Will-my-pc-run-Arma3? What cpu/gpu to get? What settings? What system specifications?

[Groove_C 267]

30 minutes ago, opus132 said:

I've been playing around with settings for a couple hours and nothing seems to matter much (except for viewing distance). On that benchmark map, i get a consistent 30 FPS on ultra and about 37 at low.

Changing anything may, at best, give me an extra FPS.

It's your CPU.

A R5 2600 is ok for Arma, but not optimal.

Your RAM is probably also not optimal, for Arma.

[oldbear 390]

Yes, switching the Shadows parameter to "Low" makes the game engine to transfer shadows rendering from the GPU to the CPU which is strange if not stupid if you have a correct graphics card and can have a counterproductive effect if the processor is struggling. Frankly the advice was especially valid for AMD FX CPUs who were desperately trying to extract FPS from a processor not cut for the job.

 

In fact, there is a 2nd parameter you can play with.

Quality > Terrain is a mixed CPU/GPU parameter because if rendering terrain is heavily CPU dependent,  ground textures and clutters computing is GPU related. Clutters are objects associated with ground textures, such as grass, small rocks, fallen branches and mushrooms. Limiting the quality will lighten certain tasks and reduce waiting times, never set it on "Low", if you get an average rig, you can set this parameter on "Very High".

 

What is important is the way the game is running in game. You have to worry about the settings before the game and never think about it in game.
Arma3 is a game fully playable/enjoyable at 30 FPS and with an R5 2600 [ATM the minimum recommended CPU for Arma3] you can play in the 35 FPS range on 1080p and that's what counts.

[opus132 3]

I'll still switch to a Ryzen 4000 if by some miracle my motherboard will be updated for it if not i may even switch to Intel. For now 30 FPS it is i guess. 

 

Damn this hobby, so expensive. 

[oldbear 390]

Not so expensive, have a look at what you need to get a playable Arma3 "Minimum Recommended" rig retrofitted to reach "APEX Standards" ™ 😎

CPU : AMD R5 2600

Mobo : B450M

GPU : AMD RX 570 / Nvidia 1650

RAM : 16 GB Corsair LPX 3000 MHz

SSD : 500 GB SATA

PSU : 550W 80Plus Bronze

 

[Groove_C 267]

Ryzen 4xxx won't be released before the end of the year.

Since ASMedia, the company that will manufacture X670 chipsets, has its capacities now occupied to produce lower cost B550/520 chipsets.

This is why they can't start to manufacture X670 right now.

Same situation as with X570 chipsets, that ASMedia should also have manufactured, but they were not ready and this is why AMD had to manufacture X570 chipsets themselves, in order to be able to release Ryzen 3xxx on time, hence the cost and the fan on the chipset.

 

But with 200 MHz more vs. Ryzen 3xxx and double the cache + unified cache, so a single core can access all the cache, Ryzen 4xxx will finally be on par with highly OC'ed Intel CPUs, right out of the box (no OC).

[opus132 3]

They say release date is this year. I'm sure it will be late but i distinctly remember 2020, unless they'll delay it. 

 

On my motherboard (MSI B450 Pro Carbon etc), they had to remove the graphics on the BIOS because they had no space to fit in support for Ryzen 3000. It is a vain hope they might add support for Ryzen 4000, but still, they did it for B350 so there's always the chance. 

 

To be honest i'm much more shocked that i can't even run Operation Flashpoint properly, hahaha. Viewing distance 900 and terrain on normal and FPS will go from 400 FPS to 50-60 in a pinch, depending where i'm panning. If i bump it to 5000, i can even go below 20 FPS and if i set distance to 5000 AND i put terrain at very high the game just crashes. I can't wrap my brain around a game from 2001 running slowly on modern hardware. Honestly Arma 3 almost runs better. 

[oldbear 390]

In fact, we have no specific information from AMD on how the transition to Zen3 will materialize, so far,  we have this screen and rumours.

There may of course be technical reasons, but the fact that AMD wants to keep control of the high end chipset design is basically linked to AMD's strategy which made it clear that it was no longer the low cost solution supplier. This is probably also the reason for the delay in releasing the B550 chipset, AMD wants to make the most of the marketing of the most expensive solutions.

 

Regarding Arma: Cold War Assault, you have to consider it for what it is, a collector's item that is still alive. 😎

 

 

[Reezo 45]

I just read the last three pages that I had missed since my last visit and I can't thank you enough, once again, for all the info.

 

I have a question, hopefully it's not that stupid. Has anyone tried to run YAAB on Ryzen CPUs by assigning the ArmA 3 process to specific cores? For example 1 thread = 1 core, all even, all odd, etc. etc.?

 

Would there be any benefit especially if one was to stream at the same time of playing? I am currently evaluating trading the motherboard+CPU to a colleague/friend (same charity startup) that is having issues with his 9700k and streaming, as he said he would need more multicore power. On the other hand there's me, playing ArmA, in which a top speed single core would be more beneficial. So I was wondering...

 

Thanks!

[opus132 3]

When i'm gaming, my Ryzen rarely exceeds 50% CPU usage. You can see the difference in the Cinebench score: 

 

 

So yes, Ryzen might be a better choice for your friend, while Intel seems to be the better choice for Arma players.  

[froggyluv 2130]

So I just fried my backup system motherboard the other daymessing around with the ethernet cable to rough - had a 4690k in their oc'ed to like 4.3 with cheap 1600 ram. My previous build in which mobo also died was a 6600k with good ddr4 3200 ram but that barely got used.

 

So now was sitting around with 2 good cpus so i decided to get a new mobo for the 6600k so i could use the good ram and get a better overclock. Went to Microcenter (retail computer parts store in States) but they told me they no longer carried 1151 boards for the 6600k as that was too old a generation (OK BOOMER!). Now I could have just mail ordered one but being impatient, decided to grab a new 9600k and Asus z390 mobo and run it with the 16 gigs 3200 ram i had sittin at home.

 

Now I used to be a mild to moderate overclocker but just wanted to see what this thing would do on its own (with turbo) and

 

 

5.0ghz?! On Turbo alone?! Holy shit this is crazy -ive always dreamed for a 5.0 clock but being an Airmen afraid of Liquid and all its evil ways...Is this normal??

 

PS: ALMOST walked out with an AMD cpu after some pretty slick talking by the geeky salesmen about the pure superiority of AMD tech (7nm) over Intel..but then remembered what my Girlfriend Arma would say and I grabbed and ran

 

 

[opus132 3]

Can't seem to find a whole lot on AMD vs Intel with current gen. Found this, for whatever it's worth: 

 

 

So now i have to decide, 3600x, wait for Ryzen 4000 and hope it's compatible with my motherboard, get a new motherboard and the 9600k (can't afford the 9700k), or get a new motherboard and wait to see what intel is going to come up with next.

 

Serves me right for going AMD, but at 1440p i always figured the GPU was going to be the problem well before the CPU, i had no idea i was going to find a game where the opposite is true, heh. 

[Reezo 45]

Finally able to contribute to this, I made a post on IMGUR gathering all my reports, along with YAAB:

 

As a quick resume, my specs at the moment are:

Spoiler

- AMD Ryzen 3700x

- G.SKILL TridentZ 2x8GB 3200 CL14

- Asus ROG Strix x470-f gaming

- Samsung EVO 850 SSD (OS), WD Black SN_750 SSD (Games)

- DeepCool Castle 240EX

- NVIDIA Asus Rog Strix 2070 Super

- Thermaltake Berlin 630W PSU

- Phanteks Evolv-X

- Thermaltake Riing 12 fans (120mm 3xIN, 1xOUT)

 

https://imgur.com/a/tXPyZE3

 

UPDATING THIS POST FROM WHAT I BENCHMARKED AFTER THIS POST WAS MADE

(Just making it easy for someone who is reading this thread to go straight to my final findings)

 

YAAB - LOW Settings, 1440p:

CPU on AUTO, RAM on XMP: 66.8fps

CPU on AUTO, RAM on 3200 FAST: 71.9fps - 7.6% increase

CPU on AUTO, RAM on 3466 FAST: 73.5fps - 10.0% increase

CPU on AUTO, RAM on 3666 FAST: 74.4fps - 11.4% increase

CPU on AUTO, RAM on 3733 FAST: 75.6ps - 13.2% increase

CPU on AUTO, RAM on 3733 FAST with CMA_x64 Memory Allocator: 80.2fps - 20.1% increase

 

YAAB - ULTRA Settings, 1440p:

CPU on AUTO, RAM on XMP: 43.1fps

CPU on AUTO, RAM on 3200 FAST: 43.8fps - 1.6% increase

CPU on AUTO, RAM on 3466 FAST: 45.6fps - 5.8% increase

CPU on AUTO, RAM on 3666 FAST: 46.7fps - 8.35% increase

CPU on AUTO, RAM on 3733 FAST: 47.2fps - 9.5% increase

CPU on AUTO, RAM on 3733 FAST with CMA_x64 Memory Allocator: 53.8fps - 24.9% increase

 

NOTES:

• I can't seem to be able to POST with RAM @ 3800 CL16 (and Infinity Fabric at 1900). It seems the end of the line for the memory I have and the motherboard/cpu combination is 3733MHz CL14, which is - to my standards - insanely high
• I was able to keep it stable by keeing MISC, TERMINATION and CAD_BUS parameters (see DRAM Calculator for what I mean) to AUTO, except for the COMMAND RATE which I kept at 1T)
• RAM Latency at the AIDA64 benches reached 64.5ns, the lowest I could get it
• I am running at 1.45v DRAM Voltage and 1.1v SOC Voltage which is pretty within safe ranges I'd say
• When you read "CPU on AUTO" it means I only tweaked the Infinity Fabric speed to match 50% of the RAM speed (e.g. 3200 RAM = 1600 IF and so on)
• I ran a benchmark at ULTRA but removing AA and PP (from the default FSAA:8x and SMAA Ultra) and the difference is a 3fps average fps, give or take. It's crazy how efficient these cards are considering the difference from FSAAx8 + SMAA ULTRA to both DISABLED
• For some weird reason Cinebench r15 scores keep going lower than my highest score with the memory running at 3466... but I don't care because YAAB consistently says that faster ram = more fps

[opus132 3]

Can't overclock my ram at all. It's 3600 but horrid 18 22 22 42 timings. The 2600 might be the culprit since i heard it doesn't handle high frequency memory all that well. Might get away with lowering the timings if i drop the frequency but at this point i might as well just wait to upgrade my CPU and try it then. 

 

I can push the 2600 to 4.1 GHz but on YAAB on Ultra i go from 29-30 to 32-33. Not worth it, also because it crushes in Cinebench R20 (lasts a lot longer on Prime95 for some reason). Voltage is up to v1.38 which is as far as i would want to take it. 

 

So the core improvement between the Ryzen 2000 and 3000 is about 13-19 FPS, which is pretty substantial, but still can't get to that magical 60 FPS. If they could squeeze more core power on the Ryzen 4000 i would argue AMD might actually take first spot, all though it's to be seen whether they can or not (rumors say there's a 17% improvement in performance but they don't say where. For all we know the 3000 might have already reached the limit for their 7nm architecture when it comes to single core performance and the 4000 might just improve other things. Maybe they got past the frequency wall). 

 

 

 

 

 

[opus132 3]

2 hours ago, Reezo said:

Finally able to contribute to this, I made a post on IMGUR gathering all my reports, along with YAAB:

 

As a quick resume, my specs at the moment are:

  Reveal hidden contents

- AMD Ryzen 3700x

- G.SKILL TridentZ 2x8GB 3200 CL14

- Asus ROG Strix x470-f gaming

- Samsung EVO 850 SSD (OS), WD Black SN_750 SSD (Games)

- DeepCool Castle 240EX

- NVIDIA Asus Rog Strix 2070 Super

- Thermaltake Berlin 650W PSU

- Phanteks Evolv-X

- Thermaltake Riing 12 fans (120mm 3xIN, 1xOUT)

 

https://imgur.com/a/tXPyZE3

 

UPDATE:

Just by tightening the ram timings (copied what DRAM Calculator had for me at 3200 FAST), I saw this jump:

 

YAAB - LOW Settings, 1440p:

CPU on AUTO, RAM on XMP: 66.8fps

CPU on AUTO, RAM on 3200 FAST: 71.9fps - 7.6% increase

 

YAAB - ULTRA Settings, 1440p:

CPU on AUTO, RAM on XMP: 43.1fps

CPU on AUTO, RAM on 3200 FAST: 43.8fps - 1.6% increase

 

YAAB - ULTRA Settings, 1440p with FSAAx2 and CMAA (can't honestly tell the difference.. at 1440p.):

CPU on AUTO, RAM on 3200 FAST: 49.9fps - 14% increase from the same BIOS settings)

 

One question, is the YAAB a worst case scenario or is that what you could expet most of the time in a normal game session? 

 

 

[opus132 3]

Ho one thing, what matters the most for memory, frequency or timings? 

[evilmedic 0]

16 hours ago, opus132 said:

Ho one thing, what matters the most for memory, frequency or timings? 

 

You need to optimise for memory access latency. Lower memory latency = higher framerates.  

You can lower memory access latency by increasing memory clock and lower/tightening timings.  

[Reezo 45]

@opus132 as EvilMedic wrote, it's a matter of making your RAM faster through faster memory clock and tightening timings.

 

I am not aware of any build at the moment -  6 years and counting after the release of ArmA 3 - that can push 60fps in 1440p on Ultra Settings, in YAAB. If I am wrong someone will eventually chime in.

 

The reason why YAAB is used it's because it is a reliable and repeatable test environment (as reliable as possible given the ArmA 3 behavior and engine). Without consistency in your tests, the test is faulty. It's science, basically.

 

YAAB produces the same amount of stress from your system and is therefore capable of measuring how better or worse your tweaks are doing. It is - at the moment - the de facto standard for testing your ArmA 3 performance.

 

I have not tried pushing my RAM further at the moment but I remember the 2700x I had was not as easy to run RAM as this 3700x.

 

Bottom line is: you see how my frames went up a lot (look at the low settings, because on Ultra obviously the game is GPU-bound more than before) and I did not even tweak my RAM too much: the difference you see is only from the XMP profile running at 3200 CL14 and the FAST settings of DRAM Calculator.. still at 3200 CL14.. I mean, CAS latency did not even go lower (it stayed at 14..) but because of the tightening of the other timings, I still got over 7% improvement. It's almost unbelievable and I promise I will try it out and report my findings when I go above what I have now.

 

Because of how better Ryzen CPUs are for multi-core activity, I'd say a streaming+gaming combo rig might benefit more from it than from an Intel, but if you use the NVIDIA Hardware Encoder you'll be using your GPU mostly for encoding the stream, so I don't think that will tax your CPU enough to see a difference between AMD and Intel (again, if I am wrong someone will chime in).

 

I might have a chance to test a 9700k (no idea on the motherboard though..) soon, then I could really compare apples to apples in ArmA 3. Same rig except for Motherboard+CPU, that's as close as I can get to test it.

 

As you might be aware I am streaming for charity through a group of friends that run a startup here in Berlin and if my stream gets enough interest and pull I might get a better advantage at taking some parts from their pool. These guys are extremely dedicated and are going the extra mile off their own pockets (kudos) but I can't subtract more parts from them than I already did at the moment. But I am confident with time this is doable (shameless plug: make sure you follow my Twitch channel to get me to Twitch affiliate soon, please 🙂 so I can show them I matter!).

 

One interesting read about memory tightening on the Ryzen is here:

https://www.reddit.com/r/overclocking/comments/ahs5a2/demystifying_memory_overclocking_on_ryzen_oc/

 

 

UPDATE:

Also don't forget to try this.. it's a great tip that often goes unnoticed:

@Groove_C wrote:

I suggest you to run YAAB 1080p standard with CMA x64 AVX2 memory allocator.

It further improves FPS (by using AVX2 instructions set of the CPU, instead of much older and not so efficient SSE instructions set).
It further improves the smoothness of the game (frametime) (by using 2 MB chunks of memory, instead of 4 KB chunks of memory).

- Download it here: http://www.armaholic.com/page.php?id=31217

- Navigate in archive to malloc -> 64-bit -> AVX2 folder

- Copy/move cma_x64.dll file to Steam -> steamapps -> Common -> Arma 3 -> DLL folder

- Navigate in archive to malloc -> 64-bit -> Config -> WithLargePages folder

- Copy/move cma.ini file to Steam -> steamapps -> Common -> Arma 3 root folder

- Type secpol in search and select Local Security Policy - Navigate in Security Settings to Local Policies -> User Rights Assignment - > Lock pages in memory

- Right click on Lock pages in memory and select Properties

- Click on Add User or Group - in the field Enter the object names to select, type your PC username and click on Check Names

- it should add your PC user name

- then click OK and OK and reboot the PC

- Navigate to Arma 3 root folder, right click on arma3_x64 and in Properties -> Compatibility, select Run this program as an administrator

- launch Arma 3 launcher

- navigate to PARAMETERS tab -> ALL PARAMETERS

- locate and check Memory allocator (64-bit) in Advanced section

- select cma_x64 memory allocator and start the game

 

Let us know if it changed something for you.

 

P.S. - should be done only on systems with not less than 16 GB RAM.

[Reezo 45]

Just FIY (I usually update but since we are on the topic of hard proof..) my tightening of the RAM keeps bringing home the results:

 

YAAB - LOW Settings, 1440p:

CPU on AUTO, RAM on XMP: 66.8fps

CPU on AUTO, RAM on 3200 FAST: 71.9fps - 7.6% increase

CPU on AUTO, RAM on 3466 FAST: 73.5fps - 10.0% increase

 

YAAB - ULTRA Settings, 1440p:

CPU on AUTO, RAM on XMP: 43.1fps

CPU on AUTO, RAM on 3200 FAST: 43.8fps - 1.6% increase

CPU on AUTO, RAM on 3466 FAST: 45.6fps - 5.8% increase

 

And this is only by tweaking the RAM according to what DRAM Calculator says.. and then bringing the Infinity Fabric value (AMD 3x00 series only) to half the memory speed (3200 > 1600, 3466 > 1733 and so on).

 

There's no further tweaking I've done both in software and hardware. Pretty impressive I'd say. 10% is a lot and it's worth saying that you can get better timings than the guidelines that DRAM Calculator reports (as already stated: you should do the tightening manually, using guides and longer tests, because DRAM Calculator is only a guideline).

 

Some people reported that sub-timings are not so important and should be left to auto.. I will try and investigate this as well, I am not totally sure, yet, maybe you people can elaborate on this.

[opus132 3]

I'm assuming you have b-die RAM or is the series 3000 that efficient at overclocking RAM?  If i touch mine even slightly the system fails to boot, sometimes on a loop where i have to clear the CMOS to get it going again. 

 

If we take your XMP value as reference, compared to the 2600x your 3700x yields 13 FPS more on Ultra and 26 on Low. That's pretty impressive. If the next Ryzen has a core jump like that AMD would really be putting Intel in a bad place. 

[Reezo 45]

@opus132 Yes I have G.SKILL TridentZ RGB 2x8GB 3200 CL14 RAM:

https://www.gskill.com/specification/165/166/1536653463/F4-3200C14D-16GTZR-Specification

 

It's b-die and I was able to run it at 3466MHz even with my 2700x, although the timings weren't probably as tight. For safety reasons, being this the rig I am streaming on, I have to tread carefully in order to keep the system reliable. From here on the real difference will be: will I be able to boot at 3600? My 2700x wasn't keeping those memory clocks in a reliable way. Anything past 3466 was either unreliable or not giving me the performance boost I was expecting. It could be the timings were looser and not loosing more than I was gaining from those 133Mhz more.

 

We'll see.. I keep reading there is not much to OC in the CPU department on this Ryzen serie so the priority is on the memory. I guess on Intel it's going to be the other way around.

 

I promise I'll do my best to get my hands on an Intel i7 9700k (again: no idea on the motherboard) in order to install it in this rig and compare how ArmA 3 runs. I know this has been done, probably in the past, a lot of times.. but I'd be happy to provide yet another consistent test environment.

 

If I can get my stream to have a bit of traction (I'm working hard on it, given the niche genre) I'll feel a bit less guilty to visit my friends and ask them to take away the motherboard and CPU they have there. They don't have many spares and the rigs are all being used so the other alternative is for me to buy it on Amazon and then maybe return it after the tests. I am not even sure you guys *do want* to see this, maybe I am just a geek? Anyway, it would be cool! *yes I am a geek*.

 

Despite spitting all these numbers, I can tell you the experience here during my streams is extremely enjoyable. No stutters, no slowdowns, it's butter smooth and I couldn't believe ArmA would become as such, so I am happy. I am playing capped at 72fps with a 75Hz G-Sync display and the stream is at 30fps, so you won't be able to "feel" it the same way as I do here standing in front of the display (it's being re-encoded at 30fps so there you have it). When I look at the recorded streams it reminds me of ArmA 2 OA when I played it on my i7 2600k and Sabertooth motherboard.. man I could push that CPU to 5.0GHz back then, I got such a golden chip! I never went over 4.3GHz, and now I regret it 😉

 

UPDATE:

I tried disabling SMT or assigning different sets of cores to ArmA 3, it did nothing at all. Maybe you can set it to "Above Normal" as priority if you want to feel safer in terms of gameplay but honestly all my CPU Affinity tests turned out to give YAAB from 50% to 75% less FPS. It's that bad!

[domokun 515]

@Reezo thanks for your analysis. What was the performance impact in YAAB of switching to memalloc to cma_x64?

[Reezo 45]

It's happening people!

 

 

@domokun I haven't tried it yet, but I will test it soon and report! I will keep tweaking my RAM first and then proceed to cma_x64 testing!

[opus132 3]

I hope the 17% core increase in Ryzen 4000 is true. I'm sick of Intel and their price gouging. 

[Reezo 45]

26 minutes ago, opus132 said:

I hope the 17% core increase in Ryzen 4000 is true. I'm sick of Intel and their price gouging. 

 

It's entirely possible but the race will keep going and going and going .. and going. That's like that game where you splat your hand on somebody else's hand and so on so forth. This one hurts the wallet instead of the actual back of the hand 🙂

 

Honestly right now I would generally go for a 3700x or - if you are heavy into video editing, renders and such, clearly a Threadripper. Gaming "alone" (100% exclusively) might still have peaks on Intel because of games not being - at the moment - multicore oriented.. but personally, all I care is that my rig doesn't give me heavy dips. Those I DO HATE with all my heart. Dips in the form of stutters or lag/sluggish slowdowns (the former being abrupt, the latter being smooth) are horrible in games, for me. For the rest I keep the fps at 60-75 on my G-Sync system so 240fps and such are not my thing.

 

High average and no dips is all I cheer for *raises glass of Moscow Mule*

[opus132 3]

I was thinking more of the 3600x. Core performance would be the same, and i don't need the extra stuff the 3700x has. 

 

But i'm holding out for a 4600x IF it turns out my motherboard will run it.