Etern4l

Sorry, I’m struggling to get impressed by YT “empires”, since they are mostly based on human weakness and stupidity. 15M subscribers? Pfft, how about 35M https://youtube.com/@SSSniperWolf I value the guys with 1-2M (or less) sub channels but who do honest work. Sadly the society rewards Linuses and Sniper Wolves of this world more.

Agreed, just a minor correction: L”MG” is a speck of dust on the broader tech sector landscape, but it seems that they are a much larger channel (15.5M subs) than GN (2M) so GN is the underdog and the party who opted to publicly defend themselves against claims of inferior testing methodology. Looks like they did their research too (no doubt lawyers must have been involved). With this new stuff coming out of the woodwork exposing the situation was the right thing to do. I doubt there was any other way, realistically. Ridiculous to hear about some external investigator coming in. LTT is basically Linus, they can’t use the investigator to kick him out (say if the wife and the new CEO dude decided to have a go on their own), and obviously Linus knows exactly what’s going on in there. it’s not like LMG employs 10k people in two skyscrapers and Linus just landed on Earth after a decade on Mars. I guess the investigator will help identify the fall people on behalf of the “exec team” who must be very busy with something else at the moment.

Disgusting. There has been a fair amount of discussion on this in the “official thread”

Ridiculous? Of course, but I’m finding it hard to laugh at the scale of this sad phenomenon which is fuelled by the vast stupidity, and social media algos designed to exploit it. The problem with social media influencing, is the underhand nature of it. You are not watching an ad, or infomercial (which would be fair enough) - you are watching a “review” or “industry news”. Here is one example, going back to what’s wrong with Linus Influencing Computer Enterprises, which ticked me off: A bunch of platitudes, zero actual benchmarking, no deep analysis, just a huge talking head trying to convince you PCs are done by dousing you with FUD. Disgusting. The message is simple: “Yes, we took money from Apple for making this, because we can - watch it b…!”

Anyone else found the fake remorse video unpalatable? (I just watched a few bits) "LMG".... OMG, a Media Group! I'm disappointed the name is not simply LC - Linus Corporation. Basically a YT site with a few goofy influencing videos a week, but now sporting a CEO, CFO etc. Did he also hire management consultants to help out with this incredibly tricky situation? I have always wondered what the hell they need all that hefty infrastructure for. Is it perhaps that the employees are unhappy that they end up having to cut corners just to pay for all those gigantic and surely unnecessary overheads? Anyway, it's Steve David vs Goliath, and it's not a too big too fail kind of situation. I would not mind if they did, since I found most of their content questionable to pointless (but obviously there is a bunch of happy clientele for that). Too much influencing BS, not enough objective in depth content, so what did they do? Attempted to bad mouth competition hoping to grab market share that way. Disgusting. Street cred is difficult to gain, and quite easy to lose.

Makes sense. Still, just looking at the flow rates, his testing suggests there are seriously diminishing returns at some point.

Please, the fearless shunt-modding, delidding, uberoverclocking bad... @Mr. Fox is concerned with running a 30 min CB loop on the grounds of a possible silicon degradation. Made my evening here sir :)) While, it's true that gamers probably never get close to 300W on a 13900K, you can look at it as running 300W for 30 min instead of 150W for an hour. Anyway, I'm not saying 30 min @ 300W or bust, but rather just suggesting it would be interesting to see how long until the temps achieve stable state (since they clearly didn't in your results). At 300W one would have thought that 10 mins would have been enough. Enough with the grilling, you ran a quick test on your setup, instead of spending half a day swapping AIOs in a case, like Tomshardware allegedly did here: https://www.tomshardware.com/reviews/lia-li-galahad-II-performance Case Be Quiet! Silent Base 802, system fans set to speed 1 setting. I am not sure what to think about their review. -6C is an awful lot for what is essentially the same type of cooling tech, especially if derBauer's data on the merits of 4000RPM pump is is correct. I suppose that 30% thicker radiator and 3000 RPM 100cfm high static pressure fans must have had some effect though... Wow. Hmm, can anyone confirm this? So, basically AIO pump speeds over 2000RPM are marketing gimmicks? Would be easy enough for me to check 2000RPM vs 3000RPM, just not now unfortunately. I wasn't concerned about the lack of PWM control in LFII because I wanted to dial the pump down, I was concerned because I wasn't sure if the pump was working correctly.

There is nothing funny about how Nvidia has been treating consumers lately.

Not particularly simple. There are quite a few variables, and there isn't an easy formula we could use - would need to solve a potentially hairy system of differential equations to completely model the thermodynamics here. As you noted, delta T of the coolant temperature is a function of time since the start of test, but a host of other variables are also involved in determining the coolant temp (which won't reach ambient under load, certainly not in an AIO): 1. Coolant volume - will affect the rate of change, but possibly the equilibrium state as well. 2. Pump speed - interesting point about faster not being better (contradicts the simplicity claim BTW). Would be good to see some supporting data. 3. The effective heat dissipation capacity of the heat exchanger (radiator(s) + fans) - difficult to quantify, since this itself depends on other variables such as T_coolant - T_air and as such it's not constant, 4. All the other factors mentioned earlier (coldplate/block design, TIM etc.) Here is an interesting test where someone looked just on the radiator/fan side of things: https://linustechtips.com/topic/282708-what-is-the-maximum-heat-dissipation-of-a-360mm-radiator/ A fairly detailed open bench like test and determined that the coolant dT (presumably at equilibrium) with 300W load is about 10C. Also interesting to see that water temps in and out are very close, so it's not clear how a fast pump could adversely affect performance. Two problems with this: 1. The testing period of 5 min doesn't really correspond to any real life usage scenario - I mean who uses their computer for 5 minutes? it's not a mobile phone. IMHO it would have been much more informative to run the test to equilibrium (and present the results as a plot in the time domain). The custom loop would then serve as a reference point for AIO manufacturers to aspire to :) 2. Yet the conclusion refers to some sort of equivalence between custom loop and AIO, which could unfortunately be misleading That's fine and clear, as long as people understand that the results won't necessarily reflect the AIO performance when mounted inside a case. Fair to assume there will be some readers who will gain that understanding the hard way - that's OK too, part of the hobby :) That thick rad is what makes that AIO work. The pump is not the fastest though, only 2000 RPM with no separate PWM control. The coldplate is also on the small side. I put mine back in the box as not being able to see the pump speed drove me nuts. That said, I think it does/did an OK job, and that VRM fan was probably helpful.

Thanks, I missed that - again, looks like a risky statement, and applies to the open bench setup at best. Looking at the specs, I don't believe this EK AIO would be a significant improvement over my DeepCool LT/S720, and there is certainly room for improvement in terms of the cooling the latter can provide. I would be hugely surprised if a large external rad would not have made a significant difference. Also, wouldn't it be fair to say that virtually no one buys either an AIO or a custom loop to cool the CPU only? Again, I would expect there to be a significant difference between setups with internal and external rads. What do you mean by normalize? Given the same load, shouldn't the coolant temp in a loop containing MO-RA3 rad be significantly lower given that it can exchange much more heat (roughly 3x) than a single 360 AIO rad? It's a rhetorical question in the context of a typical PC build setup, but should even be the case in an open bench - the question then would be: how much cooler is the coolant in the custom loop? Guess not super easy to accurately measure, other than by looking at rad or tubing temps, all I can say right now is that the coolant tubes on my AIO are warm to the touch, nowhere near ambient, I would ballpark it at 15C+ above - surely a huge external rad would bring that down quite a bit? Ca. 3x the exchange area + ambient airflow. Put another way: assuming the open bench scenario to normalize the airflow temps between a MO-RA3 and an AIO, would you expect a good AIO and a custom loop attached to the CPU only to achieve the same coolant temp which is very close to ambient? Doesn't matter if the custom loop has just a single 360 rad or something like MO-RA3? I think if we had this data (even external tubing temp as a proxy), things would have been much clearer.

Hate to jump into this peer review of your EK AIO test, but the only way the AIO would come close to the performance of your gigantic custom loop with external radiator and a dozen+ fans, is if 3 conditions were met: 1. The cold plate on the EK performed similarly to whatever state of the art water block you pitted it against - OK, I guess that’s plausible. 2. The water inflow into the AIO head had the same temperature under extended load as the water coming into the block from your gigaloop. I would say this is strictly impossible in normal conditions where the AIO rad is mounted in the case and sucking in warm air from the inside, although this effect would not be present in an open bench. The main conclusion here would be that you tested on an open bench, which would be a less realistic AIO test, but even then it would also be surprising to see that your gigantic rad makes no difference vs a single 360 unit. 3. Even if somehow the coolant temperatures entering the block/head were equal, there would be the question of flow rate. IIRC the EK AIO uses a 3200RPM pump, whereas I would presume your setup involves 2+ 4000RPM pumps, so this would suggest an additional delta in favour of your custom loop. Best case scenario here is that we are effectively saying: a custom loop, with huge external radiators, massive coolant volume, multiple pumps etc. - all that is a waste of space, time and money: people can just grab an EK AIO and get the same end result. To say this is a surprising message coming from @Mr. Fox would be an understatement.

Was initially skeptical of that Tom review, but the specs are indeed awesome. The coldplate looks very similar to my DeepCool LT720, but that’s where the contest ends: the LianLi has a whopping 4200 RPM pump and a 32mm radiator (edit: 3000RPM fans are also included...). Game over, totally explains the performance improvement. Looks like the strongest 360 AIO on the market hands down. Shame about the proprietary fan connectors though. The preinstalled fans are not ideal either for people who would rather use third party options, but indeed those fans looks so strong most people would have little reason to upgrade them. https://lian-li.com/product/galahad-ii-trinity-performance/

That’s really unfair. Penalised for well-meaning, if perhaps a little naive and incompetent, communications which may have affected fellow human beings who unfortunately turned out to be witnesses:it’s not SBFs fault prosecution got so many people talking, who could keep track?

If the strategy was for people to achieve enlightenment before they act on the immediate threats we would all be toast. Out of the countless billions of people we've only had one Buddha, 2.5 thousand years ago... The problem is that unconstrained AI research will almost inevitably lead to the development of AGI, and this will likely happen much sooner than people think. The serious issues resulting from indiscriminate deployment of lesser ML technology have been discussed for years, and as of yet very little has been done to actually address them. Most of the effort has been focused in eliminating "biases" and similar virtue signalling around woke agenda, while at the same time technology has been used to support extremism on the electoral front. Thanks. Unfortunately, works like these are inconsequential in terms of impact. The only thing that can save humanity is a very quick increase in awareness among the broader population. Unfortunately, that's a really tall order and thus we are heading towards self-made dystopian future, or lack thereof .

It’s just the weight of experience bro. Had you wasted as much time as me following more or less useless online memory OC tips, you would be firm on the matter as well 🙂

Thanks, given your massive experience with DDR5 overclocking, that opinion really means a lot. On a more basic level, looks like you missed the part where he mentioned all the BIOS tweaks. Relatively little discussion of testing though (just skipped through most of it, so apologies if I missed that). In my experience it's very difficult to get DDR5 OC hacks to work really long term stable, so this stuff is good for some light benching or gaming, and then people end up saying ridiculous things like "great setup, passed all the tests - just crashed in BF or The Last of Us". Most synthetic memory test tools (including the free, 8-core limited, version of Memtest86) are surprisingly bad at picking up anything other than gross memory faults.

Clickbait IMHO. Given the number of hacks involved in supposedly getting this to work, I bet the whole thing would quickly fold under heavy load.

Jay should stick to talking about things he has at least a remote idea about. Unlike with crypto, lower end GPUs are next to useless when it comes to the large language models (which is what is meant by “AI” in the media right now) - not least because NVidia cut out NVlink support in 4000 desktop series. That stuff needs a lot of VRAM. Clickbaits pay the bills I guess. The real negative aspect on consumers is caused by the demand on the data centre side. If the demand wasn’t there NVidia would be able to offer better products at better prices to the consumer. There is no such thing as free AI folks.

On the plus side, the fewer Windows installations the better.

The latest MSI BIOS update added support for the upcoming Intel CPUs. Will Intel do it (+33% MT performance) again? Somewhat unlikely given it's still the same socket and fab process I guess, unless the CPUs ship with LN2 cooling lol.

Thanks, I will probably struggle for time though. Any good points? The very notion of "AI ethics" is nonsensical to me. You can have ML ethics, but not AGI ethics in the human sense. AGI/ASI might adopt some sort of ethics towards us, if we are lucky. No amount of human debate or "alignment" efforts will ultimately affect that. Yeah, so this is about thinking a few steps ahead, something we quite evidently suck at as a species. We can recover from climate change, extreme weather, rare earth metals/Tesla car shortage, even smaller scale nuclear conflict, but good luck outmaneuvering a vastly superior intelligence - either enjoying complete agency, or controlled by a megacorporation, a weirdo billionaire, or a hostile totalitarian regime.

Maybe stillsuits will eventually become a thing, for now folks living in Africa have to put up with heat like that and worse (50C) on a regular basis without them. Not sure what their lifespans look like though.

Yeah, random shutdowns are really bad. Is a screw missing somewhere?

X as in “cross out”.

30,000 people evacuated in Greece. Huge wildfires on Rhodes and Corfu among 40C temps.