Jump to content
Sign in to follow this  
fabio_chavez

Benchmarking Arma with new i5-6600K and 3000mhz DDR4

Recommended Posts

Hey guys any suggestions how i should benchmark arma 3 performance with my new CPU and RAM before and after Overclock/XMP?

Share this post


Link to post
Share on other sites

The Helo benchmark can vary a bit sometimes so make multiple runs and if you're comparing stuff analyze and watch the benchmark carefully.

I've somehow missed that YAAB benchmark. Need to test it ASAP :P

 

I'd suggest give us at least 4.0GHz CPU clock results. Also what are your RAM timings? Giving that information and if you could downclock timings and/or speed then we could put the RAM in the level with some DDR3. For example in Arma 3 DDR3 10-10-10-28  2133MHz is as fast as DDR4 16-16-16-36 2800MHz.

 

That way at least I and likely some others with somewhat high end DDR3 sticks could easily compare some clock-to-clock results when the RAM is set on the same level like I said earlier.

Share this post


Link to post
Share on other sites

 LOl, im still on my 2500k which ive used for years on Arma -is there really any reason to upgrade for this game?

Share this post


Link to post
Share on other sites

 LOl, im still on my 2500k which ive used for years on Arma -is there really any reason to upgrade for this game?

 

Not really, my best upgrade arma wise was getting a SSD. CPU / GFX upgrade made only little difference.

Share this post


Link to post
Share on other sites

 LOl, im still on my 2500k which ive used for years on Arma -is there really any reason to upgrade for this game?

If he can give nice results for us, then we'll see. At least with that YAAB if he can improve fps nicely then it can be worth it but that all depends how much you want to spend money on what fps improvement.

Share this post


Link to post
Share on other sites

should i use a certain malloc (win 10x64 16bg DDR4)? i havent been running arma in a while but i think i read somewhere that the last bigger patch brought some new stuff in that department?

Share this post


Link to post
Share on other sites

The new allocator is part of 1.52... so testing stock

Share this post


Link to post
Share on other sites

Comparing i5-6600K at stock speed (3,5ghz-turbo: 3,9ghz) with different RAM Speeds (2133mhz 15-15-15-35 and 3000mhz 15-15-15-35 )


the first results are the following (​average from 3 runs)

no XMP
Altis: 31,6666 fps
Stratis: 29,3333 fps
YAAB: ​25,2 fps

with XMP
Altis: ​37 fps
Stratis: 31 fps
YAAB: ​28,8666 fps

The ram clock does have an effect on fps with i5 on stock Speed, however, the stratis benchmark hardly increases a single fps, assumingly because the load from that Mission is predominantly AI.
I could tighten the timings a bit to see how the RAM on 2133mhz compares then but im more interessted to get the CPU clock up, hopefully to around 4,7ghz, afterwards i might make another run with smaller RAM timings.

I also ​tried to see if i can see a difference in MP but it doesnt make much sense because the server bound Performance varies a lot.

p.s. my GPU is a HD 69​70 oc btw
 

  • Like 1

Share this post


Link to post
Share on other sites

Weird... I would have thought that would be higher... on my last pass of Helos Bmark after the 1.52 update I got 47fps. Was that a full ultra setting Fabio?

mine is a Mixture of Ultra and very high.

AO off, mirrors off,caustics off, VD 1750 and totalD 4000

Share this post


Link to post
Share on other sites

Weird... I would have thought that would be higher... on my last pass of Helos Bmark after the 1.52 update I got 47fps. Was that a full ultra setting Fabio?

mine is a Mixture of Ultra and very high.

AO off, mirrors off,caustics off, VD 1750 and totalD 4000

I think it's the view distance. I get 29fps on Stratis with 12000m VD on my old i5@3.6GHz/1800MHz RAM. Other settings are mostly High/Very high.

Share this post


Link to post
Share on other sites
It was allready late yesterday and i realized that i mixed regular arma3.exe with the perf Client in some of the runs, if that makes any difference in the Benchmarks, i will rerun them.

On Quality Settings:
I somehow assumed that the missions force their own Quality Settings, by the way there isnt any common Standard for Quality Settings in Benchmarks (als the reason most tech/game site Benchmarks of arma arent very meaningfull), maybe you should suggest a reference config for YAAB Greenfist?

On Benchmark Design:​
Also one thing that Bugs me about most Benchmarks is that its all or mostly AI heavy, i prefer pvp so the Hardware requirements are generally different to PvE, i know its not possible to simulate the network/server part but id like to see a Benchmark Mission that focuses on physx Simulation and particle effects etc, without AI in the mix to get an idea of what to expect from non-AI multiplayer under stable conditions.

 

Share this post


Link to post
Share on other sites

I'd like to see benchmarks that are more "realistic" in a sense. Say 5 jets fattacking a town, 3 AAs shooting at them. Thats when I notice my FPS cuts in half, just because they fire so many rounds. At least, that's what I assume it depends on.

 

It would be nice to see what the FPS is without Physx as well. For example Metro: LL sucks on AMD cards with Physx on. I get a third of my normal FPS. 90 FPS -> 30 FPS.

Share this post


Link to post
Share on other sites

On Quality Settings:

I somehow assumed that the missions force their own Quality Settings, by the way there isnt any common Standard for Quality Settings in Benchmarks (als the reason most tech/game site Benchmarks of arma arent very meaningfull), maybe you should suggest a reference config for YAAB Greenfist?

On Benchmark Design:​

Also one thing that Bugs me about most Benchmarks is that its all or mostly AI heavy, i prefer pvp so the Hardware requirements are generally different to PvE, i know its not possible to simulate the network/server part but id like to see a Benchmark Mission that focuses on physx Simulation and particle effects etc, without AI in the mix to get an idea of what to expect from non-AI multiplayer under stable conditions.

I don't have any preset preference for reference settings, perhaps the "High" preset would be reasonable...?

 

Since a mission can't force all the relevant settings, we'll have to use the vanilla presets when comparing results between users. But even then the different screen resolutions will skew the numbers.

Only thing YAAB forces is the overall view distance (1500m). I didn't want to set the object VD because the scripting command changes it permanently, which would confuse users when their settings change unannounced.

 

I've been planning to make yet another another benchmark mission without the AI, with only things like physx, ballistic physics, particle effects etc. Still in single player though.

 

edit. Funny thing I noticed just now, Physx seems to running in its own thread! Pretty nice. Now who said Arma is single-threaded?! :D

Share this post


Link to post
Share on other sites

With quality preset "High" in 1920x1080 i get the following (average of 3 runs)

Stratis: ​51,6666
Altis: 60
YAAB: ​38,3


​(all with the vanilla client, i double checked 1 run with performance client but it only varied ~1fps)

Share this post


Link to post
Share on other sites

Hmm so you've HD6970. I think it's better to low some graphics. I'd say disable every AA&PP setting at least. The important thing is that GPU shouldn't ever to get at 100% usage or it lowers the average fps. Also the graphs from the YAAB would be nice to see. So find the good video settings first and then, the results should be easily comparable. But I assume that YAAB didn't bottleneck your GPU so I'll try now with the same graphic settings.

 

 

/////

 

i7 2600K @ 3.9GHz, 9-11-10-28. This likely means I'm 0,5-1fps behind comparing to your RAM, but I7 can help bit to compensate that.

 

YAAB, High pre-set video settings, results avg 36,6fps.

 

So in a battle where is much shooting and AI stuff going on and explosion smokes are covering the screen many times, the difference between Sandy Bridge and Skylake is around ~1fps. Likely that's the increase in multiplayer when the fps in there is around 20-40fps.

 

But yeah please post the graph from the benchmark. Easier to see where the improvement is happening and how much there is that.

 

I bet your Helo benchmark has some GPU bottlenecks because I can get higher results (66fps) with the config above.

Share this post


Link to post
Share on other sites

Using YAAB I get 20.2 on very high settings, 20.6 on high settings and a measly 24 on standard settings with very little variation.

 

I swear that performance has dropped significantly since Beta for me, with small increments at the time. Seeing that my graphic settings has such a small impact, it suggests to me that it must be my CPU.

My setup is an old I7 920 oc @ 3,8 ghz w/o hyper threading, 24 gb ram and a non-SSD and a Radeon 7970. I'm able to play most games decently, and at least much better than this. I recon I've lost at least 10 fps if not more since alpha. 

 

Edit: Just wanted to mention that I recently tried playing the campaign over again, and had incredibly low fps all around, even dipping below 10 in certain scenes. I only played the first mission, and it's not particularly heavy compared to later missions where I struggled before as well.

I also tried playing through "Deliverance", and had times where the game would just simply freeze for a significant amount of time.
 

 

Tried the new Battlefront beta today, max settings with very strong performance in game (never mind that it completely froze my computer after every match though, maxing out CPU usage). I just can't believe how terrible ARMA has become, and this from someone who have defended BIS on performance for ages.

Yes, ARMA does allow for further view distances and has more complex stuff to work on than arcade shooters... But this is terribly poor. They should be so glad they don't have a real competitor, because everyone would abandon ARMA in a split second if there was a real alternative...


 

Share this post


Link to post
Share on other sites

Just benchmarked my 2500k on YAAB and got 38.6fps on high default setting, gpu was thrashing 99% for a lot of it though.

I also ran it through with my preferred settings and got 37fps, gpu usage is around 80%.

 

1920x1080

 

Sampling        100%
Texture            Very High
Objects            Very High
Terrain            Very High

Shadow            Very High
Particles        High
Cloud            Very High
PiP            Standard

HDR            Standard
Dyynamic Lights        Very High

Overall Visibility    2250
Object Visibility    2250
Shadow Visibility    100

Bloom            Off
Raidal Blur        Off
Rotation Blur        Off
Depth of Field        Off
Sharpen Filter        Off
AO            Disabled
Caustics        Enabled

FSAA            4x
ATOC            All trees + grass
PPAA            SMAA ultra
Anisotropic Filtering    Ultra
 

System is: -

i5 2500k @4.5ghz

16gig - 2100mhz 11-11-11-31-2T

gtx660

cruicial mx100 ssd

Share this post


Link to post
Share on other sites

FWIW, with YAAB on high graphics preset at 1920x1080 I get 31.5 FPS with my mildly OC I7 950 and GeForce GTX 970

 

Full specs:

  • Windows 7 64bit Professional
  • Intel I7-950 (overclocked @ 3.85GHz)
  • Asus P6X58D Premium
  • 3x2048MB Mushkin Ridgeback 998826 (PC3-12800 6-8-6-24)
  • MSI Nvidia GeForce GTX970 4096Mb
  • Creative X-Fi XtremeGamer Fatality (PCI)
  • Samsung 500GB 850 EVO SSD 

 

When comparing benchmarks for various CPUs, it may be worth using the lowest graphics settings available (including resolution), to reduce influence of the GPU and ensure it isn't a bottleneck.

Share this post


Link to post
Share on other sites

FWIW, with YAAB on high graphics preset at 1920x1080 I get 31.5 FPS with my mildly OC I7 950 and GeForce GTX 970

 

Full specs:

  • Windows 7 64bit Professional
  • Intel I7-950 (overclocked @ 3.85GHz)
  • Asus P6X58D Premium
  • 3x2048MB Mushkin Ridgeback 998826 (PC3-12800 6-8-6-24)
  • MSI Nvidia GeForce GTX970 4096Mb
  • Creative X-Fi XtremeGamer Fatality (PCI)
  • Samsung 500GB 850 EVO SSD 

 

When comparing benchmarks for various CPUs, it may be worth using the lowest graphics settings available (including resolution), to reduce influence of the GPU and ensure it isn't a bottleneck.

That is positively really weird, I have a very close CPU to you, the I7 920 oc'ed at 3,8ghz pulling around 24 frames. I do have a worse graphics card, however, I get allmost the same result with everything (except shadows) on low.

Share this post


Link to post
Share on other sites

That is positively really weird, I have a very close CPU to you, the I7 920 oc'ed at 3,8ghz pulling around 24 frames. I do have a worse graphics card, however, I get allmost the same result with everything (except shadows) on low.

 

When you account for CPU clock speeds, my results seem to compare as expected against the I7 2500k results forteh posted (if he was running at 3850Mhz, he would get ~33 FPS compared to my 31 FPS,  and 2500k is approx 10% faster than 920/950 clock for clock)

 

Some of it could be memory (my ram is triple channel 7-9-7-24-88-1T @ 875Mhz (1750MHz DDR), but otherwise I'd say something is wrong at your end sorry :(

Maybe time for an SSD and fresh windows install?

Share this post


Link to post
Share on other sites

I suspect it's the ram that is helping keep the speed up, switching from 1600mhz to 2133mhz was a big jump to the tune of around 20-30% increase on the helo benchmark; I never benched using YAAB with the slower ram but it is a well known peice of the puzzle of getting arma3 to run well.

 

I clocked my cpu back to 3.5ghz (stock 6600k speed) and got 30.8fps on YAAB, clocked it up to 4.7Ghz and got 39.5fps, both on high preset and with 2133mhz ram; so yes skylake is faster to the tune of 30% at stock speeds on YAAB.  However for the cost I'll stick to the 2500k as mp performance is still going to be on the floor :D

 

For reference I get around 68-70fps on the helo benchmark on high settings.

Share this post


Link to post
Share on other sites

When you account for CPU clock speeds, my results seem to compare as expected against the I7 2500k results forteh posted (if he was running at 3850Mhz, he would get ~33 FPS compared to my 31 FPS,  and 2500k is approx 10% faster than 920/950 clock for clock)

 

Some of it could be memory (my ram is triple channel 7-9-7-24-88-1T @ 875Mhz (1750MHz DDR), but otherwise I'd say something is wrong at your end sorry :(

Maybe time for an SSD and fresh windows install?

Thanks for your reply. I doubt it's due to my windows install, I recently reinstalled windows 7 and I have it on SSD. I'm running a very bare bones setup in terms of software installed right now, might be about 2 months since last reinstall? I did move ARMA to my SSD though, but it didn't help in terms of frames (as it shouldn't).

But since you are replying in the first place, I could swear that I have lost about 10 frames over maybe the last year or so in ARMA. Has there been a performance drop over time on your part? Back in alpha, I had allmost better performance in ARMA III compared to ARMA II (albeit it was on Stratis of course).

 

Share this post


Link to post
Share on other sites

Thanks for your reply. I doubt it's due to my windows install, I recently reinstalled windows 7 and I have it on SSD. I'm running a very bare bones setup in terms of software installed right now, might be about 2 months since last reinstall? I did move ARMA to my SSD though, but it didn't help in terms of frames (as it shouldn't).

But since you are replying in the first place, I could swear that I have lost about 10 frames over maybe the last year or so in ARMA. Has there been a performance drop over time on your part? Back in alpha, I had allmost better performance in ARMA III compared to ARMA II (albeit it was on Stratis of course).

 

I also believe Arma 3 has lost about 10fps (realistically thinking, maybe about 7fps)  for me in a year and some part of that is likely explained by the security change, that made the fred41's Large Pages and stuff useless. I remember having better fps when driving or flying with PiP.

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
Sign in to follow this  

×