20 replies to this topic

[Nazgûl]

While working on SEE's rather high poly models, and when testing them in game, I seem to get the impression that the poly count of the models matters less than the number of models on screen. High poly models doesn't seem to slow down the game in a more significant way that low poly models. Number of units however, do, reagrdless of low or high poly. I have heard and red all kinds of opinions regarding SAGE and poly count, as well as meshes, textures, etc and what matters the most in game. Now, I own a rather old laptop, that is however still tagging along the competition with 2,5 GHz CPU, 4 GB RAM, 1 GB GPU. The question is, would I gain a lot of power in game, if I update to a new stationary system with specs like the ones I have my eye on; 4 GHz CPU, 12 GB RAM, 2 GB GPU?  In other words, would I or any other player gain more power with higher frame rates, if we update our hardware, or is SAGE simply too old to make use of extra hardware power?

 

Some say poly count matters most. Others say size of textures and some (like me) say it's the number of units that lagg the game when going high. So what do you say... Can a hefty system update make a significant increase in frame rates, or has the hardware passed SAGE's limitations nowadays?

 

 

[NewErr]

I think you can update your system but the difference will be insignificant as the game will stil have framedrops and will lag during huge battles. If it is worth it? Well that's really up to you, I think is worth it just as a standard and not just for modding BFME. Your Zbrush should work better tho

Edited by NewErr, 15 December 2015 - 06:17 AM.

[Nazgûl]

I didn't mean wether it's worth it for me, I meant more like... would players with more hefty systems experience less lagg in huge battles or is the game too old to utilize more hardware power? I will most likely update my system anyways =)

[NewErr]

Yeah the game is too old...

[Kwen]

This is mostly true. You're right that at this point in the Engine's life simply throwing more hardware horsepower at it might help, but not much at all because it is in fact that the engine just cannot handle what we as modders would like it to.

However poly count does matter a good bit. When we were developing TPW our first official Hoplite model sat around 3k ponies when fully geared out. At first we figured this was a bit high but that we could get away with it. It turned out that since they were the most common unit on the screen the game would definitely bog down with severe lag. When we rehauled the model down to under 1500 polies, the game ran significantly better with the same number of units on screen at the same time. So the poly count definitely was the major factor in this case as to what was causing lag.

[Mathijs]

Granted, our PC's weren't as godly as they are now; still, that new model (which was 900 polies, if I recall correctly) fixed our problems.

 

To be honest, the main thing that bothers me about overly high poly stuff is that it actually looks too smooth compared to the rest of the game. To me, having a smooth model that is 6k polies come out of a 800 poly barracks model looks ridiculous. It's a visual integrity thing to me. You might say that replacing all models would then fix this problem, but there's still low poly structures and props, low resolution terrain textures, plus an engine that was not designed to render armies of high poly units.

 

Plus, BFME's animation sets weren't designed for high poly meshes, meaning a lot of stuff is going to clip and twist in strange ways, even if the mesh matches the skeleton (again, in my experience). 

 

To me, for BFME, a low poly model simply looks better, as contradictory as that might sound (on the surface). 

[Kwen]

I still had a pretty hefty machine back then, I believe I was using a GTX 560 at the time.

[NewErr]

Myeah.. reminds me I have to optimize a bit my elven units, some of them are a bit too high.

[Mathijs]

Plus alpha channels kill performance...

[NewErr]

Now you are being mean, I use alpha channels to optimize models.

[Mathijs]

Well that's just not a very good strategy, I'm afraid. Much better to have a few more edges than an army full of alpha channels. They require much more rendering 'power'.

[NewErr]

Most of the time is better slapping some alpha plane then a solid model.

[Kwen]

You do realize that having an Alpha effectively means the game has to do two things instead of one, right? The model is still rendered in full, plus the game has to render a transparency layer. In moderation and static objects this is fine, but in mass it can indeed cause a good bit more lag than a properly optimized opaque mesh.

[NewErr]

But instead of rendering 100 extra polies it renders 1 with an alpha on it.

Anyhow, this discussion can last forever, since the game will still run slow in any case.

Edited by NewErr, 15 December 2015 - 11:27 PM.

[Nazgûl]

Well, it's only one way to find out I guess... when I run my mod on a higher spec PC. In other aspects, I agree with Mathijs philosophy completely. However, Ican't model in the way that is best suited for this game, so... yeah.

Still, my testing shows that having 1000 low poly units on screens laggs more than 500 high poly, for some reason...

Hopefully, I'll live to see another moddable LotRs RTS game in the future where I can use my high poly models in a more suitable way. But unitil then, the visual integrity will have to suffer a bit =/

[Kwen]

Well yes, 1000 low poly units would in fact lag more than 500 high poly ones, simply because that's pushing the engine past its limitations in more ways than just graphically. With a scale of units that large path finding tends to break down, and in general the game was simply not designed to handle that many at once. This is the main reason the command point limits are by default so low, not because the engine can't handle the visual aspect. Remember the engine was built for CnC which didn't use the horde system, so the number of units on screen at once were significantly less. The BFME games attempted to bring on the scale of LotR, which they do. But the engine can't keep up when we as modders try to push it even further than the developers did.

 

Either way, it's not really a fair comparison to reference 1000 low poly units to 500 high poly ones. A better comparison would be to have the same number of units and use a few different models/textures with varying levels of detail.

 

Basically, we are in charge of whether or not we break our own game by the level of detail and amount of scale we choose to implement. The trick is to just find the right balance.

Edited by Kwen, 16 December 2015 - 07:35 PM.

[Nazgûl]

Yeah, that darn balance

 

Here's an interesting topic, from back in the days when SAGE was something fairly new and interesting...

http://www.cnclabs.c...-IMPORTANT.aspx

 

...for about a year I have been able to test the Sage engine and for grins I made units with 60,000 poly a piece and built over 50 of them, the total poly count on the map at one time was over 1.8million and I lost 2.3 frams a sec (very small) and it was not due to the poly count but the number of units. The Sage engine is faster then people think there are things that can slow it down in a hurry, poly count is not one of them unless you are talking about 100,000 poly count...

[MattTheLegoman]

Better to optimise where you can. Better safe than sorry.

[Nazgûl]

Yeah, for sure. For heroes, I don't make too much of an effort and most heroes are about 3-6k, but for horde units I'm trying to optimize as much as possible without "killing" the design. 1,5-2k is what I'm trying to reach for the Iron Hills dwarf for instance. However, it will be interesting to see how the mod runs on a new system. 

[Kwen]

One of the best things you can do is to also be conservative on horde size and command points. Bigger isn't better.