Same framerate on AMD 6970 & GTX 680. How come?

[autorotation 1]

I recently upgraded my XFX 6970 to Nvidia GTX 680 for two reasons; better framerates and less noise. In other games, like Battlefield 3, the GTX is averaging about twice the framerate as my overclocked 6970 did, with 1/3 of the noise.

But in the game I play and enjoy the most, ToH, the framerate remained almost identical, regardless of overclocking or standard 680, with the exact same settings.

Incredibly, the minimum framerate actually dropped compared to my 6970, albeit with only a few frames. The maximum framerate jumped from 94 to 147, but average came in at 78 compared to 72 in favour of the GTX.

So I have to wonder; Is the engine THAT poor in ToH? And, is there something you developers can do about it to optimize it? Because, frankly, it is a little more than a little bit silly to run a game-engine uncapable of utilizing more than 2 threads in the 4 thread/core CPU's in a modern game.

I feel a bit cheated, to say the least, seeing as ToH is the main reason I even have a computer.

I know it's not my system that's keeping the performance down, but here's the set-up anyways:

Asus Maximus IV Extreme b3

8 gb Corsair Dominator 1600mhz

Intel 2600k @ 4,6 ghz

Corsair SSD Force Series 3, 240GB

[.kju 3239]

You have to recognize and accept that a simulation mainly needs CPU power (and related tech like RAM, HDD access speed, etc).

This is this the limiting factor in your case it seems.

Multicore is very hard to do. People main split unrelated parts to different threats, yet all that needs sync is still often best done on one CPU/threat.

[jedra 11]

Take a step back. An average framerate of 78 is very good. Regardless of your disappointment in the relatively small increase you are running the game very nicely.

There comes a point with most software where throwing more hardware at it does not make a difference. The engine is and always has been CPU heavy and as PvPScene said, you can't just switch a flag to make it multithreaded. I don't know what approach they are taking with Arma 3 with regard to this (it's a new engine after all), but they would hopefully have been listening these past few years! Then maybe in the future the Take On series will get an engine update too.

In the meantime if you are bottlenecked at nearly 80fps I can only ever be jealous!

[autorotation 1]

I understand that a simulation is CPU heavy. Which is why one should expect that a new simulator would actually be able to use multithreaded CPUs, which are basically the norm among gamers today.

And while an average framerate of 78 is exceptionally good, it doesn't really say anything unless you're comparing it to the actual graphical settings you have the game running at. And these settings are what I ran the benchmark on;

Video Memory: Very high

Texture: Very high

Objects: Very high

HDR quality: Very high

Visibility: 3515

View: 2037

Shadow view: 100

Traffic: 921

Antialiasing: Disabled

Terrain: Very high

Shadow: high

Antisotropic: Disabled

Postprocess: Disabled

Cloud: Normal

I would be very interested to see other people using the same settings, or give me their settings and framerate numbers so we can compare and see how much hardware really affect the game.

[Derbysieger 11]

Did you have problems with Video Memory on Default or why is it on Very High? Usually Default setting will give you the best performance if there aren't any problems with it (default generally uses video memory better than very high).

BTW I wouldn't want to play TOH with only 3500m VD. I'm using 12000m on South Asia, ~8000m on Seattle and ~6000m on Chernarus. Obviously I never get framerates as high as you. usually around 30fps on Seattle, 40fps on SA and 25-30fps on Chernarus.

Edited March 29, 2012 by Derbysieger

[royaltyinexile 175]

I recently upgraded my XFX 6970 to Nvidia GTX 680 [...] but average came in at 78 compared to 72 in favour of the GTX.

So I have to wonder; Is the engine THAT poor in ToH? And, is there something you developers can do about it to optimize it?

As other players have noted, performance depends upon many factors. To answer your question simply, the same framerate may be due to a bottleneck in other aspects of the hardware; namely: CPU and HDD. Tweaking your video settings to get the performance you're most satisfied with is an important, but sometimes complex task. :)

Another factor may be that we're not exploiting all of the power of this card or your hardware efficiently enough. Given the diverse and shifting base of hardware that PC devs must work with, and the legacy of the Real Virtuality (RV) engine, it's a tough challenge; yet, if improvements can be made, the support offered by BI in the past should tell you we'd try our best to squeeze more performance out.

While certainly there's always more work that can be done to improve performance, the idea of 'performance' itself can very wildly depending upon the user expectations. The beauty of the RV engine and its unique qualities lie in its sheer scale. Comparing frame-rates and quality settings with BF3 can lead to a false impression, or rather, an impression that is limited to a small number of criteria; therefore, it may be reductive to label an engine as 'poor' without considering all of its strengths.

TcHfNAUExH0

Here's an example of over 100 helicopters flying in the 60x60km Seattle map at around 0530. The engine scales to handle all of this additional complexity, and it's achieved in 5 minutes from the editor. While that doesn't mean much when you just want to fly your helicopter on your own at the best possibly frame-rate and settings, it does demonstrate that the engine is enormously powerful in other areas, and 'performance' might be considered relative to player expectation and usefulness, a product of how you want to use the platform.

Best,

RiE

[Mr. Charles 22]

Next time you do such a video, look out for exploding Helicopters ;)

[autorotation 1]

Did you have problems with Video Memory on Default or why is it on Very High? Usually Default setting will give you the best performance if there aren't any problems with it (default generally uses video memory better than very high).

BTW I wouldn't want to play TOH with only 3500m VD. I'm using 12000m on South Asia, ~8000m on Seattle and ~6000m on Chernarus. Obviously I never get framerates as high as you. usually around 30fps on Seattle, 40fps on SA and 25-30fps on Chernarus.

No problems with Video Memory on default, but the framerate drops around 10 frps on default.

I don't want to play at those settings, which was my original point ;)

I've discovered that I can drastically increase settings without shutting the framerate completely down. With visibility set to 10 056 and object view to 6172, I got an average of 47, with 26 minimum and 93 max.

The problem for me is that even at those lowest settings mentioned above this post, Chernarus is basically unplayable as a sim due to framerate drops to around 20-25 frps around the main city in the south. And I want to fly at Chernarus at a smooth framerate AND be able to see where I'm going while I'm doing it :cheers:

As other players have noted, performance depends upon many factors. To answer your question simply, the same framerate may be due to a bottleneck in other aspects of the hardware; namely: CPU and HDD. Tweaking your video settings to get the performance you're most satisfied with is an important, but sometimes complex task. :)

Another factor may be that we're not exploiting all of the power of this card or your hardware efficiently enough. Given the diverse and shifting base of hardware that PC devs must work with, and the legacy of the Real Virtuality (RV) engine, it's a tough challenge; yet, if improvements can be made, the support offered by BI in the past should tell you we'd try our best to squeeze more performance out.

I understand the brain-ackingly difficult issue you as producers have when you have to develop something that has to be compatible with thousands of different set-ups. But I don't understand why one would ignore the fact that most pc gaming platforms out there today has more than two-thread-processors. Perhaps it isn't ignored, but nevertheless it's not utilized. In such a CPU heavy game, I cannot for the life of me figure out why such a call was made.

I do know though that BI is class-leading in aftermarket support and optimizing, which is why you have faithful followers that are willing to give you time to update, expand and improve on your games. It is also why I write instead of ditching the game.

While certainly there's always more work that can be done to improve performance, the idea of 'performance' itself can very wildly depending upon the user expectations. The beauty of the RV engine and its unique qualities lie in its sheer scale. Comparing frame-rates and quality settings with BF3 can lead to a false impression, or rather, an impression that is limited to a small number of criteria; therefore, it may be reductive to label an engine as 'poor' without considering all of its strengths.

Here's an example of over 100 helicopters flying in the 60x60km Seattle map at around 0530. The engine scales to handle all of this additional complexity, and it's achieved in 5 minutes from the editor. While that doesn't mean much when you just want to fly your helicopter on your own at the best possibly frame-rate and settings, it does demonstrate that the engine is enormously powerful in other areas, and 'performance' might be considered relative to player expectation and usefulness, a product of how you want to use the platform.

Best,

RiE

The comparison to battlefield was only to demonstrate the difference in performance scaling with new vs old hardware between the two games. And 90% better scaling performance vs 5% scaling performance indicates that something is "wrong", if I may use that expression. And watching the load on the CPU, I can see that it's not being utilized much compared to its potential, which means the game(engine) is badly optimized for such hardware. Or maybe I'm missing something?

What I'm really wondering is if there's a possibility for BI to tweak the engine in such a way as to make it more hardware-friendly? It seems like there's a performance-roof, where you get to this point where faster graphic-card doesn't matter too much. It could of course be some kind of driver issue or some other fancy thing that I don't know anything about.

I will continue to test though, to see if there's a significant difference to be made somewhere:cool:

[royaltyinexile 175]

And watching the load on the CPU, I can see that it's not being utilized much compared to its potential, which means the game(engine) is badly optimized for such hardware. Or maybe I'm missing something?

I understand your point is more complex then perhaps this (admittedly old) article covers, but you might find it an interesting read: http://www.bistudio.com/english/company/developers-blog/91-real-virtuality-going-multicore

Next time you do such a video, look out for exploding Helicopters ;)

Nah, I went a whole year making a civilian game. The time to blow sh*t up is nigh!

Best,

RiE

[f00 10]

*removed*

Edited March 30, 2012 by f00
wrong section

[ericinexile 1]

RiE,

Can u post that mission? I'm still learning the editor and the camera part is still mystery to me. I'd like to see how my iMac handles it.

[royaltyinexile 175]

Can u post that mission? I'm still learning the editor and the camera part is still mystery to me. I'd like to see how my iMac handles it.

That scene didn't actually have a camera - I was in the 3D world as Joe Larkin; the smooth panning is a product of using an XBox Controller!

Best,

RiE

[Groove_C 267]

I've read all the posts above and I'm wondering how some people come to say that Real Virtuality 3 engine don't make usage of more than 2 threads without any proof of that!!! :391:

Shame on you!!!

Some of you here should google a little bit befor posting such observations that would confuse or misslead other people reading this sh*t!!! :936:

I'm playing games based on this engine since 2001 (OFP) and this are the only games I've ever played (OFP/Arma/ArmA2).

Those are my prefered and that's why I'm at least following all the log changes applied through all the patches out there.

And I can tell you that BIS guys have always done their best to optimize the engine as much as possible depending on ressources (people/time) availble!

Don't forget that such studios like those working on BF 3 or CoD MW 3 have hundreds of people working on those arcade FPS shooters and are not even 1 millimeter close to what have been acheaved by such small Bohemia Interactive Studios in sense of game envirement scale and scripting (actions/scenarios) possibilities and perspectives!!! !!! !!!

Here you can take a look on the latest test using GTX 580 @ default to see how Real Virtuality 3 engine scales on more cores and more frequency!

@640x480 10.000 m view distance

http://www.computerbase.de/artikel/prozessoren/2011/test-intel-sandy-bridge-e/41/#abschnitt_arma_2_operation_arrowhead

@1680x1050 10.000 m view distance

http://www.computerbase.de/artikel/prozessoren/2011/test-intel-sandy-bridge-e/49/#abschnitt_arma_2_operation_arrowhead

[autorotation 1]

I've done some more testing, and I've found that this game is quite interesting in terms of scaling. In the first post I was using very low settings compared to what one should play at, mainly because those settings were pretty much the maximum I could afford to use while playing Chernarus at an acceptable framerate (20 frps at minimum).

It turned out that the difference between the XFX 6970 and the GTX 680 were negligible in the benchmark-test. While the numbers itself (min 43, average 71, max 94 for 6970 compared to 40, 78 and 147 for 680) wasn't bad, one should keep in mind that the settings are basically unplayable due to lack of viewdistance and no anti-aliasing, making the game look like flight simulator 98.

This negligible difference between the two cards made me quite disgusted with the game because I literally upgraded the graphics card to improve visual performance in this game alone.

But since then I've done some more testing. It turns out that drastically increasing some of the settings, the GTX 680 performs admirably compared to the 6970. It should however be noted that an overclock of my CPU to 4,6ghz is also a factor.

With these settings:

Visibility: 6780

Objectv: 5268

Shadow: 197

Traffic: 399

FXAA: high

Videomem: default

Texture: Very high

Object: High

HDR: normal

PiP: Dis

Anti-aliasing: Very high

Terrain: Normal

Shadow: Normal

Postpro: Dis

Cloud: Low

3d resolution: 100%

I get these numbers for my 680, albeit overclocked to 1200mhz core, 3086 memory: Min; 36, Average; 58, max; 105.

While I don't have the numbers for the 6970 at these settings, I can say that the 6970 was hit pretty hard while toying with the anti-aliasing settings.

Right now the game looks absolutely gorgeous in Seattle, with the rough edges completely gone and the nasty fog-like slur over the screen disappeared when turning the FXAA to high, making the colors seem clear and crisp.

I could of course increase some settings to make it look even better, but I prefer to fly around downtown with around 50 fps because fluidity while flying is the most important factor to me.

I have to point out though that I can't possibly keep these settings for Chernarus, which still bugs me. But I have a feeling that the imported ARMA map isn't optimized for TOKH. What bugs me the most about having to reduce the viewdistance is the return of the fog-like slur. The vibrant, strong colors of the beautiful island disappears in a grey mist when the viewdistance drops below 4000m.

So I have three questions for the developers;

1. Is it possible to reduce the grey mist effect in a future update, because it ruins the beautiful potential of the game. I know it's there to make the transition between what you can see and what you can't more fluid, but the loss of strong and vibrant colors is greater than the need for ultra-smooth transition.

2. Is it possible to reduce the glass-effect of the cockpit bubble? While a nice detail, it actually reduces he visual quality so much so that changing viewpoint to outside the helicopter reveals almost a different world graphics-wise.

3. When should I expect the update that includes TOKH's official Chernarus?

I would also point out that I would gladly (and THE ROCK means GLADLY) pay 50 dollars to be able to use a map-making editor! You know, the kind where you can insert buildings, roads, trees and elevation/mountains in an easy manner.

If this could be done, the community would wet themselves me thinks :bounce3:

[liquidpinky 11]

I would also point out that I would gladly (and THE ROCK means GLADLY) pay 50 dollars to be able to use a map-making editor! You know, the kind where you can insert buildings, roads, trees and elevation/mountains in an easy manner.

If this could be done, the community would wet themselves me thinks :bounce3:

You can actually already use the RTE with Take On, with some cheeky tricks during the setup.

Juts lie to it when asking for the A2 install and exe locations and point it to the Take On exe instead, it works a treat in Take On although sometimes the UI may look a little off.

Will be even better when you add A2 content officially, although I have it added as mods at the moment.

Airfield I am working on at the moment with it:

[autorotation 1]

Pardon moi for being a noobster, but what is RTE and, more importantly, how can I get it? :D

[liquidpinky 11]

Look for RTE (Real Time Editor) on Armaholic.

Sorry took so long to reply, playing a bit Carrier Command at the moment.

[Wurnman 1]

I have my settings at fairly high, but i have noticed whilst doing the tutorial that my teacher next to me his baseball cap would start off looking good but later into our training it appears almost soup like (washed out). When i restart the mission or the game it then goes back to normal. I have installed patch 1.05. Any ideas as to whey this would happen? My pc specs are:

Quad core AMD

6gb DDR 3 ram

Nvidia GTX 560Ti Top card

[jumpinghubert 48]

I

Video Memory: Very high

Texture: Very high

Objects: Very high

HDR quality: Very high

Visibility: 3515

View: 2037

Shadow view: 100

Traffic: 921

Antialiasing: Disabled

Terrain: Very high

Shadow: high

Antisotropic: Disabled

Postprocess: Disabled

Cloud: Normal

you have disabled the most gpu-heavy settings: postprocess, antialiasing, anisotropic filter.....try to log the gpu-usage with afterburner to find the ideal gfx-settings.

Edited June 12, 2012 by JumpingHubert