Groove_C

for Microsoft malloc with lock pages, 32 GB RAM is kinda mandatory, because  you need to always have more than 8 GB RAM consecutive free space, when running Arma 3, as it would lock several consecutive 1 GB large pages in your RAM and prevent it from being swapped into the paging file of your disk.

Will use Microsoft mimalloc 2.0.6 (AVX) with lock pages in Windows and see whether FPS degradation over many many hours of uniterrupted gameplay on same server/mission/round is solved or improved vs. CMA AVX malloc (Intel tbbmalloc 2016) or default BI Intel tbbmalloc (2017).

Microsoft mimalloc 2.0.6 (AVX) (April 2022) with lock pages in Windows is BattlEye compatible and much much much newer than CMA AVX malloc (Intel tbbmalloc 2016). + a lot of bugs fixed and unnecessary debug and stats codes removed, that are present in CMA AVX malloc, which causes overhead in certain situations and memory leaks. Microsoft mimalloc 2 0.6 (AVX) handles memory fragmentation better, in a long run. My tests of Microsoft mimalloc 2.0.6 (AVX) (April 2022) with lock pages in Windows, don't show any min. and avg. FPS difference (within margin of error) vs. CMA AVX malloc (Intel tbbmalloc 2016), but max. FPS is slightly higher, on Microsoft mimalloc (AVX) with lock pages in Windows. AVX2 and AVX-512 mallocs perform same as AVX (within margin of error). https://github.com/GoldJohnKing/mimalloc/releases https://github.com/GoldJohnKing/oneTBB/releases

it's also better than i3-10100, since, again - 12 instead of 6 MB L3 cache of 10100 and better per MHz performance than 10100, although both are 4/8 cores/threads.

@oldbeari3-12100 is better than stock (no OC) i7-7700K or i5-8600/9600K, even if it had same 8-9 MB L3 cache. + it has 12 MB L3 cache. So, Alder Lake architecture alone already boosts FPS in Arma, despite 4.1 GHz all cores vs. 7-9 gens i7-i5 and 1/3 more L3 cache increase FPS even more. I'm talking specifically about Arma. And since the frequency is just slightly above 4.0 GHz + only 4/8 cores/threads, it's not a problem to cool it down with a 40€ air cooler. Paired with a B660 board and some 3600 MHz RAM and you have a relatively good system, specifically for Arma 3. And everything is new - no used hardware.

haven't encountered anyone using one for Arma here or in the official A3 Discord, yet.

with only 3600 MHz 16-18-18-36 and no CMA AVX(2) malloc it already does slightly more thsn 100 FPS avg. and slightly more than 60 FPS min in YAAB 1080p standard. And that at 4.45 GHz with a just slightly more expensive than 100€ mobo, like 65 W or less and an air cooler. Add to that up to 10 FPS more alone from CMA AVX(2) malloc + RAM from 3600 to 3800 MHz and all the timings manually tuned, gives even more FPS. So I would say like 115 FPS for sure, if not slightly more.

Ok, but when you share your YAAB results, you have to mention CPU model, cores and cache frequency, Hyper-Threading on or off, E-cores on or off, BattlEye on or off, RAM frequency, primary timings and 1:1 or 1:2 mode, for everybody to be able to understand where the FPS comes from. Because sharing FPS only and 0 info is not how it works here. Nobody should have to guess.

His CPU and RAM settings aren't stable. + tested without BattlEye. + he disabled 8 cores and disabled Hyper-Threading.

He has it at 5.4 GHz cores and 4.9 GHz cache, Win 11 and his RAM is not CL15, but CL14 and most importantly - it runs in gear 1. Gear 2 ruins the performance.

https://videocardz.com/newz/intel-core-i9-12900ks-pre-binned-cpu-with-5-2-ghz-all-core-frequency-allegedly-in-the-works

Well, your 12600K has 1/3 cache less than 12900K. Cache size boosts FPS. 12900K is OC'ed to 5.4 GHz. He has disabled the E-cores, since they reduce FPS by 10-15. Your 12600K is only at 4 6-4.7 GHz. He has the best RAM there is. You RAM is below average. And he hasn't even tuned his RAM - it runs at XMP. So more than 128 FPS average is possible, like 135 or 140.

CPU and RAM tuning can result in up to 50 min and avg FPS more, in 1080p standard video settings.

@Gwent_Addict any of 10th, 11th gen Intel or 3000/5000 AMD make no sense. Also 12th gen Intel makes no sense, because of cooler requirements, motherboards prices and especially DDR5 prices and not that good performance. Intel Optane SSD makes also no point, despite marginal at best difference, considering the price and capacity + place it needs/takes on the motherboard. Something like a R5 5600X or R7 5800X, but with V-Cache + good DDR4 and a good B550 or X570S motherboard is what I would suggest, with a 240/280/360 mm AiO cooler, like EK-AiO Basic.

@Valken it's normal that with E cores enabled in Win 10, performance is worse, because Win 10 is not designed to properly distribute load between P and E cores. Win 11 performance with E cores on and off is same, since it is designed to properly distribute load between P and E cores. Win 11 performance is lower vs. Win 10, only because one needs not to forget to disable VBS (virtualization-based security) in it, that reduces performance, similar to like when an antivirus scans your PC all the time in the background while you're playing a game.

The opposite. Only 6 big cores, with 2 big cores and 4 little cores disabled, which makes 0 sense. Also only 1 run and now word whether in windowed or full screen mod. G-Sync on/off, fast sync or v-sync off, also not known. No word about Arma launcher settings, CPU cooler model, motherboard model, RAM model/capacity, Windows version, whether there were any programs running in the background...

I would have bought 12600K of course, but DDR5 motherboard price + good DDR5 RAM...

A 5600X with V-Cache (Q1 2022) + B550 board and 2x16 GB DDR4 3600 MHz 16-18-18 Crucial Ballistix would be a better choice, since DDR4 is cheaper, as well as B550 boards and a 240 mm EK-AiO Basic is fine for it.

Historically, Arma never worked particularly well with AMD GPUs, FPS and visually wise. But that's because of DX9 and DX11, that runs better on nVidia GPUs. Next Arma will be on DX12 and it should run same or better on AMD GPUs. nVidia PhysiX in Arma is done on CPU only, so no need for a nVidia GPU because of PhysiX.

DDR4 4000 CL14 costs around 500€ and there is little chance one is able to use it at its 4000 MHz. DDR5 at only 5200 MHz and relatively loose XMP timings, is not worse, if not better in majority of games, because it's quad channel vs. dual channel for DDR4. Also it would be wise to wait for DDR5 kit with close to 6000 or past 6000 MHz with sub CL40 XMP timings to then manually optimise them additionally. So a Z690 motherboard is what you would need. You would also need at least a 280 mm AiO.

Now your replies make sense, again 😉 https://forums.bohemia.net/forums/topic/187603-will-my-pc-run-arma3-what-cpugpu-to-get-what-settings-what-system-specifications/?do=findComment&comment=3448757

I hope that EK AiO will be able to keep my 11900K lower than 90°C or closer to 80°C peak in Prime95 (non-AVX). Unfortunately I have a really bad quality cores on my 11900K - too high voltage is required for a given frequency. At least its integrated memory controller is relatively good. I hope to do 5.0 GHz cores and 4.5 GHz cache with the AiO. In Arma it won't be as hot as in Prime95 (non-AVX) for sure.

The bigger the pump the less efforts it makes to do the job it has to accomplish and less noise, because a big pump doesn't have to spin at crazy rpm, all the time on its limit and will survive longer.

Corsair pump (left), EK pump (right) 😏 Corsair uses Asetek pumps, like most of other AiO manufacturers (Fractal Design, EVGA, NZXT, Lian Li, Phanteks, Thermaltake, Asus, MSI, Gigabyte etc.) Arctic Liquid Freezer II pump is better than Corsair and other AiOs with Asetek pumps, but not as good/powerful as EK pump.

EK-AiO fill port on the radiator