Neverwinter Nights 2 Blog

NEAT! Thx for the intell. Glad to see progress.

I like that there is a lot of green on that list. Bodes well for the future. I have yet to feel that my faith in the Developers has been out of place. This once again establishes that while not perfect, you are always moving ahead.

 

Now if we could only get that Max 9 Expotron, A VFX Editor Guide and a few other things while you're at it and we'd be able to at least call you demigods.

 

regards,

dunniteowl

Thanks Rob for this entry, it is nice to see the patch's progress.

 

Can't wait for this patch release.

What I am asking myself:

Will the performance issues on ATI cards be fixed? Ever?

 

There's nothing about it in the patch notes, so I am not optimistic.

 

I switched from a Radeon 1950Pro, itself not a slow card at all, to a Radeon 3850, which

is about 1.7 to 2 times at fast. High details and shadows switched on still make the game

crawl.

 

I was about to give up on seeing any kind of performance increase in NWN2, until

I saw the following on Beyond3D's forums:

 

OpenGL guy's post

 

Originally Posted by Morgoth the Dark Enemy View Post

-NWN2 and NWN2:MOTB-slow as a dog with maximum settings on ATi HW since release and up to present time, for no obvious reason, even after numerous patches/drivers.

 

No obvious reason to you, but the root cause was given to the developer some time ago.

 

Just FYI, OpenGL guy works directly at ATI, he is not one to make stuff up.

So, when can we expect our ATI hardware to run the game as it should?

What I am asking myself:

Will the performance issues on ATI cards be fixed? Ever?

 

There's nothing about it in the patch notes, so I am not optimistic.

 

Also don't see jack about fixing the BSOD caused by simple cutscenes.

 

For the past two patches the game has been unplayable.

 

I run 32-bit XP PRo Sp2, 8800 GTX with latest production Nvidia drivers, 30" Samsung allowing my game res of 2560 X 1600.

 

The cutscene crash happens in the OC, MotB, and all community modules that use them.

 

Submitted numerous bug reports to Obsidian. Don't recall a reply to any of them.

Good news Rob,

 

But... there is always a "but", can we expect, for the next NwN2 Community Update, the result of the Custom Content's Contest ?

I run 32-bit XP PRo Sp2, 8800 GTX with latest production Nvidia drivers, 30" Samsung allowing my game res of 2560 X 1600.

 

The cutscene crash happens in the OC, MotB, and all community modules that use them.

 

Occurred to me that I had added another 2GB of RAM to my PC two NWN2 patches ago.

 

Turns out, the driver for my "Creative Labs X-Fi Xtreme Gamer Fata1ty Pro" has a bug. Creative has a beta driver out with 4GB RAM issues on its fixlist.

 

I disabled the /PAE /3GB switches in my boot.ini & can no longer reproduce the Blue Screen of Death in cutscenes.

 

Pretty unique set of circumstances. Hope someone else can benefit from this post. Watch those drivers.

If there indeed is a possible fix, and ATI has shared the problem with Obsidian, I would be glad to know when a possible fix could be released?!?

Any chance we're going to see a pivot/rotate attribute put on speedtree objects in my lifetime?

 

Forests of limited seed values would look so much better if only speedtrees could be swiveled slightly offset from their neighbors.

Any chance we're going to see a pivot/rotate attribute put on speedtree objects in my lifetime?

 

Forests of limited seed values would look so much better if only speedtrees could be swiveled slightly offset from their neighbors.

 

Agreed. This one thing would make all the difference in making forests, or even simple tree lined streets. I know there's magic and all, but I just don't see the wizards of NeverWinter spending the time (nor the merchant class spending the gold) to have all the trees look exactly the same.

 

So, I suppose I would vote to have this too as it's frustrated me as well.

 

regards,

dunniteowl

 

My understanding with the trees is that rotating them is essentially the same as rendering entirely new objects, performancewise -- so the advantage we have today of 10 trees using the same seed = 1 tree rendered (kinda) is wiped out. However, that info involves reaching in my memory all the way back to toolset beta timeframe, so I could be mistaken

 

However, I've also seen numerous modules where people place only 5 or 6 different tree seeds; and using scaling and irregular placement you literally can't tell that they're the same when you're walking through the area. (In the toolset, of course you can because you get the birds' eye view -- but that view isn't available to the players)

 

Actually I have experimented a bit with trees and their seed values. I have several areas that I created with massive amounts of trees and with a very wide array of seed values for them. In general, I have made 4 medium sized areas 16x12, 8x14, 12x12 & 8x32 and just absolutely loaded them down with a variety of trees/plants and seed values for same.

 

In each area I used approximately 12 different Tree types, each with varying seed values approaching 12 seeds per tree type. This means that in some of these areas, I have 12x12 types of unique trees according to the specifications listed in Developer responses. I grant that I have not placed anything other than trees, plants and grasses in these areas (though I used a fair amount of grass types as well) and, while there are no encounters, transitions, area effect triggers, convos, etc., the movement performance of a character in the areas while testing to see how they look has been just fine. I think the original recommendations were no more than 20 unique seed/tree variants for performance related reasons.

 

I am running an AMD X2 5000+ (meaning two separate cores running at 2.3GHz each) with 2GB DDR2 RAM @ 800MHz FSB. My specs allow for a bit more lattitude compared with recommended specs, however, I am sure that what I am 'pulling off' is not that special in terms of performance at this time compared to what I perceive is 'average' system specs on most machines.

 

(Note, while 144 individually 'unique' trees are being used, there are, quite literally several hundred trees of each species, making the total number of each 'unique' plant/tree value approach 30-100 each of each type, depending on the look I was after. The larger forested areas literally have close to two thousand trees as placeables in them.)

 

Still, if I could rotate a tree a bit, that would make all the difference visually speaking (though I admit, scaling does really make some differences as well and could mean enough difference to lose the "patterned" look to some areas' trees that I have seen.)

 

Of course, I haven't been doing all this to test the system, I just didn't care about the specs when I made the areas. I was sort of doing my best to see how 'realistic' I could make an area in the outdoors. In fact, until I had made the 3rd test area, I didn't even think about the recommended performance settings of 4-5 unique tree types with 3-5 separate seed values. I just told my self, "Make this look as realistic as you can."

 

As I gain more experience with scripting and running more tests, I will revisit these areas and add in encounters triggers and more placeables to see what happens. Also, admittedly, I have been making these areas in mind with giving the adventurer a sense of scope to travel (meaning a lot of unique little areas) during an adventure, most of them having only one to three encounters of any sort within them in the first place. This being the case, I am pretty sure that the "pushing" I have been doing in terms of the performance envelopes will prove noteworthy at some time.

 

Meanwhile, if it requires a whole nother render of a 'new' model for a rotated version of one already, then forget it. I already have methods to break up the pattern of sameness and am also working on extending the regions of performance (in terms of proof of performance for amount of unique items) such that this one request is not all that important. It would be nice to have, though, just as a quicker way to make things seem unique without having to resort to the constant seed value changes, or rescaling of objects.

 

my 32 cents on the matter...

 

dunniteowl

My understanding with the trees is that rotating them is essentially the same as rendering entirely new objects, performancewise

I seriously doubt the tree is being rendered to a Z-buffer sprite and then scaled in 2D and stamped on the framebuffer, it is almost certainly being re-rendered each time anyway ... and the 3D card doesn't care if the transform matrix has an extra rotation applied. Even without that, in theory for light sources at infinity it can get you some advantages with shadowing (you can reuse shadow volumes) ... but who plays the game with environmental shadows?

 

PS. is UIObject_Input_ActionTargetScript going to get fixed? (ie. allowing nSpellTargetIndex to be a variable, not crashing on certain shapes and allowing a range limit.)

Some great content there. We've got all those integrated into our persistent world too.