36 replies to this topic

[CwStrife]

Running MSAA puts my GPU to 100% usage now along with the user.cfg file above in the link

[Goose]

CwStrife, on 19 December 2015 - 10:34 AM, said:

Running MSAA puts my GPU to 100% usage now along with the user.cfg file above in the link

Yeah: It does that.

Here: http://mwomercs.com/...ost__p__4893328

[CwStrife]

Goose, on 19 December 2015 - 12:01 PM, said:

Yeah: It does that.

Here: http://mwomercs.com/...ost__p__4893328

Goose, that topic just says someone has the same problem as I do. Any idea though why enabling MSAA anti-aliasing ramps GPU usage to 100% constantly? Its sorta the fix I'm looking for but MSAA is so demanding my fps really doesn't improve any.

[Goose]

While turning on the punishing AA types will up your cards' load, it's not gon'a change your framerate for the better, if at all.

Put "sys_budget_soundCPU = 0" in your user.cfg, and try reducing any or all of Particles, Shadows, Damage Glow (http://mwomercs.com/...th-hotfix-live/), Objects, and Environment …

Also: http://mwomercs.com/...ost__p__4801576

[CwStrife]

soundCPU is already in there and set on 0. Particles are on low, shadows low, shading low, enviroment and textures are the only thing are on very high but again, even when I tried to set them on low I didn't get more FPS, I just continued to get less GPU usage... Goose, have you had contact with PGI on this at all? I'm reading all the links you send me as well.

[Goose]

… Where is your user.cfg? "D:\GAMES\Piranha\MechWarrior Online" or the like? Next to the system.cfg, right?

Not "E:\Users\Goose\Saved Games\MechWarrior Online" ?

[CwStrife]

Goose, on 20 December 2015 - 03:12 PM, said:

… Where is your user.cfg? "D:\GAMES\Piranha\MechWarrior Online" or the like? Next to the system.cfg, right?

Not "E:\Users\Goose\Saved Games\MechWarrior Online" ?

Yes it's in the main Mechwarrior folder, not in the saved games folder.

The only difference the user.cfg file did was utilize all 8 cores on the CPU instead of 4... but still doesn't resolve piss poor GPU usage giving framerates like i'm running a super budget graphics card. :-\

[Tarl Cabot]

http://www.cpu-world...5_i5-3570K.html

The above is with a non-OC 3570k i5.

i5-3570

• In single-threaded programs, the microprocessor is 44% faster.

FX-8350

• In multi-threaded programs, the FX-8350 is 8% faster. The difference in performance is application specific. In some of them the CPU could be up to 35% faster

MWO is a single thread program spread out over several cores. It is why MWO could do well with the Intel anniversary edition g3258, OC.


Simply noting where each is at default setting. And all the benchmarks using Crysis 3 from a few years ago between the FX-8350 and if-3570k, 3770k, etc, at default settings had the benchmarks approx the same using the same video cards from 3 years ago. Add newer GPU and it would appear Intel benefits more than AMD, even more so if the Intel CPU are OC.

The other side is that the heavy modifications PGI has had to do with the CryEngine 3 may or may not have had a negative impact, especially for an engine also meant to be used on X-Box/PS3 where for each set the components are the same. And it was noted elsewhere that PGI may not have an AMD system in place. It would not be the first team who did not have a cross-section of systems to use for testing.

I am definitely not an expert, simply someone who has seen the results due to lack of a cross-section and trying to work out why stuff that may appear obvious gets missed. /shrugs

Edited by Tarl Cabot, 20 December 2015 - 07:29 PM.

[CwStrife]

Tarl Cabot, on 20 December 2015 - 07:27 PM, said:

http://www.cpu-world...5_i5-3570K.html

The above is with a non-OC 3570k i5.

i5-3570

• In single-threaded programs, the microprocessor is 44% faster.

FX-8350

• In multi-threaded programs, the FX-8350 is 8% faster. The difference in performance is application specific. In some of them the CPU could be up to 35% faster

MWO is a single thread program spread out over several cores. It is why MWO could do well with the Intel anniversary edition g3258, OC.



Simply noting where each is at default setting. And all the benchmarks using Crysis 3 from a few years ago between the FX-8350 and if-3570k, 3770k, etc, at default settings had the benchmarks approx the same using the same video cards from 3 years ago. Add newer GPU and it would appear Intel benefits more than AMD, even more so if the Intel CPU are OC.

The other side is that the heavy modifications PGI has had to do with the CryEngine 3 may or may not have had a negative impact, especially for an engine also meant to be used on X-Box/PS3 where for each set the components are the same. And it was noted elsewhere that PGI may not have an AMD system in place. It would not be the first team who did not have a cross-section of systems to use for testing.

I am definitely not an expert, simply someone who has seen the results due to lack of a cross-section and trying to work out why stuff that may appear obvious gets missed. /shrugs

Oh i'm not going to doubt that the Intel systems are faster, but even people with Intel systems are reporting the same issue as well... just not as many it seems.

I'm not sure sure it exactly comes from the CPU particularly; because if I use a tool to monitor each core individually none of them ever go above 65% give or take a couple % points. With the user.cfg file that I added now the game is using all 8 cores, and the load is never more than like 50% on any given core.

Again, the thing is that if you look at my screenshots, when I first load in a match the GPU will ramp up to 100%, it's after I am able to start moving and the mech has "powered up" that the GPU drops to around 40-50% load.

By enabling MSAA anti-aliasing I get between 75-99% GPU usage, but that doesn't help me so much because MSAA obviously won't increase my FPS... it can actually hurt it in some cases, but at least the GPU gets a workout.

That leaves me thinking it really can't be the processor, i'm really thinking it's something else. I've been looking through Goose's posts and things that can be put in user.cfg file for the cryengine but so far i'm not really seeing anything GPU related, just alot of processor related stuff.

[Tarl Cabot]

What has not discussed is the global GPU settings, before even getting into the game settings. Player A may use Quality settings vs Player B using Performance vs Player C using Standard, then applying same thinking to the actual game settings can have a greater effect. Unfortunately coding to cover a spectrum of systems can be daunting, and having the staff with different testing systems for both Intel and AMD simply adds another layer to it.

AA Off, Temporal Filtering, MSAA 2x, MSAA 4x, MSAA 8x, NVIDIA TXAA 2x, NVIDIA TXAA 4x, and PPAA Off, FXAA and T-AA (Temporal Anti-Aliasing).

Add that MWO loves cpu speed more than cores, with 4.0ghz being approx the starting point and 4.2ghz being the sweet spot, at least with Intel. The other is if PGI does not have an AMD testing system, they have nothing in-house to test, not simply actual changes but to debug coding that could improve AMD performance. Even though Intel/AMD does approx the same work, they do it slightly in different ways, taking different routes. What a good route is for a bike may not be a good route for a sport car, or a jeep/4x4, vis versa

/shudders. Just thinking of two other games where the devs had local players bring in their systems to debug a few critical items cause they did not have a testing system that fell within a specific parameter that was affecting a considerable amount of players. And there was no real option to "upgrade" the system except to purchase an entirely new system.

edit - I have not seen this mentioned, but during beta Nvidia users needed to turn off/disable Ambient Occlusion. Either set it up to turn it off Globally or setup MWO as a program that turns it off.

Edited by Tarl Cabot, 20 December 2015 - 08:44 PM.

[CwStrife]

Tarl Cabot, on 20 December 2015 - 08:34 PM, said:

What has not discussed is the global GPU settings, before even getting into the game settings. Player A may use Quality settings vs Player B using Performance vs Player C using Standard, then applying same thinking to the actual game settings can have a greater effect. Unfortunately coding to cover a spectrum of systems can be daunting, and having the staff with different testing systems for both Intel and AMD simply adds another layer to it.

AA Off, Temporal Filtering, MSAA 2x, MSAA 4x, MSAA 8x, NVIDIA TXAA 2x, NVIDIA TXAA 4x, and PPAA Off, FXAA and T-AA (Temporal Anti-Aliasing).

Add that MWO loves cpu speed more than cores, with 4.0ghz being approx the starting point and 4.2ghz being the sweet spot, at least with Intel. The other is if PGI does not have an AMD testing system, they have nothing in-house to test, not simply actual changes but to debug coding that could improve AMD performance. Even though Intel/AMD does approx the same work, they do it slightly in different ways, taking different routes. What a good route is for a bike may not be a good route for a sport car, or a jeep/4x4, vis versa

/shudders. Just thinking of two other games where the devs had local players bring in their systems to debug a few critical items cause they did not have a testing system that fell within a specific parameter that was affecting a considerable amount of players. And there was no real option to "upgrade" the system except to purchase an entirely new system.

edit - I have not seen this mentioned, but during beta Nvidia users needed to turn off/disable Ambient Occlusion. Either set it up to turn it off Globally or setup MWO as a program that turns it off.

I can't believe PGI didn't take any of this into account then. It's just bad on a development point. I play in a gigantic TS server and there are so many of us with this problem running all different kind of systems. I mean i'm running my FX-8350 at 4.4GHz at the moment (i'm only on air cooling) and my friend who lives down the street from me has watercooling on his at 4.9GHz stable and he runs a GTX 980 and even he has the same problem. Then I have another friend who lives about 5 miles away and he runs an i7 with 2x GTX980 and he also has the same problem. GPU usage is so low.

The fact the game even runs 100% GPU in the beginning and end of the match goes to show something is missing here, because in game it just isn't the same. I just don't understand why supposedly this game wants so much CPU usage... I mean games like The Witcher, GTA V, etc. or anything else that is demanding still runs really well on AMD or Intel system or AMD or NVidia video card. It's just awkward.

[Goose]

Tarl Cabot, on 20 December 2015 - 08:34 PM, said:

edit - I have not seen this mentioned, but during beta Nvidia users needed to turn off/disable Ambient Occlusion. Either set it up to turn it off Globally or setup MWO as a program that turns it off.

That got fixed back with the GeForce driver 314, or there-abouts …

[Gabbatek]

this is insane. I have this problem too.

[Gabbatek]

I have this problem!!

My 980ti does exactly what the o.p says, even in sli both cards do it!

help meeeeee

[Gabbatek]

Found a fix for this.

Vsync was being forced on, even though it was off in the game!

[Star Dust]

I have this problem with a GTX 970, driver version 384.94 (and the previous version i was on) and with an i5 3570k (@ tested at both 3.8 and 4.2 Ghz).

In the testing grounds (I chose terra therma, seems to be the most load) there is an occasional stutter, but for the most part I get nearly constant 80 FPS (which is the limit I have on Riva). My GPU load varies from ~70-90%.

In an online match my GPU seldom gets loaded above 60% and my FPS (map dependent) ranges between 30-65 EXCEPT at the end of match screen, and beginning of match screens. Occasionally, but very seldom I will get very brief periods where GPU usage goes above 65, and at those same times my FPS improves.

Edited by raarz, 04 August 2017 - 08:13 PM.

[NARC BAIT]

the reason that you get a much more consistent result in the 'testing grounds' is because a lot of the game systems are not functioning, and generally speaking, the slowdowns are occouring from within those systems ... things like the battlegrid, network communication, and character animation ... sure the testing grounds might have 12 mechs in it to kill plus you, but none of them are moving unless you shoot them and there is basically no load generated from any of it ... if you really want to go through a lot of things, to possibly end up with some improvement, you should probably start a new thread ...