How this 120$ Noctua Cooler beats a 360 AIO

More about Hetzner AX52 Server:
derbauer.hetzner.com/en/AX52-en/
---------------------------------------------------------
Support me on Patreon:
/ der8auer
---------------------------------------------------------
Music / Credits:
Outro:
Dylan Sitts feat. HDBeenDope - For The Record (Dylan Sitts Remix)
---------------------------------------------------------
Paid content in this video:
- Hetzner Spot
Samples used in this video:
- All Noctua Products
Timestamps:
0:00 Intro
1:07 Hetzner ( Advertising )
1:44 Test setup & initial situation
2:45 Noctua NM-DD1 & NM-DDS2
4:05 Changing the cooler & delidding the CPU
4:56 7000X3D: Incompatibility
6:03 Direct Die Frame & KryoSheet
7:40 Noctua NM-AMB15
8:33 Switching to liquid metal
10:14 Temperatures
11:10 AM5 High Performance Heatspreader
12:19 Summary/Conclusion
14:48 Outro

[@Lishtenbird]

The way any air cooler "beats" any AIO for me is by being dead reliable. Don't need to stalk news channels to check if your AIO is gunking or leaking. Don't need to worry about pump whine or bubbles or whatever. Don't need to worry about it degrading in a couple years, and becoming e-waste. Unless you're a competitive overclocker or in production where any time is money, you don't _require_ an AIO.

[@DustyCruz]

Even on chips that get really hot, like the 13900K?

[@MikeKitchenman]

there are plenty of justifications for AIOs. Starting with shipping: by having almost no weight on the socket, there's no nearly risk of damage during shipping. It also leaves a lot more open space in the case to allow for work later.

[@SgtBluntman]

But theyre trendy and have pretty lights.🤣

[@Cynic_6489]

it just works

[@blakes8901]

You're right for about every situation, except small form factor enthusiasts. The space efficiency of being able to move the cooling solution is invaluable for small form factor builds.

[@mirrorsandstuff]

"you can always sell the delidder afterwards" - honestly, I respect that so much for a tool your team designed.

[@jilherme]

true!!

[@guts1016]

That's just common sense lol...

[@g.4279]

They may as well have a rental service.

[@VeryIntellijent]

@@g.4279 or a straight up de-lidding service

[@erkinalp]

@@VeryIntellijent shipping as unlidded might cause some trouble as cargo workers sometimes handle stuff carelessly

[@CaveyMoth]

I wish that more CPU heatsinks used a four point mounting system. There are four holes available on the motherboard, after all. Gamers Nexus has proven with their testing that you get much more consistent mounting pressure when using four points vs two points for heatsink mounting.

[@shanent5793]

Screw consistency. If I'm building a PC it should be optimal without being constrained by consistency. Two point mounting is obviously kinematically superior, so GN's results need checking

[@bosstowndynamics5488]

​@@shanent5793I fail to see how it's "obvious" that 2 point mounting is superior, and even if there's an intuitive argument favouring it the funny thing about intuition is that it's often wrong.

[@shanent5793]

@@bosstowndynamics5488 Even in this very video he complains about inconsistent mounting because the cooler shifts side to side. This is a result of an additional lateral constraint caused by the friction between the strap and the cooler. A roller would eliminate that constraint, allowing for more deterministic placement. Excess constraints on a any degree of freedom will introduce inconsistency with removal and replacement. This has been well-understood in optical design, where repeatability is measured in nanomètres. There is an alternative in the form of elastic averaging which is also very repeatable, but since it relies on elasticity it is less effective on small systems. If you look closely at the stock CPU retainer, it is also a two-point mount. This is a cheap way of getting consistent clamping without needing a torque wrench, though the extreme length of Intel CPUs highlight its limitations. The latest SP5 socket illustrates a combination design where the retention frame has two longitudinal mounts, a screw at one end and the other end held down by a hinge connected to a beam that is clamped to the board by two screws. The frame is very rigid, with a tall web along the sides, while the six or more contact points to the EPYC heat spreader are just stamped protrusions, allowing for elastic averaging. Then the cooler itself is secured with six small studs around the perimeter, also using elastic averaging. Direct-die is a very different situation, with two or three small, asymmetrically distributed contact surfaces. And since it's DIY I think it's more important to keep costs down instead of focusing on repeatability when the system is infrequently assembled. I think it's more than just an intuition and I would appreciate it if you could point me to that GN article or video

[@glebglub]

@@bosstowndynamics5488 the shifting side-to-side is potentially beneficial when using a thermal paste as opposed to liquid metal or graphene sheets, allowing it to spread out as well as push the excess out in doing so, as opposed to a 4 point mount where excess may be "swirled" around the middle. of course when going to direct-die as opposed to a heat-spreader and/or using liquid metal/graphene sheets, these points are moot, since direct-die is a less homogenous mounting surface, LM requires a thin-as-possible application, and graphene sheets are extremely fragile. the "obvious superiority" is with homogenous flat surfaces and thermal paste, which is what >90% of people will do

[@glebglub]

@@shanent5793 GN's pressure mounting testing uses a pressure-sensitive sheet of... something (I don't know exactly what it is) that changes colour depending how firmly its pressed in that exact area. as such, considering the side-to-side shifting tore apart Roman's graphene sheet, it's possible that the results of GN's testing are not entirely accurate (however the way to get it more accurate would involve some stupidly expensive processes, such as X-Ray Fluorescence Photospectrometry by mounting the cooler to a glass substrate separated by a liquid whose refractive index changes readilly with pressure, but since XRF can only penetrate so deep, it may necessitate electron microscopy, which is a lot more expensive in turn afaik)

[@totodoro]

I love how the perfomance wasnt on par with your expectation, but showed the results and shortcomings either way. That's what people have come to respect about you and thermal grizzly.

[@GOLTURBO555]

Respect?

[@garryson856]

@@GOLTURBO555 Respect acquired from honesty and not trying to sugarcoat it

[@AeroEngr69]

When deliding my 7950x I found that soaking the glue covering the caps with 99.5% ISP softened it and made it very easy to remove. I applied a thin coating of MG Chemicals 422C on all of the caps and bare contact areas. The direct die frame contacts this thin coating, but doesn't seem to affect anything negatively. So this is an option to those wanting a conformal coating still. Also, if you have an airbrush, you could tape off the dies and rear contacts and spray this conformal coating onto the caps and contacts for an even thinner coating. I did this with an r-pi a long time ago when developing a smart solution for an HVAC company. It works perfectly and no thinner is required, but definitely clean the airbrush immediately after spraying. Learned this the hard way, lol.

[@jameshackett9992]

Would conformal coating rated for transformers be any good I have 180c spray air dries etc?

[@EliteRock]

In my experience the plates on most coolers need to be flattened to work optimally (or even properly) with direct die. Noctua mill theirs to an arched profile and a lot of AIO's are domed to help them confirm better to the shape of most IHS's. Otherwise, on the almost optically flat surface of the die(s) you end up with only a line or point of direct contact. Problem then is you've removed the nickel layer from the plate and it will react with the gallium (LM), but you can pre-treat it with gallium (cheaper than using LM itself) by agitating it into the surface while using a heat-gun or hairdryer (I mask off the plate with Kapton tape to make the treated area precise which soothes my OCD) effectively creating a 'barrier layer' of gallium/copper amalgam which will keep the LM application stable for longer. ETA >> I have an NH-D15S (single fan) which keeps my i9-9900KF (5.0/1.32V) under c.75°C in the summer while it pulls c. 165W in CB R23.

[@bligh1156]

Noctua air coolers are awesome. Even the old D15 is still great depending on application - plus no messing around with tubes, water, pumps and all the unreliability that is introduced with those things.

[@mikem9536]

The first thing you do when troubleshooting is ditch the water cooler and slap on an air cooler.

[@pickleadaykeepsthedoctoraw5542]

@@mikem9536 No it's not.

[@lilpain1997]

@@mikem9536 If you are troubleshooting a PC that is not first thing you do...

[@zakelwe]

@@lilpain1997 That's a massive generalisation. Most troubleshooting in PCs is just down to connections

[@randomguydoes2901]

@@lilpain1997 if you want to shoot some troubles before they appear, that's definitely what you do

[@HCGonzalezJr87]

Interesting demonstration. What's even more impressive is the level of honesty and transparency in this video. You showed us the test cases you ran and included outcomes for the products even if they were not favorable. Integrity. I'm sure the marketing department reached out afterwards though. :P

[@JSmith73]

Comparing stock on AIO to delidded on Air though? Scientific value: Zero, except that delidding is probably a cool thing.

[@Zecuto]

​@@JSmith73don't mistake AIO for custom loops, people purchasing them often want a plug-and-play solution. Comparing delidded Noctua tower cooler to a custom water cooling that runs without delidding - that would actually be a bad set up in my opinion.

[@HCGonzalezJr87]

It’s an interesting demonstration of options. Comparing apples to oranges of course but some people might not want to get an AIO for whatever reason.

[@Zerod-rn3ye]

You mean the lack of honesty and transparency. Issues? #1 Comparing only a very low 1200 RPM normalized fan speed between the two which is not only insanely low and not realistic for a real load hence his asinine thermals even at the end when no proper fan curve to manage noise vs thermal needs, but such low RPM fans will have issues pulling air through the grill of many types of AIO radiators (some worse than others). Yet he doesn't even really acknowledge the normalized vs max cooling potential which is misleading. #2 He is comparing a 170 Euro price setup vs a 200 Euro AIO when 45 - 120 Euro 280/360 AIO's destroy this setup with proper fan curves when allow to run higher, often by a very considerable margin. Thus this is not saving money, being actually considerably more expensive in fact... and it has worse results and you have to deal with liquid metal which simply isn't for everyone.

[@lxfguits]

I have always just used pretty much random paste for my builds. Has worked well enough for me. I recently serviced my custom loop ITX build and didn't bother to even apply new paste because it was still soft and I haven't had any thermal issues. The paste I used was the one that came with my alphacool water blocks and it's three years old now. While putting it all back together I just smeared all the previously squeezed out paste back over the heat spreader and haven't seen any negative impact compared to what measurements I did when I first built it.

[@a120068020]

I have 2 i9 13900k systems. One had an NH-D15 the other an Aorus Waterforce X280. Both have similar cooling capability but the fans would run up and down all the time on the AIO and the fans were noisier so I put the NH-D15 in the 2nd system too.

[@InnocentiusLacrimosa]

That is an interesting real world test you have made there 🙂

[@theHardwareBench]

If you mount the rad externally the Waterforce would be superior, rads in cases soak up lots of heat from other components really easily.

[@cemsengul16]

I learned the hard way that AIOs are not actually more quiet than air cooler. Sure your fans can run at really low rpm but you have a screaming pump which gives you a headache. I would take loud fans all day instead. My ears get a ringing from this AIO I bought and I start to develop a headache a few minutes after booting up.

[@brianrobinson3961]

Would be interested in seeing a demo of the KryoSheet and what temps/performance you can expect with it vs liquid metal and thermal paste under normal mounting conditions.

[@kBitre]

He already has a video showing that

[@arespirit]

Carbonaut is the new one I want to see tested

[@shadow7037932]

And compare it against things like PTM 97950/Laird TPCM which both have long long service intervals.

[@brianrobinson3961]

@@arespirit KryoSheet I believe is the new product. Carbonaut is the older version. This new version uses Graphene but I haven’t seen a video really showing what types of temps you get vs high end thermal paste.

[@floodo1]

“Interestingly this is still isnt quite as good as I expected”

[@benjaminsmekens2344]

I've recently gone through the process of delidding my 7950x, used the TG delidding tool and TG LM. My heatkiller pro block didn't make good contact with the die (the metal standoff's on the backplate where preventing it from doing so). So I took out the dremel and went to work. After that I used some smaller z-height washers and the block mounted perfectly on the 3 dies. All in all, lost about 10-15c, and gained performance in the process (chip can do all core 5.8Ghz at 80-84c max temps), bringing my CB23 scores to over 40k. I used nail polish over the capacitors to avoid shorting, and while this was a bit of a nerve wrecking process, I'm very happy I did it. The heavy full copper heatkiller pro block doesn't have any issue with absorbing the heat load dissipated by the dies. So while definitely not recommended for the first timers, it was my next step in going even more extreme on my system and all said an done, happy camper! Thanks for the great tooling provided to make the process so much easier, even if I did still have to do quite a bit of diy (but that had todo with the waterblock, not the method in general). 👍

[@youtube7076]

those thermal grizzly products are, imo, pretty darn good, i am happy to see someone who, in their video, is completely correctly utilizing the products, well done! you earned a sub from me, and now i look forward to more videos from you. thanks.

[@daveg4417]

I switched to Noctua air coolers more than a decade ago. My newest system, an Intel Xeon W7-2495X 24-Core, ASUS W790 ACE, 512GB DDR5, also uses a Noctua NH-U14S 4677. I stopped using water cooling about a decade ago when I had a $6000 system leak and corrode the main board and case. I'm too paranoid now, I would be checking a water system daily.

[@itspodin]

Would have been interesting to see the difference between stock configuration of the Noctua compared tot the delidded situation. The comparison seems a bit flawed since we don't know how much performance gain is due to the cooler itself.

[@user-yy2bk2pg1l]

That's the key trick in this video 😂

[@__aceofspades]

That wasnt shown for obvious reasons, as its a marketing video. While Noctua makes good products, the cooler here isnt outperforming or even matching 360mm AIO. Instead we are just seeing how bad the AMD IHS is, and how you can significantly lower your temps by removing it and going direct die.

[@JSmith73]

Indeed. My thoughts exactly.

[@-opus]

Surprised he did that. I stopped watching as soon as he mentioned the dodgy douche Linus. Sounds like I didn't miss much though.

[@EricFixalot]

@@__aceofspades To be fair, you can go anywhere and compare Noctua to an AIO, the question here was "What if you spend less on an air cooler, but de-lid it?". The fact of the matter is that the direct die air cooler did perform better than a 360mm AIO with the heat spreader. If you're budget is $200 for cooling and you want the reliability of an air cooler, this could make sense.

[@CodeMonkeX]

The fact they are releasing this for so cheap, and giving out the STL for printing makes me feel really good about spending extra for Noctua.

[@Zerod-rn3ye]

Not cheap. This is so much more expensive a setup at around 170 Euro than a superior 45 - 120 euro AIO that can trounce this. He really was not honest when he said 200 Euro AIO.

[@Rakadis]

Had a bad run with my water cooling setup in 2016. Killed about 8000 USD worth of equipment. After that I have air-cooled everything and I will probably never do water-cooling again. So high performance air-cooling makes me happy. (Edit: Typo)

[@overtrist]

Lots of people learn their lesson that way. Watercooling is just not safe.

[@phenomenologicalparadox5216]

What happened to your setup? Something in a custom loop break and spill while running or something?

[@soragranda]

​@@overtristIt heavily depends... take in mind tons of people are currently running water cooling setups and they are fine. A good company support and maintenance matters and made a difference...

[@s13shaka]

Nobody cares.

[@Cynic_6489]

@@s13shaka you clearly cared enough to comment, you mouthbreather.

[@52Sierra]

there are vivid pros and cons for both air and water cooling. but asides that, noctua coolers are really good. shame that some ppl really hate their products without reason.

[@SnazzieTV]

The reason is because the colour is ugly as hell. they dont help themselves

[@gordoncheng9819]

From what I have heard before, there are lots of people hate Noctua's stuff simply due to the color (poop brown as they called), but I actually like that brown color a lot lol

[@52Sierra]

@@gordoncheng9819 there’s chromax for ppl who don’t like brown, but they still hate noctua for some reason unknown, calling them ‘poop’ like you said…. if they like rgb that much they can just buy something else but they choose to swear at noctua and ppl who buy their products

[@ReinaldoGonzalezreix2x]

just asking, who hates on Noctua? i mean, their products are a little expensive in comparison with other brands, but the quality is above all of them i think they even offer a lot of value, i have a server with an ancient nh-d14 that have like 10 years working and it stills perform like the first day

[@Humanaut.]

​@@mikeycracksonNot half the price, the PA is 1/3 of the price. Only question is longevity and warranty but given that you can buy 3x PA or 1x Noctua I'd still get a PA unless you the new Noctua beats it by a large margin. Under heavy loads the Noctua is quiter too but since I game with headset I don't care.

[@nonejust2269]

Interesting information, enjoyed watching

[@rick5078]

mg chemicals 422C would also be a great option to protect those components around the dies. Just apply a layer (with a brush) thick enough to protect the components while thin enough to not interfere with the direct die frame.

[@TRC_WA]

I’m running the NH-D15 on my 7950x3D… and just added the new offset mounting bars. I ran a 30 min all core burn test and peaked at 85C! 😂 Everyday use temps are much lower. Encoding peaks around 70C. Gaming even lower. Definitely didn’t need an AIO! 👍👍

[@username8644]

What did you use to stress test? That's pretty hot tbh. That's what I get after 12 hours of prime95 (which is by far the hottest stress test for me), on my 7980xe which I've overclocked quite significantly. And that's a notoriously hot CPU that's been overclocked. Cinebench is 10 degrees less hot too. Edit: with a 360 mediocre performance AIO. Also 30 minutes isnt long enough. Prime95 won't even hit peak temps in 30 minutes, takes hours.

[@commanderoof4578]

@@username8644the CPU is literally designed to push itself till it hits a limit Normally thermal limit is hit and you peg at 95c but sometimes power limit be the limiter For Ryzen 7000 series anything under 90c is seen as optimal cooling

[@username8644]

@@commanderoof4578 Right. And my cpu can go to 110 degrees. So I'm technically running significantly cooler than this guy with a significantly hotter CPU that is overclocked. Which is my entire point. Edit: and also informing that a 30m stress test isn't really relevant.

[@commanderoof4578]

@@username8644 the Ryzen 7000 series are designed to FORCIBLY TRY TO RUN AT 95C and that is the limiter most of the time Also Ryzen 7000 series has a super bloody tiny die and a monstrously thick IHS Again its by bloody design that its doing those temperatures and poor choice too keep AM4 cooler compatibility play plays a large role in it

[@TRC_WA]

@@username8644 I used Aida 64 and Furmark… and in both cases hit 85C in about a min and stayed there for 29 mins. Pretty good indication the NH-D15 is working as designed because the temps didn’t climb any higher… not even 1C. You’ll never convince me running for 10 more hours would change anything. If they don’t move for 29 minutes they won’t move for 290 minutes either. Additionally… this is the only time my 7950x3D has even sniffed 80C+… I don’t consider it running hot at all especially when I can set up 3 hours of encoding with Handbrake and stay in the 65-70C range. Did I mention gaming temps are even lower? (4K Ultra with 4090 doing all the work)

[@justagamer1985]

will the heat spreader and micro waterblock still be compatible with x3d cpu? Also for the kryo sheet, will it also perform better for non-delid cpu as well as gpu (which is technically all direct-die)? Have been assembling pc for 5 years but never mastered thermal paste application lol.

[@wonderlustgaming8977]

Great video! Im very curious when the am5 high performance heatspreader will be available to purchase?! Im waiting patiently :(

[@lrmcatspaw1]

I think I would really like Derbauer to design a cooling solution for laptops.... I have this strange feeling that he could colab with a company and make some of the sickest cooling solutions that would still be practical.

[@Jrfeimst2]

Did I miss something? Are you comparing a 360 aio with a non direct die vs noctua with a direct die? If that is the case it really isn’t a comparison since you have different variables.

[@ssl3546]

He's keeping cost fixed. i don't know how he picked the AIO but it costs significantly more than the Noctua cooler.

[@Jrfeimst2]

@@ssl3546 oh ok 👍

[@FireCestina1200]

18$ Assassin King 120 SE cooling 7500F to 7800X without any sweats. Cooling 7800X3D with no much struggle. And his Peerless Assassin cooling 420W processor for 35$. So literally freaking anything. And for 35 bucks and no risk of any failure. Huge W

[@quirkyMakes]

Fascinating. Although I cannot afford these things I still run a enlarged air cooler. But I don't really overclock too much instead I clock for electical/thermal stability. I try not to allow my components to get over 70c. Memory still runs around 50c to 60c depending on the day. Afternoons are the worst as the temperature here gets up to 43c

[@denisnatera3633]

where you live?

[@HarleyN93]

You should come out with a rubber gasket that you can put round both dies that can be squashed slightly by a heatsink to make liquid metal even more safe. hope that makes sense

[@F34RTEHR34PER]

I wonder if you could stack two kryosheets, and if there is a benefit to that. I cut to fit one for the ROG Ally, so I'd be super interested.

[@Null_Experis]

The kryosheet is a graphite pad. It's not optimal for cooling unless both sufaces have an absolutely perfect mirror finish on them, as it would produce air pockets in any imperfections as well as the graphite being not the best conductor compared to a silver based compound or even liquid metal. I use graphite pads in retro computing because older CPUs were designed with crappy lithium compound as the standard tim, but modern components are designed for higher performance tim.

[@tiggertsk]

It's a graphene one, and it has higher conductivity than any thermal paste. Liquid metal is better. You are correct here. But most people find its application quite difficult.

[@Null_Experis]

@@tiggertsk This isn't graphene, it's graphite. Actual graphene is a single atom in height. I own a few of these pads, and they are NOT for heavy use. Thermal grizzly doesn't even list the W/m*K rating for them, and even if it were comparable to a paste, it's still a solid sheet and cannot fill microscopic faults and cracks in the heat spreaders of the CPU and heatsink, these create air pockets that act as insulators. Even in a direct-on-die solution, the heatsink would need to be a perfectly smooth surface with no imperfections for this to work as efficiently as TIM.

[@naimah92]

Is there going to be an update to the direct die frame to make it compatible with the X3D CPUs?

[@MrHakisak]

I really want that high performance heat spreader for lga-1700, looks so cool!

[@TechLevelUpOfficial]

Couple points about the video. I believe using a higher end CPU such as 7700X would be the minimum target for people willing to delid or invest in these high end parts. Secondly you don't need 200$ to get a good AIO, best example is the Arctic Freezer II it sells for half that amount so the conclusion based on an overpriced cooling solution is a bit misleading tbh. Without forgetting the fact that delidding isn't for everyone and it can still break your CPU, all in all this was a good experiment.

[@Argo16x]

After many years, I built a desktop PC with a Core i9 13900K. For the first time in my life, I realized that the limit is not the cooling solution (now I've a custom liquid coooling solution based on Heatkiller products), but the delta between the silicon and the IHS. CPU makers should definitely work a solution for that.

[@arvopenaali896]

I just want a lowish profile horizontal cooler with like a 14cm fan. There's plenty of room on most boards esp if you use a riser. Would be a big upgrade for sandwich cases.

[@cthree87]

I would have liked to see a comparison between the air cooler normally installed vs delid rather than AIO vs delid air cooler. I just changed from an H100i AIO to a single fan Noctua 12S cooler and saw similar thermal performance but with one less fan and 12 dBA less noise. I can’t tell from your test whether I would get any additional thermal performance from deliding the CPU which what I was hoping to know.

[@spuchoa]

Great video, I still have high hopes on maintenance free solution.

[@Serphilith]

Now try an Arctic Freezer 360 with liquid metal, vs the Noctua

[@paypur8193]

I took a look at the compatibility list for the NM-DD1 and didn't see any low form factor coolers like the NH-L9. Is it not possible to design a direct die spacers for these coolers or are the benefits negligible with smaller coolers so you didn't bother to make them?

[@nazgu1]

Roman, any updates on when the direct die blocks/heatspreaders will become available for general public? :)

[@Ghennesph]

NH-U12A is just really good. It outperforms the D15 at lower TDPs, iirc, by more than it loses to it at higher TDPs.

[@TykeMison_]

Air cooling has a weakness with heatsoak though, so while Direct Die-ing it definitely improves perf, one has to ask "for how long at how much power output does the benefit sustain?" If you're aircooling a 65-90 watt (true watts not "TDP") CPU on a single-stack heatsink it probably won't be a problem. Personally I would try to find a way to get direct airflow to the hottest part of a heatsink, the gap between the baseplate and fin stack where the pipes are the hottest.

[@theHardwareBench]

So does liquid cooling if the radiator is in the case.

[@nerophine]

hi Der8bauer what were your thoughts on the 7600 ryzen ? good overclocker ? was it any better as a tuner cpu than the 7600x ?

[@Riotmakertk]

I wonder if the KryoSheet work well with the older game consoles, like say PS3 and PS4 1st gens? In any case love the frank and honest coverage as always. On a side note my Noctua NHD-15 has been going strong since Skylake, currently running on a 5800x. Plan to avoid AIO's but I can see some strong appeal in delidding now...

[@Riotmakertk]

@@elcactuar3354 corrected 👍

[@tandlose]

Curious why you didnt compare it to the arctic liquid freezer 360 instead 🤔🤔🤔🤔

[@michal1693]

i hope for the more test of the "AM5 High Performance Heatspreader" with different heat transfer solutions. Hopefully normal thermal paste, Thermal Grizzly Kryonaut

[@daigriffiths399]

'that should be quite entertaining' - I laughed out loud. Thank you.

[@doingbartthings]

Ahh.. did I miss where you tried the direct die with the AIO to see the difference? Maybe not possible in this video due to not having a way to reliably lower the mounting position? I would of like to see more results with your heat spreader that thing looks great. I bet it does really well with some liquid metal.

[@boidsonly]

What difference would would it make, if any, by taping the CPU before placing the cover on?

[@kelownatechkid]

This looks like a great option. With the kryosheet air cooler combo you get very high reliability as well, no liquid cooler maintenance required

[@teethashaquan5395]

Don't bother. I bought one recently and it breaks apart too freaking easily with an AIO. I'm expecting a revision very soon for this expensive junk.

[@hinchlikescake7592]

Those are some excellent thermals. I wouldn't do it as moving a PC on a direct die with that much weight sounds to be recipe for disaster but very interesting results.

[@SB-mr2nk]

Damn I needed this two months ago Lmao! I’m moving to Florida tho so maybe I can remount

[@Targetlockon]

Where to buy quality Kapton tape and what dimensions? Recommendations please thank you Der Bauer 👌

[@fluffi8161]

Moin alter bekannter! Kann man heutzutage beim OC noch Strom sparen? Also gegenüber Stock. Also mit Sweetspot oder läuft die CPU immer höher als unter Stock wenn man Werte festsetzt?

[@anub1s15]

aw it would have been an excellent moment to move from cryosheet to carbonaut :P at least...i think the carbon flap is slightly less sensitive. (been using one for about 2 years now, so far so good as I mainly used it to not have to worry about the thermal paste rather then have the absolute best possible goo between the heatspreader and cooler)

[@Lucatoni22]

This is why i love Noctua and Thermal Grizzly. Respect

[@pprietojuez]

Is the Direct Die Frame compatible with the Arctic Freezer AIO mount system, or is there a problem withe new heigth.

[@Kiki-lc3ng]

Did you use the NH-U12A with 1 fan (as in B-roll) or with 2?

[@arthurabraham3271]

what happens if we have a cpu running at a lower clock but at stock voltage? (like if we go into power management and reduce minimum processor state)

[@darkmanure]

Awesome how you got a collab with Noctua.

[@Belgaer]

Came to this video because I started getting a pump failure error light, then noticed the pump whine (hearing loss).... Now I'm researching for my next air cooler. 😂 According to my order history, the 3 year anniversary of when I ordered this Corsair h150i Elite Capellix 360mm will be 2 days from now. 🤔

[@Hito343]

I love air coolers there just look awesome, big heap of metal with a fan on it, makes the PC look more ''tech'' then just some small block with tubes.

[@anaveragehuman2937]

Would be interesting to see some curve optimizer results from a delidded cpu. CO alone (no pbo or other optimizations) on my 5900x dropped multicore CBR23 temps by about 15 degrees C, upped boost clocks 100-150mhz, and gave 4-5% performance boost.

[@tiggertsk]

That is not an indication for any cpu. It depends on silicon quality solely (plus maybe a tiny bit on LLC). Imagine overclocking guide written specifically for 'golden sample' CPU that just straight up doesn't work for any other processor. Is that helpful at all?

[@GhostMotley]

Is there an update on when the AM5 and LGA1700 high performance heat spreaders will be available?

[@barstool2985]

The only issue with air coolers is case compatibility, unfortunately my pc0-11 D is 155mm max, being 3mm too short otherwise I would have gone with a Noctua cooler for my recent upgrade. The Scythe Fuma 3 fits but isn't out for a few months, i'll probably swap out the corsair H100i I got once it's released though for the Fuma because the pump sound is atrocious on the new corsair AIO's.

[@kliibapz]

My case is also 155mm max and yesterday I installed NH-U12A on my 7600. Still there is 2-3mm gap between glass. I think there will be no problem about your case too. It's Zalman N3 btw

[@waynetuttle6872]

THIS, this is why I have no problems buying the products you have that fit my use case. You have zero issue with calling it what it is. I’ve watched other manufacturers force the issue and you had zero issue with explaining why you wouldn’t use your product in this situation once you figured out that it wasn’t going to work with ease.

[@fpshooterful]

Am i missing something here? How come you didn't test the 360 with the delid CPU? I just saw you test the CPU at stock settings with the 360mm AIO.

[@norkris8729]

thats how this clickbait works :) good catch

[@Z4KIUS]

I wonder how much heat could P1 take away from 7950X with direct die, liquid metal and the offset mount surely it won't reach the 230W PPT, but with some CO the performance drop could be negligible

[@streife8539]

soo cool :) punn power

[@Wrackey]

6:34 wouldn't modifying the direct-die frame cause un-anodized aluminium to be exposed, making it conductive ?

[@hallsofvalhalla1749]

This is cool, although I find that the my 7800x3D run quiet with my noctua NHD15. I get temps around 80C when I push the processor. Note that I didn't bother overclocking. I can get single core speeds of around 5ghz, which is plenty. I get why folks may want to de-lid the ryzen 9s and some intel chips though.

[@garystinten9339]

Could you do a NH-L9a AM5 review with a 7800x3D? Thanks.

[@Jango_D]

i run a 360mm alphacool aio on my 7800x3d only because i screwed up my GPU's cooler and ended up just watercooling it to i figured alphacools quick disconnects made it the easiest option for me. 7800x3d + 3080 in one 360mm loop runs perfect neither part gets too hot at all.

[@chrisr2543]

I love that.... "which we will de-lid, obviously...." welcome to der8auer's world of "obvious" XD

[@Razor2048]

Please do a test with just the offset mount and stock IHS with an NH-D15.

[@Synaps4]

Amazing video. Gives me all kinds of ideas. When you're talking about the air cooler being at a disadvantage in the case remember you can always build a custom fan duct to pull air direct from the outside rather than inside the case, and this does have good results, but the variety of cases often means it needs to be custom.

[@PainterVierax]

yup! that's exactly what I did decades ago on some socket A build (a weird time when AMD chose direct die on desktop). Now, most of the time I just remove the side panel of the case.

[@EliteRock]

Last two builds, I've mounted the air cooler back-to-front, with the rear fan blowing IN with a ghetto-mod filter on it (a frame cut out of ABS sheet, attached with 3M magnetic strips, can be detached/reattached by feel), with a centre/top exhaust. CPU cooling is completely unaffected by heat from the GPU, the hot air from which is also very effectively exhausted by the top/centre fan. Of course there are intake fans (two front and one centre/bottom).

[@ViktorJaneba]

I installed Kryosheet last night (was tired of periodical repasting) to my i9 13900k - cooler is 360 Zalman Reserator with 3x push 3x pull fans - and honestly, I am disappointed and going back to paste. It starts throttling and goes to red numbers after one minute in CB R23 multithread test (OCed CPU, all cores, no limits, 285W max), while paste kept it at max 96 degrees, no throttling during 15 minutes test. Ambient is the same, the room is A/Ced. Will move that pad to the other rig though, thermals there are not such insane.

[@MadMathMike]

6:24 Instead of filing away the frame, could you file (or sand) away some of the glue instead?

[@eleventy-seven]

I love air cooling. Great video

[@kungfuwitcher7621]

On my next build I am thinking about a de-lidded 7600 or 7700. Thing is I don’t have a great deal of engineering experience for want of a better term, so is this a process that really a bloke like der8 should do, not a enthusiast like myself?

[@JPWestmas]

I love it when cheaper solutions beat out the performance of the more expensive and stylish options.

[@unifairsum21]

Always wondered if you could just rent the delidder out like you can with other tools just pay a fee and shipping and then ship it back once done but I respect you saying you can just sell it after

[@matasa7463]

Could you use Honeywell PTM 7950 phase change thermal pads on these?

[@rtswinxp]

So if i don't want to use liquid metal at all, deliding has no point using with AIO or Air regardless?

[@Plutodeli]

Any news on the AM5 High Performance Heatspreader availability?

[@Sumpula]

Id like to see direct die tests of ITX system with like max 50 and 70mm high coolers.

[@busshead]

Can I do this mod with the corsair aio? I already have it so I am only interested in the cooling gains.

[@stanley3647]

Bear in mind - new revisions of NH-U12A has offset mount in standard (in box) Works well with my "stock" 5800x 3d

[@jasonmunoz6494]

Does the offset mount work with air coolers from other brands?

[@deathrager2404]

i still got a kryonaut and conductonaut kit from you guys from my older rig where i delidded a 4670k. ran like 15-22 degrees cooler after that. i still use the kryonaut ngl, and it still helps alot. currently on am4 and my temps have not passed 62 degrees celcius (5600x with cooler master hyper 212), i did undervolt to like 1.082 or 1.088. either way, temps are great.

[@Aonobaka29]

I do have a NH-U12A so very interesting video to look at

[@minecraftmike5193]

What a weird video lol Let's take another look at his final tally: Noctua Cooler: €110 Delidder: €70 Direct Die Frame: €40 Offset Bars: €5 Grand Total: €225 In his calculation, he intentionally excludes the delidder's cost, arguing that you can resell it at a minor loss of about €10-20. To me, this seems like quite a leap. Given that I regularly browse the second-hand market, I can attest to how infrequently these items appear; It's important to highlight that these are HIGHLY SPECIALIZED ITEMS, NOT something that the average hobbyist or even many enthusiasts tinker with. If this video intended to craft a compelling argument for these products, it stumbled significantly in my view. It not only portrayed the KryoSheets as unreliable, but it also made the whole process seem fraught with an unwarranted layer of complexity and inherent risk. To top it off, all it managed to present were marginal improvements when compared to a standard AIO.

[@madb132]

Guessing not for you then, eh. The point Roman is making is that, if you already own a Noctua cooler, then you can still use it if you were to upgrade to 7000 series cpu and still benefit without buying a AIO. Simples. So your grand total should be 110 less or even less if you were to sell the Delidder,. Not for you, that's fine but their will be plenty wanting the option.👍

[@kaminekoch.7465]

@@madb132 110 + lost warranty + potentially broken CPU when delidding. You will probably pay twice as much in power bill after the OC in a month than what you might have saved by just buying a proper CPU for your use case.

[@marcotomiri3440]

I don't remember where i hear it, maybe GN, or maybe the Linus video's to noctua direct dye stand, i remember they said that delid with normal termal paste won't work, the die it's too small and too much heat will come from dies, that termal past get saturated. Direct die works only with liquid metal.

[@paxxo1985]

An AIO pump can fail, but an air cooler cannot. Noctua also provides free mounting kits for new sockets and delivers them to your home for free. I have been using a Noctua U12-P for 15 years, changing sockets from Intel 2600k to various other Intel and AM4 AMD sockets, and currently, it is mounted on my AM5 socket. It still functions perfectly, just like the first day. The fans work flawlessly, and it remains silent. Let's see if an AIO can last for 15 years and if AIO companies provide free mounting kits for new sockets. Noctua is the best 70 euros I've ever spent since 2008.

[@JohnSmith-pn2vl]

quality is always cheaper :) so get an Apple Silicon which doesnt even need cooling :D

[@madb132]

@@JohnSmith-pn2vl " so get an Apple Silicon which doesn't even need cooling" 🤣🤣 This is PC, "We" do as we like, Not told what to do by a company. 😁👍

[@BuzzKiller23]

Fans fail too

[@kimnice]

@@BuzzKiller23 Very very rarely..at least with Noctua. Fans on my NH-D15 are now 8-9 years old and they work just as well as they did when they were new.

[@paxxo1985]

@@BuzzKiller23 fans can be replaced, aio pump cannot. Also aio has fans too lmao

[@saltypepesugar]

This is cool! Is there something similar for intel 13th gen cpu?

[@kaizenbg]

My Noctua U14S is taking care of a third processor now. I'm never giving up on that...

[@wdowa94]

What would happen if you replace kapton tape with something with average thermal conductivity?

[@Kefka2b]

Would that be compatible with low profile coolers? Interested in building a Fractal Design Ridge with an i7 13700k.

[@cristianroth8524]

I'm really curious what would happen with even more powerful fans. The Noctua Industrial 3000 is what I'm talking about. How much would the extra airflow and pressure improve cooling?

[@No1x3N]

not much, specially if noise normalized

[@I_Really_Enjoy_Not_Seeing_You]

You really got me thinking. I have both the nh-12ua and the industrial 120mm. Only one, though. So, I just went and swapped them out. My not so scientific test shows at best it was a 1-2 degree difference and like, a 50 decibel difference. The 3000 is normally my case fan to expel hot air from after the cpu cooler typically spinning @~1400 rpm and the curve is set to immediately rev up to full speed when something reaches a temp I don't like as both a makeshift air raid siren to let me know to back off and to assist further in cooling things off in the process. Maybe with 2 it would be 3-4? Possibly more without a slower fan ahead or behind? At comparable speeds to the 12ua, performance was actually worse with a LOT more noise up until you start hitting ~2600 RPM. Then even more noise, but ever so slightly cooler. I then went back to my normal config. Case used is a Coolermaster HAF XB EVO with 2 industrial 2000rpm 140mm intake fans, and that 3000 120mm exhaust in the back as previously mentioned. On a side note, yes, the 3000 is significantly louder than the 2000 at any speed. All 3 also seem make an odd chittering noise when spinning slow on PWM that doesn't happen when running DC.