27 réponses à ce sujet

[B_Harrison]

I've just posted the first few (of many) NWN development posts over at my blog, now that I'm modding the game again. There'll be a lot of theory and project-specific stuff on there, but I'm also going to be releasing any and all functional resources for the community's use; that might include tutorials, techniques, re-makes of vanilla content & old mods, etc. as well.

Neverwinter Mod Madness

I thought I'd put any/all updates on my NWN content here in one thread, to keep things like "release" posts in one place and for any comments or suggestions that would never reach my blog.

[Jenna WSI]

Glad to have you back, Ben.

[Bard Simpson]

Welcome back, Mr. Harrison!!

Those are some nice preliminaries. I really like the axes and spear 2h, but I love blade 1h! Heh, maybe I'll even run into a plaguesword someday. Well I don't want to literally run into one, of course, but........well, you know what I mean.

[B_Harrison]

Thanks guys!

#002: World Scale

If anyone wondered why it was NWN mod madness, the madness part is starting to kick in as of #002. But I'll also be posting some actually useable content soon, hopefully.

[B_Harrison]

Bard Simpson wrote...
Heh, maybe I'll even run into a plaguesword someday. Well I don't want to literally run into one, of course, but........well, you know what I mean.

Ha! I could convert the Plaguesword to (read: remake it for) NWN without much effort, though it'd necessarily lose all its normal and spec detail. Same goes for any of my higher-poly/newer game stuff.

[WebShaman]

Ben is back!

Glad to have you back, Ben!

[Rolo Kipp]

<thinking things...>

B_Harrison wrote...

#002: World Scale

Ben! Glad you are back. Glad *I* am back, for that matter :-)
I am really glad you are exploring this fork. It's something I have been thinking about for a while and really want to work on.

As soon as I get up to speed, again :-P

<...are bigger than they are>

[Borden Haelven]

Very interesting. If you could scale the PC it would make a Giant's Lair more interesting... Movement speed would have to be halved to carry the effect properly though.

[Bannor Bloodfist]

I guess we need a sticky somewhere about the HARD CORE limits of the NWN engine.

It is designed and ONLY uses 16 bit memory space. The toolset/engine REQUIRES 10x10 meter tile space. Sure you can create a 5 meter sized tile, but the engine paints that tile into a 10x10 meter space, leaving gaps. You can NOT get the engine to place a 2nd 5x5 meter tile into the gaps between.

The toolset/engine does NOT really recognize more that 640k of ram. Sure, there are options to increase stack space etc, but the core doesn't really use it much. It can NOT run on more than one core on a multi-core processor nor can it truly recognize 32/64 memory space. The ram extensions that it can recognize are just 16bit mem space that is linked together. (Wording is wrong, a programmer would be able to expand the description better, but the facts are the facts). NWN core engine is written, hard coded, as 16 bit so it will run on a win98 16 bit OS. The extensions added by vista, xp, win7, et al, can't change the core coding.

The latest version of 1.69 STILL runs on an old win98 computer.. slow as hell of course, but it runs. Without having to change anything. It will also run in 32bit mode on a win7/8 box, but it never utilizes the power of the later operating systems. It can't, if it did, it wouldn't run on win98 without major duplicated hard code. Something never done by Bioware.

Great ideas, but not practical for such an old engine that just can't handle that sort of power.

Sure, you can create a 32x32 sized area, but once you fill it with creatures, and placeables, a server has issues, and the client/dm client crashes due to lack of ram, and a trashed stack space and even if it loads, it runs slower than molasses, even on a truly powerful computer.

Multi-cores? Nope, the only option you gain with that is the ability to force nwn to use a core other than core zero, so your os can run in that space. Some savings, but truly very little, and the core goes fully utilized. NWN can not actually use the extra cores, it was not written to use multi-threading which is required to use the extra power of the newer computers. It was written, hard coded, for win98. Check the min requirements on your box or the download page wherever you purchased the online-download version.

[B_Harrison]

Just responded in the blog comments, but I will briefly here too-

NWN either isn't limited to, or doesn't need more than 1MB(!) of system memory. The recommended minimum system memory to run Win 98 is stated by MS as 16MB...

In any case, 32x32 areas were brought up by Sherincall in the comments. I've no intention of using areas even half that size, nor are my tiles sized differently to vanilla NWN tiles (rather modeled to a different scale). The NWN engine can easily handle this for my purposes. For the purposes of gigantic placeable-cluttered action PWs using vanilla art, probably not.

Modifié par B_Harrison, 23 septembre 2012 - 01:26 .

[Rolo Kipp]

<nods at all the good points...>

@BB: Yup, I know that. but what I was proposing and what it looks like Ben is already experimenting with is changing the scale of the geometry. The tile will still be (for engine purposes) 10mx10m, but with all the geometry scaled down (by 50% in his case, to 10% in mine), the *illusion* will be of a much larger area (with, yes, a lot of geometric detail).

The thing I'm hoping will make a difference is how much of the 3D stuff is put off onto the GFX card - notably the textures and vertexes. I don't *know* this will work the way I want, but if I can create a "Skies above the forest" tileset scaled so each 10m tile represents 100mx100m, then a 12x12 area ill be 1.2kmx1.2km. With an alternate set of (scaled down) appearances with appropriate flying phenos, we might just have an interesting alternative to overland maps.

Or, as ben demonstrates, a method of making *really* impressive boss fights ;-)

<...while dancing around the edges>

[Failed.Bard]

I'd been considering altering the base races in my mod to just use the human and dwarf base race models. Halfling, elf, half-elf, gnome, and half-orc would just be used for scaling purposes. 50/75/100/150 in the human chain, and 50/100/150 in the dwarf chain.

I was just planning to use it to allow enlarging/shrinking spells on PCs, but it looks like there could be some potential for it beyond that.

[NWN_baba yaga]

very interesting thought. I wonder why bio never used it from the get go so the whole game would have been much larger in scale;)

[s e n]

I think making tiles of double the scale, or even more its very cool idea, but I think its not doable, basically because there is an inverted proportion: the higher the scale, the lower the chance to customize tesselation: imagine the use of crossers: its already too squarish at 10x10, i cant imagine being bound with a 20x20 or even bigger scale... yes, you have more room to bend a river crosser, but you will need a lot of tile variations to make it look "different", and still you'd be bound to the 20 meters large edges, this would make the area just like a poorly cooked seamless texture: in a good one, you dont spot the seams, and everything will look unique, in a poor one well, its ugly.... by my tastes, i would do the opposite: making 5x5 tiles to increase the tile randomization!

[Rolo Kipp]

<chalking one up...>

I see what you're saying, Sen. The smaller the scale of textures the worse it looks. But that is (IMO) a cosmetic challenge, rather than a technical one. And larger textures (2048x2048) *do* work fine on modern machines...

The other issue, of pathing, is rather moot for my aerial tilesets as the pathing will be wide open - i.e. you can fly over just about anything that would be considered an obstacle at normal scale/ground travel. On something like your pretty new tileset (scaled down)... ouch.

It rather reinforces the importance of hand-crafted tilesets. Indeed, at the scales I'm talking about, I'd like to create my areas en toto (perhaps even with aerial images mapped on them) and then tileslice them for unique tilesets. Again, a laborious, custom-crafted approach that may or may not be worth it.

<...for the craftsmen>

[B_Harrison]

Rolo Kipp wrote...
It rather reinforces the importance of hand-crafted tilesets. Indeed, at the scales I'm talking about, I'd like to create my areas en toto (perhaps even with aerial images mapped on them) and then tileslice them for unique tilesets. Again, a laborious, custom-crafted approach that may or may not be worth it.

That's the main idea, as far as my NWN work is concerned. I probably should have clarified that; the concept really doesn't work when one is creating tilesets for general use, not to mention implementing [mods like this] into existing modules or modules based on existing resources.

That said, thanks for all the insight on the idea's broader applications (or lack of), guys!

[Bannor Bloodfist]

NWN_baba yaga wrote...

very interesting thought. I wonder why bio never used it from the get go so the whole game would have been much larger in scale;)

Primarily because you are actually LOSING space for detail, by converting that space to distance between things.  If you make a photograph using a micro-dot camera, sure the data is there, but you can't see it without englaring the view anyway, and by having that much more so called detail, you are crapping all over the ability of the game to move fairly quickly. (I am not even talking about running). 

IE, you have a scaled down PC, he has this wonderfully designed WWF style champions belt around his waste, but you can't see it in game.  So, complete waste of effort on the artist side for the design and texture work, and a huge lag inducing mess for an old engine.

As far as base texture sizes go, 2048x2048 is typically wasted in most pre-2010 games, much more so in a zoomed out view in such a scaled down version.  Most 2048 textures are just huge duplicated fairly flat, boring, things.  Going 2048 for an entire creature's texture can make sense, but since the game already breaks that creature into multiple pieces each assigned it's own texture OR a UVW mapped section of the same texture, the engine completely re-loads the full texture each time is has to grab a piece of it, overlaping the same memory space so no extra memory loss, but it is typicaly a waste to go 2048x2048 anyway unless you are mapping onto a very large object.

I know, I know, I typically sound like a nay-sayer to these sorts of ideas.  Folks put silly counter arguments abouit win98 only requiring 1mb, but that is ludicrous, most specifically because we are not truly discussing the OS, but the way this game was designed.  It can, does, will, use up to 640, beyond that, it pages ram in/out.  Where if the Aurora engine had been written in/for a true multi-threaded OS, it could and would use however much ram you provide. 

The game also has limitations in what it will offload to the GPU.  It just doesn't use the newer systems to even a 10th of their possible power.  It was coded as a 16bit app, with vid cards typically only having 128mb of ram at the time, or a max of 256.  Why?  The OS, remember we are talking win98?,only recognizes something on the order of 968 meg of total ram space, it reserved either 16 or 64 meg (can't remember) for bios functions, and reserved everything above 640 for other purposes, only part of which was ram space for a vid card.  It doesn't matter what OS your PC or Mac is using, the app is 16 bit, with all of a 16 bit apps limitations.  There is no way around that.

All you have to do is have a util running that shows proper core usage rates... then watch whatever core on your multi-cored processor, that you have assigned NWN to run under... it takes FULL control of that core, pages in and pages out sections of ram continuously.  If you also have a util to watch the vid card gpu usage you will find that it seldom even reaches 50% usage, and ram usage for the gpu doesn't really matter all that much because the gpu can pre-cache all it wants to pre-cache, but NWN reloads and overwrites it continuosly regardless.


There is no economy of scale here.  In fact, you are running a very high risk of overloading the engine to the point that it crashes because it just can't clean up after itself well enough.


Gawd, I wish it were otherwise, you have no idea how much I wish it were otherwise,  But you can't get an ostrich egg to hatch an elephant, even if you have an incubator and some guy claiming his bird nest will allow gene enrichment.  You can take a magnifying glass and hold tiny paintbrush to create a fully detailed map of the united states on something the size of a dime, but when you reach 50, your eyes go and you can't see the darned thing anyway. 

Again, I am NOT knocking you guys for the ideas/thoughts etc, but you have to realize that this old engine just can't handle what you want to acomplish.  If we had the ability to truely re-write the code, ie if Bioware were to release the code, we might stand a chance.  Injection modding the way nwnx works can improve some features, but it still can not make a 16 bit app work like a multi-threaded 32/64 bit app.  Nwnx folks have tried for years to improve the memory limits, but it is just not possible without having the real engine code and the ability to completely re-compile that engine.  Overlaying something on top still leaves the limits in place in 99% of the cases.  Besides, if we were to get an option to add a new feature, I would want multi-Z-level access/usage. 

Ok, I have likely bored you all to tears and overstepped the boundries here by quite a bit, but I hate to see folks waste valuable custom content creativity on attempting something that truly would have very little usage in game, meaning that time and energy could be used to create something new and exciting within the limitations of such an old game and that would be more likely to be used by more builders.

There have been many folks attempting to create a more usable map type system... something more creative and usable than what was accomplished by the Darkness Over Daggerford team.  That was basically what you are attempting here, but using a different scale type of system etc...No team has ever really finished that overland map system though, and I think you will likely end up in the same sort of boat as you hit the frustrating limitations of this engine.

Shoot.... there I go again...

[Pstemarie]

The best compromise I found was making a custom tileset using as few tiles as possible - all stripped down to just have ground and/or water. I then laid in a narrow river and made some scaled down placeables to represent features such as mountains, towns, forests, etc. The placeables are simple by design and carry only the minimal detail necessary to render something that doesn't look like a pixel.

I opted to keep the PC full size and just make him invisible. In effect the camera does all the work and since the PC isn't visible it's easy to trick the eye into thinking you're looking at something much larger than it really is. The PC has full range of movement, giving the illusion that you're panning over an area. Although it moves quickly for its perceived scale, without a visible PC for perspective, you hardly notice the speed with which you're moving across such a "large" area.

[Rolo Kipp]

<definitely...>

I agree very much with the first sentence of that last paragraph (before you shoot ;-) in my rather special case. Well, not so much more usable as more interesting (I want aerial combat with a unique aerial combat system ;-) And I agree very much with the limitations you detail.

But I *like* to see people come up with far out ideas that *might* enrich our tools (like all that neat stuff Virusman/OTR were playing with not so very long ago). Specifically, what Ben is playing with - exaggerating the size difference in certain crafted encounters using non-conforming scaled models & tiles - sounds both doable and something that could really make a strong statement (I've never been all that happy with large, but not awesome, dragons :-P ).

Re the uber-texture: I only mentioned that because of the tessellating nature of scaled down textures could be compensated for by, as an example, taking the image of a normal 10mx10m floor and composting a 2x2 version (20mx20m) and hand-crafting the seams to remove the most obvious tessellation. Further processing to make minor variations to the 4 "inset" tiles would make the "half-scale" tile more visually interesting. In this case the single uber-texture would be replacing a set of standard nwn textures and would be loaded with the tileset in lieu of them.

My point was that large textures load with a lot fewer problems than they did back, oh, 5 years ago. (personally, with the exception of the marble base for that aweful altar I did, I don't need the detail from such large textures. I generally work with atlases of 512 or 1024...).

But then again, I don't really want to hijack Ben's thread ;-) Wanna continue this on my aerial travel & combat thread? <yep. coffee's kicking in nicely now>

<...not bored>

[Tarot Redhand]

@Pstemarie, I just want to double check that you are aware of Demangels Dragon riding system as an alternative to the idea you are exploring. There was actually supposed to be a YouTube video of this but I am afraid I just can't seem to find it.

TR

[Pstemarie]

My ideas already been explored, but thanks for the link

[B_Harrison]

Hijack (the thread) at will; it's good to see some information-sharing happening even if it's moved away from what I want to accomplish, which at this point is essentially a proof-of-concept - and certainly possible inside the NWN engine. When I present something that's not theoretical/project-specific, I'll also release it, so I do realise which things don't have much use to the wider community.

[Just a ghost]

You may want to check NWN2 for the repercussions of shrinking/enlarging player characters. I remember this was a strategy to get more square meters into an area, and the folks doing that must have run into gameplay issues as well.

[B_Harrison]

#003: Item Art

Ranged weapons, again not textured yet. Will have some more tile-related stuff up soon, and some useable resources at last.

Modifié par B_Harrison, 05 octobre 2012 - 01:11 .

[Rolo Kipp]

<quite impressed...>

Your standard high-quality job :-)
How you fit such gorgeous curves into such low polys... *shakes head in wonder*

<...and quite jealous>