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

[Groove_C 267]

Buying a CPU with a lot of cores, for gaming, is an option, only to a certain extent.

I would say not going past 16 threads, on current architectures, since past 8 cores, inter-core data latency increases with Intel's Ring Bus.

Even with 6 cores, ring bus already shows negative effects (vs. 4/8 cores/threads) on inter-core data latency.

 

Intel's Mesh scales a lot better than ring bus, providing more uniform inter-core data latency, but it's much slower than Ring Bus.

 

As you can see here, until 4/8 cores/threads, Intel's Ring Bus is just perfect (inter-core data latency in nanoseconds).

Past 4/8 cores/threads, Intel's Ring Bus is not that good anymore, although thanks to monolythic DIE it's still much much better than AMD with its CCXs, CCDs and Infinity Fabric.

AMD is a clear looser here.

 

i7-7700K (4/8 cores/threads)

 

Ryzen 3 3300X (4/8 cores/threads)

 

Ryzen 3 3100 (4/8 cores/threads)

 

i7-9750H (6/12 cores/threads)

 

i5-10600K (6/12 cores/threads)

 

Ryzen 5 3600 (6/12 cores/threads)

 

i7-10700K (8/16 cores/threads)

 

Ryzen 9 4900HS (8/16 cores/threads)

 

i9-10900K (10/20 cores/threads)

Ryzen 9 3950X (16/32 cores/threads)

 

[oldbear 390]

@Valken You don't need to wait next Bohemia move to plan to build up your "Next PC", as a minimum it must be on

As a minimum, it will have to be at the same level as the upcoming consoles.

CPU : R7 3700X / i7 10700K

GPU : RTX 2070

RAM : 32 GB

SSD : 1To

 

[Groove_C 267]

@oldbear @Valken's current build is more than enough to run A3 at decent enough settings, so he has no problems to wait until next gen of CPUs (and GPUs).

If next gen CPUs/GPUs won't be worth to buy because of price/performance vs. current gen, he still can buy current gen CPU/GPU for less $$$.

[oldbear 390]

About the future config, I was speaking generically, in correct English I probably should have said "no one needs to wait", but as some have already noticed, English is not not my first language.

Nonetheless, based on my own experience with my spare PC game ...
CPU: i7 4970
GPU: GTX 1060 3GB
RAM: 16 GB DDR3 1866 MHz
SSD: 500GB [System] + 500GB [Arma *]

... the performances are at the level of an R5 2600 that I also have the opportunity to test.

I see that there is a real performance gap between this Intel gen and the AMD Ryzen 2.

My next focus will be on the AMD Zen3 + AMD RDNA2 combo.

[Groove_C 267]

Well, you should consider that Valken has a K model of i7-4790, which is OC'ed to 4.6 GHz and not like 3.8 GHz of a non-K i7-4790 that you mention.

+ he has best of the best DDR3 RAM 2400 MHz and not 1866 MHz.

And the performance is higher than that of a R5 2600, in Arma.

 

I've tested my i7-4790K @ 4.8 GHz with same 2400 MHz CL10 DDR3 RAM as Valken against a friend's R5 2600X 4.2 GHz 3600 MHz 14-15-15-28 RAM and it performed worse than my 2014 confing. So...

 

1 hour ago, oldbear said:

I see that there is a real performance gap between this Intel gen and the AMD Ryzen 2.

You mean Zen 2 maybe and not Ryzen 2 (2xxx)?

Yes, there is a FPS and smoothness difference (in Arma) between Intel 4th gen and Ryzen 3rd gen, where Ryzen 3xxx with some good DDR4 3600-3800 MHz RAM (without CMA AVX2 malloc) is slightly better than Intel 4xxx (with CMA AVX2 malloc), but it's still not worth the price to pay for mainboard + CPU + RAM combo, considering the difference is minimal.

 

Ryzen 4xxx should finally make it worth enough to finally upgrade, even from your i7-7700K.

[oldbear 390]

Well, ATM, my test rig running R5 3600X + RTX 2060 + 16 GB DDR4 TridentZ Neo 3600MHz C16  performs better than my game rig featuring i7 7700K [slightly OC] + RX 5700 + 16 GB DDR4 TridentZ 3200MHz C16.

I am still waiting for Next Gen Intel for planning my gaming rig upgrade  😎

[Groove_C 267]

@oldbear what's the core and cache frequency of your i7-7700K?

[oldbear 390]

@Groove_C CPU is always running @ 4.5 GHz on 4 cores after an error on my part, in fact, I did not wish to have this setting but despite all my manipulations, I was not able to return to standard 4.2GHz/4.5 GHz Turbo.

It works well like that without overheating, so I stopped fighting the ASUS MoBo/BIOS tricks.

 

 

[Groove_C 267]

@oldbear it's really easy to fix your "problem".

In the BIOS, in Ai Tweaker tab, you have an option called ASUS MultiCore Enhancement and if it's on Auto or Enbaled, it forces the CPU out of its stock/original specifications, by making all cores work at boost frequency that is normally allowed by Intel only for 1-2 core(s) and not for all + this option also boosts the voltage to much higher/exagerated values.

Your CPU all cores max boost, as specified by Intel is 4.4 GHz. 4.5 GHz is only for 1-2 core(s).

So just Disable ASUS MultiCore Enhancement to make your CPU work as specified by Intel.

Also make sure, that CPU Core Ratio (under ASUS MultiCore Enhancement) is set to Auto.

 

But there are also other things to check for, so when you're not playing or not using any other CPU heavy programs, the CPU can clock down to just 800 MHz on desktop, to heat/consume/degrade less, when not needed.

1. In the BIOS, in Ai Tweaker tab, scroll down to Internal CPU Power Management and there, make sure that Intel SpeedStep is Enabled.

Also make sure that Turbo Mode (under Intel SpeedStep) is Enabled.

2. In the BIOS, in Advanced tab, in CPU Configuration, scroll all the way down to CPU Power Management Configuration and there, make sure that Intel SpeedStep is Enabled.

Also make sure that Turbo Mode (under Intel SpeedStep) is Enabled.

3. In the BIOS, in Advanced tab, in CPU Configuration, scroll all the way down to CPU Power Management Configuration and there, make sure that CPU C States is set to Auto or Enabled.

4. In Windows, in Alimentation et mise en veille, in right upper corner click on Paramètres d'alimentation supplémentaires -> Modifier les paramètres du mode -> Modifier les paramètres d'alimentation avancés, scroll down to Gestion de l'alimentation du processeur and there in état minimal du processeur, enter 19%, since 18.18% is 800 MHz out of 100% that represent 4400 MHz stock boost all cores.

Because if you don't set the energy plan in Windows to 19%, everything you've done in the BIOS will be ignored, since you'll be in Windows and it's Windows power plan that overtakes it all.

 

Having done all of this, you will have your CPU resotred to its original frequency and voltage, being cooler and also consuming/degrading less + no 4.5 GHz on the desktop for no reason.

[Groove_C 267]

Actually your i7-7700K can win vs. R3 3300X, R5 3600(X) or R7 3700/3800X, in Arma.

It's just that you have it at pretty conservative core and cache frequency of only 4.5/4.2 GHz + only 3200 MHz 16-18-18 RAM.

 

Your i7-7700K can do 5.0-5.1/4.7-4.8 GHz core/cache without problem at more than acceptable temperature and voltage (that will work for years), all of this while still being air-cooled (Noctua) + very good 3600-4000 MHz RAM with very low timings.

The only thing that must be done, is that it must be delidded, because of cheap Intel thermal paste, between the CPU DIE and its copper cover (IHS), that must be already rock hard now or transformed into dust.

This cheap thermal paste from Intel used from 3rd until 8th genereation of CPUs included makes the CPU temperature ~20°C worse than with solder, like it was until 2nd generation and came back starting from 9th generation.

 

So once your i7-7700K is OC'ed to the moon, there is 0 benefit, in Arma, from an i7-9700K, i9-9900K, i5-10600K, i7-10700K or i9-10900K, even if also OC'ed to the moon.

Since maybe 100 MHz more (5.1-5.2 GHz) than your i7-7700K and maybe with RAM that's 133-400 MHz faster (4133-4400 MHz) won't be a big deal. Especially not worth the price of a new mainboard + CPU + RAM.

Not even talking about Ryzen 3x00(X).

[oldbear 390]

@Groove_C Thanks for the tips but I had nothing to change in BIOS and Windows Parameters ... and I am still @4.4/4.5 MHz, but that's not a "problem".

I'm no more interested in overclocking than in delid, I'm happy with the way my gaming rig allows me to play Arma3 every day and I don't see any reason to bother with changes that ultimately does not concern me.

This allows me to share my experience with noobs and late converts to gaming in Arma in the most basic way possible.
I understand that for some this is a subject of interest and even passion, in the same way that over 50 years ago I spent a lot of time polishing pistons and adjusting valves, but now I prefer spending my time in mission editing and video gaming 😎.

[Groove_C 267]

Yeah, I'm also strating to spend less and less time searching for latest tech info/specs/capabilities. It must be due to my age and my preferences chaning more towards spending time on more useful things.

So if one doesn't want to spend months/years researching in order to spend less $$$ and OC for free, for added perofrmance, one sure can buy much better stock hardware, but this can cost considerably more.

But time is money. So you pay more, to be able to spend your time one things you like and not on things you have to do, because you decided to spare money on hardware and have to tinker a lot afterwards.

 

Otherwise, if comapring stock vs. stock, sure, newer CPUs are much better, since almost 5.0 GHz stock all cores and more cores/threads and accepting better RAM makes a noticable difference vs. older stock hardware.

[RAGP13 2]

Just build a new PC and was wondering if my performance is in line with what's expected. I honestly was expecting smooth sailing in this game with everything maxed, but in cities (especially when in vehicles) I get massive drops to the 30-40 fps range. I've already disabled PiP because that's a 10fps hit.

 

Ryzen 5 3600 (PBO + Auto OC)

Sapphire RX 5700 XT Pulse (slight undervolt + VRAM tuning)

16GB Crucial running at 3600CL16

 

YAAB on ultra is showing around 40fps.

 

3DMark Time Spy is ±9500 points for reference.

 

I guess not having an extremely fast single core performance is affecting performance the most?

 

[Valken 622]

Set your video objects distance down to 500, then adjust your video settings until you get good FPS. Then slowly increase the objects back up until your FPS starts to move around. 

 

The terrain and objects can affect both GPU memory and CPU.

 

Or use the PRESETS under video and start with normal, high and adjust to balance between quality vs speed.

 

 

[oldbear 390]

Playing at 40 FPS on "Ultra" settings in YAAB is OK with a Ryzen 5 3600.

As long as the graphics card does not create a bottleneck, a GTX 970 being the minimum, whatever the graphics card, performance will be relatively limited, because of the maximum boost frequency of the R5 3600.

Here what I am getting with my AMD test rig ...

 

Arma* engine is displaying a very large terrain area.

The General Visibility parameter is a 100% CPU parameter.

It defines the area that will be calculated and render by the CPU before being sent to be displayed.

Please have look at the Old Bear method ™  😎

 

[RAGP13 2]

1 hour ago, oldbear said:

Playing at 40 FPS on "Ultra" settings in YAAB is OK with a Ryzen 5 3600.

As long as the graphics card does not create a bottleneck, a GTX 970 being the minimum, whatever the graphics card, performance will be relatively limited, because of the maximum boost frequency of the R5 3600.

Here what I am getting with my AMD test rig ...

 

Arma* engine is displaying a very large terrain area.

The General Visibility parameter is a 100% CPU parameter.

It defines the area that will be calculated and render by the CPU before being sent to be displayed.

Please have look at the Old Bear method ™  😎

 

 

Any tweaking you have done to get 44fps on Ultra? I assume that you just hit "Ultra" for video setting and left everything else the same?

 

I slightly overclocked my 3600 to 4.3Ghz and now get 41.4fps, but expected to be closer to your 3600X. Or does ARMA prefer NVIDIA over AMD and is your 2060 faster in ARMA than my 5700XT? I thought GPU didn't really matter for YAAT because GPU usage is very very low when running it.

[Valken 622]

Nvidia is still better in DX9 - DX11 games because their drivers use the CPU to do some calculations. I have no idea what it is but you can always see the CPU usage is higher with Nvidia than AMD GPU on the same CPU, and it seems Intel CPU usage is HIGHER than AMD CPU when using Nvidia GPUs for the same game getting slightly better FPS. Probably some kind of optimizations.

 

On DX12 and Vulkan, it is almost API optimized based on the game and maxes out most GPU.

 

[oldbear 390]

I am  getting the same score playing a RX 5700 or a RTX 2060, but the GTX 970 is just behind.

The only difference with the GTX 970 is that it runs at 70/98% howling like a turbo in Ultra, the RX 5700 or the RTX 2060, are quite quiet running with a 40/60% load.

Over the "GTX 1060 6GB limit" there are not much differences between GPU

 

Nevertheless, from my point of view, Nvidia GPUs today are just better than AMD's, it has nothing to do with hidden PhysX secrete operation on CPU, they are just plain better.

 

With the R5 3600/GTX 1650 rig I had built for one of my grandsons for late Xmas, I was getting ...

... out of the box, just DDR4 XMP switch on and CPO allowed. Here are the configuration details :

• AMD RYZEN 5 3600
• Gigabyte GA-B450I AORUS PRO WIFI
• Corsair Vengeance LPX 16GB (2 x 8GB) 3200MHz C16
• Inno3D GTX 1650 Compact
• SAMSUNG 970 EVO Plus NVMe M.2 500GB
• Samsung SSD Interne 860 EVO 2.5" (500 Go)
• Corsair TX550M 80PLUS Gold
• Windows 10 Pro

The only special feature in this config is that I replaced the cooler delivered as standard by an AMD Wraith Prism bought as a spare part.

[RAGP13 2]

On 8/18/2020 at 9:21 PM, oldbear said:

I am  getting the same score playing a RX 5700 or a RTX 2060, but the GTX 970 is just behind.

The only difference with the GTX 970 is that it runs at 70/98% howling like a turbo in Ultra, the RX 5700 or the RTX 2060, are quite quiet running with a 40/60% load.

Over the "GTX 1060 6GB limit" there are not much differences between GPU

 

Nevertheless, from my point of view, Nvidia GPUs today are just better than AMD's, it has nothing to do with hidden PhysX secrete operation on CPU, they are just plain better.

 

With the R5 3600/GTX 1650 rig I had built for one of my grandsons for late Xmas, I was getting ...

... out of the box, just DDR4 XMP switch on and CPO allowed. Here are the configuration details :

• AMD RYZEN 5 3600
• Gigabyte GA-B450I AORUS PRO WIFI
• Corsair Vengeance LPX 16GB (2 x 8GB) 3200MHz C16
• Inno3D GTX 1650 Compact
• SAMSUNG 970 EVO Plus NVMe M.2 500GB
• Samsung SSD Interne 860 EVO 2.5" (500 Go)
• Corsair TX550M 80PLUS Gold
• Windows 10 Pro

The only special feature in this config is that I replaced the cooler delivered as standard by an AMD Wraith Prism bought as a spare part.

 

 

These results are quite a bit lower than what's in the YAAT thread on this forum though:

 

There it's around 50 fps with Ryzen 5 3600 and RX570 or GTX 2070 ti

[Valken 622]

@RAGP13 Those benchmarks more than likely out of date and only RELATIVE to the reviewer at that time because the build of ARMA 3 was different.

 

To MAX out ARMA 3, you need to set your

 

CPU to the highest stable speed (OC or PBO), all cores, well at least 2 cores

Memory at the Highest Speed and Lowest Latency and ensure 2x Channels fully working

GPU running without other background load.

ARMA 3 the game installed on SSD or RAM drive if possible due to all the near continuous asset loading during game play. 

I did the ABOVE in ARMA 2, AMAZING performance.

Close all the background applications or run a "game booster" program to do it for you before you launch STEAM and ARMA 3.

 

Then run the benchmark at LOW, NORMAL, HIGH, VERY HIGH and ULTRA settings to see where you sit with the current build of ARMA 3.

 

Run it 3 times each to get an average with NO MODS loaded to ensure consistency.

 

Also, in the Arma 3 Launcher, you should check use 64 Bit, and more threads as all of us would have these settings.

You can test the memory allocator but the current default or Intel one is good.

There is an AMD optimized "dll" file you can find to push further but it may crash.

 

Lastly, many of the ARMA guys here do a MANUAL OC on the CPU to lock the CPU frequency to the highest stable.

 

I am unsure of Ryzen PBO as I do not have such a system, yet.

 

My next system would more than likely be an 8 core Ryzen seeing how good they have been improving but right now, I am on an older Intel 4790K 4x Core CPU which is really close to performance to your system. Except I have 32GB of DDR3 RAM vs you with DDR4 RAM.

 

GPU you should be way faster than me on a 1060 GTX.

 

[oldbear 390]

@RAGP13 I will not start a discussion about YAAB results from other rigs.

YAAB is "In short: About 2.5 minutes long scripted intro for measuring your average FPS. Emphasis on AI fighting performance, bullets, and explosions.", it's not really a benchmark, however, it gives the player a good idea of what to expect in game.

There are large and many variations in FPS level due to the handling of the AI and the environment around the camera.

 

Frankly, a Ryzen 3000 isn't the best option for having the highest possible level of FPS in Arma3. The reason is simple, because of the architecture of the RVEngine, a now obsolete engine, there is an absolute dependence on the frequency and efficiency of the CPU, the best option is probably an Intel i9-9900K slightly OCed running over 5.0 GHz.

 

But you have got an AMD RYZEN 5 3600, a rather good efficient and versatile mid range asset.

We must try to use it at best :

1° in the launcher, as you are using Windows 10 64, let by default the 64-bit client and Enable Large Page Support.

2°in game do not use Presets, but tweak parameters one by one. Always remember that the Overall Visibility parameter is not a video parameter, but a 100% CPU parameter.

Use this Visibility parameter to adjust the CPU load and therefore the FPS level.

 

You can also help the CPU by associating it with fast memory close to the sweet spot designated by AMD with the release of Ryzen 3000.
There is a real interest in using 3600 MHz DDR4 such as the Trident Z Neo 16GB (2x8GB) 3600MHz CL16-16-16-36.

 

 

[RAGP13 2]

11 hours ago, Valken said:

@RAGP13 Those benchmarks more than likely out of date and only RELATIVE to the reviewer at that time because the build of ARMA 3 was different.

 

To MAX out ARMA 3, you need to set your

 

CPU to the highest stable speed (OC or PBO), all cores, well at least 2 cores

Memory at the Highest Speed and Lowest Latency and ensure 2x Channels fully working

GPU running without other background load.

ARMA 3 the game installed on SSD or RAM drive if possible due to all the near continuous asset loading during game play. 

I did the ABOVE in ARMA 2, AMAZING performance.

Close all the background applications or run a "game booster" program to do it for you before you launch STEAM and ARMA 3.

 

Then run the benchmark at LOW, NORMAL, HIGH, VERY HIGH and ULTRA settings to see where you sit with the current build of ARMA 3.

 

Run it 3 times each to get an average with NO MODS loaded to ensure consistency.

 

Also, in the Arma 3 Launcher, you should check use 64 Bit, and more threads as all of us would have these settings.

You can test the memory allocator but the current default or Intel one is good.

There is an AMD optimized "dll" file you can find to push further but it may crash.

 

Lastly, many of the ARMA guys here do a MANUAL OC on the CPU to lock the CPU frequency to the highest stable.

 

I am unsure of Ryzen PBO as I do not have such a system, yet.

 

My next system would more than likely be an 8 core Ryzen seeing how good they have been improving but right now, I am on an older Intel 4790K 4x Core CPU which is really close to performance to your system. Except I have 32GB of DDR3 RAM vs you with DDR4 RAM.

 

GPU you should be way faster than me on a 1060 GTX.

 

 

10 hours ago, oldbear said:

@RAGP13 I will not start a discussion about YAAB results from other rigs.

YAAB is "In short: About 2.5 minutes long scripted intro for measuring your average FPS. Emphasis on AI fighting performance, bullets, and explosions.", it's not really a benchmark, however, it gives the player a good idea of what to expect in game.

There are large and many variations in FPS level due to the handling of the AI and the environment around the camera.

 

Frankly, a Ryzen 3000 isn't the best option for having the highest possible level of FPS in Arma3. The reason is simple, because of the architecture of the RVEngine, a now obsolete engine, there is an absolute dependence on the frequency and efficiency of the CPU, the best option is probably an Intel i9-9900K slightly OCed running over 5.0 GHz.

 

But you have got an AMD RYZEN 5 3600, a rather good efficient and versatile mid range asset.

We must try to use it at best :

1° in the launcher, as you are using Windows 10 64, let by default the 64-bit client and Enable Large Page Support.

2°in game do not use Presets, but tweak parameters one by one. Always remember that the Overall Visibility parameter is not a video parameter, but a 100% CPU parameter.

Use this Visibility parameter to adjust the CPU load and therefore the FPS level.

 

You can also help the CPU by associating it with fast memory close to the sweet spot designated by AMD with the release of Ryzen 3000.
There is a real interest in using 3600 MHz DDR4 such as the Trident Z Neo 16GB (2x8GB) 3600MHz CL16-16-16-36.

 

 

I think I've done all of the above, including the Launcher parameters: Ryzen 5 3600 @ 4.3Ghz. Memory at 3600CL16 (2x8GB). GPU overclocked. ARMA on SSD. Nothing else running in the background.

That nets me 41fps on Ultra 1080p in YAAB.

What I still find surprising is that, even though YAAB mostly focussed on CPU, the GPU usage is really really low during the YAAB with only 40watts of it being used (of 200+ maximum), and GPU speeds of ±200-300Mhz (can do 2000mhz max). I do get max GPU usage when in the main menu though (200+ fps).

 

Guess that's normal and just the CPU bottlenecking the GPU in this test?

 

During gameplay it's around 60-80fps on Ultra, but not when in a city. Then it drops quite drastically.

 

 

 

 

[Valken 622]

@Groove_C Renoir APU is out and do you know the intercore bounce on those 6 and 8 core cpus? Would be good to know.

 

Something happened and I might NEED to upgrade faster then  I thought. So I would look at something that can hold me on both ARMA 3 and ARMA 4 Enfusion engine at the minimum.

 

@RAGP13 The benchmark is sometimes  misleading for one other reason: The AI is running on YOUR computer.

 

In most online games, the AI is running on the server so your FPS would go up further as the CPU load would be lower.

 

But on some maps, there are optimization problems that will result in fixed FPS no matter what.

 

Tanoa is GPU heavy due to the trees and high poly buildings in ULTRA.

 

I was surprised that Chernarus 2020 from the CUP map pack is now GPU heavy due to all the upgrades they did. It looks great so I believe it was worth it so long as I get ~ 30 FPS constantly in Ultra.

 

On my system, if I want more FPS or a better snappy feeling online, I just set

 

Shadow, Clouds, PIP and WATER to HIGH or VERY HIGH.

Then drop the OBJECTS distance down to 500 or 1000.

 

I leave the DRAW distance at 3800.

 

Everything else is on Ultra including 8xAA with CSAA.

You can drop AA down to 4xAA with FXAA Very High to see if that speeds. 

AA is ROPS heavy and Nvidia usually has more ROPS per GPU than AMD at the same performance levels. 

I did this when I had an AMD GPU but  your 5700XT with 64 ROPS should be doing 8xAA fine.

My 1060 GTX only has 48 ROPs and I do run 8xAA most of the time.

[RAGP13 2]

4 hours ago, Valken said:

 

 

@RAGP13 

Everything else is on Ultra including 8xAA with CSAA.

You can drop AA down to 4xAA with FXAA Very High to see if that speeds. 

AA is ROPS heavy and Nvidia usually has more ROPS per GPU than AMD at the same performance levels. 

I did this when I had an AMD GPU but  your 5700XT with 64 ROPS should be doing 8xAA fine.

My 1060 GTX only has 48 ROPs and I do run 8xAA most of the time.

 

2xAA vs 8xAA is maybe 1fps difference. But 0xAA vs 2xAA is about 10fps difference.

 

I'm tempted to leave AA off as I'm now playing on a new 1440p monitor.

[Groove_C 267]

@Valken need to wait for real tests of Renoir APUs.

 

They have better inter-core latency than Zen 2, accept even higher clocked RAM (in 1:1 mode), but they have even less cache than Intel CPUs.

 

Need to see the frequency they can do.