Context:

• I have been building PCs for about 15~ years now as a side-job; with about 100-200 under my belt (mostly for gaming and lite production work).
• I have limited OC experience, mostly on the much older Ivy CPUs. I apologize if parts of the information below are redundant or not required; I am a bit outside my comfort on this.
• I only intend to use this for 'Intense' gaming and streaming; as well as lite Music Production via FLStudio.

PC Build:

• CPU: i7-12700KF
• Motherboard: MSI Z690A-PRO
• CPU Voltage: 1.29V (CORE)
• CPU Bus Speed/Multiplier: 100*49
• Clock Speed: 4.8ghz (on Boost; without Boost it seems to hover around 3.8-4.2)
• RAM: GSKILL DDR4 3600 16-19-19-39
• Cooling: Watercooling Loop (MSI MAG P240)
• GPU: XFX 6800XT
• OS: Windows 10 PRO 64 Bit

Whats Happening:
- Prior to BIOS UPDATE, the system would immediately throw CPU OVERHEATING error when booting into BIOS. After updating to most recent BIOS and enabling XMP (without MemoryTryIt override) it seems to be 'useable' in the sense I was able to install windows and drivers and some test applications. However, when doing PRETTY MUCH ANYTHING (loading a YT video, launching a game, downloading updates/steam items, etc) it will surge to 90C+ and start 'toggling' the Throttle as it will slingshot from 85>100+>90>100* etc; in XTU my graph almost looks like a heart-rythm... which is appropriate as I am 1 step short of a heart-attack if this thing cooks itself.

What I've Tried:

• After watching a few YT videos on the 12700K (or anything 12th Gen+) it seems Intel threw TDP out the window? Many people seem to be experiencing power/voltage overdraw on these and advise Undervolting.
• After attempting to Undervolt a bit, I set to 1.29V with Adptive+Offset (Offset of -.12) via XTU; this helped a bit; dropped 5~C ; but I still get insane spiking as mentioned above that often throws it into throttle; and I cannot launch ANY game without it sitting at 95-105 consitently.
  • Its important to note; the CASE is not the BEST; and I recognize that as an airflow problem, which Ive purchased 4 Noctua NF-A12x25 to update the default water-loop fans and the case fans; but for the purposes of my testing above; and for moving forward to get this stable; I've removed the glass side of the case to open and allow airflow. (this did not resolve the issue and barely moved the avg 1-2*C; which makes leads me to believe this issue leans more on voltages ) - I also intent to re-paste the CPU+Pump Interface with some Noctua NT-H1; instead of the default paste the AIO came with (but was all I had at the time).

What I've Noticed:

• The CPU/GPU are not taxed AT ALL on any of these application launches; even when I launched some test games, my CPU Utilization is only around 5-20% and my GPU is under 40-50% (on desired settings).
• AIO FAN + PUMP SPEEDS are all pretty good. PUMPs at 4k+ and fans are around 1500-2000 RPM on the Radiator; and the case-fans (although case is wide open) are hanging around 900-1200 RPM.

My Questions:

1. Should I be adjusting the MIN/MAX WATTS allowed for the Boosts? in XTU I see : "Turbo Boost Short Power Max (P2) & Turbo Boost Power Max (P1)" which are both set to 241W (which is seemingly insanely high compared to the TDP I thought I was getting into; 125W~ )
2. Should I be lowing any of the Clock Ratios? I see a few people with x48; whereas mine is x49 (I am unsure how much this actually impacts).
   • I saw a YT video where someone had some success dropping the ratios for the IF USING X AMOUNT OF CORES logic.
3. Should I continue to underclock/undervolt? Based on HARDWARE MONITOR, this thing isnt pulling more than 1.26V; but that is not counting runs on Cinebench 23R and XTU STRESS (which I prefer to not run; compared to IBT; but perhaps thats outdated thinking on my part?)
4. Is there anything else I should be looking at in BIOS or Intel XTU? Any and all advice is greatly appreciated. I would love to get this rig up and operational fully before the Holiday.

Here is a picture of the most recent temp-run; only having CHROME + STEAM DOWNLOADS running. (1 Hour Lookback)

 

[jkflipflop98]

I would start with the most simple and most likely cause. Your thermal block isn't seated on the CPU correctly. Take it all apart, clean it, apply new paste, and put it all back together carefully.

 

[Aeacus]

Your thermal block isn't seated on the CPU correctly. Take it all apart, clean it, apply new paste, and put it all back together carefully.

For a 1st time builder - perhaps. But i don't think this kind of mistake would happen to someone who does it for living.

 

[jkflipflop98]

Everyone makes mistakes. Sometimes a bracket gets bound and you think it's tight but it's not.

The most probable cause is as stated.

 

[Karadjgne]

For a 1st time builder - perhaps. But i don't think this kind of mistake would happen to someone who does it for living.

Linus mounted an aircooler that had pre-applied paste in a pc build video. Without first removing the clear plastic. Didn't catch that error until booting put it into thermal overload.

I've made my fair share of noob blunders. It happens to most and I chalk it upto complacency.

 

[Karadjgne]

Can't vouch for the A12, since I never tried them on the D15S heatsink, but the other 2 do buzz when used as pull fans on the heatsink. Starts doing it around 1000rpm.
If spacers, or something to act as one, are used in between, the sound can be reduced.
Didn't hear that sound at all on the Celsius+ S28.

The D14 could make noise at higher rpm, but that was due to the mismatched fans. But that happens to all fans. Dunno lol.

 

[ForHonor13]

Wow!
Thanks everyone for your replies.
I will be working through the feedback and responses today..

==

@Karadjgne & @jkflipflop98 ; you are not wrong. I should go back to basics and ensure I have a solid foundation for troubleshooting.
I did reseat it; but I will reseat, repaste, and not mount on the front to leave out of case for testing with XTU and Cinebench so we arent as concerned with airflow due to case limitations (I personally think I should be atleast system stable, even in the case with its current airflow limitations).

Although I accept the feedback around the AIO; I am inclined to think based on intuition alone...
Even if the AIO was garbage and had corrupted coolant, etc -- I would ATLEAST expect to be system-stable on idle, right? or am I crazy.
The fact that P4/P5 are +10-14*C on average, feels like I am fighting some type of "Smart" MOBO Feature for an initial overvolt? or a bad die? (although I realize thats a copout in 99% of calls).

==

@Aeacus - Thanks for the vote of confidence! Although, I have definitely made my share of mistakes Thanks for confirming you like the NH-15/S - "Btw, NH-D15S chromax.black only has one, middle fan. So, how could that interfere with RAM clearance?" -- I typed that weirdly; words escaped me that late in the evening. I know the "S" is better for RAM clearance due to lack of a fan; but, I saw people be successful with the non-S; and double-fan (they just moved the fan that was on RAM side to IO side; or moved to case). If you feel I've made a terrible mistake, I will cancel the Newegg Order and move to the "S" - although I am hoping to get this AIO "working as intended" today to be ATLEAST system stable.

==

@Phaaze88 - Thanks! I will look at the Fan Curve for the FANS vs PUMP. To be honest, I am NOT a fan of this BIOS; it feels like im fighting multiple profiles when I make any edits? Its the oddest thing Ive ever encountered with a build; let alone a MOBO BIOS. (More of an ASUS guy myself BIOS wise). I will watch another video on navigation, perhaps I am doing something wrong. I noticed I have "CPU" and "XMP Profile 1" selected sometimes. Its as if they are seperate profiles? but I dont have XMP Enabled on "CPU" whereas its enabledd on "XMP1" -- Wondering if I should just flash backwards to the Factory/Reset CMOS/BIOS?

==

Whew~
I apologize for the chunks of texts everyone; but greatly appreciate your input. Will follow-up more today as I work through the items.. hopefully dont have to RMA this AIO (for now).

 

[Phaaze88]

Any fan becomes audible over 1000rpm, but this is a specific characteristic of the A14 and A15 when used on the NH-D heatsinks this way:

Spoiler

View: https://www.youtube.com/watch?v=UdmmLj9A1qQ

(Skip to 1:35. After 2:45, it's the usual push profile.)

Thanks! I will look at the Fan Curve for the FANS vs PUMP. To be honest, I am NOT a fan of this BIOS; it feels like im fighting multiple profiles when I make any edits? Its the oddest thing Ive ever encountered with a build; let alone a MOBO BIOS. (More of an ASUS guy myself BIOS wise). I will watch another video on navigation, perhaps I am doing something wrong. I noticed I have "CPU" and "XMP Profile 1" selected sometimes. Its as if they are seperate profiles? but I dont have XMP Enabled on "CPU" whereas its enabledd on "XMP1" -- Wondering if I should just flash backwards to the Factory/Reset CMOS/BIOS?

There's variances between bioses for sure. Some are a bigger headache than others. There should be 3-4 presets on fan control for each header(except the liquid cooling ones), you can try those, and if not satisfied create a manual curve... IDK, the manual one on this Asus board seems easy enough.

The liquid cooling headers, AIO_PUMP and W_PUMP, are like liquid cooling for dummies.; they run at 100% out of the box, unlike the other headers, but can be manually adjusted just the same.
W_PUMP, is rated for up to 3Amps. Every other header is 1-2Amps.

XMP Profile 1 is probably Gear 1. XMP Gear 1 and 2 was introduced with Rocket Lake(11th gen).
Gear 1 means CPU Memory Controller & Memory Speed are equal.
Gear 2 means CPU Memory Controller operates at half the Memory Speed (such as CPU Memory Controller is at 1600MHz while Memory Speed is at 3200MHz when operating as Gear 2).

More on that here:

Spoiler

View: https://www.youtube.com/watch?v=LU_w9fZvSso

 

[ForHonor13]

Any fan becomes audible over 1000rpm, but this is a specific characteristic of the A14 and A15 when used on the NH-D heatsinks this way:

Spoiler

View: https://www.youtube.com/watch?v=UdmmLj9A1qQ

(Skip to 1:35. After 2:45, it's the usual push profile.)



There's variances between bioses for sure. Some are a bigger headache than others. There should be 3-4 presets on fan control for each header(except the liquid cooling ones), you can try those, and if not satisfied create a manual curve... IDK, the manual one on this Asus board seems easy enough.

The liquid cooling headers, AIO_PUMP and W_PUMP, are like liquid cooling for dummies.; they run at 100% out of the box, unlike the other headers, but can be manually adjusted just the same.
W_PUMP, is rated for up to 3Amps. Every other header is 1-2Amps.

XMP Profile 1 is probably Gear 1. XMP Gear 1 and 2 was introduced with Rocket Lake(11th gen).
Gear 1 means CPU Memory Controller & Memory Speed are equal.
Gear 2 means CPU Memory Controller operates at half the Memory Speed (such as CPU Memory Controller is at 1600MHz while Memory Speed is at 3200MHz when operating as Gear 2).

More on that here:

Spoiler

View: https://www.youtube.com/watch?v=LU_w9fZvSso

Hello @Phaaze88
Thanks for the response!

I have "Good News" for those following.
I have managed to get this somewhat "system stable" by doing the Reseat + Repaste as mentioned above (sometimes, It is truely the basics I suppose.. but I am completely unsure how I wouldve messed that up 2x) and going to LKG Config (which has seemingly some of these weird profile edits Im fighting?)

I was able to stay under 90*C on IDLE; compared to previously overheating in idle or light load.
In comparison, temperatures are quite good I think (32-40C on Idle, 60-85 on Single-Core Load) but Throttled Hard on Multi-Core.

CINEBENCH is providing the following :

• SINGLE CORE: 1907 (Feels low?)
• MULTICORE: 14068 (Definitely Low, seems to be throttling, Core P5 and P3 are culprits as outliers.. with a 10-20* delta)

From what I can tell now..
I am struggling to fight something in the BIOS and XTU
XTU now says "UNDEFINED" for Core Voltage; which in BIOS is undervolted to 1.225 STATIC; XTU is saying CORE is at 1.3~ DYNAMIC (You can see this in the log-snips below as VID MAX and CPUV arent being respected in the BIOS and are way above intended.)

This must be my ignorance in the BIOS; although I am unsure what Ive done (as mentioned above it seems I am fighting multiple 'profiles' when I update/save changes, yet, I never make changes? (Save the FAN CURVE mentioned above.. which my FAN CURVE seems to be tied together for all 3 (PUMP+AIO FAN1+2) meaning I cannot keep the pump at Max (although per HWMonitor it seems to already be @ 4K RPM?) but the FANs do appear to be hitting 1.5-2k RPM as expected. In BIOS I have the setting for "WATER COOLER" on; and toggled between that and "BLOCKCOOLER" with little to no effect.

It must be the 1.3V CPU killing me on Multi, I imagine? but Im banging my head trying to understand why its not respecting the BIOS or XTU heirarchy.

Spoiler: LOGS & SETTINGS

Below is the SINGLE THREAD RUN

Below is the MULTI THREAD RUN

Below is the XTU Settings (UNDEFINED, not respected; yet BIOS is also not puting the desired V?)

PS: I will get into BIOS and make a list of settings currently in place.

 

[Phaaze88]

In BIOS I have the setting for "WATER COOLER" on; and toggled between that and "BLOCKCOOLER" with little to no effect.

OK. Now this here affects your cpu's power limits. I think that's why the cpu Vcore gets so dumb under all core.
I can't remember off the top of my head what the power limit is for block cooler, but for water cooling, it's 4096w - it'll NEVER hit that, but it's basically telling the cpu:
"Yo 12700K! You can suck up as much power as you need to maintain max boost clocks..." even though it might only max out around 280w or something in a Cinebench load.


Do you have Msi Center installed? Previously, that app was applying a 'performance optimized' overclock without notifying the user, but it was too aggressive.
I'm wondering if that has changed?

 

[ForHonor13]

@Aeacus - I appreciate your vote of confidence for me; but I would not say I do this for a living; just a Hobby+ lol - I am a bit outdated on these generations for CPU and MOBO limits; so I am definitely not above asking for wisdom

OK. Now this here affects your cpu's power limits. I think that's why the cpu Vcore gets so dumb under all core.
I can't remember off the top of my head what the power limit is for block cooler, but for water cooling, it's 4096w - it'll NEVER hit that, but it's basically telling the cpu:
"Yo 12700K! You can suck up as much power as you need to maintain max boost clocks..." even though it might only max out around 280w or something in a Cinebench load.


Do you have Msi Center installed? Previously, that app was applying a 'performance optimized' overclock without notifying the user, but it was too aggressive.
I'm wondering if that has changed?

Hi @Phaaze88

Understood the Cooler Behavior default, that would definitely be a factor here then.

I will, I assume, disable the Cooler Setting in the BIOS or move back to BLOCK or TOWER?
Or, is there a location I can just set this for something reasonable instead of relying on that weird pattern setting?

I noticed in XTU I had UNLIMITED for both Power Short and Boox Max. (My understanding of this is it sets the LOWER and UPPER limits of when boost will stop via the PL1 and PL2? Please correct if Im off base, Id love to learn this properly)

I do NOT have MSI Center installed, I had read previously to the build that it was causing those issues and avoided. I wanted to start as 'blank slate' as possible to reduce any variables. Thanks for asking. Helps me feel sane lol.

Side Note: as Im getting this more stable and starting to push it; I am noticing P5 is just a complete outlier, +8-15* in some cases and instantly causing throttle on a MultiThread run of Cinebench. Other cores do get steamy, at around 90-95; but dont throttle.

I managed to get the BIOS reflashed although its reflashed to the same version it was on (newest update)and its now atleast respecting the Core CPU Voltage requests (somewhat) and XTU is not "UNDEFINED"; although I am not using XTU moving forward for settings, just monitoring - there is definitely a weird BIOS vs XTU thing going on.

I am asking STATIC 1.225V on BIOS; with Core CPU respecitng that but VID at 1.35+ still? Perhaps I dont have a good understanding here.
Do you have any advice on how I can start correcting for these temps (outside the continued undervolt), esp on Core P5?

I am atleast confident now it is less of a CASE / Cooler/ AIO Issue.
(Thanks so much all for your feedback, the P240 MAG AIO seems to be doing "OK" but I will definitely be replacing it once budget allows)

 

[Phaaze88]

I noticed in XTU I had UNLIMITED for both Power Short and Boox Max. (My understanding of this is it sets the LOWER and UPPER limits of when boost will stop via the PL1 and PL2? Please correct if Im off base, Id love to learn this properly)

This here, can override...
"I will, I assume, disable the Cooler Setting in the BIOS or move back to BLOCK or TOWER?
Or, is there a location I can just set this for something reasonable instead of relying on that weird pattern setting? "
That, so there is no real need to touch cooler setting.

https://www.intel.com/content/www/u...-25m-cache-up-to-5-00-ghz/specifications.html
Intel Processor Base Power is 125w and Maximum Turbo Power is 190w. Tau, or Max Turbo Power Duration is 56 seconds on a steady load.
The mobo manufacturer might, or might not follow the Intel spec in their board's base package, leading to users of the same cpu - and cooling - seeing much worse thermal operation than others.
If you don't want the cpu to clock down after Tau expires, then you can set PBP and MTP at the same value, 190w.
This of course, comes with higher thermals, but 190w may come out to be less than whatever the 12700K was pulling at UNLIMITED POWER settings.
If playing games or doing other light stuff, you can just run the Intel spec; they aren't steady loads. If it's work that tasks all the cores, you'll prefer PBP and MTP = 190w.

I am asking STATIC 1.225V on BIOS; with Core CPU respecitng that but VID at 1.35+ still? Perhaps I dont have a good understanding here.
Do you have any advice on how I can start correcting for these temps (outside the continued undervolt), esp on Core P5?

I might not be of too much help here. I think this still applies: VID is the voltage the cores are requesting from the motherboard and Vcore is what the cores are actually getting.

 

[ForHonor13]

This here, can override...
"I will, I assume, disable the Cooler Setting in the BIOS or move back to BLOCK or TOWER?
Or, is there a location I can just set this for something reasonable instead of relying on that weird pattern setting? "
That, so there is no real need to touch cooler setting.

https://www.intel.com/content/www/u...-25m-cache-up-to-5-00-ghz/specifications.html
Intel Processor Base Power is 125w and Maximum Turbo Power is 190w. Tau, or Max Turbo Power Duration is 56 seconds on a steady load.
The mobo manufacturer might, or might not follow the Intel spec in their board's base package, leading to users of the same cpu - and cooling - seeing much worse thermal operation than others.
If you don't want the cpu to clock down after Tau expires, then you can set PBP and MTP at the same value, 190w.
This of course, comes with higher thermals, but 190w may come out to be less than whatever the 12700K was pulling at UNLIMITED POWER settings.
If playing games or doing other light stuff, you can just run the Intel spec; they aren't steady loads. If it's work that tasks all the cores, you'll prefer PBP and MTP = 190w.



I might not be of too much help here. I think this still applies: VID is the voltage the cores are requesting from the motherboard and Vcore is what the cores are actually getting.

Thanks alot @Phaaze88
I will play around with those over the next day or so while I start dialing in with the cooling options I have.
Seems I have it somewhat stable at the moment for pretty much anything but a Cinabench MultiCore run (and once again only P5. causing the throttle)

PS: You are definitely of help.

 

[Aeacus]

I know the "S" is better for RAM clearance due to lack of a fan; but, I saw people be successful with the non-S; and double-fan (they just moved the fan that was on RAM side to IO side; or moved to case).

Besides the lack of 2nd fan, S version is also off-centered, giving you better GPU clearance, IF you need it.
As said by Karadjgne;

The NH-D15 and NH-D15S are actually 2 different coolers, not just a single or dual fan difference. The 'S' has a double offset, it's physically offset towards the rear slightly, giving better clearance for ram as it sits behind instead of over the 2nd ram slot, and it also has an offset towards the top of the case, away from gpus that use a heavy backplate/stand off backplate design. That was often a bone of contention with the D15 as it is wide enough to touch some gpus.

Would there be a GPU clearance issue - i can't tell that for sure.

-----

If you feel I've made a terrible mistake, I will cancel the Newegg Order and move to the "S" - although I am hoping to get this AIO "working as intended" today to be ATLEAST system stable.

Since D15 has two fans, one of the fans sits higher on the cooling tower, resulting in a height increase from 160mm to 165mm. This much is also said by Noctua:

3. Total height: While the front fan of the NH-D15 must be installed slightly higher in order to provide sufficient clearance for standard height (32mm) RAM modules, the single center fan of the NH-D15S can be installed lower on the fin stack. As a result, the total height of the NH-D15 including fans is 165mm whereas the total height of the NH-D15S including its single center fan is only 160mm.

To get around this, yes, one option is to install the front fan (that sits above RAM sticks) to the back of the heatsink, at the I/O side. This should also enable you to mount the fan lower, thus giving you a bit more clearance in terms of CPU cooler height.

On the worst case scenario, you can't install the 2nd 140mm fan and close your side panel. In this case, you can replace the 140mm fan with smaller, 120mm fan.

-----

Only issue NH-D15 might have, is GPU clearance issue. If GPU gets into the way, return NH-D15 and get NH-D15S.

Oh, might want to install M.2 NVMe SSD drive and RAM sticks prior installing the tower cooler, since both would be below it once cooler is mounted.

 

[ForHonor13]

Thansk @Aeacus
I think I am stable for now on the MSI MAG, although I know we dislike this, I am just pushing for stable on MultiCore Cinebench" at this point.
1900 on Single and MultiCore at 14000~
Outside of Cinebench, I have stablized the temps for all 'use-case' applications (although Im sure I could grab a few easy degrees)
This is on STATIC 1.25V with the PBP @ 150 and MTP @ 190w w/ Boost Duration @48s

 

[Aeacus]

I think I am stable for now on the MSI MAG, although I know we dislike this, I am just pushing for stable on MultiCore Cinebench" at this point.
1900 on Single and MultiCore at 14000~

Some words about AIOs in general and also why i think replacing AIO with air cooler is better idea;

Spoiler: Click here to expand and read

As far as AIOs vs air coolers go, you won't gain any cooling performance if you go with AIO over air cooler since both are cooled by ambient air.

For equal cooling performance between AIOs and air coolers, rad needs to be 240mm or 280mm. Smaller rads: 120mm and 140mm are almost always outperformed by mid-sized air coolers. Single slot rads are good in mini-ITX builds where you don't have enough CPU cooler clearance to install mid-sized CPU air cooler.

Here are the positive sides of both (air and AIO) CPU cooling methods;

Pros of air coolers:
less cost
less maintenance
less noise
far longer longevity
no leakage risks
doesn't take up case fan slots
additional cooling for the RAM
CPU cools down faster after heavy heat output

Pros of AIOs:
no RAM clearance issues*
no CPU clearance issues
CPU takes longer time to heat up during heavy heat output (about 30 mins)
* on some cases, top mounted rad can give RAM clearance issues

While how the CPU cooler looks inside the PC depends on a person. Some people prefer to see small AIO pump in the middle of their MoBo with tubing going to the rad while others prefer to see big heatsink with fans in the middle of their MoBo.

Main difference between AIO and air cooler is that with AIO, you'll get more noise at a higher cost while cooling performance remains the same.
Here's also one good article for you to read where king of air coolers (Noctua NH-D15) was put against 5x high-end AIOs, including former king of AIOs (NZXT x61 Kraken),
link: http://www.relaxedtech.com/reviews/noctua/nh-d15-versus-closed-loop-liquid-coolers/1

Personally, i'd go with air coolers every day of the week. With same cooling performance, the pros of air coolers outweigh the pros of AIOs considerably. While, for me, the 3 main pros would be:
1. Less noise.
Since i like my PC to be quiet, i can't stand the loud noise AIO makes. Also, when air gets trapped inside the AIO (some AIOs are more prone to this than others), there's additional noise coming from inside the pump.
2. Longevity.
Cheaper AIOs usually last 2-3 years and high-end ones 4-5 years before you need to replace it. While with air coolers, their life expectancy is basically unlimited. Only thing that can go bad on an air cooler is the fan on it. If the fan dies, your CPU still has cooling in form of a big heatsink. Also, new 120mm or 140mm fan doesn't cost much and it's easy to replace one. While with AIOs, the main thing that usually goes bad is the pump itself. And when that happens, your CPU has no cooling whatsoever. Since you can't replace pump on an AIO, you need to buy whole new AIO to replace the old one out.
3. No leakage risks.
Since there's liquid circling inside the AIO, there is always a risk that your AIO can leak. While it's rare, it has happened. It's well known fact that liquids and electronics don't mix.

 

[Karadjgne]

Some rebuttal to that spoiler...

Fans make noise. So doesn't matter if aio or aircooler, the noise will be reflected by the choice in fans. Everyone claims aircoolers are quieter, but that's bogus, only aircoolers made by Noctua, beQuiet etc are actually quieg, but that's balanced by corsair maglevs, Lian li, Arctic etc. So a generalized statement saying aircoolers are quieter is totally bogus. My cryorig R1 Ultimate is Far louder at All levels than the nzxt x63 it replaced (the fans died after 6 years) and gets worse temps, even if equitable to the NH-D15.

Pros of aircoolers : less cost. Not really. Most ppl using an aircooler will also have to purchase additional case fans, which argb case fans can easily cost $30 or more Each. Which happen to come with the aio. So overall costs for cooling equivalence often actually favors aios now.

Main difference between AIO and air cooler is that with AIO, you'll get more noise at a higher cost while cooling performance remains the same.

That's not correct. Every cooler, whether air or liquid has 2 factors. Capacity and efficiency. Capacity is the wattage range, efficiency is temp per watt. In the 240mm / NH-D15 range, the Noctua is often more efficient, so will end up with slightly lower temps. However, the NH-D15 cannot compete with a 280mm or 360mm in either capacity or efficiency, both offering lower temps. You won't be using an NH-D15 to try and cool a 13900k that sees heavy usage, that 300w cpu is well beyond the 250w limit of the Noctua and will be in throttle temps constantly.

Noise will be upto the aio in question, but seriously, stop assuming all aircoolers are using Noctua fans, they don't and aircoolers are often louder than aios.

 

[Aeacus]

Some rebuttal to that spoiler...

How about the most important aspect between air coolers and AIOs - reliability?

Since what good is of 280/360mm rad AIO, when it only lasts for few years? And when pump dies, forking out full price of the new AIO as well. Given that you actually catch the point of thermal throttle and don't fry your chip beforehand.

And of course, the leakage issue as well. Last i checked, air coolers doesn't leak.

However, the NH-D15 cannot compete with a 280mm or 360mm in either capacity or efficiency, both offering lower temps.

And cryogenic fluids (e.g LN2) offer even lower temps, without being dependent on ambient air. Yet, none are commercially available for end users. At least, not yet. Perhaps in the future.

 

[Karadjgne]

You may as well ask about heatpipes leaking, I had a Noctua do that on me. Or how about getting a Phanteks PE14 with a base so warped that it rocked on a level surface, or how about the amount of motherboards that warped so badly I had to replace them when those heavy aircoolers got mixed with higher humidity levels.

Both have drawbacks, both have advantages, both have failure rates. Neither kind is any better or worse. Almost all leaks are caused by the installer. Almost all reports of very high noise because it's an aio are 20 year old bs written about Corsair SP's on the ancient H100, which Was a noisy beast that was prone to leaks when the corrugated plastic tubes cracked.

The failure rate of aios is less than 0.01%. Leaks from the factory take up maybe 10% of that figure. The other 90% is caused by the installer flexing the hoses incorrectly, not really any different than dropping an aircooler and denting the cpu.

And then there's full custom loops, in which it's almost guaranteed at first turn on there will be a leak, not getting one is an absolute fluke.

Saying aios are louder than aircoolers is false. There's aios that even at full blast are still quieter than a majority of aircoolers

The 280mm Arctic at 100% is 1dbA louder than the NH-D15, essentially the exact same because nobody on the planet can tell a 1dbA difference, yet look at all the aircoolers that are louder. And the Arctic isn't using 900rpm or lower fans either, like some of the aircoolers like the Fuma2 do.

 

[Aeacus]

Saying aios are louder than aircoolers is false.

Yet, your linked chart shows otherwise. Top 8 quietest CPU coolers are air coolers. While bottom of the chart is taken up by AIOs mostly.

Added better distinctions of which are air coolers and which are AIOs to the chart you linked:

As i see it, sure, some AIOs are quiet, just like some air coolers are loud, but in general, AIOs are louder than air coolers. Even the TechPowerUp statistics show as much.

-----

Even the noise issue aside, one can not ignore the fact, that when AIO pump dies, CPU effectively has 0 cooling. And since you can not replace pump on an AIO, completely new AIO must be bought, easily costing $100+.

While when using air cooler, only thing that can go bad, is the fan, while CPU still retains some form of cooling. And buying a new fan is peanuts, 10 bucks (e.g Arctic P12 at $9.99), if even that much.

It is far more cost effective to fork out 10 bucks every 6-10 years, than 100+ bucks every 2-4 years. And without the fear of CPU burning up when AIO is toast.

 

[Karadjgne]

While when using air cooler, only thing that can go bad, is the fan, while CPU still retains some form of cooling. And buying a new fan is peanuts, 10 bucks (e.g Arctic P12 at $9.99), if even that much.

True. If the pump fails. Which isn't common by any means. I replaced my nzxt x61 kraken after almost 7 years of 24/7/364 usage because the Fans finally failed and I didn't feel like spending $60 for some Noctuas.

But. Take a closer look at those quieter aircoolers. Excepting the NH-D15, DRP4 and scythe Fuma2 (which has absurdly low rpm max fans) All of them are Far below the ability of that 280mm aio. That chart compares coolers in the 130w-150w range vs a 300w aio. That's an insane difference in ability. Even the 140w class Old Goat Hyper212 is technically louder and you'd not put that on Any Intel x700k/x900k cpu.

And on the bottom of the list is the Asus ROG, which uses Noctua 2000rpm industrial fans. Put those same fans on a NH-D15, and it'd be on the bottom of the list too.

But here's the kicker. You put any 240mm/280mm aio on a cpu rated for any cpu that qualifies for those 'quieter' coolers, at the same temps or Lower, the aio is quieter, simply because the fans will not be running at anything close to similar rpm.

At 250w (9900k or better) , the NH-D15 is running maxed fans. The Arctic isn't, closer to 50% fans for the same temps. If you talk about a 200w cpu or lower, it's a toss up but a larger aio will generally be be quieter because the fans are nowhere near the equivalent rpm.

 

[ForHonor13]

@Karadjgne & @Aeacus -
The parts were delayed; but should have it tomorrow afternoon.
Will be updating with the following :

• Noctua NH-D15S chromax.Black (Replacing the MSI MAG AIO to see differences)
• 4x Noctua NF-A12x25 (to replace the Case Fans and add 2 front intakes once I remove the AIO)
• Thermalright Intel CPU Bracket (to fix some of the insane flex going on with the lever brackets)

As far as settings; it is 'stable' but not great at the moment.
I am wondering; once I get the above installed, do I continue to just undervolt this until I am seeing sub-80 on load?
It seems I cannot even game lightly on this CPU without it peaking around 85-95*
Ive never gotten such poor performance on an AIO; which I usually see them around 65-75* or less under load.

Any advice or 'processes' to be shared when trying to further undervolt this; but keep the boost around 5GHz? Is "Intel Overclock Thermal Velocity Boost" a good idea to enable? I have no experience with it.

If I see some "problem" cores or outliers; is there a recommendation to use the Per-Core Tuning to either drop the multiplier or voltage and/or add offset?

 

[Aeacus]

I'm not that knowledgeable about CPU OC, since i'm more specialized on hardware issues. Here, it's better that someone who has better knowledge of CPU OC, answers your questions regarding CPU OC.

 

[Phaaze88]

It seems I cannot even game lightly on this CPU without it peaking around 85-95*

Games typically do not see much power used by the cpu - there are some exceptions out there though.
https://www.techpowerup.com/review/intel-core-i9-13900k/22.html (scroll down to the section, 'Gaming Power Consumption')
Gpus can beat the pants off those kinds of numbers with ease.

If the fluid loop is strained, that's worse than the fans on the radiator being choked. The fluid is what is carrying the heat; you don't want that being hindered by some organic material in the loop.

NZXT's H-series cases weren't AIO-friendly - until the Flow models came out. In the H500's scenario:
-front mounted has access to cooler room air, but fans are heavily choked by the front panel and narrow side vent.
-the rear and top only support 120mm units. 120mm are, IMO, TERRIBLE for exhausts, as they can be easily overwhelmed depending on the rest of the hardware. They should only be used as intakes.

The case is intended for negative pressure setups, but is too restricted at air intake and exhaust both; the higher system power consumption is, the more difficult managing system thermals becomes.

do I continue to just undervolt this until I am seeing sub-80 on load?

Undervolting can work ok, until bios updates come into play. Bios updates can break UVs.
In time, the vendor will provide updates that improve performance or stability with tightened voltage curves.

Lower power limiting is an alternative, but is more effective on gpus or all core cpu workloads. As games usually don't use much power, it's not quite as effective.

 

[Roland Of Gilead]

@Karadjgne & @Aeacus -
The parts were delayed; but should have it tomorrow afternoon.
Will be updating with the following :

• Noctua NH-D15S chromax.Black (Replacing the MSI MAG AIO to see differences)
• 4x Noctua NF-A12x25 (to replace the Case Fans and add 2 front intakes once I remove the AIO)
• Thermalright Intel CPU Bracket (to fix some of the insane flex going on with the lever brackets)

As far as settings; it is 'stable' but not great at the moment.
I am wondering; once I get the above installed, do I continue to just undervolt this until I am seeing sub-80 on load?
It seems I cannot even game lightly on this CPU without it peaking around 85-95*
Ive never gotten such poor performance on an AIO; which I usually see them around 65-75* or less under load.

Any advice or 'processes' to be shared when trying to further undervolt this; but keep the boost around 5GHz? Is "Intel Overclock Thermal Velocity Boost" a good idea to enable? I have no experience with it.

If I see some "problem" cores or outliers; is there a recommendation to use the Per-Core Tuning to either drop the multiplier or voltage and/or add offset?

Hey there,

Have you tried Throttlestop from Unclewebb for undervolting? It's very useful for setting undervolts. Lots of profiles. I'm pretty sure it works with 12th gen CPU's

ThrottleStop (9.5) Download | TechPowerUp

The ThrottleStop Guide (2022): How to Lower Temperatures, Increase Performance and Boost Battery Life (ultrabookreview.com)

Basic Guide to Throttle-Stop - YouTube

 

[ForHonor13]

I'm not that knowledgeable about CPU OC, since i'm more specialized on hardware issues. Here, it's better that someone who has better knowledge of CPU OC, answers your questions regarding CPU OC.

Ah, well, thanks alot for all your help thus far. I will have the AIO vs. TOWER Competition this week and get back to you haha.

Hey there,

Have you tried Throttlestop from Unclewebb for undervolting? It's very useful for setting undervolts. Lots of profiles. I'm pretty sure it works with 12th gen CPU's

ThrottleStop (9.5) Download | TechPowerUp

The ThrottleStop Guide (2022): How to Lower Temperatures, Increase Performance and Boost Battery Life (ultrabookreview.com)

Basic Guide to Throttle-Stop - YouTube

Wow! Thanks so much Roland. That is amazing content. I will give it a more thorough review when I can later this evening. But this is exactly the kind of details I was looking for.