Low CPU utilization & Low FPS

[stk2008 14]

i agree BIS keep fixing things that should not be a priority at this time.

As I have said before

Why build a house on unstable foundation when the house will inevitably fall down sooner or later.

BIS has been aware of this problem since ARMA2 and yet release ARMA3 with the same issue but this time strapped more stuff on to the engine which struggled before hand :(

[Sneakers O'Toole 2]

they must have like 0.5 of a man for the hardcore engine stuff and 50 people for artwork and soft engine stuff or w/e. so all they can do is the other 'changed helmet colour in whatever mission etc...' stuff. maybe they should have spent that 500k on a few more engine guys.

I can't believe they have still never mentioned this problem in any meaningful way yet. The were going to do a blog, then they say we haven't got the time because we are so busy fixing the problem, lol.

[jumpinghubert 49]

http://dev.arma3.com/sitrep-00051

"....Multiplayer optimization attempts are looking optimistic. The server-side algorithm that determines what messages to resend and to which clients, has been optimized and rewritten to use additional CPU cores. Together with a client-side optimization of network message computations, these changes should provide a noticeable boost to framerates in multiplayer sessions. The changes have to be carefully tested but they should be ready for the update after Zeus (1.18). You can already try them out on devbranch...."

sounds good

[Instynct 1]

i agree BIS keep fixing things that should not be a priority at this time.

As I have said before

Why build a house on unstable foundation when the house will inevitably fall down sooner or later.

BIS has been aware of this problem since ARMA2 and yet release ARMA3 with the same issue but this time strapped more stuff on to the engine which struggled before hand :(

Because they can still sell the house and save money, even though the buyer is angry.

---------- Post added at 21:18 ---------- Previous post was at 21:16 ----------

http://dev.arma3.com/sitrep-00051

"....Multiplayer optimization attempts are looking optimistic. The server-side algorithm that determines what messages to resend and to which clients, has been optimized and rewritten to use additional CPU cores. Together with a client-side optimization of network message computations, these changes should provide a noticeable boost to framerates in multiplayer sessions. The changes have to be carefully tested but they should be ready for the update after Zeus (1.18). You can already try them out on devbranch...."

sounds good

If this is what dwarden has been working on it's already been tested and has very little improvement.

[tonschuh 3]

I think the devs should stop fixing unneccessary things like controls of vehicles and other not needed stuff. First the player should have an acceptable performance, then u can adjust controls of vehicles and soldiers. I dont need good control when my frames go into basement. Yesterdays Driver 337.50 performs very good in a lot of games, of course in Arma not.

Here are some quick benches of the 337.50 Nv Driver

StarSwarm Demo , 1920x1080 , Follow , Extreme (5 Runs Average)

R335.23 WHQL ~43.73fps

R337.50 Beta ~64.24fps ( +46,9 Percent )

-------------------------------------------

Hitman Absolution "Internal-Benchmark" , 1920x1080 , FXAA , Ultra Details , DirectX 11

58,4fps vs. 81,8fps +40 Percent

-----------------------------------------------------------

Battlefield3 "Strike of Karkand" 1920x1080 , 4xMSAA , Ultra Details , DirectX 11

76fps vs. 106fps +39,5Percent

-----------------------------------------------------------------------

Battlefield4 "Siege of Shanghai" 1920x1080 , Medium Details , DirectX 11

122fps vs. 155fps +27 Percent

----------------------------------------------------------------------------

THIEF "Internal-Benchmark" , 1920x1080 , FXAA , Maximum Details , DirectX 11

80,5fps vs. 93,1fps +15,6 Percent

-----------------------------------------------------------------------------

Arma 3 = 0%

That is so ridiculous.... I am highly disappointed...

Looks like that you have a CPU-Bottleneck under normal circumstances, or maybe not ?!?!

I get no improvements when running 3DMark-11 or 3DMark-Vantage (3770k @4700MHz + 2x GTX680-SOC-2GB in 2-Way-SLI and further OC + 16GB (4x4GB) G.Skill TridentX @2400MHz (10-11-11-28-2T)), but when I compare my StarSwarm-Bench-Results from the 1st of February 2014 (can't remember which Driver-Version I used back then) with the ones from today, then I get the following outcome:

01/02/2014:

Follow:

Test Duration: 360 Seconds

Total Frames: 15923

Average FPS: 44.23

Average Unit Count: 4141

Maximum Unit Count: 5322

Average Batches/MS: 697.54

Maximum Batches/MS: 2005.35

Average Batch Count: 17077

Maximum Batch Count: 150035

09/04/2014:

Test Duration: 360 Seconds

Total Frames: 25590

Average FPS: 71.08

Average Unit Count: 4566

Maximum Unit Count: 5736

Average Batches/MS: 1158.16

Maximum Batches/MS: 4418.10

Average Batch Count: 18164

Maximum Batch Count: 149202

01/02/2014:

Attract:

Test Duration: 360 Seconds

Total Frames: 13958

Average FPS: 38.76

Average Unit Count: 4172

Maximum Unit Count: 5361

Average Batches/MS: 730.20

Maximum Batches/MS: 1905.15

Average Batch Count: 21078

Maximum Batch Count: 178249

09/04/2014:

Test Duration: 360 Seconds

Total Frames: 20750

Average FPS: 57.64

Average Unit Count: 4335

Maximum Unit Count: 5511

Average Batches/MS: 1436.87

Maximum Batches/MS: 3820.79

Average Batch Count: 26258

Maximum Batch Count: 169406

01/02/2014:

RTS:

Test Duration: 360 Seconds

Total Frames: 3727

Average FPS: 10.35

Average Unit Count: 3951

Maximum Unit Count: 5722

Average Batches/MS: 1055.62

Maximum Batches/MS: 1741.23

Average Batch Count: 95791

Maximum Batch Count: 178239

09/04/2014:

Test Duration: 360 Seconds

Total Frames: 8467

Average FPS: 23.52

Average Unit Count: 3797

Maximum Unit Count: 5349

Average Batches/MS: 2006.39

Maximum Batches/MS: 4152.98

Average Batch Count: 82394

Maximum Batch Count: 170550

01/02/2014:

Shmup:

Test Duration: 360 Seconds

Total Frames: 14448

Average FPS: 40.13

Average Unit Count: 3976

Maximum Unit Count: 5217

Average Batches/MS: 721.75

Maximum Batches/MS: 1937.17

Average Batch Count: 20018

Maximum Batch Count: 107416

09/04/2014:

Test Duration: 360 Seconds

Total Frames: 23915

Average FPS: 66.42

Average Unit Count: 4313

Maximum Unit Count: 5592

Average Batches/MS: 1079.65

Maximum Batches/MS: 3903.77

Average Batch Count: 20091

Maximum Batch Count: 102088

Maybe someone knows why there is such a different outcome between the 3D-Mark-Benchmark's and the StarSwarm one.

If the new StarSwarm-Scores are real-world improvements, then Mantle is pretty much dead.

I did some ARMA-3 Benchmarks, but only on Ultra-Settings and it doesn't look good:

Dev-Build v1.05.111581 (23/10/2013) + GeForce v331.58 WHQL:

- Stratis = 45fps

- Altis = 41fps

Dev-Build v1.17.116549 (09/04/2013) + GeForce v337.50 Beta:

- Stratis = 41fps

- Altis = 39fps

:)

Edited April 9, 2014 by TONSCHUH
ARMA-3-Bench-Results added.

[jumpinghubert 49]

I did some ARMA-3 Benchmarks, but only on Ultra-Settings and it doesn't look good:

Dev-Build v1.05.111581 (23/10/2013) + GeForce v331.58 WHQL:

- Stratis = 45fps

- Altis = 41fps

Dev-Build v1.17.116549 (09/04/2013) + GeForce v337.50 Beta:

- Stratis = 41fps

- Altis = 39fps

:)

thanks for the testing! Seems arma3 is the only game takes disadvantage from the new driver :(

[nikiforos 450]

I can't stop laughing , so typical for Arma :D

[Greenfist 1863]

Isn't TONSCHUH's test done in two completely different versions of Arma, six months apart? That's doesn't really tell anything about anything.

[jumpinghubert 49]

Isn't TONSCHUH's test done in two completely different versions of Arma, six months apart? That's doesn't really tell anything about anything.

yep thats not an ideal apple to apple thing..

[Sneakson 1]

Isn't TONSCHUH's test done in two completely different versions of Arma, six months apart? That's doesn't really tell anything about anything.

Glad you noticed ;)

The tests in this thread are competing to be the worst. Synthetic benchmarks, different ARMA versions...

I'm going to give benchmarking the new drivers a go soon.

[tonschuh 3]

yep thats not an ideal apple to apple thing..

Not ideal, but on the other hand the 6 month old version was one of the few versions which performed a bit better back then.

Now, 6 month later after a lot of "improvements" plus a new driver which should give you some more fps for free, we get actually even less fps.

If I find some spare time, then I will re-install the previous WHQL-Driver and run another benchmark, but I can't really see that there will be a major difference in the outcome.

We didn't even gain 5fps in that period, but lost another 2-4fps ?

I really start to think that Mantle would not be able to give us more fps as well.

:butbut:

[LSD_Timewarp82 21]

Looks like that you have a CPU-Bottleneck under normal circumstances, or maybe not ?!?!

lol, sure Bottleneck^^ i7 @4,2Ghz is a bottleneck for sure :rolleyes: I can go up to 5200Mhz, but for this i need to setup my good old MachII GT from my basement. But i wont do it, not in mood for that.

But to stay on topic: yes really wierd that Arma is really a game which doesnt have a noticable performance increase at all. Thats why i mentioned that the devs "should" start to work on this crappy (sorry, but i cant say it otherwise) Engine

[tonschuh 3]

lol, sure Bottleneck^^ i7 @4,2Ghz is a bottleneck for sure :rolleyes: I can go up to 5200Mhz, but for this i need to setup my good old MachII GT from my basement. But i wont do it, not in mood for that.

But to stay on topic: yes really wierd that Arma is really a game which doesnt have a noticable performance increase at all. Thats why i mentioned that the devs "should" start to work on this crappy (sorry, but i cant say it otherwise) Engine

I only mentioned the "possible" CPU-Bottleneck, as we should only gain massive amounts of fps like with Mantle if there is a bottleneck, but I doubt that this is the case with our CPU's etc., at least not outside ARMA-3, but on the other hand my StarSwarm fps go down, if I reduce my CPU-OC. @4200MHz vs. @4700MHz resulted in a ~5-6fps difference (Follow).

:butbut:

[LSD_Timewarp82 21]

The only thing what i didnt understand is that how other devs are able to increase the performance for games driver to driver, yes that thing what till today never happened in Arma. Arma would be so fantastic with legitable frames, just to think about if it would run with 50-60fps constantly, man that would be great...

But thank god they fixing helmets for Soldiers in mission xy...

[jumpinghubert 49]

Not ideal, but on the other hand the 6 month old version was one of the few versions which performed a bit better back then.

Now, 6 month later after a lot of "improvements" plus a new driver which should give you some more fps for free, we get actually even less fps.

If I find some spare time, then I will re-install the previous WHQL-Driver and run another benchmark, but I can't really see that there will be a major difference in the outcome.

We didn't even gain 5fps in that period, but lost another 2-4fps ?

I really start to think that Mantle would not be able to give us more fps as well.

:butbut:

Same here. From older (roughly 1 month ago) arma-version to newest I lost 2-4 fps too. From 47fps in helo´s bench to 44.

[nikiforos 450]

Unfortunately this is the true, instead of making Arma run better they managed to worsen the situation and lower the FPS.

And the worst part is that we all know that they have other priorities than improving game performance.

Edited April 9, 2014 by Nikiforos

[Sneakson 1]

NVidia Beta drivers 337.50 (7 April) don't increase ARMA3 framerates.

4770K + 770 w 16GB 1600 MHz on SSD here.

Stratis: (empty Stratis hill outside Agia Marina)

322.10: 61

335.00: 61

337.50: 62

Altis: (empty Altis highest hill)

322.10: 65

335.00: 66

337.50: 66

Infantry: (60s)

322.10: 46

335.00: 46

337.50: 46

Helicopters: (60s after takeoff)

322.10: 47-48

335.00: 46

337.50: 48

Combined Arms: (60s after touchdown)

322.10: 42

335.00: 45

337.50: 45-46

Support: (60s of attack)

322.10: 37-38

335.00: 37

337.50: 39

Intro: (60s of campaign intro, heli)

322.10: 36

335.00: 37

337.50: 38

I believe 335 is WHQL?

Anyways I did 332.10 tests twice, others once and the intro only once on all drivers.

The only significant increase is Combined Arms really... gonna try that again though.

Edit: did another CA and it ended up at 45... so 335.00 may have done something but 337.50 probably not.

Basically all differences are within 1-3 fps and can be ignored.

Edited April 9, 2014 by Sneakson

[LSD_Timewarp82 21]

Lol, 1-3 fps are just ridicolous. Unbelivable...

[Dwarden 1125]

so you blame us for NVIDIA making the performance worse in driver xyz?

since when we are in charge of NVIDIA driver development ?

[LSD_Timewarp82 21]

i didnt blame anyone, i already told that Arma is a fantastic game (i got over 2200hours on Steam just for Arma3) and its too sad that the performance isnt higher. I didnt know very much about the Engine you guys are using but it seems the engine is hard to tweak for you devs. Otherwise a dev should imagine how a player see this situation, release is a while ago, tons of patches and fixes were released, but performance stays the same or is getting lower. Even with very expensive pc´s you wont be able to get constantly legit frames, maybe with low details, but i have a good pc and low details are no option for me.

I still hope and pray that this performance will raise someday :)

[Ckrauslo 12]

so you blame us for NVIDIA making the performance worse in driver xyz?

since when we are in charge of NVIDIA driver development ?

Dwarden Arma 3 don't have optimization neither in Nvidia nor ATI

[Dwarden 1125]

ofc it does have, but that's optimization done by theirs driver developers not us ...

[Ckrauslo 12]

ofc it does have, but that's optimization done by theirs driver developers not us ...

Just their optimization wont allow the game to run smoothly like other games....it's the company responsibility, you're game not theirs

---------- Post added at 22:30 ---------- Previous post was at 22:25 ----------

And another thing, their drives will be update every time, this is a GPU optimization, you need a in game optimization

[Th4d 10]

Just their optimization wont allow the game to run smoothly like other games....it's the company responsibility, you're game not theirs

---------- Post added at 22:30 ---------- Previous post was at 22:25 ----------

And another thing, their drives will be update every time, this is a GPU optimization, you need a in game optimization

"your"

This new driver has great gains only when there´s a cpu bottleneck caused by the driver that slows the game down, but ArmA´s bottleneck isn´t driver related, the bottlenecks are inherent to how the game was designed and lie on it´s own code. Blaming gpu drivers is not feasable as an excuse for the games performance anymore, what saddens me is that this somewhat shows that even if they introduced Mantle or optimized for the upcoming DX12, i fear that it would wield no performance gains, just clearly shows that the engine and the design approach used is simply too outdated for current hardware.

[Ckrauslo 12]

Sorry i misspelled