GPU Killing beast engines ..or... shady unskilled jr programmers?

[MisterMac]

Reading the Max Payne 3 thread got me thinking.

I'm not too into the world of direct game engine programming or graphics, i've programmed a "bit" on the serverside of mmo's and understand the need to design systems with throughput\simplicity in mind(Don't use no vectors if an array will do!).

But how do we, as geeky gamers find out and determine wether an awesome game like Crysis 1 - has a coded engine for using all the throughput of a cpu\gpu without bottlenecking on thread codepaths, for some subsystem.

Obviously there's different pragmatic tests.
There's the good ol eye too.

...what i'm getting at - how do we as gamers define a beast requiring game from a sloppy shitty console port?

Take the highest IQ game, compare new game and performance\IQ to it?

Crysis 1 was amazing for it's time.
Still sort of is.
But what if it just had the high IQ and would be a choppy game on

Would you still say Crysis 1 is an awesome engine job - if id still had problems maxing on a gtx 480 ? or gtx 280 for that matter.


Discuss!

 

[Jaydip]

Reading the Max Payne 3 thread got me thinking.

I'm not too into the world of direct game engine programming or graphics, i've programmed a "bit" on the serverside of mmo's and understand the need to design systems with throughput\simplicity in mind(Don't use no vectors if an array will do!).

But how do we, as geeky gamers find out and determine wether an awesome game like Crysis 1 - has a coded engine for using all the throughput of a cpu\gpu without bottlenecking on thread codepaths, for some subsystem.

Obviously there's different pragmatic tests.
There's the good ol eye too.

...what i'm getting at - how do we as gamers define a beast requiring game from a sloppy shitty console port?

Take the highest IQ game, compare new game and performance\IQ to it?

Crysis 1 was amazing for it's time.
Still sort of is.
But what if it just had the high IQ and would be a choppy game on

Would you still say Crysis 1 is an awesome engine job - if id still had problems maxing on a gtx 480 ? or gtx 280 for that matter.


Discuss!

I would provided its worth it.To me a beast is not just about eye candy but some innovation in game development as well.Preferably a beast should have good scalability while the ability to challenge even the most advanced hardware.

 

[Ancalagon44]

Given that Crysis 1 never came out for PC (or if it did, it was years later), its not a shitty console port. Crysis Warhead apparently ran a little better - this is likely due to engine optimization.

Considering what Crysis 1/Warhead looks like it, for such an "old" game, I think its a good job. They would have had some of the best guys in the industry looking at the code for months to optimize it, using an in house engine.

However, I've always felt (and I dont think I'm alone here) that the GTA games are always terrible ports. They always run badly on PC, compared to other games with similar graphics.

 

[LtMikePowelll]

Given that Crysis 1 never came out for PC (or if it did, it was years later), its not a shitty console port. Crysis Warhead apparently ran a little better - this is likely due to engine optimization.

Considering what Crysis 1/Warhead looks like it, for such an "old" game, I think its a good job. They would have had some of the best guys in the industry looking at the code for months to optimize it, using an in house engine.

However, I've always felt (and I dont think I'm alone here) that the GTA games are always terrible ports. They always run badly on PC, compared to other games with similar graphics.

Crysis 1 was release as PC only in November 7, 2007 http://en.wikipedia.org/wiki/Crysis

 

[toyota]

nothing has ever been as demanding as the beginning of Clear Sky on full max settings.

lol here is all my system can manage at the opening campsite.

1920x1080 ALL max setting including 4x AA

Min, Max, Avg
10, 27, 17.151

 

[BrightCandle]

There is no way to tell if something is programmed well without the source code, and a profiler. A badly written program still uses the CPU/GPU heavily, its just doing so at a lower FPS because the algorithm used isn't very efficient for the result.

Its not really about maxing out the resources, its using them well so that you achieve high IQ with the given resources. Its almost impossible for any console port to do so.

 

[Ancalagon44]

Gah I meant Crysis 1 never came out for consoles (did it?). My bad.

 

[toyota]

Gah I meant Crysis 1 never came out for consoles (did it?). My bad.

Crysis 1 later came out for the 360.

http://www.youtube.com/watch?v=xQ_1mcnH2yQ

 

[Ancalagon44]

Okay. Point is, it was written first with PC in mind, and ported to Xbox later. Not the other way around.

 

[BrightCandle]

Okay. Point is, it was written first with PC in mind, and ported to Xbox later. Not the other way around.

There have been plenty of games targeting the PC that were resource hogs for no good reason. There might be a correlation to console ports being resource hogs on the PC but it doesn't seem to apply universally as there are also console ports that have been very good.

I think we have to look at each game on its merits and determine how well its ported. Would be nice if a games review site told us about these sorts of concerns.

 

[3DVagabond]

It's much easier to make something that looks good when you are given unlimited resources. The real skill is to make it look good and not require $1000's of dollars worth of gaming PC to play it.

 

[BFG10K]

nothing has ever been as demanding as the beginning of Clear Sky on full max settings.

lol here is all my system can manage at the opening campsite.

1920x1080 ALL max setting including 4x AA

Min, Max, Avg
10, 27, 17.151

The campsite is not the most demanding place in the game. Places like the Red Forest are far more demanding.

Not to mention that even Clear Sky pales in comparison to Cryostasis. I run the latter on a GTX680 at 1920x1200 with no AA and everything on minimum (except texture quality) and it still drops into the low 20s in places.

 

[BFG10K]

There is no way to tell if something is programmed well without the source code, and a profiler. A badly written program still uses the CPU/GPU heavily, its just doing so at a lower FPS because the algorithm used isn't very efficient for the result.

Exactly; and to add to that, a taxing game doesn’t necessarily mean it’s badly programmed. During the Quake days it was a taxing game (Pentium minimum requirement), yet it had one of the finest examples of hand-optimized x86 assembly code ever created.

Another point of note is the implementation of features that have a low performance/IQ benefit ratio like ambient occlusion, DoF, motion blur, excessive tessellation and even some extreme AA modes. In general such features have a huge performance hit but provide very little visible gain in IQ in return.

Is the game badly programmed by including them? That’s more of a subjective opinion, but for me personally they’re the first to go if I need more performance.

 

[Arachnotronic]

I like the idea of including this awesome thing called "settings". So that people with insane hardware can turn on the settings with horrible IQ/performance and the more tempered folk can pick and choose what to disable.

PC <3

 

[Bobisuruncle54]

No game has ever had the luxury of "unlimited resources" at its disposal and real skill can only go so far with certain limits. And why is a $1000 gaming PC a bad requirement? Many PC gamers would be happy to see more titles push hardware again. The stagnant GPU market is due to the aging console generation and the lack of graphical progress over the last 3 years, there is nothing to drive progress.

 

[PrincessFrosty]

...what i'm getting at - how do we as gamers define a beast requiring game from a sloppy shitty console port?

It's a good question, I just ranted in the MP3 thread about this issue because it gets on my nerves reading about people complaining about optimisation in games where it has never been conclusively demonstrated optimisation is a problem.

For the most part, you can't, certainly gamers cannot. Part of the problem is that small differences in the fidelity of the calculation (graphics or simulation) can cause a huge difference in demand of processing power. Games are all about approximation both graphically and world simulation to save on performance, it's unfair for gamers to compare 2 game engines if one is modelling something more accurately than another.

If you want to prove something is unoptimised, some function lets say f(x)=y where we run a function f, on input x and get some result y, we need to demonstrate some code which executes faster and produces the same output y for all valid values of x. That is to say, the function needs to do the exact same thing but faster. That function could be calculating a new frame of physics simulation or AI routine, or whatever.

However, I've always felt (and I dont think I'm alone here) that the GTA games are always terrible ports. They always run badly on PC, compared to other games with similar graphics.

The observation that "They always run badly on PC, compared to other games with similar graphics." is a perfectly valid observation, at least specifically with GTA IV.

HOWEVER...it's more complicated than that, GTA IV was very demanding on the CPU and was the cause of most peoples bad frame rates, most of the graphical functions are executed on the video card and not the CPU so graphics didn't really have a lot to do with the performance issues.

The game is more demanding on the CPU that most people realise, there is real time physics and AI running on hundreds of objects in the scene, cars all drive around with real time physics effecting their suspension and movement, animation for pedestrians is blended in real time with physics collision using the euphoria engine which is a demanding thing to simulate. Pedestrians AI logic looks for things for them to grab on to when they're disturbed such as car door handles, and they try and right themselves as they tumble, the result of which is blending canned animation with real time physics to produce a type of "smarter" ragdol. The drunken states of Niko and other NPCs in the game are the best example of blending canned animations with real time physics.

Now you could argue that most people couldn't run GTA IV well at launch and I'd agree with that. You could argue that they aimed their min spec for the engine to be too high for the time it was released and I'd maybe agree with that as well. But if you want to talk about optimisation then you'll have to provide some kind of evidence that what they did could be significantly improved upon, and no one ever does this!

Ask anyone crying lack of optimisation what their reason for saying it is, and it's almost that it runs slowly. but so what? If you put a trillion explosive barrels into a map and ignite them all at once it runs slowly on any system, that's an issue of load, and not optimisation.

 

[ZeroRift]

I think the subjective elements definitely mask the underlying game engine performance, but that it is still possible to draw conclusions based off of approximate comparisons between engines.

Take, for instance, the Unreal Engine (version 3) and the CryEngine (version 2). UT3 was released ~1 year after Crysis 1. Both have approximately the same texture resolution / poly counts. Both have similar scene complexity (number of actors, objects, view distance, etc.). Both have similar AI complexity (UT3's AI may be slightly more complex, but it's hard to judge). Yet UT3 produces an order of magnitude more FPS than Crysis. From this, it seems clear that the Unreal Engine performs much better for a given task than the CryEngine. As long as the difference in detail levels is much smaller than the difference in performance, I think the comparison is valid.

Another thing to consider in this, however, is level design. I've seen a couple of Unreal Engine games (ME 1&2, Batman AC & AA) that do very poorly performance wise compared to their contemporary UT release. Unfortunately, this is not as straightforward, but does still need to be accounted for when comparing game performance.

 

[Grooveriding]

The entire argument that Crysis was unoptimized is really getting old. I am a complete laymen to the tech that goes into these engines, but that game is still arguably the best looking game that has ever been released. It's also five years old now.

That along with that just now do we have a game that is somewhat on its level in 2012 (BF3) that seems to have about the same hardware demands, says to me that Crysis was just that far ahead of its time.

I look forward to the next Crysis leap we get. I want a game that is practically photo realistic with new levels of visuals; that GTX680 SLI gets 20fps in and I have to wait four years to be able to play at 60fps to come out.

Just wondering if it will happen. It was Crytek who was pushing us forward, first with Far Cry then with Crysis. Now they have gone console. Crysis 3 has some cool new tech, but it looks a lot alike to Crysis 2. I still find the first Crysis looks better than the second, DX11 or not. With mods the original Crysis is pretty insane. It's also one of the few games that performs worse on my current setup vs my previous 480 tri-sli

 

[toyota]

The entire argument that Crysis was unoptimized is really getting old. I am a complete laymen to the tech that goes into these engines, but that game is still arguably the best looking game that has ever been released. It's also five years old now.

That along with that just now do we have a game that is somewhat on its level in 2012 (BF3) that seems to have about the same hardware demands, says to me that Crysis was just that far ahead of its time.

I look forward to the next Crysis leap we get. I want a game that is practically photo realistic with new levels of visuals; that GTX680 SLI gets 20fps in and I have to wait four years to be able to play at 60fps to come out.

Just wondering if it will happen. It was Crytek who was pushing us forward, first with Far Cry then with Crysis. Now they have gone console. Crysis 3 has some cool new tech, but it looks a lot alike to Crysis 2. I still find the first Crysis looks better than the second, DX11 or not. With mods the original Crysis is pretty insane. It's also one of the few games that performs worse on my current setup vs my previous 480 tri-sli

when modders can take the game and make it look better while also being less demanding then its pretty clear it was not well optimized.

 

[MisterMac]

when modders can take the game and make it look better while also being less demanding then its pretty clear it was not well optimized.

This.


IQ is 1 thing.
But when amateurs can improve by a significant percentage on IQ & Performance - i'd argue some things got a little less attention than they deserved.


The question is - why does every producing studio seem to settle?

Is it the engine creators? is it the studio?


Why can't we have both IQ & Performance?


Diablo 3 is a good example - decent little below average graphics.
However... it requires next to nothing to run.

Makes me think blizzard spent all those years on the engine, not the metagame or story

 

[2is]

I think the subjective elements definitely mask the underlying game engine performance, but that it is still possible to draw conclusions based off of approximate comparisons between engines.

Take, for instance, the Unreal Engine (version 3) and the CryEngine (version 2). UT3 was released ~1 year after Crysis 1. Both have approximately the same texture resolution / poly counts. Both have similar scene complexity (number of actors, objects, view distance, etc.). Both have similar AI complexity (UT3's AI may be slightly more complex, but it's hard to judge). Yet UT3 produces an order of magnitude more FPS than Crysis. From this, it seems clear that the Unreal Engine performs much better for a given task than the CryEngine. As long as the difference in detail levels is much smaller than the difference in performance, I think the comparison is valid.

Another thing to consider in this, however, is level design. I've seen a couple of Unreal Engine games (ME 1&2, Batman AC & AA) that do very poorly performance wise compared to their contemporary UT release. Unfortunately, this is not as straightforward, but does still need to be accounted for when comparing game performance.

Except that UT3 at high looks like Crysis at Medium, on a map that doesn't particularly look great.

 

[Grooveriding]

when modders can take the game and make it look better while also being less demanding then its pretty clear it was not well optimized.

Eh, the only mod I know of that did that was the one that enabled some features under DX9 that were only available in DX10 with the default game.

The mods I have on Crysis make it more demanding, but look a lot better.

 

[bononos]

.....
Why can't we have both IQ & Performance?

Diablo 3 is a good example - decent little below average graphics.
However... it requires next to nothing to run.

Makes me think blizzard spent all those years on the engine, not the metagame or story

You can't possibly compare D3 with its top down isometric view with Unreal/Crysis/Witcher2. Isometric view and parallax scrolling type games are going to be easier on hardware requirements.

 

[Ben90]

It still blows my mind how the majority of games released today have problems with 1:1 mouse input. Hell, even the once coveted BF3 had negative acceleration on release (I don't know if they fixed it)

 

[MisterMac]

You can't possibly compare D3 with its top down isometric view with Unreal/Crysis/Witcher2. Isometric view and parallax scrolling type games are going to be easier on hardware requirements.

I suppose the vectorization\rendering of the view is a factor.
But you have to admit, the requirements are extremely low for the IQ D3 provides.