Groove_C

Despite only 27 mm radiator, EK-AiO performs same or marginally better than Arctic Liquid Freezer II with 42 mm radiator, because of more powerful pump and better heat transfer thanks to higher pressure on the CPU and more even contact with the CPU. Arctic Liquid Freezer II also can't be refilled contrary to EK-AiO. And I consider this as important criteria, because some of the coolant evaporates through tubing over time and air in the system can make the pump fail or significantly shorten its lifespan and also negatively impacts temperature.

When I opened the box, even without touching anything, I could directly see that the finish quality was rather mediocre/cheap, be it the radiator or waterblock-reservoir-pump unit. I also searched for info before buying it and seen a lot of reports about the pump failing relatively quickly and fans being really cheap (no good pressure/volume and a lot of noise). But I thought that fans can easily be replaced by Noctua and as one can buy the pump separately any time, it wouldn't be a problem for me. I really wanted to have everything copper and have the ability to service it and modify/expend. But it's not even the pump or the fans in the end. I installed it and despite much more coolant than inside Noctua NH-D15 and more radiator fins, it was hotter than Noctua and the temp increased even faster than on Noctua. It was a real pain to install it, because the mechanism contains too many unnecessary parts that also make it less solid, more flexible, which is of course bad. The worst are 4 springs that must apply pressure for good contact with the CPU and allow for good heat transfer. And this is the biggest problem, before pump and fans. The springs are really thin/cheap and can be easily squeezed together with just 2 fingers. So you can forget about good pressure on the CPU and about efficient heat transfer. Fans and pump spinning at their max for nothing, since heat can't really be transfered from the CPU - the coolant stays cool, because of thin/cheap pressure springs + really not solid fixation mechanism. Now ordered EK-AiO Elite 280 D-RGB, because it has the most powerful pump among all existing AiOs, can be refilled through an additional opening in the radiator and it has a very good coldplate - not too thick and microchannels covering a big area. + the pressure on the CPU is very even and strong/good and the mounting mechanism is the simplest and most reliable among all existing AiOs. I don't have place for 4 of its fans (push-pull), I don't need RGB. It's only because of the pump I bought it. EK-AiO Basic and D-RGB, so all non-Elite EK AiOs, have "same" pump, but with lower max rpm. Elite has 3300 rpm pump, Basic and D-RGB (non-Elite) have only 2600 rpm physically "same" pump.

Don't buy Alphacool Eisbaer Aurora AiO, its quality and performance are really bad. Only on paper it seems to be good. Will see if I can return it and buy EK AiO instead.

Me buying AiO wasn't a choice, but rather a necessity.

I don't like Corsair AiO. It's cheap (production), only RMA or throw away, with aluminium radiator, relatively small coldplate with microchannels only in the center and can only be used as is + crap and necessary software always running in the background.

...

...

...

EK-AiO Elite 280 D-RGB. Alphacool Eisbaer Aurora 360 I bought first, was of very bad build quality and performed very poorly, actually worse than my Noctua NH-D15S.

Unfortunately no time now. Bought a 280 mm AiO, since on air it's impossible to cool 11900K, because its memory controller gets very hot, due to crazy RAM OC. And since the memory controller is inside the CPU itself, it's "boiling" ))) With lower RAM frequency and higher timings it's possible to cool it with Noctua NH-D15. Will see if I can do 4000 MHz 13-14-14 instead of 14-15-15, when I have time. But avg. FPS is around 110 or slightly higher in YAAB 1080p standard.

Now I have my RAM at 4000 MHz 14-15-15. Arma loves it 😏

If you can do 3900 MHz CL14, it's even better than 3866 MHz CL14. But both are better than 4000 MHz CL15.

Well 4000 and CL15 is only the frequency and only the first of several dozens of timings. Also, just so you know, 3866 CL14 outperforms 4000 CL15, requires less voltage and only 29x multiplier (with 100:133) instead of multi 40 (100:100), which is harder for the memory controller. So it's just a nice number )))

Well, you don't tell us your RAM temp, room temp, RAM timings and frequency. RAM voltage alone means nothing. Even 65°C is a perfectly working temp for RAM, but Samsung chips can start to have errors even past only 35°, if you go for very high frequency and very low timings.

@Valkenyes, a fan than cools RAM to not higher than 40°C. And less than that when ambient temp 18-22°C. Each of my RAM modules have temp sensors. I also downclocked my RAM from 3960 MHz to 3866 MHz, so it requires much less voltage, RAM and memory controller inside the CPU reach much lower temperature and FPS difference is only marginal. You can measure it, but can't feel it. So not worth the stress for the hardware.

Just don't forget that for such performance you would need such hardware and even having all off this, won't be enough to reach such FPS, without a lot of knowledge and patience to tune everything to the last MHz and lowest doable timings - it's on the limit of doable even if it's still on air. So expectations have to be lower than that.

Yes, >30 FPS and also not less than 30, is what's needed, to enjoy the game and no need for very expensive hardware to achieve this. You are correct. But if much much higher FPS can be achieved, even if only through more €€€/£££/$$$ and with a lot of knowledge/hard work/time and the person wants that or at least feels the "need", why not. Nothing wrong with that.

So your wife/girlfriend shouldn't approve the upcoming upgrade, since you already have 30+ FPS 🤣

No, as @oldbearsaid, 30 FPS is plenty, for Arma - no need for 60+.

)))

very high temperature of 5600X was 100% not a problem with your Noctua cooler, because a CPU like 5600X, considering its TPD, is nothing for your Noctua cooler.

If R5 5600X was throttling, then i7-10700K would have been the better choice. It's more powerful than R7 3700X, per core, because at stock it has 4.7 GHz boost all cores, instead of 4.3-4.4 GHz of 3700X. And it has also 16 MB L3 cache, like 3700X, but in one piece, not 2x8 MB + much much better RAM responsiveness. And your Noctua wouldn't have had any problems to cool it, because of one big central DIE. But I imagine you already had AMD motherboard, so a new Intel motherboard wasn't really an option.

That's the problem with Ryzen CPUs construction - they can't effectively transfer heat from DIE to heatspreader, because the DIE is very very small, because of 7 nm and not located in the center, but the power that's inside it is very high. It's not so obvious on Ryzen 3000, only because their frequency is not that high. That's why I prefer Intel. Because they have their DIE in the middle and it's much much bigger, so more surface to spread heat across and better heat transfer to the cooler + much better memory responsiveness. But yes, 14 nm makes them consume more electricity.

Well, that's to be expected, because it's a laptop CPU and its frequency is relatively low, compared to your 9900K, despite it also having 16 threads and 16 MB L3 cache. + 3200 MHz CL16 in a laptop is not the same as 3200 MHz CL16 in a desktop PC, because it's only the frequency and the first of many timings. The rest of the timings are way higher on laptop RAM than on desktop PC RAM, which further reduces the performance. So one shoud never directly compare laptops vs. desktop PCs. A 5800X desktop CPU on the other hand has 32 MB L3 cache, instead of only 16 MB and also boosts higher than your 9900K, as exemple.

Well, if you have built a system just for Microsoft Flight Sim and unless you didn't already have 3700X, then it was a really bad choice, because it's as single core as Arma and works best with highest frequency possible, instead of threads count. A 5600/5800X would have been the way to do it, if that is when you've build your current Flight Sim system. But if you've built it when Ryzen 5000 weren't even released or available/in stock, then a 10600K/11700K would still have been a better choice for it than 3700X.