Jump to content
Placebo

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

Recommended Posts

On 1/12/2020 at 9:13 PM, SpartanOfficial said:

The system still performs fine in all the other games I've tested it on. I don't know what to do, I've scanned for viruses with my antivirus, CCleaned everything several times, clean installed several times, cleaned my computer of dust (after all this happened), validated steam files, launched as an admin, changed graphics, updated graphics drivers fully, optimized the game using nvidia experience, I would defrag but it seems pointless since I tried it on two different SSDs and same issue also it would damage them. Should I contact arma devs - if so how? Should I be contacting my computer manufacturer since it is under warranty and is 6months old? How do I properly examine both software and the hardware to find any answers? 

 

Sidenote: I get 200+ fps in the menus but I guess that's just a still image usually. 

 

Skylake-X systems need to be tuned in order to get any decent performance.  Given the 9920x , 2080ti and you said it was a manufactured machine? Is this machine a Corsair One 180i  ?

1.) Has anyone done any tuning for you or are you running on stock settings from corsair ?

2.) Do you have CPU-Z installed? If so, can you share the output of the CPU tab and the Memory Tab ?

 

CPU

AzyNsuN.jpg

 

Memory Tab.

fXIkCim.jpg

Share this post


Link to post
Share on other sites

I've just gone through the entire prologue/bootcamp thing, and the game went from being mostly above 60 FPS to dropping to 25 FPS at times during the last scene, the one where you drive around the dead bodies etc (with the GPU usage dropping to 17% at times, which i found hilarious). 

 

System is Ryzen 2600, 5700 XT with 16GB 3600 C18 memory. 

 

Guess i'm upgrading my CPU, but i'm wondering whether even the Ryzen 3600 will be enough for this game. Maybe i could go intel, or wait for the Ryzen 4000 and hope my motherboard will be made compatible for it. 

 

But one thing i'm wondering, is there any graphic setting that drops the load off the CPU at all? I tried fiddling around with the visuals settings but all i managed to do is lighten the load of the GPU, which is pointless since it is rarely taxed to begin with. The only thing i noticed that makes a difference is viewing distance. If the game is dropping below 60 during the prologue, how bad can i expect things to get elsewhere? 

 

 

 

 

 

 

Share this post


Link to post
Share on other sites
3 hours ago, opus132 said:

I've just gone through the entire prologue/bootcamp thing, and the game went from being mostly above 60 FPS to dropping to 25 FPS at times during the last scene, the one where you drive around the dead bodies etc (with the GPU usage dropping to 17% at times, which i found hilarious). 

 

System is Ryzen 2600, 5700 XT with 16GB 3600 C18 memory. 

 

Guess i'm upgrading my CPU, but i'm wondering whether even the Ryzen 3600 will be enough for this game. Maybe i could go intel, or wait for the Ryzen 4000 and hope my motherboard will be made compatible for it. 

 

But one thing i'm wondering, is there any graphic setting that drops the load off the CPU at all? I tried fiddling around with the visuals settings but all i managed to do is lighten the load of the GPU, which is pointless since it is rarely taxed to begin with. The only thing i noticed that makes a difference is viewing distance. If the game is dropping below 60 during the prologue, how bad can i expect things to get elsewhere? 

 

 

 

 

 

 

 

Its best to try the YaaB benchmark to see where you are at now compared to other systems.

=> Getting hands on a tuning the system settings could get you 25% more FPS compared to good parts on stock.

 

More details on YaaB can be found on the following thread.

 

Share this post


Link to post
Share on other sites

I did a quick test for now (will try to match the conditions specified in that thread later),  got 37 FPS on standard,  and 29 FPS on ultra, 1440p. 

 

CPU is OC at 4.1GHz, but the memory is C18 and not C15:

 

YAAB v.0.98 | patch 1.88

R5 2600X (6C/12T) @ 4.2 GHz | 16 GB DDR4 3600 MHz 15-15-14-28-160-1T (single rank) | RX 470 (most certainly not killed unnecessary background tasks / no info on AI LEVEL)
58.3 FPS 1080p standard

 

Not even the i9-9900K feels all that great though, i find this a bit disconcerting. 

 

Is there a guide that specifies which graphic settings is heavy for the CPU specifically? Tried fiddling a bit by lowering some stuff but i gained nothing, so i'm assuming i haven't hit anything that actually taxes the CPU, only the GPU. 

 

Share this post


Link to post
Share on other sites
50 minutes ago, opus132 said:

I did a quick test for now (will try to match the conditions specified in that thread later),  got 37 FPS on standard,  and 29 FPS on ultra, 1440p. 

 

CPU is OC at 4.1GHz, but the memory is C18 and not C15:

 

YAAB v.0.98 | patch 1.88

R5 2600X (6C/12T) @ 4.2 GHz | 16 GB DDR4 3600 MHz 15-15-14-28-160-1T (single rank) | RX 470 (most certainly not killed unnecessary background tasks / no info on AI LEVEL)
58.3 FPS 1080p standard

 

Not even the i9-9900K feels all that great though, i find this a bit disconcerting. 

 

Is there a guide that specifies which graphic settings is heavy for the CPU specifically? Tried fiddling a bit by lowering some stuff but i gained nothing, so i'm assuming i haven't hit anything that actually taxes the CPU, only the GPU. 

 

 

ARMA3 performance is CPU  / Memory bound.  It is extremely sensitive memory latency. You will need to optimize for that, either by increasing memory speed , or tightening timings, infinity fabric speeds.  

3600c18 is about 10ns

3600c15 is about 8.33ns 

- Ultra setting increases object draw distances, which increases the work the CPU needs to be done and hence lowers frame rates. 

 

1.) The platform choice affects it the most. Its easiest to get good performance on intel mainstream platform.

- Intel Mainstream (z390)

- Intel HEDT but only if its tuned well and OCéd  ##  But the value proposition of Intel HEDT in 2020 is extremely questionable.  

- Ryzen 3xxx

- Ryzen 2xxx

- Ryzen 1xxx

- Ryzen Thread Ripper

 

2.) Next is memory latency.  If you looked at the top Ryzen 3 results from the other thread at > 

 

YAAB v.1.00 | patch 1.92

R5 3600 (6C/12T) @ 4.1/1.6 GHz core/infinity | 16 GB DDR4 3800 MHz 14-15-14-32-266-1T (single rank) | GTX 1070 Ti (Windows 10 1903)
63.8 FPS 1080p standard

59.7 FPS 1080p high

48.4 FPS 1080p ultra

 

YAAB v.1.00 | patch 1.92

R9 3900X (12C/24T) @ 4.225/1.867 GHz core/infinity | 32 GB DDR4 3733 MHz 14-15-14-32-262-1T (dual rank) | GTX 1070 Ti (Windows 10 1903)
62.3 FPS 1080p standard

58.7 FPS 1080p high

 

YAAB v.1.00 | patch 1.92

R5 3600 (6C/12T) @ 4.1/1.6 GHz core/infinity | 16 GB DDR4 3200 MHz 14-14-14-34-560-1T (single rank) | GTX 1070 Ti (Windows 10 1903)
57.3 FPS 1080p standard

51.8 FPS 1080p high

 

YAAB v.1.00 | patch 1.94

R5 3600X (6C/12T) @ 4.250/1.6 GHz core/infinity | 16 GB DDR4 3200 MHz 16-18-18-36-1T (single rank) | RX 570 (Windows 10 1903)
53.1 FPS 1080p standard

49.9 FPS 1080p ultra

 

The memory latency in the above configs are. You will see a pattern.

3800c14 ~ 7.36ns

3733c14 ~ 7.5ns

3200c14 ~ 8.75ns

3600c16 ~ 8.88ns

 

You can use the following page to calculate rough memory settings.  https://notkyon.moe/ram-latency.htm  

 

Share this post


Link to post
Share on other sites

In order to understand the situation here, you have to remember the fundamentals for RV game engine and therefore for Arma3.
Arma3 is based on the  "aged 10 years" Real Virtuality Engine design, by many aspects, in particular, the priority given to simulation, the ability to manage an almost infinite open world, the modularity and the original structure of multiplayer game management, the game engine has no equivalent.

 

Since the release of Operation FlashPoint in 2001, the engine has evolved, it has turned multicore with Arma2: Operation Arrowhead but is not much multithreaded.

RV Engine does not allows any kind of parallelism in the processing, all is strictly dependent on the simulation running on a limited number of cores on the processor.

The initial job for the engine  is the calculation of the terrain, it's a 100% processor job. The processor load is easy to understand, it progresses like the square of the radius of the circle in the center.

In game, the player is in the center and the radius of the circle is the global visibility parameter that's why the game is CPU dependent and the 12 000m Visibility an "insane" setting.

 

This explains why, it is useless to have a high level graphics card. From the moment the graphics card will be able to improve the images with the best effects, it is sufficient for the task.

In my own tests with graphic quality in "Ultra", both an RTX 2060 and a RX 5700 are only used at 40% maximum in play on a 1080p monitor. According to my own experience, a GTX 1060 6 GB has allowed the best balance.

 

In my tests with an R5 2600, an R5 3600 and an R5 3600X,* I could see a real performances gap between the Ryzen 2000 and the Ryzen 3000. On the other hand, I did not find any really significant difference between the 2 Ryzen 3000, this stems from the R5 3600X's inability to maintain a high boot frequency.

 

*Note : those tests are hosted on French CanardPC Forums, with an excessive use of French but also some images 😉

Share this post


Link to post
Share on other sites

For those of you with Ryzen systems whose performance is almost solely dependent on getting rare, fast RAM, this kit might be useful and is a good price in the UK. It wouldn't hurt Intel users who have the right motherboard either. DDR4 4000 at C15, I'd be interested to see if anyone actually gets it to run at 4000. It's Samsung b-die so is more likely to be fast-Ryzen compliant.

 

https://www.ebuyer.com/875837-patriot-viper-steel-series-ddr4-16gb-2-x-8gb-4000mhz-memory-kit-pvs416g400c9k

Share this post


Link to post
Share on other sites

Ryzen 3000 users, be cautious with this Patriot Viper Steel Series DDR4 16 GB kit @ 4000 Mhz for it seems it doesn't allow stable operation. 

Patriot Memory PVR416G360C6K 16 GB Viper Gaming Series DDR4 @ 3600 Mhz seems OK but is currently unavailable.

 

Share this post


Link to post
Share on other sites
1 hour ago, oldbear said:

Ryzen 3000 users, be cautious with this Patriot Viper Steel Series DDR4 16 GB kit @ 4000 Mhz for it seems it doesn't allow stable operation. 

Patriot Memory PVR416G360C6K 16 GB Viper Gaming Series DDR4 @ 3600 Mhz seems OK but is currently unavailable.

 

I read that it was stable, but I prefer your opinion over some random child on a review site.

Share this post


Link to post
Share on other sites

That stuff was tested at 2133 C15  ...and at 4000 @ C19    which isn't particularly good.  

Share this post


Link to post
Share on other sites
4 hours ago, oldbear said:

In my tests with an R5 2600, an R5 3600 and an R5 3600X,* I could see a real performances gap between the Ryzen 2000 and the Ryzen 3000. On the other hand, I did not find any really significant difference between the 2 Ryzen 3000, this stems from the R5 3600X's inability to maintain a high boot frequency.

 

I see your results on the 2600 aren't that different from mine, so i guess it's normal. 

 

I think i'll wait for the Ryzen 4000 if not even the 3600 can get to 60 FPS. For now i'll just play single player since there low FPS doesn't matter as much. 

 

For record, i was fiddling around Arma 1 earlier and when i extended the drawing distance i ended up going below 60 FPS in the bootcamp! Amazing not even a game from 2006 is safe from this, haha. 

Share this post


Link to post
Share on other sites

BTW, aside from drawing distance, there's nothing else i can do to light the load from the CPU, right? Because at this point i'll just keep everything else to ultra, and simply lower the distance. 

Share this post


Link to post
Share on other sites

Well ... , for a given CPU, it is necessary to seek all the means to allow it to release its power by reducing the weight of the additional tasks.

Here, in a previous post you can find how I try to manage this at the moment on my R5 3600X/B450 test platform.

- never use "SAMPLING" over 110 % even if you want to count all the bolts from the enemy tank hull before firing,

- never use "SHADOW" : Low settings if you get a real GPU

- setting "VISIBILITY" > "OVERALL" to 12000m is "insane" as a well known Veteran Arma Dev officially declared a long time ago. Please use 3800m as a maximum if you are playing as a grunt.

- disable VSYNC : it is extremely unlikely to be confronted with the image tearing, personally I have never encountered the phenomenon while playing Arma.

You will have to play a bit with parameters in the AA&PP section

- use balanced parameters in the Antialiasing part such as those I am using myself for a start as I has suggested in a previous post.

The combination of a  FSAA: 8 and a PPAA: SMAA Ultra can turn out to be a FPS killer.

The effect of the parameters in that part depends on many factors, model of the card and operation of the drivers in particular, so here you are going to have to find the right balance by yourself.


 

 

Share this post


Link to post
Share on other sites

I was just fiddling around Arma Cold War Assault, and that one performs poorly too. Visibility above 900 and it drops below 60 FPS when panning at certain points. This engine is something else. 

 

 

Share this post


Link to post
Share on other sites

No, Operation Flashpoint: Cold War Crisis rebranded ARMA: Cold War Assault in 2011 is based on Real Virtuality 1 , the first public version of the game engine.

It was running on a configuration with incredibly low performance by our current standards, it is single core and needs only a minimum display.

 

Recommended system requirements:

  • CPU: Pentium III 600 CPU or better
  • RAM: 128 MB RAM
  • GPU: Compatible 3D Graphics Card with 32MB RAM
  • DX: DirectX 8
  • OS: Windows 95 / 98 / ME / 2000 / XP
  • Store: 450MB Hard-Drive space
  • Sound: 16Bit Direct Sound Compliant Sound Card
  • ODD: 24x CD-ROM
  • Note: Supported Chipsets: 3Dfx: Voodoo2, Voodoo Banshee, Voodoo3, Voodoo4, Voodoo5 ATi: Rage 128, Radeon. Matrox: G400, G450. Nvidia: Riva TnT, Riva TnT2, Geforce 256, Geforce 2 MX, GTS, Ultra, GeForce 4, GeForce 5 Series, GeForce 6 Series. S3: Savage 4, Savage 2000 3D Labs: Permedia 3. PowerVR: Series 3 (kryo)

Source : https://gamesystemrequirements.com/game/operation-flashpoint-cold-war-crisis

 

Even if I had spent probably more than 20 000h playing Operation Flashpoint on my AMD XP 2000/ATI 9500 Pro gaming rig, I think we must now consider this game as a collector's item or as a historic milestone even if from my point of view the campaigns developed for this version are without equivalent. 😢

 

Share this post


Link to post
Share on other sites

Need to modify my guide on YAAB benchmark settings to use and what data to privde as result, since providing only average FPS doesn't reflect the reality.

One can have 32 min. FPS and have 66 avg. and one can also have 42 min. and have 66 avg. FPS.

 

Will rewrite my guide and people will be asked to provide min. and max. FPS as well and also test at low settings, to exclude the GPU as much as possible as possible bottleneck.

 

Also most of results collected until today in my guide are from before AMD releasing BIOSes and Windows updates that increased AMD CPUs' frequency and also before Intel and Windows released fixes for Intel CPUs, that have a hit on their performance.

 

While one will spend most of time during game session around the avg. FPS, min. FPS is also very important indicator of your hardware performance.

Share this post


Link to post
Share on other sites

Not sure if anyone finds it interesting, but we started putting our 'clan' results of Yet Another ArmA Benchmark scenario found in the Steam Workshop into a Google Sheets table. Here you can see the first few results, more to come. Would this kind of sheet be interesting for rest of you to insert your scores? We ran YAAB at Ultra settings (not touching anything manually), 1920 x 1080, no mods and no Nvidia ShadowPlays or similar on.

 

Updated chart with MIN & MAX: https://drive.google.com/open?id=1yoizcbQ1WZTAltVACygDgPA5OJOyqp1X

Share this post


Link to post
Share on other sites

Without indicating windows and bios version and min and max FPS and also if some people run systems with low frequency AMD BIOS and/or windows or run Intel system with or without security fixes that lower CPUs' perofrmance, the results are worthless.

Share this post


Link to post
Share on other sites
6 minutes ago, Groove_C said:

Without indicating windows and bios version and min and max FPS and also if some people run systems with low frequency AMD BIOS and/or windows or run Intel system with or without security fixes that lower CPUs' perofrmance, the results are worthless.

 

Worthless? Not for example 'not completely accurate', but worthless? Sure.

  • Like 1

Share this post


Link to post
Share on other sites

Yes, exactly, worthless, since not comparable and not reflecting the reality.

Also running just ultra is really not what one should do.

Share this post


Link to post
Share on other sites
1 minute ago, Groove_C said:

Yes, exactly, worthless, since not comparable and not reflecting the reality.

🤗

Share this post


Link to post
Share on other sites

One can run at ultra for additional results, but otherwise one should run at standard and even better if at low, to exclude the GPU as much as possible.

Because the person with a GTX 1050 Ti will have less FPS at ultra than the other with a RTX 2080 Ti, but this doesn't necessarily mean, that the person with the GTX 1050 Ti has worse CPU and RAM, which is what Arma still relies mostly on.

 

One can run a AMD CPU with outdated BIOS and Windows, which will result in less FPS, beause lower frequency and one can run Intel CPU with outdated BIOS and Windows, without security patches and have more FPS, but not because it's super powerful, but because of missing updates.

Share this post


Link to post
Share on other sites
9 hours ago, oldbear said:

Recommended system requirements:

  • CPU: Pentium III 600 CPU or better

 

Decades later, i still can't max the viewing distance with a far more powerful CPU than that, haha. 

 

 

Share this post


Link to post
Share on other sites
1 hour ago, Groove_C said:

One can run at ultra for additional results, but otherwise one should run at standard and even better if at low, to exclude the GPU as much as possible.

 

Aren't there some settings that actually start putting more stress on the CPU if set too low, like shadows if i remember correctly?

 

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. GPU consumption remains low at all times, the fan doesn't even break past 1000 RPM. Changing anything may, at best, give me an extra FPS. 

 

Guess i'll have to get used to the game dropping to 30 FPS before upgrading my CPU. Luckily, sometimes it goes well above 60, so i guess it only drops to 30 during intense moments. 

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now

×