UT2004, Far Cry - CPU and VPU (nv40, R420)

[hovz]

typo, i meant 40. and pretyt much every server is 10 people, so the closest he can get to that is 9 bots.

 

[Ostsol]

also dropped into the 30s much less than ut did. even tho its far more advanced.

Didn't we already ruled out GPU stress as a cause for framerate drop? As such, that FarCry has more advanced effects is irrelevant.

also the far cry chart was MUCH less jumpy than the ut one.

Both are equally jumpy. You might notice that the UT2004 bench's running time is close to twice the length of the FarCry benches.

also what are the specs of ur system, i dont usually read beyond3d reviews.

There's a link to reviews at the top of this page. . . Because I'm so nice, here's a direct link.

 

[hovz]

also dropped into the 30s much less than ut did. even tho its far more advanced.

Didn't we already ruled out GPU stress as a cause for framerate drop? As such, that FarCry has more advanced effects is irrelevant.

also the far cry chart was MUCH less jumpy than the ut one.

Both are equally jumpy. You might notice that the UT2004 bench's running time is close to twice the length of the FarCry benches.

also what are the specs of ur system, i dont usually read beyond3d reviews.

There's a link to reviews at the top of this page. . . Because I'm so nice, here's a direct link.

its processing more physics too, not just effects. also theres far more objects on screne that have to be sent to the cpu. and this is just a guess, but id say the ai in far cry is also more advanced since the enemies have to be aware of all their surroundings to be able to duck for cover behind the objects around them. so even on a 3 ghz systemwith top of the line video card, extremely frequent drops to the 30s which will happen with every setting at its lowest. thats not what id call a good engine.

 

[Ostsol]

its processing more physics too, not just effects.

Like what?

also theres far more objects on screne that have to be sent to the cpu.

Most objects are static and merely have to rendered in place and maybe collided with active objects in the immediate area. Scene complexity can be argued in favour of UT2004 as well. Not only might there be static meshes all over the place, but also projectiles flying around.

and this is just a guess, but id say the ai in far cry is also more advanced since the enemies have to be aware of all their surroundings to be able to duck for cover behind the objects around them.

There's two ways that this can be accomplished. Either there's a navigation point behind the object (marking a position from which the object can be used as cover) or the object has some parameters that dictate that it can be used as cover and the distance to stand from it. Neither of which are particularily new or advanced.

One can also argue that UT2004 bots' dodging and manoevering is equivalent in terms of decision making complexity, and also more of a load on physics. While a FarCry character is standing in place behind cover it does not have to have collision tests with static objects, but a UT2004 bots' jumping and dodging requires constant collision detection with everything in the immediate area.

so even on a 3 ghz systemwith top of the line video card, extremely frequent drops to the 30s which will happen with every setting at its lowest. thats not what id call a good engine.

Even assuming both are doing the same amount of work or that FarCry is doing more, I'll once again state that the two engines may have content coded entirely differently. I know that FarCry uses Lua for mission scripting, but I don't know what it uses for defining other game-specific content and logic. If it uses C++ then it will certainly be inherently faster than the Unreal Engine, which uses a scripting language interpretted by a virtual machine.

 

[hovz]

its processing more physics too, not just effects.

Like what?

also theres far more objects on screne that have to be sent to the cpu.

Most objects are static and merely have to rendered in place and maybe collided with active objects in the immediate area. Scene complexity can be argued in favour of UT2004 as well. Not only might there be static meshes all over the place, but also projectiles flying around.

and this is just a guess, but id say the ai in far cry is also more advanced since the enemies have to be aware of all their surroundings to be able to duck for cover behind the objects around them.

There's two ways that this can be accomplished. Either there's a navigation point behind the object (marking a position from which the object can be used as cover) or the object has some parameters that dictate that it can be used as cover and the distance to stand from it. Neither of which are particularily new or advanced.

One can also argue that UT2004 bots' dodging and manoevering is equivalent in terms of decision making complexity, and also more of a load on physics. While a FarCry character is standing in place behind cover it does not have to have collision tests with static objects, but a UT2004 bots' jumping and dodging requires constant collision detection with everything in the immediate area.

so even on a 3 ghz systemwith top of the line video card, extremely frequent drops to the 30s which will happen with every setting at its lowest. thats not what id call a good engine.

Even assuming both are doing the same amount of work or that FarCry is doing more, I'll once again state that the two engines may have content coded entirely differently. I know that FarCry uses Lua for mission scripting, but I don't know what it uses for defining other game-specific content and logic. If it uses C++ then it will certainly be inherently faster than the Unreal Engine, which uses a scripting language interpretted by a virtual machine.

dont even compare the scene compelxity of ut to far cry...lol......even if they are static, which half of them arent, especially in indoor levels, they still have to be sent to the cpu to register the according collision area. ut doesnt even compare in this aspect. theres what, a few poles or stands here and there?

as far as physics go you can interact with about half of the objects on screen, they arent all just static crates or barrels. the enemies have hit detection, they arent just a square hit box. the water actually has physics of water when somehting falls into it. the guards use a form of ragdoll physics at all times, not just when they die. the guns accurracy changes dynamically. the calculations done to make the sound and visibility in regards to stealth work.

and why do i as an end user care if its coded in c++ or scripting? the fact is the performance is not what you can classify as good for such an aged game on a top end system. as an end user i dont care about why they chose to code in a slower language, all i know is that the performance is lacking, and there is NO WAY to remedy it. in far cry when u turn down details ur lows and highs go up, not just ur highs.

 

[Ostsol]

dont even compare the scene compelxity of ut to far cry...

If I had UT2003 and UnrealEd installed, I'd comment on this.

lol......even if they are static, which half of them arent, especially in indoor levels, they still have to be sent to the cpu to register the according collision area. ut doesnt even compare in this aspect. theres what, a few poles or stands here and there?

as far as physics go you can interact with about half of the objects on screen, they arent all just static crates or barrels. the enemies have hit detection, they arent just a square hit box. the water actually has physics of water when somehting falls into it. the guards use a form of ragdoll physics at all times, not just when they die. the guns accurracy changes dynamically. the calculations done to make the sound and visibility in regards to stealth work.

Depending on how scene partitioning is accomplished, most objects can be discarded for collision detection. At the most basic level, only objects in a specific partition of the map (or partitions if the object is straddling more than one) are compared. Even then it's only dynamic (read: moving or with a changing collision hull) that are checked against other dynamic and static objects. Also, just because an object is dynamic doesn't mean that it needs to check against other objects for collision. If it is not moving, it's treated as static.

Also, collision between meshes is probably always at first performed using the most primative system possible. Usually this is a collision sphere or cylinder. When collisions are detected within this primative, the more complex collision hull is used to confirm whether or not if a collision really took place and in some cases to find out exactly where. As such, even FarCry's models probably use something even more primative than "a square hit box" -- at least at first. After initial collision is resolved, then collision is resolved with the more accurate collision volume (which almost always has much less polygons than the original model) or major-bones in the case of a character model. It certainly does not perform per-poly collision detection -- no game does (unless it's entirely made up of a BSP tree, like maps in the old Unreal Engine or some Quake engines could be).

Another good point is that UT2004 also detects collision with character models on a per major-bone basis. This can be noted not only with the ragdoll effect, but in the gore produced. Sometimes you can blast off only an arm or a leg or a head, for example (still killing the enemy, of course).

and why do i as an end user care if its coded in c++ or scripting? the fact is the performance is not what you can classify as good for such an aged game on a top end system. as an end user i dont care about why they chose to code in a slower language, all i know is that the performance is lacking, and there is NO WAY to remedy it. in far cry when u turn down details ur lows and highs go up, not just ur highs.

Then why do you claim that the Unreal Engine is badly programmed? As an end user all you care about is performance, so why bother to try and explain it as bad programming? Since you said it was, I am simply trying to explain that the engine is designed in such a way that may make it inherently slower in some areas. This does not make it badly programmed at all. It is merely a design decision.

 

[hovz]

if a design decision gives the end user worse performance, then as far as we go how is it not programmed badly? it certainly isnt giving us advanced features or options to negate the bad performance

and even if it is as simple as you say, the fatc is far cry is still doing much more of these simple calculations than ut. dnyamic objects still take more overhead than static ones. 100 trees, ruins of buildings, rocks, buildings, tents etc. still take more overhead then walls and platforms. and simplified hit detection is better then none at all. the only thing you can shoot off specifically is heads. the rest is completely random gore effects.

 

[Heathen]

Where's the Ignore button when you desperately need it?

 

[London Geezer]

Where's the Ignore button when you desperately need it?

There is one, it's called the don't read his posts button.

 

[Ostsol]

if a design decision gives the end user worse performance, then as far as we go how is it not programmed badly? it certainly isnt giving us advanced features or options to negate the bad performance

Epic is in the buisness of making 3d graphics engines that provide developers with a good framework to make games. This framework includes utilities to aid in making the development and integration of content relatively quick and easy. UnrealScript is such a utility. Technically, if a developer wanted they could code everything in C++, practically ignoring UnrealScript. After all, when they license the engine they also get the full source-code.

In any case, design is a totally separate issue from programming.

and even if it is as simple as you say, the fatc is far cry is still doing much more of these simple calculations than ut. dnyamic objects still take more overhead than static ones. 100 trees, ruins of buildings, rocks, buildings, tents etc. still take more overhead then walls and platforms. and simplified hit detection is better then none at all. the only thing you can shoot off specifically is heads. the rest is completely random gore effects.

As I said, if those 100 trees, ruins of buildings, rocks, buildings, and tents are not moving, they are no different than static objects. Physics does not even touch any of those objects until a moving dynamic object is in the same partition.

 

[hovz]

even if, its still doing alot more work than ut. after all, as you said earlier, every shot has to be traced compeltely to detect whether or not it will collide with an object. all those details in the levels have to be ccounted for when shots are flying everywhere

 

[Colourless]

even if, its still doing alot more work than ut. after all, as you said earlier, every shot has to be traced compeltely to detect whether or not it will collide with an object. all those details in the levels have to be ccounted for when shots are flying everywhere

Actually they don't, if the engine is designed well. You'll store the objects in a way that you can grossly cull them when doing traces so you don't have to check against every object.

 

[hovz]

ok so then ut would be doing it the same way if it was designed well. which would still mean far cry is keeping track of much more culled objects. i dont even see how you guys can argue that far cry isnt doing much more wortk than ut.

 

[see colon]

i dont even see how you guys can argue that far cry isnt doing much more wortk than ut.

i don't see how you can argue that farcry is doing more work, given your previous arguments. either...
a)ut2004 is poorly coded and suffers in performance (because it is doing more work behind the scenes)
or
b)farcry is coded well and has consistant performance (because it only does what it needs to do behind the scenes)

you can't have farcry doing more work and running better, it just doesn't make sense. even if the work UT2004 is doing isn't seen, isn't used by the game (remeber the engine was designed first to be a licensable product), or is a result of poor coding requiring the engine to do more work to achieve similar results, it's still doing more work.

farcry is pretty, but UT2004 does some things that farcry does not. for example...

1.) there is no physics interaction with the water in farcry (boats, people walking through water, ect don't create real wake, only a texture) but some maps in ut do.

2.) draw distance for objects is longer in UT2004, especialy noticable in onslaught maps. notice in farcry that people fade in and out in the distance while you can see someone running as a tiny speck in ut2004. (a nice optimization, i might add, but more proof that farcry is doing less work)

3.) weapon and vehichle variation is higher in UT2004. vehiches can be functionaly damaged (breaking the blades off the scorpion, for example) instead of just destroyed. farcry has pretty much 3 types of projectiles (bullets, rockets, and grenages), UT has plenty more. (again, more variation means more work. think of the physics calcs needed to follow grenades, rockets, flack and flack balls around when they are bouncing through a hallway, all while someone is poaching in a corner shock combo-ing)

4.) UT2004 has built in voice chat and voice recognition including voice commands to bots. farcry does not. not to mention text to speech.

5.) UT2004 has dolby digital support for competent hardware.

6.) UT2004 has tactile force feedback on competent hardware

7.) ut2004 has more variation in bot ai, wich is required for the variety of game types. bots actualy attempt to accomplish objective, not just "get him" and hide behind a box.

8.) texture detail is higher in many cases in UT2004 (farcry has some pretty shit textures here and there, and they really stand out)

9.) UT has destructable character models. arms, legs, heards, and torsos can all be destroyed, in any combination. you can use as many explosives as you want in farcry, the body will still be intact.

10.) UT has more partical effects, most of them controlled by the physics engine (turning down physics detail will remove may of them). while i'm talking about physics, ut2004 has better ragdoll effects. several times in farcry i've killed people, ususaly with hedshots, and they die but remain standing, or fall to their knees and remain upright. another shot knocks them over, but it's not exactly accurate.

and if farcry is so well programed why do i take a 15-20FPS hit when i turn on the flashlight inside a building?
c:

 

[hovz]

i dont even see how you guys can argue that far cry isnt doing much more wortk than ut.

i don't see how you can argue that farcry is doing more work, given your previous arguments. either...
a)ut2004 is poorly coded and suffers in performance (because it is doing more work behind the scenes)
or
b)farcry is coded well and has consistant performance (because it only does what it needs to do behind the scenes)

you can't have farcry doing more work and running better, it just doesn't make sense. even if the work UT2004 is doing isn't seen, isn't used by the game (remeber the engine was designed first to be a licensable product), or is a result of poor coding requiring the engine to do more work to achieve similar results, it's still doing more work.

farcry is pretty, but UT2004 does some things that farcry does not. for example...

1.) there is no physics interaction with the water in farcry (boats, people walking through water, ect don't create real wake, only a texture) but some maps in ut do.

2.) draw distance for objects is longer in UT2004, especialy noticable in onslaught maps. notice in farcry that people fade in and out in the distance while you can see someone running as a tiny speck in ut2004. (a nice optimization, i might add, but more proof that farcry is doing less work)

3.) weapon and vehichle variation is higher in UT2004. vehiches can be functionaly damaged (breaking the blades off the scorpion, for example) instead of just destroyed. farcry has pretty much 3 types of projectiles (bullets, rockets, and grenages), UT has plenty more. (again, more variation means more work. think of the physics calcs needed to follow grenades, rockets, flack and flack balls around when they are bouncing through a hallway, all while someone is poaching in a corner shock combo-ing)

4.) UT2004 has built in voice chat and voice recognition including voice commands to bots. farcry does not. not to mention text to speech.

5.) UT2004 has dolby digital support for competent hardware.

6.) UT2004 has tactile force feedback on competent hardware

7.) ut2004 has more variation in bot ai, wich is required for the variety of game types. bots actualy attempt to accomplish objective, not just "get him" and hide behind a box.

8.) texture detail is higher in many cases in UT2004 (farcry has some pretty shit textures here and there, and they really stand out)

9.) UT has destructable character models. arms, legs, heards, and torsos can all be destroyed, in any combination. you can use as many explosives as you want in farcry, the body will still be intact.

10.) UT has more partical effects, most of them controlled by the physics engine (turning down physics detail will remove may of them). while i'm talking about physics, ut2004 has better ragdoll effects. several times in farcry i've killed people, ususaly with hedshots, and they die but remain standing, or fall to their knees and remain upright. another shot knocks them over, but it's not exactly accurate.

and if farcry is so well programed why do i take a 15-20FPS hit when i turn on the flashlight inside a building?
c:

1) if ti is indeed doign work it doesnt have to, then it IS poorly coded

1) the water in ut onlyreacts to the flak gun primary if im correct. peopel jumping in or grenades that land in dont have any effect.

2) thats probly because the draw distance for ut is 1/4th of far cry. of course there no fading in of small objects, the entire sceen is just fogged. btw enemies dont fade out in far cry, just small bushes and other objects.

3) the vehicle damage isnt dynamic at all, its pre animated and calculated. the only vehicle that exhibits damage is the scorpian, and only in 1 instance. hardly an arguement. far cyr has grenades too, it also has to factor in the sight range of the flashbang grenade. also the entire sound field has to be monitored to alert guards to your presence.

4) i turn them off, again if they are still doing all the backwork of it with it turned off, thats just another example of bad coding.

5) see 4

6) see 4

7) not applicable to online matches

8) purely a video card feature, dotn even try to compare video effects of ut to far cry

9) farcrys hitscan and reaction system balances this out

10) just plain wrong

11) purely a video card feature. i believe the flashlight uses 2.0 per pixel lighting? again a pretty advanced feature of an already taxing engine.

 

[AlNom]

1) the water in ut onlyreacts to the flak gun primary if im correct. peopel jumping in or grenades that land in dont have any effect.

incorrect. fact.

7) not applicable to online matches

alright.... turn off bots and UT should run very well. It does for me. (edit: this applies to all 4 computers I have)

9) farcrys hitscan and reaction system balances this out

how do these balance if they're not equivalent to gibbing?

10) just plain wrong

do particles react to projectiles in farcry?

11) purely a video card feature.

Sure, the whole game is a video card feature. Let's just turn all graphical features to lowest and run UT without bots. Now do the same for farcry. Welcome to 1999.

In fact, let's shut off the monitor. It's purely a video card feature to have this on. Now we have real time black screen with an almost infinite frames per second.

 

[hovz]

thing is, i could turn every option to the lowest, hop in an online match and still get the frequent drops to 30s.

he said ut has MORE particle effects than ut. he never mentioned weapons affecting them. which only the rl does anyway, and only in specialized areas.

even with no bots it still drosp to 30 frequently in online games.

checking to see which 'bone' was hit and giving the approriate reaction is similar in complexity to gibbing off a body part

oh and way to take the video card statement completely out of context genius

 

[AlNom]

well, there must be something wrong with your computer because I hardly drop below 30fps online (not including onslaught and assault with 16 people going at it on screen though ).

oh and way to take the video card statement completely out of context genius

edit:

 

[hovz]

in the 30s, not below 30

 

[CMAN]

What are you talking about hovz? UT2004 runs much smoother than Far Cry for me.

P4 3.0 Ghz, 9700 Pro, 512 MB RAM, Audigy 2

I max out all settings and set resolution at either 1024x768 or 1280x1024. I also use EAX 3.0.

I barely notice slow downs in UT2004, I'm sure they are there, but I can't see them as much as I do in Far Cry (due to shader effects and other graphics effects not present in UT). Even if UT2004 did slow down more than Far Cry, it's still a hugely fun game to play and I give congrats out to Epic for programming the hell out of the game and making it one of the funnest games I've ever played. Then again, Far Cry is hugely fun also, but in a different way.