RTS Starter Kit Questions
by Ubutopia · in Torque Game Engine · 02/03/2005 (7:57 am) · 8 replies
Is there an downloadable demo available?
How about a higher resolution demo movie?
The RTS kit is very interesting to me, but I've got quite a few questions:
What type of character models are you using to get high framerates with lots of models on screen? Rigidly jointed (sausage link) models or continously skinned (skin/bones) models?
If you are using skin/bones, is the skinning and animation blending being done all in the CPU or is it being offloaded to the GPU via vertex shaders?
What about vehicles? Can the characters mount / dis-mount vehicles?
You don't mention the type of path-finding which is supported. Do you have A-Star support? Grid/Cell style or graph/node style?
What about interiors? Does the RTS kit support characters entering interior spaces? If so how does the camera handle this? What about path-finding in interiors?
How easy will it be to integrate changes from the Torque Head version into the RTS code?
What about integration of other changes / features / tool kits / art kits? (lighting kit, TSE, .etc?)
How about a higher resolution demo movie?
The RTS kit is very interesting to me, but I've got quite a few questions:
What type of character models are you using to get high framerates with lots of models on screen? Rigidly jointed (sausage link) models or continously skinned (skin/bones) models?
If you are using skin/bones, is the skinning and animation blending being done all in the CPU or is it being offloaded to the GPU via vertex shaders?
What about vehicles? Can the characters mount / dis-mount vehicles?
You don't mention the type of path-finding which is supported. Do you have A-Star support? Grid/Cell style or graph/node style?
What about interiors? Does the RTS kit support characters entering interior spaces? If so how does the camera handle this? What about path-finding in interiors?
How easy will it be to integrate changes from the Torque Head version into the RTS code?
What about integration of other changes / features / tool kits / art kits? (lighting kit, TSE, .etc?)
#2
2) I haven't checked out the movie myself, so no answer here, sorry. I don't think anything "better" is available though.
3) All the models are standard .dts shapes, simply reduced poly count (600-950 IIRC). You can use any poly level model you want of course, but you will have to limit your number of units respectively.
4) The rendering pipeline is TGE 1.3, with just a few modifications (selection circles, a couple of performance tweaks). No shaders at all.
5) A lot of the FPS based functionality such as mounting objects to .dts shapes, and possibly (I haven't actually checked this out myself) has been trimmed down or removed to maximize performance. It is, of course, an SK designed for large armies, and tradeoffs were made. Of course, you still have stock 1.3 to compare from, and can bring back and features you like from stock 1.3 (or add your own) as you wish, but you have to keep in mind the performance tradeoffs.
6) There is not currently any innate pathfinding algorithms in place. There are a couple of projects of various levels working on them, and very faint hints of new releases well in the future that may include some basic pathfinding of some form. NOTE: This is my personal "feeling", not at all any inside information from GG.
7) Again, the RTS SK is focused on RTS games. Historically in that genre, there are very few true "inside interior" games, and the RTS SK doesn't support them out of the box. That being said, you have the source code when you purchase it, as well as stock 1.3 to compare against--you can do anything you like! You would need to re-work the camera (it's a bird's eye view, and won't take into account "ceilings" for example), as well as modify the selection circle rendering which IIRC is based on the terrain.
8) RTS SK is Torque Head, with the modifications required for the head. In fact, a lot of stuff done in 1.3 was done in prep for the RTS SK. GG appears committed to keeping the RTS SK (as well as other packs) aligned with "head" in the future--in fact, they are working very hard at making it even easier to integrate various packs to "head".
9) It took us about 3 hours to put in the lighting pack, for both windows and linux. It's not that difficult at all. I'm not sure if anyone has tried to push the RTS SK into TSE yet, but there is nothing specific to the SK that would make this harder (or easier). The SK is designed to work with .dts shapes primarily, so .dif based environment packs are probably not your best direction to go, but again, you have the source code!
02/03/2005 (3:23 pm)
1) There is no official demo at this time--GG made the decision to not hold back the SK based on the extended time it would take to make a strong/worthy demo, IMO a very good decision.2) I haven't checked out the movie myself, so no answer here, sorry. I don't think anything "better" is available though.
3) All the models are standard .dts shapes, simply reduced poly count (600-950 IIRC). You can use any poly level model you want of course, but you will have to limit your number of units respectively.
4) The rendering pipeline is TGE 1.3, with just a few modifications (selection circles, a couple of performance tweaks). No shaders at all.
5) A lot of the FPS based functionality such as mounting objects to .dts shapes, and possibly (I haven't actually checked this out myself) has been trimmed down or removed to maximize performance. It is, of course, an SK designed for large armies, and tradeoffs were made. Of course, you still have stock 1.3 to compare from, and can bring back and features you like from stock 1.3 (or add your own) as you wish, but you have to keep in mind the performance tradeoffs.
6) There is not currently any innate pathfinding algorithms in place. There are a couple of projects of various levels working on them, and very faint hints of new releases well in the future that may include some basic pathfinding of some form. NOTE: This is my personal "feeling", not at all any inside information from GG.
7) Again, the RTS SK is focused on RTS games. Historically in that genre, there are very few true "inside interior" games, and the RTS SK doesn't support them out of the box. That being said, you have the source code when you purchase it, as well as stock 1.3 to compare against--you can do anything you like! You would need to re-work the camera (it's a bird's eye view, and won't take into account "ceilings" for example), as well as modify the selection circle rendering which IIRC is based on the terrain.
8) RTS SK is Torque Head, with the modifications required for the head. In fact, a lot of stuff done in 1.3 was done in prep for the RTS SK. GG appears committed to keeping the RTS SK (as well as other packs) aligned with "head" in the future--in fact, they are working very hard at making it even easier to integrate various packs to "head".
9) It took us about 3 hours to put in the lighting pack, for both windows and linux. It's not that difficult at all. I'm not sure if anyone has tried to push the RTS SK into TSE yet, but there is nothing specific to the SK that would make this harder (or easier). The SK is designed to work with .dts shapes primarily, so .dif based environment packs are probably not your best direction to go, but again, you have the source code!
#3
2) Too bad. The movie is very low res. They should use FRAPS to capture a high res one.
3) I guessed that they are DTS shapes :-) But a weighted skin is more work to render than a rigid jointed mesh. You can do either with animated DTS characters.
4) Unfortunate. You can get a lot done on the GPU with vertex shadiers for animation blending, morphing and skinning. Especially if you have lots of characters to render. (Instancing would also be useful, but that is new with DX 9.0c.)
5) Too bad. Vehicles (animals) that units could ride in an RTS would be nice. You can of course just fake this with model swapping, but it would have been cool and saved memory to have a common 'horse' that all 'knights' could ride.
6) Having written one 2D RTS in 1996 and the path-finding used in X-Men Legends RPG, I can say that an RTS without pathfinding is sort of like a car without tires. :-) Luckily I can write my own. I was just hoping I wouldn't have to.
7) Well RTS can mean a lot of things. Warcraft II had 'interior' missions as do many other RTS games. You can think of Dungeon Siege as an RTS with RPG elements. DS also has indoor / outdoor sequences which are handled quite nicely.
8) I meant future versions of the HEAD. In other words are the changes well marked? Or do you have to diff vs. 1.3 to figure out the change, then add that change into a future version (1.4 or whatever).
Well I guess for $50 it might be worth the time-savings, but it doesn't really sound like a complete solution out of the box.
I guess that is pretty much par for what I've seen with Torque. It is a good solid start, but it leaves a fair number of "must have" features out.
02/03/2005 (4:40 pm)
1) What I wanted to get was some idea of performance. Even an FPS counter and a machine spec on a demo movie would be helpful. You know graphics card, screen res, character #, poly count, bone count, cpu speed, memory.2) Too bad. The movie is very low res. They should use FRAPS to capture a high res one.
3) I guessed that they are DTS shapes :-) But a weighted skin is more work to render than a rigid jointed mesh. You can do either with animated DTS characters.
4) Unfortunate. You can get a lot done on the GPU with vertex shadiers for animation blending, morphing and skinning. Especially if you have lots of characters to render. (Instancing would also be useful, but that is new with DX 9.0c.)
5) Too bad. Vehicles (animals) that units could ride in an RTS would be nice. You can of course just fake this with model swapping, but it would have been cool and saved memory to have a common 'horse' that all 'knights' could ride.
6) Having written one 2D RTS in 1996 and the path-finding used in X-Men Legends RPG, I can say that an RTS without pathfinding is sort of like a car without tires. :-) Luckily I can write my own. I was just hoping I wouldn't have to.
7) Well RTS can mean a lot of things. Warcraft II had 'interior' missions as do many other RTS games. You can think of Dungeon Siege as an RTS with RPG elements. DS also has indoor / outdoor sequences which are handled quite nicely.
8) I meant future versions of the HEAD. In other words are the changes well marked? Or do you have to diff vs. 1.3 to figure out the change, then add that change into a future version (1.4 or whatever).
Well I guess for $50 it might be worth the time-savings, but it doesn't really sound like a complete solution out of the box.
I guess that is pretty much par for what I've seen with Torque. It is a good solid start, but it leaves a fair number of "must have" features out.
#4
No one ever claimed it was a complete solution out of the box--in fact, quite the opposite: It's a STARTER KIT.
GG could have placed a basic A* pathfinding functionality into the RTS SK, but guess what? Each and every purchaser would want to modify it to do what their project needs, and in many cases wind up re-writing it from scratch. Pathfinding in general is so finely focused on a specific game's mechanics that it makes no sense to try to present a generic solution for a $50 starter kit. You said it yourself in your response to 7): RTS can mean a lot of things.
Especially in the future, but more and more even now, GG is aiming towards making sure that any functionality that is released officially as part of a GG product be as non-game specific as possible, or at worst extremely configurable and flexible. As far as I am aware, the decision was made that with the amount of time and resources required to provide that level of capability for pathfinding in the RTS SK, it would have pushed release well past the desired timeframe, as well as inflated the price point far above what would have worked for the community.
Honestly though, how can you "guess" that the time savings is worth $50? Even at minimum wage, it would take 3 days of work to pay for the starter kit. Can you honestly say that in 24 hours of coding you could duplicate all of the functionality of the kit as it exists today?
TGE and TGE starter/content kits have never been targetted/advertised as "make your game in a day/week/whatever" products...intentionally. As it should be, especially for the price to feature ratio they provide!
02/04/2005 (9:35 am)
Quote:Well I guess for $50 it might be worth the time-savings, but it doesn't really sound like a complete solution out of the box.
No one ever claimed it was a complete solution out of the box--in fact, quite the opposite: It's a STARTER KIT.
GG could have placed a basic A* pathfinding functionality into the RTS SK, but guess what? Each and every purchaser would want to modify it to do what their project needs, and in many cases wind up re-writing it from scratch. Pathfinding in general is so finely focused on a specific game's mechanics that it makes no sense to try to present a generic solution for a $50 starter kit. You said it yourself in your response to 7): RTS can mean a lot of things.
Especially in the future, but more and more even now, GG is aiming towards making sure that any functionality that is released officially as part of a GG product be as non-game specific as possible, or at worst extremely configurable and flexible. As far as I am aware, the decision was made that with the amount of time and resources required to provide that level of capability for pathfinding in the RTS SK, it would have pushed release well past the desired timeframe, as well as inflated the price point far above what would have worked for the community.
Honestly though, how can you "guess" that the time savings is worth $50? Even at minimum wage, it would take 3 days of work to pay for the starter kit. Can you honestly say that in 24 hours of coding you could duplicate all of the functionality of the kit as it exists today?
TGE and TGE starter/content kits have never been targetted/advertised as "make your game in a day/week/whatever" products...intentionally. As it should be, especially for the price to feature ratio they provide!
#5
Torque $100... RTS Starter 50$... Don't even begin to complain about the prices.
I don't even plan on working on a starter kit at the moment, the selection and camera improvements are worth many times the cost alone.
02/04/2005 (9:47 am)
At 50$ this is a no brainer. I'm finally going to be purchasing this, been wanting to for months, since I saw the first snapshot of it. Torque $100... RTS Starter 50$... Don't even begin to complain about the prices.
I don't even plan on working on a starter kit at the moment, the selection and camera improvements are worth many times the cost alone.
#6
Any word from an official GG employee on these points?
02/05/2005 (6:53 am)
I'm not trying to provoke anyone. Just trying to get some question's answered. No need to get all defensive.Any word from an official GG employee on these points?
#7
-Jeff Tunnell GG
02/05/2005 (3:35 pm)
@Ubutopia: Sounds like you have a lot of experience writing professional RTS engines. You are a Torque owner, so maybe writing your own RTS package would be easier. There are not many GG community members that have the kind of experience that you have. There has been a LOT of call for starter kits, which help point people in the right direction. Stephen Zepp has more experience with the RTS SK than anybody in the community. His answers were great.-Jeff Tunnell GG
#8
Characters are skinned meshes, all blending is done on the CPU. The animations are also seperate, not linked (IE they aren't all lock-step walking etc). You can get a lot of performance gain by offloading to the GPU, for sure. However Torque is targeted at fixed function, and we will not use tech that is higher than it's target in a starter kit for it. If we did a version for TSE, however, that would be a different story.
You could actually pretty easily put the mounting/unmounting back in. The RTS unit class is actually derived from Player right now, which is not actually the optimal solution, but the reason we went this way was for familarity with the current code base, and easier integration for functionality like interaction with water, mounting vehicles, etc.
There actually are a bunch of pathfinding things laying around that we could have thrown into RTS, however two things were keeping us from that, most importantly was time. We have a ton of projects we all split time between and this needed to get out of the door. We pushed off AI until the next update.
You can enter interiors, there is no special camera work for this. There is one issue, however. The selection circle code is based off the Terrain because otherwise, when zooming in and out, the only thing that knows about the LOD geometry of the terrain is the terrain. The next revision I was planning to add support for the selection circles in and on DIFs (interiors) using the shadow code.
The changes are pretty well isolated. RTS is a seperate download and has it's own copy of Torque, this is primarly because we were not working off vanilla 1.3, though we did the merge against that. There are a lot of little changes that are bug-fixes etc for 1.4 that made it in to RTS. Most of the actual RTS code is isolated from the Torque code. You would have to run a diff, but it wouldn't be horrible at all.
The RTS unit is using a simplified skeleton, and lower poly count. I don't know if you've seen this, but this is Josh's .plan from a while back. If you scroll down to the shots of a crapton of units, it's pretty impressive. Those are also all networked objects. You can use whatever art-pack model you want, however it will impact performance.
Hope that answers some questions.
02/06/2005 (12:29 pm)
The RTS kit is getting an upgrade when me and Ben can give it the time, which will be no sooner than after GDC.Characters are skinned meshes, all blending is done on the CPU. The animations are also seperate, not linked (IE they aren't all lock-step walking etc). You can get a lot of performance gain by offloading to the GPU, for sure. However Torque is targeted at fixed function, and we will not use tech that is higher than it's target in a starter kit for it. If we did a version for TSE, however, that would be a different story.
You could actually pretty easily put the mounting/unmounting back in. The RTS unit class is actually derived from Player right now, which is not actually the optimal solution, but the reason we went this way was for familarity with the current code base, and easier integration for functionality like interaction with water, mounting vehicles, etc.
There actually are a bunch of pathfinding things laying around that we could have thrown into RTS, however two things were keeping us from that, most importantly was time. We have a ton of projects we all split time between and this needed to get out of the door. We pushed off AI until the next update.
You can enter interiors, there is no special camera work for this. There is one issue, however. The selection circle code is based off the Terrain because otherwise, when zooming in and out, the only thing that knows about the LOD geometry of the terrain is the terrain. The next revision I was planning to add support for the selection circles in and on DIFs (interiors) using the shadow code.
The changes are pretty well isolated. RTS is a seperate download and has it's own copy of Torque, this is primarly because we were not working off vanilla 1.3, though we did the merge against that. There are a lot of little changes that are bug-fixes etc for 1.4 that made it in to RTS. Most of the actual RTS code is isolated from the Torque code. You would have to run a diff, but it wouldn't be horrible at all.
The RTS unit is using a simplified skeleton, and lower poly count. I don't know if you've seen this, but this is Josh's .plan from a while back. If you scroll down to the shots of a crapton of units, it's pretty impressive. Those are also all networked objects. You can use whatever art-pack model you want, however it will impact performance.
Hope that answers some questions.
Torque Owner Eric Lavigne