One more note: I tested the TR CPU with another 360 AIO just to be safe and I still had 70-80°C in Windows idle without doing anything. So something must be wrong with that CPU or maybe the VRM Edit: After all your feedback I will make sure to work on an update video to investigate the issue further :)
The 3960x has a 4x6 CCD configuration. And CCDs are 8-cores with disabled ones to get to 6-cores in a CCD. And if you look in HWMon, you'll see that 1 CCD is at 90+C while the other 3 are only at 70 and 80. So it's clear that this one CCD is the source of the heat. But, the fact that this thermal issue became progressively worse over time, WITHOUT giving you computational errors (which you would get from a bad core, memory or mainboard), makes me think one of these "disabled" cores wasn't cut properly and over time made a progressively worse short, creating the heat and power draw. Something I could see confusing the thermal and power protection since none of the actually active CPUs is reporting any sort of problems other than getting temperature from somewhere. I'd say get in contact with AMD about this, but the CPU is relatively old, so they may have already been aware of this possibly happening and have adjusting how severely they cut disabled/bad cores in a CCD years ago. If it's not a disabled Core on the CCD, it could still be the CPU, mainboard or memory (although very rare, I have had a system have a ram stick pump excessive power into a memory channel, still functioning yet causing heating issues) making diagnosing this issue a case of "replace everything with spares and see what happens". If it's a disabled core causing the issues, I'd be really interested in hearing that, since while it's a theorized possibility, it would be the first case I know where it actually happened.
For any non-german speakers that are wondering why flüssig is not a good translation for liquid: Flüssig is an adjective, its noun is Flüssigkeit. Thus, the LCD should say Flüssigkeit, which was probably shortenend to fit on the screen. This is lost in translation, since liquid is both adjective and noun.
I made a video reviewing this Kryosheet, and have gotten a lot of negative feedback, not on the Kryosheet itself, but rather how I handled it. People are complaining that I touched it with my fingers. Can you please clarify if this is an issue or not? Thank you Roman!
Would love to see the chip put into another board. I thought at first it was the board not reading temp correctly, but it was heating up on thermal cam. So, if the chip is getting hot and the cooler is working, and power is not spiking up then it must be something within the chip broken. (seems very odd as usually chips good motherboards bad).
@@vailpcs4040 DRMOS has short circuit protection,OCP,OVP,OTP, so if some of this happens they will send a signal to the PWM controller that will make the system not power on at all.
jesus just in bios 96C..its like the cooler is not even making any contact lol. Also big thank you for all the effort you put into your content, especially for us EN people
i think the first step would be testing the cpu on another motherboard, its more likely the motherboard's VRM has failed for some reason than the CPU failure, although it did seem like a contact issue given what the thermal paste looked like after removing the AIO water block, possibly the settings in bios have changed for what ever reason and increased the power draw, maybe the Power supply isnt delivering the correct current, voltage etc...
I had a very similar issue with an Asus rampage 3 extreme I had on my main PC for about 2 years. Near the end the system couldn't stay on for longer than 5 minutes, it'd just shut down in the same fashion with your threadripper. I couldn't figure out what it was at the time, so I eventually got sick of it and switched my platform to ryzen, which was new back then, and didn't look back. It was only 6 years later that the curiosity got the better of me and tried to diagnose the old mainboard. All that time it was the northbridge overheating, as soon as the IOH got to 100° the system would shut down. I put on some fresh thermal paste on both the northbridge and the Southbridge but the problem persists, the only way to keep the IOH temps in check is to have a fan blowing directly at the northbridge. As soon as I did that the system was stable. It's actually my home server now and it has no problems staying on 24/7.
Thermal engineer here: My guess would be a malfunctioning pump or something clogged, here's my reasoning: Water has a specific heat capacity of 4,200 Joules per kilogram per degree kelvin. At one kilogram (or liter) per second flow rate it would take 4.2kW to raise the water temp by 1°C. The difference is more like 45°C? Which with a CPU power consumption of 420w would suggest 0.1°C rise at 1l/s. Meaning to hit that 65°C we're talking a flow rate of 1.5mL per second. or 0.09L/min, or 5.4 l/hr. The corsair H100 has known issues with slow flow rate apparently. Not sure exactly what model you're using Put intuitively: If the flow rate was a miniscule amount above zero then the water temp out of the CPU block would be pretty much the CPU max temp, with good fans and rad the return would be about room temp. This is vaguely what we're seeing. If the flow rate was infinite with good fans and rad, then the water temp in and out of the CPU would be almost identical. This is the opposite of what we're seeing. If the fans and rad weren't functioning then the return to the CPU would also be hot. If the thermal interface was bad then the flow from the CPU wouldn't be hot and roughly the same as the input. This means that it must be a water flow rate issue. Check pump rpm, flow rate. Check for clogs/kinks/flow restrictions.
Yeah even without that conclusion, just trying a different cooler would seem to be a logical troubleshooting step. Confirming the 12 V power draw and 240 V power draw would also have been interesting.
Yeah, der8auer did some strange conclusions. The pump definitely did not move the water enough, even if the pump rpms looked okay. You can also see that in the deformation of dust filter mesh. Heavy deformation in the first third where the water is the hottest and then less deformed where the water is expected to be colder. The dT is far to big.
Yeah that makes sense really as to why one tube is hot and the other is cold because of the flow rate is way below optimal. Personally i would have swapped to an air cooler for testing that's why i always have one at hand. However it doesn't explain high idle load. So a reset of the BIOS maybe in order also since the CPU is not thermal throttling like it should.
@@wiedehopf9068 Great point. he said it wasn't thermal throttling so maybe it was blowing through it's power limit? Knowing the actual power draw would be useful diagnostic information. But to get 65°C at a ~20°C ambient is a dT of 45k and would require 189 kJ/L. 30 l/hr is 0.008 l/s and the absolute minimum I'd expect. And that's >1.5kW of heat output through the CPU! With a more normal 120 L/hr it'd be 6kW. No way that's the case because a German household socket can only output 3.6kW at 230v 16A max. What even is the power rating of the PSU? No way it's putting out more than... 1500W?
Really good solution for the Kryosheets to stop them moving about. I'm planning to use them in my upcoming build and was slightly nervous about biffing the alignment while getting the cooler on, so the fact you have a solution for this already is great.
Roman, was this the 3960X you delidded? Could it be that the LM dried out after diffusing into the IHS for 3 years? If not, I still think it might have something to do with the IHS bowing after uneven pressure was applied to it with an undersized coldplate. Wouldn't also surprise me if A'SUS lived up to their name again. I've had problems on every AMD board I bought from them for some reason. Really annoying.
@@WayStedYou thermal runaway. The hotter a chip is, the more internal resistance it has, meaning the more power it'll consume doing the same thing. And even when widows is "idle", it has over 100 processes going on in the background. It's one of the reasons why laptop battery life will be subpar on windows, no matter what processor you use. Also, TR3000 already pulls 50W+ at idle because of the massive IO die. 100W wouldn't be impossible if the memory was running quad channel at 3600 or whatever. But, as I said, ASUS having screwed something up wouldn't surprise me. I eventually had problems with every single AMD board I've bought from them. Can't say the same thing for other vendors.
I was hoping you're gonna at least put a different AIO on the CPU and check whether the issue persists, I was almost, hm, screaming, 'PUT A DIFFERENT COOLER ON' :D That'd be the next logical step honestly.
At least make sure the pump was running. The hot/cold thermal camera tubing means nothing is the pump was only moving a tiny amount of liquid. Just checking the cold plate and making sure there was liquid but not ensuring the pump is actually spinning but reporting an erroneous RPM to its sensor was kinda dumb.
Il dit avoir essayé avec un AIO 360, vous étiez ou pendant la vidéo ? Il n'a pas montré chaque étape du diagnostique sinon la vidéo aurait duré beaucoup plus longtemps mais il dit clairement j'ai essayé avec un 360 est le problème est exactement le même ! Pourquoi reprenez vous pratiquement tous la seule raison écarté et de manière sure à 100% depuis 15/20 commentaires depuis qu'une personne n'ayant rien compris et pas suivi la vidéo et surtout l'audio à incriminé l'aio !! C'est inquiétant Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
@@zybchIl a essayé un aio en 360 c'est quoi qui fait que vous n'arrivez pas à comprendre Oubliez l'AIO il n'y est pour rien Mettez les sous titre si vous êtes malentendant, je sais pas mais ça devient pénible depuis le premier qui a sorti ça sans avoir suivi, vous êtes nombreux à c'est la faute à l'aio ! NON! NON ! et NON ! c'est même dit et expliquez pourquoi NBON, car il a essayé avec un aio 360 au lieu du 240 est le problème est exactement le même Le prochain qui parle d'AIO je vais lui dire de se faire soigner l'audition ou d'aller voir un psychiatre car c'est dramatique de ne pas voir suivi la vidéo et de persister dans un avis qui n'est pas le bon !!!! J'ai mis suffisamment de point d'exclamations pour que vous ne fassiez pas autres choses comme pendant que Der8auer vous expliquez ou il faut que j'en rajoute davantage !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Vous avez un sérieux problème intellectuel j'en ai bien peur pour ne pas vouloir comprendre
Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Et moi j'aurais espérez que vous liriez ce que met la personne qui publie la vidéo avant de passer pour un con ! Mais visiblement j'ai moi une raison d'etre décu par vous et vous une bonne raison de passer pour un con !
My idea is that the return tube somewhere was gunked up, and water could not circulate. That's why one of the tubes were hot, but they should have been around the same temperature (+- 5-10 °C compared to each other)
Look 1:40 fans RPM are all over the place. There really can be only few things, but as it is actually hot, sensors are probably ok. Pump might be rotating, but are blades ok, as rad seems to be "Cold". Radiant/conducting heat, as heat rises, so one tube is hot but nothing else, water is not circulating --> blocked somewhere or pump. Blockade has gotten worse as time goes on, as you said radiator has pushed super-hot air before!
Now this is so coincidental, I have the same MOBO-CPU and had the same liquid cooler. Pump speed was at max and CPU was overheating. In fact, this was my second liquid cooler as the first one leaked. Started to suspect the cooler so I replaced it with a dual-fan Noctua threadripper cooler. Problem solved, CPU now running at normal temps from idle to max. All within spec. I am done with liquid coolers, no way to determine flow in the cooling loop so obviously no flow going thru the cooler. I have read of internal blockages with this cooler on some forums. At least with the Noctua you can actually see if there is any dust buildup on the fans / fins and clean.
Cpu heating? This may sound strange, but it happened to me, after changing all the components, except the PSU, nothing changed. Soon, I discovered the component: PSU. Either the PSU or some cable, or even connector, I don't know because I didn't have anything to test so I simply replaced the PSU and everything was normal, both on the old and new components.
Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Vous avez un sérieux problème intellectuel j'en ai bien peur pour ne pas vouloir comprendre Lire avant de dire n'importe quoi est également une solution mais pas au même problème !
The AIO has to be the issue. If the CPU got to 100c, that means the water is getting hot, and theirs no way the return would be that cold like it shown on the thermal video when the water in he returning loop back to the pump should have been a lot warmer than what it was showing on the termal video. This video might best a test for his audience to see if we caught it or how knowledgeable we are. CPU hot at idle, First check and test cooler, termal paste. Then check HWIfo for power draw, then reset Bios. If this doesn't work, put CPU in another motherboard, or it might be too late, you fried your CPU due to a bad AIO. Like to know if the CPU works on another setup, cheers!
Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) NON l'aio n'est pas le problème, mais vous l'auriez su en lisant avant de poster ou mieux en regardant attentivement la vidéo ! Mais apparemment comme pour un certain nombre trop pressé de poster il vaut mieux risquer de passer pour un con et si jamais la personne qui lis est asse stupide pour un génie
many people knew from the very beginning that it was definitely not AiO. Now der8auer himself recorded a follow-up, saying that he was lazy, that he did the diagnostics wrong and that it was his fault from the beginning and forgot that he delidded this CPU
Sir, that does not seem right that the hot side of an AIO loop is *that* much hotter than the cold side. That's either a *really* good AIO being able to dump all the heat and cool the water to ambient, *or* there's a clog in the radiator. I was a little disappointed you didn't put the thermal camera on the radiator to see how much of it was hot or if the heat was only really radiating out from one spot, signifying a leak. Can't wait for the follow-up. Maybe try to blow a little compressed air into the AIO and see if it coughs out a hairball.
IHS partially splitted from the core. There might have been bad contact from the start and after many cycles of heating up and cooling down it finally cracked.
the cold plate doesn't cover the threadripper which would explain the issue - I suppose the solder has cracked inside, due to temp difference/expansion
Blocked radiator. Easy one. One searing hot pipe. One cold pipe. There was no flow. Your thermal imaging camera showed that it was blocked as it entered the radiator. Chances are it’s been hot too long at some point and also damaged itself due to high temperature for a consistent time. That deformation shows the CPU was in burn out mode for quite some time.
*Disclaimer - I have not read through the comments - Apologies if this it is a repeat.* 20 years ago (22 really) I had a capacitor the half failed on my motherboard. it was top right corner of cpu. It was so hot, one couldn't touch it. This lead to the CPU over heating at idle and if started from cold would get hot very quick. My suggestion and thought for this issue is similar to that, something in the power circuit on the motherboard, not holding power back on chip, or die. Disabling cores to get it to run, also lead me to think this is the case. Though the cores / CPU could be damaged now. So hard to say. Look forwarded to more details
Rassure toi tu es plus intelligent et moins à coté de la plaque que 50% des commentaires qui incrimine l'aio sans savoir de quoi il parlent, et encore moins savoir lire ou regarder une vidéo ! Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Je fait plus que me répéter mais je pense que mis à part les dizaines de dauphins qui en ont suivi un autre pour crever sur une plage cet été. j'avais rarement vu un phénomène de groupe avec autant d'attardé qui pensent tous savoir (sauf lire et ce qui concerne l'informatique)
Pstates settings? Seems strange to be running hot when just sitting in BIOS. Perhaps a reset of all settings and a CMOS reset followed by a thorough manual configuration and checking of all settings?
Merde enfin quelqu'un qui a un cerveau et qui ne bloque pas sur l'AIO alors qu'il a été mis hors de cause dans la vidéo ! Je commençais à désespérer !
หลายเดือนก่อน +18
You are supposed to use liquid cooling blocks made for threadripper cpu's, the round surface coolers leave half of the processor surface empty. you know but you missed it.
That wouldnt be the reason. In BIOS the CPU should NORMALLY barely using power. As seen by the thermal imaging the cooler is transferring a ton of heat, and in Windows the AMD cpu is pulling 120w+ doing nothing.
You should to back to the coverage about how people blew this out of proportion and the companies responded by making giant cold plates. Spoiler: they performed within 2% of the smaller cold plates. Heat -> Heat spreader -> Cold plate (That's the purpose of heat spreader... to move the heat to where it can be dissipated.)
Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Et vous êtes supposé regarder et écouter la vidéo ainsi que ce s'il poste avant de passer pour un con à moins qu'il y avait urgence de montrer à la planète que vous ne vous préoccupez que de vous même, y copris sur le contenu d'une autre personne qui a déjà préciser que l'aio est hors de cause! et puis après plusieurs années si l'aio était un pb depuis le début c'est pas que maintenant qu'il poserait problème surtout qu'un 360 a les même symptôme que le 240 qui n'a jamais posé pb jusqu'à maintenait avec le treadripper cherche encore le géni pas en lecture en tout cas ni même en visionnage de vidéo et encor moins en informatique !
When seing the thermal images of the loop I immediately thought: That radiator has basically no flow. In no circumstance would 70+c water turn into a room temp radiator at the inlet port... It would start really hot and then spread out from there. Edit: Really disappointing that you did not verify the flow in the radiator as it was absolutely not working well...
Le plus décevant est que vous soyez incapable de lire correctement avant de poster pour passer pour un con... mais bon c'est pas comme si : Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
@@emuletmoi I did watch the entire video and I have seen the follow up video. I still say the thermal camera images make NO sense. There is no way there should have been a very hot and a very cold radiator tube. You'd need a very low flow to get a high dT on the water through the cpu block, that can then be dropped to the ambient in the radiator. This is not normal. Maybe the system was not running the pump at full rpm; I do not know. But it did not lok right.
OK, so you had to disable some of the cores beforehand to keep it working at all at some point, right? so why didn't you check the convexity of the CPU and/or motherboard itself? maybe some of the pins got loose and had no contact whatsoever and yet the CPU was powered on so it might have been some overpowering due to loose contacts. put it on a table, take a look from the side and see if the IHS had been bent.
Yea that is interesting.. I wanna see update on that if you figure out what is causing that. Somehow it's pulling more power than it's reporting? Did you check how much it's pulling from the wall?
With one of the AIO tubes being very hot and the other being not hot at all, don't you typically expect the coolant to not have such huge temperature differences? The temperature of the coolant usually only goes down gradually after repeated passes through the loop, doesn't it? I know you ruled out the AIO in this video, but this still seems pretty sus to me. Maybe there was a partial blockage in the radiator or something.
@@Rmobylera Coolant should only vary by a couple degrees through a single block custom loop. An AIO shouldn't be that much worse. Heck, 60C is roughly at pump temperature limits.
I suspect the CPU has a fault, causing a high idle power draw. The AIO does also appear to have an issue, most likely low water. But the AIO issue will be a result of age and the consistent heat coming from the CPU. The high CPU idle might be dodgy ECC ram, faulty ram controller, bad CPU pin or VRM. My guess is faulty ECC ram causing the ram controller to work harder, and increasing CPU package wattage and temps. Look forward to see the actual reason for the failiure.
@@Rmobylera i think he's sponsored by the AIO and doesnt want to cast blame, but i saw evidence of a corsiar AIO failed on a bestbuy prebuit just 2 months ago.
Don't be lazy and try a different cooler. The difference in temperature between the intake tube and outtake tube suggest a broken or slow pump. Ideally they should have a temperature delta under 10-15c MAX, as fast flowing water doesn't heat up much
Very unusual from you not to nail down the issue, especially since it might be something to do with the degradation of the CPU/Bios hardware that might lead to CPU getting pumped full of voltage without MB knowing about it. I would love to see further investigation, I am sure someone would gladly let you borrow another Threadripper CPU/MoBo combo to act as control group.
Have you tried a threadripper air cpu cooler to just confirm it’s not a cooler issue. CMOS reset? Memory controller may not handle having all slots filled (AMD always sucks with memory)?
Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Vous avez un sérieux problème intellectuel j'en ai bien peur pour ne pas vouloir comprendre ou m^me juste lire ou regarder la vidéo avant de poster
The older TR chips might've gotten by on a smaller cold plate, but I think the 3000 and newer have chiplettes that extend beyond that AIOs cold plate. There's only one AIO that would give adequate coverage, and it's the unreliable Enermax (I have one).
I have a question about the direct die cooler. Will the 7000 series cooler work for the 9000 series?, if not do you plan on production for that series?
I had an AIO pump fail on me once. It made me so much paranoid that I set my bios so that it takes me to temps dashboard first before anything else. Also I got ALF2 since and been happy with it ever since.
Something may be causing an aggressive clock/power consumption. Lots of causes. I’d try the next steps to try and determine if there is a config allowing the CPU to be full throttle at idle. Reset BIOS. Try setting the cpu power on in Windows to balanced. If PBO is on, try turning it off. Set the max TDP to something the cooler can handle. Try static clocks. If it’s not software then my guess is there is a short in the CPU and it’s dying.
cant u tell its the cat man. she couldnt sleep on top of it due to all the heat from cpu. i recon she just got annoyed with this and sabotaged the rig. just go and ask for her highnesses forgiveness and u will see that ur rig works in no time
i'm pleased you will be adding something to the graphite sheet product. i was wondering about buying it and using a drop of thermal paste to secure it while i fixed the cooler.
I miss using my NZXT Kraken G10. I had it on my 1080 Ti, and it was amazingly cool, always. But can't use it on the 6900 XT. I did place heat sinks on basically all parts on the PCB of the graphics card, and it worked so well.
As far as forgetting to remove the peel off the thermal pad on the integrated nvme heatsink, I too came to that realization one day... funny thing is I never saw a difference 'til this day. That poor nvme ssd is doing just fine and probably didn't really care one bit lol Also, I'm thinking it's CPU related... the moment the AiO is ruled out was the main culprit.
My GTX 1070 with Arctic Accelero Xtreme 3 was bent like that as well, especially the die area had a bulge in the middle on the back, but it worked just fine.
You might have baked the CPU core chiplets by using a standard-sized coldplate. I've seen this happened several times including my old Athlon 64. Replacing the degraded CPU is the only option.
Derp. Your Athlon didn't have protections for overheating scenarios. CPUs have had automatic thermal throttling and shutdown standard for almost 20 years. A heat spreader works by... you know... spreading the heat to the surface of the CPU. That allows you to use a much smaller cold plate because... you know... the heat transfers to the cold plate through the heat spreader. There was a huge song-and-dance about using a smaller AIO with Threadripper so they made dedicated solutions which performed within 2% of the smaller ones. (I know, right? It's almost like the companies making TR-compatible solutions evaluate if they're fit for purpose, or something.)
@@tim3172 Heatspreader is a misleading name. It's there only to protect the CPU chip(s). To be fair, it does spread heat marginally, but it's maybe a millimetre on each side, tops. And Athlon 64 does have overtemp protection. Your comment is invalid.
You need custom loop: more robust pump, better cooling performance,easier to debug. I cool a 3090 and a 5950x with only 240+120 rads. And it’s almost silence. Just maybe. I don’t know what is exactly going on here..
Probably some voltage regulation/reporting issue? The normal voltage in use on that CPU should be around 1.3-1.4V. The BIOS showed half of that and still near it's max temp and staying there. That combination seems impossible? Only way to know for sure is test it on another motherboard.
Wait... what? The temperature was too high and it powered off?????????????? We're forever being told from every outlet that some arbitrary value below the junction temperature is "bad" when a lower arbitrary value is "good". You mean to tell me a CPU will preserve itself when it gets (actually) too hot? Mind. Blown.
I read somewhere that for the NZXT Kraken G12 bracket that you would have to use the AMD brackets to use it with 20 series. Can you confirm if this is true?
Actually, it might signal the fault is restricted / limted flow. Usually, the flow rate is high enough that the radiator doesn't get to dissipate much heat before the water is heading back down to the cpu. That high temperature difference is unusual, and might be pointing to the problem being restricted flow (whether pump not working right, or something obstructing a tube, or ????, is for Roman to look for). Also, we don't know what scale the colors were set to. While the camera was PROBABLY autoscaling, it might not have been. If it was, then the return tube at near ambient temperature is absolutely a problem. If not autoscaling, then the return pipe might have been just 5C cooler (for example) than the "hot" pipe.
If it's drawing that much power while doing nothing in bios, then the defect is definitely not with the cooler. You may have access to other threadripper motherboards or CPUs to test if it's either one, but most people don't. I do imagine AMD would want to analyze it for what's causing the problem. If it's a design flaw or something slipped by in their validation testing that they could improve on that side.
My assumption is that the BIOS could be causing issues with auto-overclocking the TDP or clock speed. Could you consider updating it to see if this resolves the problem? For example, you could update to PRIME TRX40-PRO S BIOS 2102. second, could be the mobo some part of chipset failure. but it is too expensive to replace the mobo just for this testing.
Seems like clogged radiator, it's impossible for the AIO to achieve that much ΔT between the hot pipe and the cold pipe. So the flow should be severely hampered through the radiator.
And this is why keeping extra or "leftover" parts are important. I would have swapped another cpu into this system just to see if I get the same results. Much easier to start there vs tearing down the entire system. If you receive the same results, probably a MOBO issue or cooling issue. If you do not receive the same results, you can deduce the current CPU is okay. You could also drop the TR CPU into another working system and see if that CPU again achieves the same negative result.* Troubleshooting is absolutely a pain to do but always best, if possible, to move the suspect (CPU or other part) to a working system . IE monitor seems defective. Hook monitor to another system. If same result mon is bad or cables etc. But I wouldn't start by removing my GPU or running more invasive tests if I can just hook it up to something I know works as intended. Happy to see you will do some more investigating. Cheers 🖖 *Edit after rereading my post. Forgot TR do not use the same MOBO as say a 3800 cpu ... 😮💨 If you have another TR setup then my advice is valid. If not, then you could reach out to someone you trust and can assist. OR if you just happen to have an additional TR mobo you can still test that CPU just to see if the results are similar. Though that does cause a more intensive tear down. The joys of computers. 😳lol
First of all, the thermal camera shows the year 1970. 😂 The processor has an internal problem in three ccds, because there is a current leakage inside the paths themselves. it happens on ccd1, ccd2, ccd4. Vram is under heavy load, because it constantly feeds the demand of the ccds. If you manage to lower the temperatures for a while, do the AIDA64 test for the L3 cache, if their bandwidth is significantly reduced, it is definitely due to a bad processor. GL+HF
its the AIO i had the exact same issue with a corsair aio years back with a fx 9950 i swapped the aio and reset the bios to defaults and it fixed the issue
Did you check if the radiator has flow? It could have corroded to the point of blockage. I think checking the radiators flow with a different pump would determine whether there's an issue with the pump or with radiator clog. I had a very similar issue with my custom loop that is about 7 years old.
As tu seulement regardé la vidéo ?? Il a essayé avec un AIO en 360 au lieu de l'AO en 240 est le problème est exactement le même !!! Vous vous êtes concerté pour ne pas suivre la vidéo et ce qu'il y est dit et avaez décider de rendre l'aio responsable alors que der8auer dit lui même qu'il n'y est pour rien ? Le premier avec sa connerie de rendre l'aio responsable et qui fait que la plupart répète comme des moutons me fait penser qu'il aurait mieux fait de sauter d'une falaise que vous en fassiez de même car c'est consternant de ne pas vouloir comprendre à un tel niveau !!! A se demander pourquoi il perd son temps à expliquer et carrément à faire des vidéo si vous n'écoutez pas !! Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Vous avez un sérieux problème intellectuel j'en ai bien peur pour ne pas vouloir comprendre
Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) lol vous êtes au moins la 100eme personne a parler du fait qu'il n'a pas changer d'aio pour tester alors qu'il l'a fait et même marqué noir sur blanc en plus de le dire car à force il a compris que les ignares commenter sans rien comprendre !! Épinglé par der8auer EN @der8auer-en il y a 6 jours (modifié) Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM. Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :) Et alors lol c'est quoi ton explication mr lol je sais tout !!!???
Only thing i could think of is a bad psu/ motherboard or corrupted bios, but thats just so odd to happen randomly w.o changing software unless its a faulty psu.
I was lucky to be able to get the 2950X when the next generation of Threadrippers was released, but now it's that much faster than my gaming PC CPU, the 5600X 😄 I still want a Threadripper for the other PC cause of the PCIe lanes for capture cards and stuff.
My OCD was triggered by not using the top CPU-connected NVMe slot. Also found it amusing that the plastic over the NVMe thermal pad was not removed. Proves no one is perfect. XD
@@Travelfast To my understanding of the 600 series chipset design and layout, there are only 4 lanes of PCIe 4.0 going between the CPU and the chipset. This means if you put a PCIe 4.0 SSD on one of the chipset NVMe slots, wouldn't that saturate the chipset bandwidth? That could then mean that any further load on that bus from things like USB, audio or LAN would slow down the NVMe drive's performance...
I dont know what is actually wrong, however I also has a strange issue with a 5950x and it crashed windows after some time with a ACPI error. I think this case is probably very similar, the power regulator in the soc io die being broken.
I agree with you after checking the cooling solutions that the problem has to be with either the CPU or the motherboard. Personally I'm leaning more towards the motherboard. I get why you ultimately decided to just go build a new system, but I would've at least tried clearing the CMOS and testing again first (maybe even updating the BIOS) just in case there was some kind of errant setting I was unaware of nor saw when checking settings (Did you overclock that Threadripper? I doubt you did, one because there's not much point and two because you were only using a 240 AIO that didn't cover the full IHS. If you were overclocking it, you surely would've been using a much beefier cooling solution, but I have to ask as a matter of never making assumptions. Honestly, I would personally use the Noctua NH-U14S TR4 edition over that 240 AIO. Threadripper really benefits from using a cooler that covers the full IHS and an air cooler that does cover the full IHS has been found to perform better than an AIO that doesn't.). I might've even tried specifically configuring it to run on a lower power profile, but that doesn't exactly prove which is the problem until you can get ether a different CPU or motherboard to test. But it might get you stable enough to use it in the meantime to do more testing. 120W at idle does seem really high. Way too high. Up until the beginning of this year, I was running a Threadripper 1950X and that thing rarely ever pulled more than 60W at idle. Which oddly enough is the same wattage my currently 7950X pulls at idle most of the time. You're 3960X is pulling double that at idle and it's not even double the core count. I know not to assume linearity in power requirements, but still my gut feeling is 120W at idle seems way too high. I strongly feel like that idle wattage should be much lower. I've never used a 3960X, but I would be really surprised if it was normal for that CPU to idle at anything more than 80W at max. So that's what I would be looking at next. Starting with clearing the CMOS. Then a firmware update. Possibly even trying a new motherboard entirely. Honestly, I never say this because it's almost never the case, despite how many people almost always default to blaming this component, but in this case, I really am leaning heavily towards the motherboard being the issue.
One more note: I tested the TR CPU with another 360 AIO just to be safe and I still had 70-80°C in Windows idle without doing anything. So something must be wrong with that CPU or maybe the VRM
Edit: After all your feedback I will make sure to work on an update video to investigate the issue further :)
What could be wrong with the VRM to cause this? If the voltages would be too high, wouldn't this show up somewhere?
It is windows
Something fishy with the BIOS doing something funky?
It would be a nice video to deep dive into the issue to see whether it was caused by mobo or cpu.
The 3960x has a 4x6 CCD configuration.
And CCDs are 8-cores with disabled ones to get to 6-cores in a CCD.
And if you look in HWMon, you'll see that 1 CCD is at 90+C while the other 3 are only at 70 and 80.
So it's clear that this one CCD is the source of the heat.
But, the fact that this thermal issue became progressively worse over time, WITHOUT giving you computational errors (which you would get from a bad core, memory or mainboard), makes me think one of these "disabled" cores wasn't cut properly and over time made a progressively worse short, creating the heat and power draw. Something I could see confusing the thermal and power protection since none of the actually active CPUs is reporting any sort of problems other than getting temperature from somewhere.
I'd say get in contact with AMD about this, but the CPU is relatively old, so they may have already been aware of this possibly happening and have adjusting how severely they cut disabled/bad cores in a CCD years ago.
If it's not a disabled Core on the CCD, it could still be the CPU, mainboard or memory (although very rare, I have had a system have a ram stick pump excessive power into a memory channel, still functioning yet causing heating issues) making diagnosing this issue a case of "replace everything with spares and see what happens".
If it's a disabled core causing the issues, I'd be really interested in hearing that, since while it's a theorized possibility, it would be the first case I know where it actually happened.
For any non-german speakers that are wondering why flüssig is not a good translation for liquid:
Flüssig is an adjective, its noun is Flüssigkeit. Thus, the LCD should say Flüssigkeit, which was probably shortenend to fit on the screen.
This is lost in translation, since liquid is both adjective and noun.
Thank you! The more we learn with youtube comments 🙌
"In any case, not a fan of this, literally" LOL
very sad and dissapointed, the problem seemed very cool would love to know what acutally caused it. Yeah an educated guess can be made but still.
yea I realize that after your comments :/ Will make an update video to track it down. Sorry for the disappointment
🫡
@@der8auer-en At a complete guess something not wired right around the socket or an overlooked setting in bios throwing too much voltage into the cpu.
It has to be something with the CPU, but what? This is a very strange problem to have.
@@der8auer-en Wooooh! More content! Looking forward to the update video! (This thus made a great teaser!)
I made a video reviewing this Kryosheet, and have gotten a lot of negative feedback, not on the Kryosheet itself, but rather how I handled it. People are complaining that I touched it with my fingers. Can you please clarify if this is an issue or not? Thank you Roman!
not an issue at all. You won't be able to tell a difference if you touched it or not.
@@der8auer-en Thank you for settling this!
how dare you touching the holy sheet
@@Riztard 🤣🤣🤣 Thats the reaction I keep getting. Its crazy.
That's some crazy sheeet.
Would love to see the chip put into another board. I thought at first it was the board not reading temp correctly, but it was heating up on thermal cam. So, if the chip is getting hot and the cooler is working, and power is not spiking up then it must be something within the chip broken. (seems very odd as usually chips good motherboards bad).
Perhaps the onboard VRM / power delivery has failed?
@@vailpcs4040 DRMOS has short circuit protection,OCP,OVP,OTP, so if some of this happens they will send a signal to the PWM controller that will make the system not power on at all.
jesus just in bios 96C..its like the cooler is not even making any contact lol.
Also big thank you for all the effort you put into your content, especially for us EN people
Without the cooler the system would probably do a thermal shutdown nearly instantly. It's thanks to the cooler he could even see this happening.
@@enlightendbel while this is true, the liquid cooler still doesn't seem to be working.
@@scarecrow5848 the liquid cooler was clearly working fine because it was transferring heat away, making top of the case too hot to touch.
@@scarecrow5848 That's a strange conclusion when there's about a 30°C difference in water temperature before/after the radiator.
@@a.x.w true, cuz in a working AIO the temp diff is not 30c.. its 5-6c so AIO not working
i think the first step would be testing the cpu on another motherboard, its more likely the motherboard's VRM has failed for some reason than the CPU failure, although it did seem like a contact issue given what the thermal paste looked like after removing the AIO water block, possibly the settings in bios have changed for what ever reason and increased the power draw, maybe the Power supply isnt delivering the correct current, voltage etc...
Id really love a deep dive into why the threadripper was over powering so badly.
I had a very similar issue with an Asus rampage 3 extreme I had on my main PC for about 2 years. Near the end the system couldn't stay on for longer than 5 minutes, it'd just shut down in the same fashion with your threadripper. I couldn't figure out what it was at the time, so I eventually got sick of it and switched my platform to ryzen, which was new back then, and didn't look back.
It was only 6 years later that the curiosity got the better of me and tried to diagnose the old mainboard. All that time it was the northbridge overheating, as soon as the IOH got to 100° the system would shut down. I put on some fresh thermal paste on both the northbridge and the Southbridge but the problem persists, the only way to keep the IOH temps in check is to have a fan blowing directly at the northbridge. As soon as I did that the system was stable. It's actually my home server now and it has no problems staying on 24/7.
Thermal engineer here:
My guess would be a malfunctioning pump or something clogged, here's my reasoning:
Water has a specific heat capacity of 4,200 Joules per kilogram per degree kelvin. At one kilogram (or liter) per second flow rate it would take 4.2kW to raise the water temp by 1°C. The difference is more like 45°C? Which with a CPU power consumption of 420w would suggest 0.1°C rise at 1l/s. Meaning to hit that 65°C we're talking a flow rate of 1.5mL per second. or 0.09L/min, or 5.4 l/hr. The corsair H100 has known issues with slow flow rate apparently. Not sure exactly what model you're using
Put intuitively:
If the flow rate was a miniscule amount above zero then the water temp out of the CPU block would be pretty much the CPU max temp, with good fans and rad the return would be about room temp. This is vaguely what we're seeing.
If the flow rate was infinite with good fans and rad, then the water temp in and out of the CPU would be almost identical. This is the opposite of what we're seeing.
If the fans and rad weren't functioning then the return to the CPU would also be hot.
If the thermal interface was bad then the flow from the CPU wouldn't be hot and roughly the same as the input.
This means that it must be a water flow rate issue. Check pump rpm, flow rate. Check for clogs/kinks/flow restrictions.
Yeah even without that conclusion, just trying a different cooler would seem to be a logical troubleshooting step.
Confirming the 12 V power draw and 240 V power draw would also have been interesting.
Yeah, der8auer did some strange conclusions. The pump definitely did not move the water enough, even if the pump rpms looked okay. You can also see that in the deformation of dust filter mesh. Heavy deformation in the first third where the water is the hottest and then less deformed where the water is expected to be colder. The dT is far to big.
Yeah that makes sense really as to why one tube is hot and the other is cold because of the flow rate is way below optimal. Personally i would have swapped to an air cooler for testing that's why i always have one at hand. However it doesn't explain high idle load. So a reset of the BIOS maybe in order also since the CPU is not thermal throttling like it should.
@@wiedehopf9068 Great point. he said it wasn't thermal throttling so maybe it was blowing through it's power limit? Knowing the actual power draw would be useful diagnostic information.
But to get 65°C at a ~20°C ambient is a dT of 45k and would require 189 kJ/L. 30 l/hr is 0.008 l/s and the absolute minimum I'd expect. And that's >1.5kW of heat output through the CPU! With a more normal 120 L/hr it'd be 6kW. No way that's the case because a German household socket can only output 3.6kW at 230v 16A max.
What even is the power rating of the PSU? No way it's putting out more than... 1500W?
i think the AIO failed and the other one is leaking too but internet armchairs dont fix AIO's
The CPU shunt-modded itself😁
Really good solution for the Kryosheets to stop them moving about. I'm planning to use them in my upcoming build and was slightly nervous about biffing the alignment while getting the cooler on, so the fact you have a solution for this already is great.
Roman, was this the 3960X you delidded? Could it be that the LM dried out after diffusing into the IHS for 3 years?
If not, I still think it might have something to do with the IHS bowing after uneven pressure was applied to it with an undersized coldplate.
Wouldn't also surprise me if A'SUS lived up to their name again. I've had problems on every AMD board I bought from them for some reason. Really annoying.
bend chip could be, not all memory sticks where detected in his bios. Unless half of them where dummy sticks to just fill the slots.
@@f0x4nn3Half of the sticks are in fact dummy sticks....
Doesn't explain it pulling 100w at idle
@@WayStedYou thermal runaway. The hotter a chip is, the more internal resistance it has, meaning the more power it'll consume doing the same thing. And even when widows is "idle", it has over 100 processes going on in the background. It's one of the reasons why laptop battery life will be subpar on windows, no matter what processor you use.
Also, TR3000 already pulls 50W+ at idle because of the massive IO die. 100W wouldn't be impossible if the memory was running quad channel at 3600 or whatever.
But, as I said, ASUS having screwed something up wouldn't surprise me. I eventually had problems with every single AMD board I've bought from them. Can't say the same thing for other vendors.
@@lennard9331 More resistance means less power draw.
Can it be that heatspreader separated from chip?
That was my first hope
Both hoses should be warm on the water cooling.
I was hoping you're gonna at least put a different AIO on the CPU and check whether the issue persists, I was almost, hm, screaming, 'PUT A DIFFERENT COOLER ON' :D That'd be the next logical step honestly.
At least make sure the pump was running. The hot/cold thermal camera tubing means nothing is the pump was only moving a tiny amount of liquid. Just checking the cold plate and making sure there was liquid but not ensuring the pump is actually spinning but reporting an erroneous RPM to its sensor was kinda dumb.
Il dit avoir essayé avec un AIO 360, vous étiez ou pendant la vidéo ?
Il n'a pas montré chaque étape du diagnostique sinon la vidéo aurait duré beaucoup plus longtemps mais il dit clairement j'ai essayé avec un 360 est le problème est exactement le même !
Pourquoi reprenez vous pratiquement tous la seule raison écarté et de manière sure à 100% depuis 15/20 commentaires depuis qu'une personne n'ayant rien compris et pas suivi la vidéo et surtout l'audio à incriminé l'aio !!
C'est inquiétant
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
@@zybchIl a essayé un aio en 360 c'est quoi qui fait que vous n'arrivez pas à comprendre
Oubliez l'AIO il n'y est pour rien
Mettez les sous titre si vous êtes malentendant, je sais pas mais ça devient pénible depuis le premier qui a sorti ça sans avoir suivi, vous êtes nombreux à c'est la faute à l'aio !
NON! NON ! et NON ! c'est même dit et expliquez pourquoi NBON, car il a essayé avec un aio 360 au lieu du 240 est le problème est exactement le même
Le prochain qui parle d'AIO je vais lui dire de se faire soigner l'audition ou d'aller voir un psychiatre car c'est dramatique de ne pas voir suivi la vidéo et de persister dans un avis qui n'est pas le bon !!!!
J'ai mis suffisamment de point d'exclamations pour que vous ne fassiez pas autres choses comme pendant que Der8auer vous expliquez ou il faut que j'en rajoute davantage !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Vous avez un sérieux problème intellectuel j'en ai bien peur
pour ne pas vouloir comprendre
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Et moi j'aurais espérez que vous liriez ce que met la personne qui publie la vidéo avant de passer pour un con !
Mais visiblement j'ai moi une raison d'etre décu par vous et vous une bonne raison de passer pour un con !
Absolutely try another cooler first. Especially with one hot tube, one cold. I would have gone with a 360.
My idea is that the return tube somewhere was gunked up, and water could not circulate. That's why one of the tubes were hot, but they should have been around the same temperature (+- 5-10 °C compared to each other)
The pump would have been screaming.
Look 1:40 fans RPM are all over the place. There really can be only few things, but as it is actually hot, sensors are probably ok. Pump might be rotating, but are blades ok, as rad seems to be "Cold". Radiant/conducting heat, as heat rises, so one tube is hot but nothing else, water is not circulating --> blocked somewhere or pump. Blockade has gotten worse as time goes on, as you said radiator has pushed super-hot air before!
Now this is so coincidental, I have the same MOBO-CPU and had the same liquid cooler. Pump speed was at max and CPU was overheating. In fact, this was my second liquid cooler as the first one leaked. Started to suspect the cooler so I replaced it with a dual-fan Noctua threadripper cooler. Problem solved, CPU now running at normal temps from idle to max. All within spec. I am done with liquid coolers, no way to determine flow in the cooling loop so obviously no flow going thru the cooler. I have read of internal blockages with this cooler on some forums. At least with the Noctua you can actually see if there is any dust buildup on the fans / fins and clean.
Cpu heating? This may sound strange, but it happened to me, after changing all the components, except the PSU, nothing changed. Soon, I discovered the component: PSU. Either the PSU or some cable, or even connector, I don't know because I didn't have anything to test so I simply replaced the PSU and everything was normal, both on the old and new components.
An easy check would have been to put a different cooler on the CPU and see if anything changes
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Vous avez un sérieux problème intellectuel j'en ai bien peur
pour ne pas vouloir comprendre
Lire avant de dire n'importe quoi est également une solution mais pas au même problème !
The AIO has to be the issue. If the CPU got to 100c, that means the water is getting hot, and theirs no way the return would be that cold like it shown on the thermal video when the water in he returning loop back to the pump should have been a lot warmer than what it was showing on the termal video.
This video might best a test for his audience to see if we caught it or how knowledgeable we are. CPU hot at idle, First check and test cooler, termal paste. Then check HWIfo for power draw, then reset Bios. If this doesn't work, put CPU in another motherboard, or it might be too late, you fried your CPU due to a bad AIO.
Like to know if the CPU works on another setup, cheers!
yeah it's very likely a pump issue and water flowing too slow, the intake tube was too fucking hot at 67c while the outtake was at worse 30c.
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
NON l'aio n'est pas le problème, mais vous l'auriez su en lisant avant de poster ou mieux en regardant attentivement la vidéo !
Mais apparemment comme pour un certain nombre trop pressé de poster il vaut mieux risquer de passer pour un con et si jamais la personne qui lis est asse stupide pour un génie
many people knew from the very beginning that it was definitely not AiO. Now der8auer himself recorded a follow-up, saying that he was lazy, that he did the diagnostics wrong and that it was his fault from the beginning and forgot that he delidded this CPU
Sir, that does not seem right that the hot side of an AIO loop is *that* much hotter than the cold side.
That's either a *really* good AIO being able to dump all the heat and cool the water to ambient, *or* there's a clog in the radiator. I was a little disappointed you didn't put the thermal camera on the radiator to see how much of it was hot or if the heat was only really radiating out from one spot, signifying a leak.
Can't wait for the follow-up. Maybe try to blow a little compressed air into the AIO and see if it coughs out a hairball.
assuming what temperature range on the IR cam setting ?? You jump to conclusions.
I was looking for this comment, that's also my assumption. Either way, I'm a bit disappointed with Roman for not investigating this further...
That's how a hot side works when there is a lot of heat...
Also, he tested with a bigger AIO and the idle temps were still terrible
IHS partially splitted from the core. There might have been bad contact from the start and after many cycles of heating up and cooling down it finally cracked.
the cold plate doesn't cover the threadripper which would explain the issue - I suppose the solder has cracked inside, due to temp difference/expansion
@@stanimir4197 Thats what i thought... IHS is soldered to the core and it splitted.
Blocked radiator. Easy one. One searing hot pipe. One cold pipe. There was no flow. Your thermal imaging camera showed that it was blocked as it entered the radiator.
Chances are it’s been hot too long at some point and also damaged itself due to high temperature for a consistent time. That deformation shows the CPU was in burn out mode for quite some time.
try to measure vcore with a multimeter.
*Disclaimer - I have not read through the comments - Apologies if this it is a repeat.*
20 years ago (22 really) I had a capacitor the half failed on my motherboard. it was top right corner of cpu. It was so hot, one couldn't touch it. This lead to the CPU over heating at idle and if started from cold would get hot very quick.
My suggestion and thought for this issue is similar to that, something in the power circuit on the motherboard, not holding power back on chip, or die. Disabling cores to get it to run, also lead me to think this is the case. Though the cores / CPU could be damaged now. So hard to say.
Look forwarded to more details
Rassure toi tu es plus intelligent et moins à coté de la plaque que 50% des commentaires qui incrimine l'aio sans savoir de quoi il parlent, et encore moins savoir lire ou regarder une vidéo !
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Je fait plus que me répéter mais je pense que mis à part les dizaines de dauphins qui en ont suivi un autre pour crever sur une plage cet été.
j'avais rarement vu un phénomène de groupe avec autant d'attardé qui pensent tous savoir (sauf lire et ce qui concerne l'informatique)
Pstates settings? Seems strange to be running hot when just sitting in BIOS. Perhaps a reset of all settings and a CMOS reset followed by a thorough manual configuration and checking of all settings?
He Did 3 bios versions so he Had his settings reseted by default
I noticed when he was in windows CPU not going into C6 power state which Ryzen CPU's heavy depend on for power savings.
Merde
enfin quelqu'un qui a un cerveau et qui ne bloque pas sur l'AIO alors qu'il a été mis hors de cause dans la vidéo !
Je commençais à désespérer !
You are supposed to use liquid cooling blocks made for threadripper cpu's, the round surface coolers leave half of the processor surface empty. you know but you missed it.
That's what I was thinking, the water moves over a different spot on the chip right?
That wouldnt be the reason. In BIOS the CPU should NORMALLY barely using power. As seen by the thermal imaging the cooler is transferring a ton of heat, and in Windows the AMD cpu is pulling 120w+ doing nothing.
You should to back to the coverage about how people blew this out of proportion and the companies responded by making giant cold plates.
Spoiler: they performed within 2% of the smaller cold plates.
Heat -> Heat spreader -> Cold plate (That's the purpose of heat spreader... to move the heat to where it can be dissipated.)
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Et vous êtes supposé regarder et écouter la vidéo ainsi que ce s'il poste avant de passer pour un con à moins qu'il y avait urgence de montrer à la planète que vous ne vous préoccupez que de vous même, y copris sur le contenu d'une autre personne qui a déjà préciser que l'aio est hors de cause!
et puis après plusieurs années si l'aio était un pb depuis le début c'est pas que maintenant qu'il poserait problème surtout qu'un 360 a les même symptôme que le 240 qui n'a jamais posé pb jusqu'à maintenait avec le treadripper
cherche encore le géni pas en lecture en tout cas ni même en visionnage de vidéo et encor moins en informatique !
When seing the thermal images of the loop I immediately thought: That radiator has basically no flow.
In no circumstance would 70+c water turn into a room temp radiator at the inlet port... It would start really hot and then spread out from there.
Edit: Really disappointing that you did not verify the flow in the radiator as it was absolutely not working well...
Le plus décevant est que vous soyez incapable de lire correctement avant de poster pour passer pour un con...
mais bon c'est pas comme si :
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
@@emuletmoi I did watch the entire video and I have seen the follow up video. I still say the thermal camera images make NO sense. There is no way there should have been a very hot and a very cold radiator tube. You'd need a very low flow to get a high dT on the water through the cpu block, that can then be dropped to the ambient in the radiator. This is not normal. Maybe the system was not running the pump at full rpm; I do not know. But it did not lok right.
OK, so you had to disable some of the cores beforehand to keep it working at all at some point, right? so why didn't you check the convexity of the CPU and/or motherboard itself? maybe some of the pins got loose and had no contact whatsoever and yet the CPU was powered on so it might have been some overpowering due to loose contacts. put it on a table, take a look from the side and see if the IHS had been bent.
Yea that is interesting.. I wanna see update on that if you figure out what is causing that. Somehow it's pulling more power than it's reporting? Did you check how much it's pulling from the wall?
With one of the AIO tubes being very hot and the other being not hot at all, don't you typically expect the coolant to not have such huge temperature differences? The temperature of the coolant usually only goes down gradually after repeated passes through the loop, doesn't it? I know you ruled out the AIO in this video, but this still seems pretty sus to me. Maybe there was a partial blockage in the radiator or something.
That's what happens with any AIO, it's normal behavior for everyone
@@Rmobylera Coolant should only vary by a couple degrees through a single block custom loop. An AIO shouldn't be that much worse. Heck, 60C is roughly at pump temperature limits.
I suspect the CPU has a fault, causing a high idle power draw. The AIO does also appear to have an issue, most likely low water. But the AIO issue will be a result of age and the consistent heat coming from the CPU.
The high CPU idle might be dodgy ECC ram, faulty ram controller, bad CPU pin or VRM.
My guess is faulty ECC ram causing the ram controller to work harder, and increasing CPU package wattage and temps. Look forward to see the actual reason for the failiure.
@@Rmobylera i think he's sponsored by the AIO and doesnt want to cast blame, but i saw evidence of a corsiar AIO failed on a bestbuy prebuit just 2 months ago.
@@Rmobylera a good aio should manage to keep the water in and out about the same temp
Love the update on Kryosheet.
@15:12
Right on 🤜🤛🏾
Makes me wonder if any silicone hydraulic oil could be used with pads that didn't come with it 🤔
Don't be lazy and try a different cooler. The difference in temperature between the intake tube and outtake tube suggest a broken or slow pump. Ideally they should have a temperature delta under 10-15c MAX, as fast flowing water doesn't heat up much
Very unusual from you not to nail down the issue, especially since it might be something to do with the degradation of the CPU/Bios hardware that might lead to CPU getting pumped full of voltage without MB knowing about it. I would love to see further investigation, I am sure someone would gladly let you borrow another Threadripper CPU/MoBo combo to act as control group.
I've had my H100i since 2020 and amazed it's still keeping my 5600X cool.
Have you tried a threadripper air cpu cooler to just confirm it’s not a cooler issue. CMOS reset? Memory controller may not handle having all slots filled (AMD always sucks with memory)?
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Vous avez un sérieux problème intellectuel j'en ai bien peur
pour ne pas vouloir comprendre ou m^me juste lire ou regarder la vidéo avant de poster
I'm curious to know myself. I'm even scratching my head going ... what is happening here .. Cannot wait to see the deep dive into this.
I had one do same thing, aio was plugged off. One tube would get hot and the other was cold, changed aio and running great.
Multiple issues. Clogged AIOs always look like that on flir. It is undersized coldplate, doing heavy rendering, it cooked and bent IMHO.
Plus, if this is a delidded CPU from back then, bad contact with the IHS is a contributor as well.
W Noctua Fan!
The older TR chips might've gotten by on a smaller cold plate, but I think the 3000 and newer have chiplettes that extend beyond that AIOs cold plate. There's only one AIO that would give adequate coverage, and it's the unreliable Enermax (I have one).
Peerless Assassin - can't go wrong.
I have a question about the direct die cooler. Will the 7000 series cooler work for the 9000 series?, if not do you plan on production for that series?
Half of the stick of memory are not detected in the bios, if you watch carefully
Good catch. Might be 1 of the memory controllers that's shorted and causing the heat.
These are dummy sticks :) Corsair called it light enhancement or something like that
@@der8auer-endo you think the dummy sticks did this?
@@der8auer-en looking at the thermal image they are not dummy sticks
@@de4ler RGB emits more heat than ram it self.
I had an AIO pump fail on me once.
It made me so much paranoid that I set my bios so that it takes me to temps dashboard first before anything else.
Also I got ALF2 since and been happy with it ever since.
Something may be causing an aggressive clock/power consumption. Lots of causes. I’d try the next steps to try and determine if there is a config allowing the CPU to be full throttle at idle. Reset BIOS. Try setting the cpu power on in Windows to balanced. If PBO is on, try turning it off. Set the max TDP to something the cooler can handle. Try static clocks.
If it’s not software then my guess is there is a short in the CPU and it’s dying.
cant u tell its the cat man. she couldnt sleep on top of it due to all the heat from cpu. i recon she just got annoyed with this and sabotaged the rig. just go and ask for her highnesses forgiveness and u will see that ur rig works in no time
the warping filter from the heat is insane lol
Yeah, that's way more than 100W.
i'm pleased you will be adding something to the graphite sheet product. i was wondering about buying it and using a drop of thermal paste to secure it while i fixed the cooler.
I miss using my NZXT Kraken G10. I had it on my 1080 Ti, and it was amazingly cool, always. But can't use it on the 6900 XT. I did place heat sinks on basically all parts on the PCB of the graphics card, and it worked so well.
As far as forgetting to remove the peel off the thermal pad on the integrated nvme heatsink, I too came to that realization one day... funny thing is I never saw a difference 'til this day.
That poor nvme ssd is doing just fine and probably didn't really care one bit lol
Also, I'm thinking it's CPU related... the moment the AiO is ruled out was the main culprit.
My GTX 1070 with Arctic Accelero Xtreme 3 was bent like that as well, especially the die area had a bulge in the middle on the back, but it worked just fine.
You might have baked the CPU core chiplets by using a standard-sized coldplate. I've seen this happened several times including my old Athlon 64. Replacing the degraded CPU is the only option.
you are right that might be it
3.5 years though, sure is some slow baking. Not saying it couldn't be it though.
Derp. Your Athlon didn't have protections for overheating scenarios.
CPUs have had automatic thermal throttling and shutdown standard for almost 20 years.
A heat spreader works by... you know... spreading the heat to the surface of the CPU.
That allows you to use a much smaller cold plate because... you know... the heat transfers to the cold plate through the heat spreader.
There was a huge song-and-dance about using a smaller AIO with Threadripper so they made dedicated solutions which performed within 2% of the smaller ones.
(I know, right? It's almost like the companies making TR-compatible solutions evaluate if they're fit for purpose, or something.)
@@tim3172 Heatspreader is a misleading name. It's there only to protect the CPU chip(s).
To be fair, it does spread heat marginally, but it's maybe a millimetre on each side, tops.
And Athlon 64 does have overtemp protection.
Your comment is invalid.
Did you try clearing your CMOS? Too bad you didn't have another CPU or motherboard on hand to try swapping out
You need custom loop: more robust pump, better cooling performance,easier to debug. I cool a 3090 and a 5950x with only 240+120 rads. And it’s almost silence.
Just maybe. I don’t know what is exactly going on here..
Probably some voltage regulation/reporting issue? The normal voltage in use on that CPU should be around 1.3-1.4V. The BIOS showed half of that and still near it's max temp and staying there. That combination seems impossible? Only way to know for sure is test it on another motherboard.
At the same time in windows you can see 1.3V, so both can't be true at the same time 🤨
The normal Vcore for that CPU could be in range from 0.7V (idle, 2.something Ghz) to 1.3-1.4V (full boost, 4.5+Ghz)
Wait... what? The temperature was too high and it powered off??????????????
We're forever being told from every outlet that some arbitrary value below the junction temperature is "bad" when a lower arbitrary value is "good".
You mean to tell me a CPU will preserve itself when it gets (actually) too hot?
Mind. Blown.
Have you considered the broken connection between the chiplets and IHS?
I think you should look at the IHS and see if it separated or not a tight bond.
I'd like to see a power measurement from the wall, 100w shouldn't be enough to warp the dust filter.
I bought nzxt c1200 2 months ago, also had wiggly connections on psu side
Get a FLEXIBLE SCREWDRIVER SHAFT.
Awesome tool for tight spacers.
I read somewhere that for the NZXT Kraken G12 bracket that you would have to use the AMD brackets to use it with 20 series. Can you confirm if this is true?
Very nice to see the tubes of the aio in such a different temperature! Shows the efficiency of the heat transfer 🔥
Actually, it might signal the fault is restricted / limted flow. Usually, the flow rate is high enough that the radiator doesn't get to dissipate much heat before the water is heading back down to the cpu. That high temperature difference is unusual, and might be pointing to the problem being restricted flow (whether pump not working right, or something obstructing a tube, or ????, is for Roman to look for). Also, we don't know what scale the colors were set to. While the camera was PROBABLY autoscaling, it might not have been. If it was, then the return tube at near ambient temperature is absolutely a problem. If not autoscaling, then the return pipe might have been just 5C cooler (for example) than the "hot" pipe.
the graphic cat 😁 08:26
It would be interesting to measure power draw with an elmor PMD or a power meter at the wall. This temperature doesn't feel in line with 120W
Solution.
LN2 Video Editing.
Have a dedicated POT MANAGER, who continuously tops up the LN2 POT, until the Editing is done.
If it's drawing that much power while doing nothing in bios, then the defect is definitely not with the cooler. You may have access to other threadripper motherboards or CPUs to test if it's either one, but most people don't. I do imagine AMD would want to analyze it for what's causing the problem. If it's a design flaw or something slipped by in their validation testing that they could improve on that side.
Nice to know the oil will be included in future batches of Kryosheet
you can fix that warped mesh filter with iron set on synthetic..
You should be double checking the motherboard. Switch motherboard and check.
I am surprised you did not install the am5 offset mount
When the kryosheet oil will be include with the kryosheet (approximatively) ? I need to repaste my 7900xtx and this will be very usefull
Also those seasonic fans....pretty cool, magnets...would it eventually run out? and just fall down?
Not if all the fans are screwed into the case.
My assumption is that the BIOS could be causing issues with auto-overclocking the TDP or clock speed. Could you consider updating it to see if this resolves the problem?
For example, you could update to PRIME TRX40-PRO S BIOS 2102.
second, could be the mobo some part of chipset failure. but it is too expensive to replace the mobo just for this testing.
I thought you needed a threadripper specifically sized block?
Seems like clogged radiator, it's impossible for the AIO to achieve that much ΔT between the hot pipe and the cold pipe. So the flow should be severely hampered through the radiator.
And this is why keeping extra or "leftover" parts are important.
I would have swapped another cpu into this system just to see if I get the same results. Much easier to start there vs tearing down the entire system. If you receive the same results, probably a MOBO issue or cooling issue. If you do not receive the same results, you can deduce the current CPU is okay. You could also drop the TR CPU into another working system and see if that CPU again achieves the same negative result.*
Troubleshooting is absolutely a pain to do but always best, if possible, to move the suspect (CPU or other part) to a working system . IE monitor seems defective. Hook monitor to another system. If same result mon is bad or cables etc. But I wouldn't start by removing my GPU or running more invasive tests if I can just hook it up to something I know works as intended.
Happy to see you will do some more investigating.
Cheers 🖖
*Edit after rereading my post. Forgot TR do not use the same MOBO as say a 3800 cpu ... 😮💨 If you have another TR setup then my advice is valid. If not, then you could reach out to someone you trust and can assist. OR if you just happen to have an additional TR mobo you can still test that CPU just to see if the results are similar. Though that does cause a more intensive tear down.
The joys of computers. 😳lol
To me, it seems like a motherboard issue with probably malfunctioning power profile. Try checking power consumptions when in bios.
First of all, the thermal camera shows the year 1970. 😂
The processor has an internal problem in three ccds, because there is a current leakage inside the paths themselves. it happens on ccd1, ccd2, ccd4. Vram is under heavy load, because it constantly feeds the demand of the ccds. If you manage to lower the temperatures for a while, do the AIDA64 test for the L3 cache, if their bandwidth is significantly reduced, it is definitely due to a bad processor.
GL+HF
The date is the least important info if hes already recording the image and 1979 is a default
It looks like a firmware problem but I hadn't heard that there was a TRX40 AGESA, driver, or BIOS drama
its the AIO i had the exact same issue with a corsair aio years back with a fx 9950 i swapped the aio and reset the bios to defaults and it fixed the issue
Talking about the asus GPU at the end.
Seems like the card sagged a bit.
Zoom at the pcie riser. Looks like about to fall right side.
Yes... but what was the actual problem?
Did you check if the radiator has flow? It could have corroded to the point of blockage. I think checking the radiators flow with a different pump would determine whether there's an issue with the pump or with radiator clog. I had a very similar issue with my custom loop that is about 7 years old.
As tu seulement regardé la vidéo ??
Il a essayé avec un AIO en 360 au lieu de l'AO en 240 est le problème est exactement le même !!!
Vous vous êtes concerté pour ne pas suivre la vidéo et ce qu'il y est dit et avaez décider de rendre l'aio responsable alors que der8auer dit lui même qu'il n'y est pour rien ?
Le premier avec sa connerie de rendre l'aio responsable et qui fait que la plupart répète comme des moutons me fait penser qu'il aurait mieux fait de sauter d'une falaise que vous en fassiez de même car c'est consternant de ne pas vouloir comprendre à un tel niveau !!!
A se demander pourquoi il perd son temps à expliquer et carrément à faire des vidéo si vous n'écoutez pas !!
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Vous avez un sérieux problème intellectuel j'en ai bien peur
pour ne pas vouloir comprendre
I have a Threadripper 3960x and I'm having random shutdowns also. I thought at first it was due to gaming, but then it happened while idle.
I love the pro art parts if they weren't so expensive
Lol! You are the first person I see to open AiO to check it rather than put some other cooler and test with it instead.
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
lol vous êtes au moins la 100eme personne a parler du fait qu'il n'a pas changer d'aio pour tester alors qu'il l'a fait et même marqué noir sur blanc en plus de le dire car à force il a compris que les ignares commenter sans rien comprendre !!
Épinglé par der8auer EN
@der8auer-en
il y a 6 jours (modifié)
Encore une remarque : j'ai testé le processeur TR avec un autre 360 AIO juste pour être sûr et j'avais toujours 70-80°C sous Windows au ralenti sans rien faire. Donc quelque chose ne va pas avec ce processeur ou peut-être avec le VRM.
Edit : Après tous vos commentaires, je m'assurerai de travailler sur une vidéo de mise à jour pour approfondir le problème :)
Et alors lol c'est quoi ton explication mr lol je sais tout !!!???
This is why I set the power manually in the BIOS and avoid auto 💀
Only thing i could think of is a bad psu/ motherboard or corrupted bios, but thats just so odd to happen randomly w.o changing software unless its a faulty psu.
I was lucky to be able to get the 2950X when the next generation of Threadrippers was released, but now it's that much faster than my gaming PC CPU, the 5600X 😄 I still want a Threadripper for the other PC cause of the PCIe lanes for capture cards and stuff.
Why is the term "Liquid" in the NZXT screen? Will there be a Solid too? Or maybe even Solidus? Is it the La Li Lu Le Lo?
Those nice white pre installed fans. I could use them...
My OCD was triggered by not using the top CPU-connected NVMe slot. Also found it amusing that the plastic over the NVMe thermal pad was not removed. Proves no one is perfect. XD
No point using ssd on that top slot if ssd is not pcie5. Runs perfectly fine through chipset.
@@Travelfast To my understanding of the 600 series chipset design and layout, there are only 4 lanes of PCIe 4.0 going between the CPU and the chipset. This means if you put a PCIe 4.0 SSD on one of the chipset NVMe slots, wouldn't that saturate the chipset bandwidth? That could then mean that any further load on that bus from things like USB, audio or LAN would slow down the NVMe drive's performance...
I dont know what is actually wrong, however I also has a strange issue with a 5950x and it crashed windows after some time with a ACPI error. I think this case is probably very similar, the power regulator in the soc io die being broken.
I agree with you after checking the cooling solutions that the problem has to be with either the CPU or the motherboard. Personally I'm leaning more towards the motherboard. I get why you ultimately decided to just go build a new system, but I would've at least tried clearing the CMOS and testing again first (maybe even updating the BIOS) just in case there was some kind of errant setting I was unaware of nor saw when checking settings
(Did you overclock that Threadripper? I doubt you did, one because there's not much point and two because you were only using a 240 AIO that didn't cover the full IHS. If you were overclocking it, you surely would've been using a much beefier cooling solution, but I have to ask as a matter of never making assumptions. Honestly, I would personally use the Noctua NH-U14S TR4 edition over that 240 AIO. Threadripper really benefits from using a cooler that covers the full IHS and an air cooler that does cover the full IHS has been found to perform better than an AIO that doesn't.).
I might've even tried specifically configuring it to run on a lower power profile, but that doesn't exactly prove which is the problem until you can get ether a different CPU or motherboard to test. But it might get you stable enough to use it in the meantime to do more testing.
120W at idle does seem really high. Way too high. Up until the beginning of this year, I was running a Threadripper 1950X and that thing rarely ever pulled more than 60W at idle. Which oddly enough is the same wattage my currently 7950X pulls at idle most of the time. You're 3960X is pulling double that at idle and it's not even double the core count. I know not to assume linearity in power requirements, but still my gut feeling is 120W at idle seems way too high. I strongly feel like that idle wattage should be much lower. I've never used a 3960X, but I would be really surprised if it was normal for that CPU to idle at anything more than 80W at max. So that's what I would be looking at next. Starting with clearing the CMOS. Then a firmware update. Possibly even trying a new motherboard entirely.
Honestly, I never say this because it's almost never the case, despite how many people almost always default to blaming this component, but in this case, I really am leaning heavily towards the motherboard being the issue.