Etern4l

Yes, what I wasn't sure of is what the signal pin is for - this is actually RPM info sent back to the mobo, so all good, given that the pump has a separate header. Will do, just really wanted to make sure I can see pump RPMs - it's idiotic LFII doesn't have this, especially given that they advertise PWM pump control as a feature.

Ha, neat hack. Definitely worth considering it if it turns out that the AIO is OK and I am not going to RMA it. For now, I am about to pull the trigger on the reasonably priced DeepCool LS720. Given free choice EKWB Nucleus would be better (a bit more premium and PWM pump control) but availability is poor, and would work out at 2x the price with delivery who knows when. The LS720 pump is 2x more powerful than the one in LFII (in terms of power consumed) - it actually requires a SATA power connector. The backplate is also larger (supports Threadripper as well). Unfortunately the radiator is standard 27mm, although with high fin density. The pump doesn't support PWM - only connects to the mobo via a 3-pin header. Now, if I understand correctly, even though the pump would only be controlled via DC, I would still be able to see RPMs in the BIOS (that's what the 3rd pin is for)?

I'm worried the amount of care billionaires have for the court of public opinion roughly halves with every $20B or so of net worth. In another beautiful story, the sociopath mocked a disabled employee whom apparently Twitter kept in a limbo to avoid a contractual severance paynent: https://fortune.com/2023/03/07/elon-musk-twitter-ceo-office-space-clip-haraldur-thorleifsson-layoff/ https://www.bbc.co.uk/news/technology-64871183 How about just letting people you no longer need go in a civil and respectful manner, and in accordance with prior legal arrangements you bought into?

While I don't have an easy way to test this exhaustively right now, in Gnome settings (on X11) scaling is very clearly a property of the display, not the system. Also apparently one can enable fractional scaling in Gnome too.

I don't use this day to day but just tested in Gnome. Just two issues: 1) Only coarse scaling, e.g. 200, 300, 400% - I understand KDE supports fractional scaling down to 6.5% or something like that 2) Some apps needed to be restarted to redraw with new scaling. Other than that, ran through several, including Libre Office - with no material issues. No surprise since this is based on X Windows which is scaling / DPI aware, and was running on serious workstations at a time when DOS ruled PCs, and Bill Gates was busy copying Mac OS :)

If he said Windows apps are something you would spend most of your time interacting with, then Indeed unless you manage to run them in Wine, the rationale looks weaker. But it's not completely pointless. If you run just some apps on a VM, that's an isolated setup: full image backups are trivialised, any exposure to Windows system or security issues are limited - you are just losing or struggling with a VM, not the main system. Case in point: Windoze 11 auto-launched 22H2 update in background which killed my Windows installation just now, probably because I wasn't ware of this and restarted... If you're going to force background system upgrades on people, implement this robustly (somehow), morons. How difficult would it have been to display a dialog "Windows upgrade is in progress, are you sure you want to restart?".

Of course. I haven't, but I'm almost certain there was no pump out with the phase-change TIM. Well, the problems are significant. 60W less power draw in CB23. It could be that the pump is just not running at full speed. We don't know because Arctic didn't think it would be useful to have RPM monitoring on the pump. Well, speaking of updates: looks like Redmont Reprobates have tried to punish me for abandoning their spyware product by deciding for me that 22H2 update is not optional and starting an unsolicited stealth background upgrade. I guess I may have restarted a few times in the middle of the update I wasn't aware of and now Windoze is unusable. Starts showing some artefacting and reboots automatically shortly after logon, even though there is unlikely to be an issue with the video card. If I boot in safe mode, the taskbar looks different, and things like network and search don't work. Network worked for a moment and windows update showed 22H2 update 65% complete or something.... <facepalm> Thanks for the advice bro. I mean the fit is good. Why would it sort of be good initially: always 310W - 330W on an AIO (whenever I saw 310W my reaction was to repaste), but then suddenly deteriorate? I guess let's see if AIO replacement helps first, and maybe I will order that delid kit too.

@Mr. Fox @Papusan Yep, something else is going on. I tried repasting twice, tightening and loosening the contact frame - absolutely no difference. 260W max, cores 5 and 7 at 100C. We can exclude thermal paste as being the culprit. Another thing that happened is that now Windows started rebooting without warning straight after logon, after briefly showing some artefacts on the screen. Linux is fine. I managed to boot into safe mode and noticed Wingross 11 kicked off unsolicited 22H2 installation in background, could have something to do with that. Given that the degradation occurred quite suddenly, I assume the AIO developed some sort of an issue. I don't have any RPM reading for the pump ever since I disconnected the bundled fans, so hard to say what's going on (as bro Papu mentioned a while back: poor design from Arctic with putting the pump and fans on a single header). The VRM fan is working. I'm not sure how to verify or reject this hypothesis other than by trying another AIO. Another possibility would be that of some internal issue between the die and the IHS, that a delid might fix. I guess that's far less likely though. Thoughts? BTW this PTM7950 is almost like an adhesive. The AIO pump came off with relative ease, however, what was left of the TIM took some rubbing and scrubbing to get off the IHS and coldplate. Wonder what it would be like after 2 months or 2 years. Still, can't fault the initial results.

No forced updates on Linux BTW. Installation of 22 H2 Windows decided to kick off for me in an untimely manner should have been my decision.

With all due respect, I don't these comments and Linux in general are in any way comparable to the Win 11 situation. Those are just one or two viewpoints by deeply opinionated people of which there are thousands in the Linux community - consequently, there will always be an opportunity to identify and point out critical opinions regarding virtually any area of the system, and Gnome has always been divisive. This is kind of similar with Windows I guess, but this is where the similarities end. Linux users have far more choice in terms of customisation of the system (don't like Gnome, use one of the several other window managers available). The experience of Gnome (and others, KDE in particular) can be fairly deeply adapted to users' liking via Tweaks and Extensions, something Microsoft doesn't really provide beyond bare basics and the ability to turn off eye candy. Moreover pretty much anyone can get directly involved in the development to some extent (including testing, feature requests etc.), which is obviously not possible in the Windows world to any meaningful degree apart from beta testing in which all users do involuntarily participate lol

I can assure you the difference between 4k and 1080p on proper 4K content is very clear to me at 6 feet on a reasonable size screen (50 inch plus). It may be a question of the actual content (here I am considering content that would clearly exhibit the extra resolution, i.e. the vast majority of cases), the person's eyesight, or other visual acuity related factors etc.

Yeah, I have the contact frame. One thing is clear, the LFII wasnt developed with 13th gen in mind, simply because it was designed much earlier, and the coldplate barely fits the IHS. Looking at those comments on OC, It's amazing what I managed to achieve with it lol BTW apparently the coldplate on the LS720 is slightly convex, which is probably what you want. I am pretty sure the IHS is slightly concave, so that would match. https://www.techpowerup.com/review/deepcool-ls720-aio-360-liquid-cpu-cooler/3.html

Thanks for pointing me in the right direction bro. I have no idea why you had to do the mods you did. Did the bottom cover get in the way of anything? Nothing I noticed on the MEG ACE, however, I have been wondering if the standoffs aren't too tall... What do you mean by Z690/Z790 bending mess? Actually, according to this thread, the exact high pump RPM AIOs I identified are the ones people recommend (Deepcool LT/LS720 and the EK Nucleus): https://www.overclock.net/threads/13900k-ks-cooling.1803906/ The problem is that it's unlikely any of those folks have had a chance to test the setup the way I would (think a 2 week CB23 run). Will keep digging around. Edit: One guy wrote: "The LT720 360 keeps my 13900ks below 100c at up to 315 watts. Most aio can’t even come close to 315 watts." Too bad he didn't specify the TIM. My modified AFII setup did 330W then settled at 325W, at 25C ambient, but that's kind of apples to oranges comparison in terms of fans I guess, if he just used 3 stock ones. Still, I would I be able to achieve that if I had a contact issue? My problem is that then the paste gets pulverised and performance collapses after a random period of time under load. Guess I just need to try and see if I can lower contact temps with a different AIO + maybe LM, then some compromise on the power will be required. What a mess indeed.

Yeah, BlueStacks has been around for quote a while.

That explains why the link wasn't shown inline by the forum... Here is the underlying page: https://www.kitguru.net/components/cooling/luke-hill/deepcool-ls720-360mm-aio-liquid-cooler-review/2/ Agree the fit could be a problem, but not sure what to do about that. Actually another AIO could be a solution, especially given that the fitting mechanism on the AFII is not rated too highly. I'm not sure if it's worth lapping the coldplate. Logically the pump speed (flow, to be more precise - which very few manufacturers quote) should play a role, right? If you have 50% more water flow, that should materially affect the heat exchange especially if the coolant temp is around ambient. Also, from what I can see the pump block on this and similar AIOs is literally 2-3x the size of the pump block on the AFII. There is also the matter of coldplate size - the one on AFII is quite small, probably further reducing the heat exchange. There is a lot of variance in the reviews, as you can see from the link above. I have to say, I am a little suspicious when I see reviews where the Corsair H150i (nothing special specs-wise) takes all the top spots lol Besides that DC LS720 there are also a couple other 3000RPM pump AIO options which seem to perform similarly: * EKWB Nucleus I believe, the current model * Thermaltake Toughliquid Ultra 360 I will have a think. A bit of unwelcome painful guesswork and hassle to be honest.

Yeah, I suspect the low pump speed plays a role here, precisely because I use custom fans meaning that the water coming out of the rad is really cool, by AIO standards. I looked at Tom's reviews and they said, wow - this AIO (another DeepCool model with 2550rpm pump) is awesome because it can handle 315W with the 13900K. Well, my setup as can handle 320-330W with a 2000RPM pump (with fresh paste). I was sizing up this AFII AIO for the 12900KS, not the 13900K, and didn't pay attention to the pump speed. Arctic managed to get away with a slow pump due to the oversized radiator. Ideally I would like something with a thick radiator and a fast pump lol. If no such product exists, then I guess it will be time to start putting together a custom loop setup. Edit: yeah the AFII 360 doesn't cut it anymore these days. 9C worse than the DeepCool LS720 (at £120 it's tempting): https://www.kitguru.net/wp-content/uploads/2022/07/temps-oc.png

Thanks, appreciate your advice, although in my mind the challenge is still not improving CPU thermals per se, but solving the TIM degradation (which is caused by insufficient IHS cooling and/or flawed contact, obviously). I really don't see how rising IHS temperature would help with that. For now I'm reviewing this AFII 360 AIO. For starters, it has a 2000 RPM pump... whereas some other offerings on the market come with a ~3000rpm pump. As @tps3443 was saying, pump speed is key. For example, DeepCool LS720 comes with a 3100RPM pump, which could provide material improvement potentially. Need to read up more.

Many thanks. Solid advice from the benching perspective, however, I'm not sure deliding first would be helpful in my case. Here is my reasoning as to why not: The nature of the problem is that, for a combination of reasons, the TIMs lose properties or pump out after a period of few weeks under heavy load. Likely contributing factors include: * Intel, maliciously (to prevent the gaming CPU from competing with Xeons) or idiotically, used insufficiently large IHS for the 300W CPU - nothing we can do about that * Potential IHS concavity or other material surface imperfections - sanding could provide an improvement, at a loss of warranty * Ditto for the AIO coldplate - sanding this is pretty much risk-free, although some coldplates are slightly convex on purpose and sanding could make things worse * The AIO itself could be lacking in terms of coldplate dimensions or pump performance (there is no pump/flow monitoring on this) - could try another AIO * There is a more durable TIM to try still - LM is the only option left, although opinions on durability seem to vary * The overall system can't handle the given power on a sustained basis - give up and limit the power That's it as far as I can see, short of rebuilding the system with a custom loop. Now, what I understand deliding would achieve, is it would improve heat transfer between the die and the IHS, further increasing the IHS temperature and compounding the problem. Great option for short-duration loads, benching and gaming, and for people with extreme cooling, but not necessarily in my case, if I understand correctly. Yes, it would potentially allow me to save the original IHS, though.

Yeah, that could be part of the problem. 1. I'm using Arctic LFII 360 AIO with 6 Noctua IPPC 3000rpm fans. A sensor placed on the radiator consistently shows exactly 2C over ambient, so I guess not too bad. That said, the cold plate on that AIO is just long enough to cover the CPU, one would hope for some margin. 2. I'm using the Thermaltake contact frame, very gently tightened about 90 deg over the point of resistance. No problems with the same setup (actually with just silent 3x Arctic and 3x Noctua 2000 RPM fans) on the 12900KS. The difference was the power draw: about 220-230W vs 270-280W (over quite extended periods). Also, with the PTM7950, things were working marvelously for about 2 weeks, then BOOM: two cores at 100C and power draw down to 240W, in a matter of a day or two. Thanks for the lapping tip, will start reading on that. I also have a couple of pics of the IHS and coldplate, after previous similar paste failures (Phobya Nanogrease lasted the longest, about a month, this is second-best at over 2 weeks). I mean, it's not the end of the road yet, a few more things to try: * Lapping, now that you mentioned it * Different AIO * LM on the IHS * Delid * New case and custom WC loop * Throw in the towel and power limit - will still handsomely outperform the 12900KS.

Well the noname PTM7950 failed as well. Angry at Intel for producing a practically uncoolable CPU at stock clocks, unless used occasionally for gaming, or cracked open for direct die cooling. At least now it's clearer what Sapphire Rapids are for.

I'm not clear on why bears would be allowed anywhere outside fenced parks / zoos compounds.

The noname PTM7950 from eBay failed after 2 weeks or so. I received another noname sheet from China, I might try that, otherwise I'm a bit out of ideas. I think I will just have to impose a power limit on the CPU. Many thanks to Intel for producing an uncoolable chip for occasional gaming use with an insuffcient IHS and based on crap 10nm process. Actually, it's not clear, or even unlikely, that PTM7950 was the issue. Quite possibly the AIO failed, as repaste didn't help.

Even Bill Maher has had enough lol I learned a new word: 'presentism" ...

"4070 is an IQ test card" lol

If the goal is 100% stability, Memtest86 in parallel mode over several hours may not catch everything. I have found Google GSAT effective but again needs to be run over very long periods.