[DAO:deeab2d] Is it worthwhile to reduce avatar tris and colliders to improve loading speed?

by 0x1b35f4d22b2c8bc706bcfcc383a1ca77fb4fc600 (punkpink)

Linked Pre-Proposal

Is it worthwhile to reduce avatar tris and colliders to improve loading speed?

Summary

This proposal suggests reducing avatar triangles and limiting colliders in buildings to improve Decentraland’s loading speed.

Abstract

In anticipation of a possible bull market causing an influx of users, this proposal aims to optimize the performance of Decentraland. By reducing avatar triangles and advocating for fewer colliders in buildings, we aim to improve loading speeds during busy events, thus offering a more fluid and accessible experience for users, in turn making it easier for users to move around without tripping. constantly with elements in the way.

Motivation

The current challenge faced by Decentraland is the slow loading speed during high-traffic events, which could hamper the platform’s capability to handle mass adoption. This proposal addresses the problem by reducing the graphical complexity of avatars and simplifying building designs, potentially leading to significant improvements in platform performance.

Specification

This proposal includes the following changes to both the documentation and the hands of the base avatars.

  • Creation of simpler hands with a maximum of 150 each hand (between the two hands 300 tris).
  • It is not necessary that they have defined fingers.
  • Implementation of these hands in the base avatars.
  • Update documentation with corresponding images.
  • Indicate in the documentation that colliders cannot be used in buildings except on opaque walls, so other elements such as trees, streetlights, minor objects and similar must not have colliders.

Clarifications:

  • Creators will be able to continue creating in the Handwear category with the current requirements.
  • The wearables in the Handwear category that exist in the marketplace will continue to exist and can continue to be marketed.
  • Users will be able to buy and wear Handwear with the above characteristics, this only affects the base models that have not obtained more elaborate handwear.

Conclusion

While recognizing the potential trade-offs in terms of community sentiment and artistic freedom, this proposal seeks to strike a balance between performance optimization and creative expression. The goal is to make Decentraland more accessible and enjoyable, especially during large-scale events.

Vote on this proposal on the Decentraland DAO

View this proposal on Snapshot

Interesting proposal as it is always good to push the platform further and increase performance.

Do you know for certain that these two changes would drastically improve load time and performance for avatars? Can you share any quantifiable evidence that this will improve performance? I am not familiar with how colliders impact load times, and before we make procedural changes like this, I would want to know for certain that the artistic freedom we would give up would be worth it. To me the biggest draw of Decentraland was that I could build anything I wanted on my land. Also, the moment we start telling people they “cannot” use something, we’ll need a plan, process, and sigh probably another committee to make sure they are adhering to the rules. I’d rather find other ways of optimizing our platform. (But if this would create a noticeable change, I’m all ears)

Do you know if there are other ways of reducing triangles to improve load time? Take this fern for instance… this fern appears on most default scenes, and sometimes grows twice in once plot. While they look innocent enough, this asset requires 280 triangles per fern. So if we modestly say that 60% of Decentraland is still native default scenes, that would be 54,000 ferns growing rampant in the wild, contributing a whopping 15,120,000 triangles to Decentraland.

But wait, there’s more!! These beautiful Sweet Peas add 428 triangles per plant, and the Green Acacia tree adds an astounding 1284 tris per tree, and occur at least once (sometimes twice) per parcel. This means that if we took out 1 fern, 1 sweet pea, and 1 acacia tree per parcel, we could effectively reduce 107,568,000 MILLION triangles from Decentraland. :astonished:

Additionally, a quick walk around default parcels will show you that each rock and log on the ground has a collider (EVEN THE SMALL ONES! Lol). Parcels average 5-8 rocks/logs per parcel. If we just removed the colliders for those rocks I think we could expedite this process and find out very quickly if removing colliders will improve performance in world.

I realize this is off the topic of what you are proposing, I just think it’s an approach that might work well with what you are looking into and would be an approach that won’t infringe on builder rights on their own property.

This was a brief study of the natural flora in Decentraland, I’m sure a more in depth analysis would show more invasive plant species that could be eradicated from our beautiful (triangle heavy) world. <3

2 Likes

Triangles are not the main raison of performance issues (at least at Decentraland scale, where we are talking about 10k tris per parcels).

WebGL Aquarium I can get 60fps with 20,000 fishes in a web browser on a small laptop
WebGL Field

3 Likes

Thanks HP, So I don’t need to cut down all the ferns and acacia trees?

But more importantly, are colliders more performance heavy than other assets? Would reducing colliders make a significant impact? Thanks :slight_smile:

1 Like

I agree with Canessa and HP about tris not being the main issue and that there are better ways to enhance performance.

In the pre proposal, it was stated: “The current avatar base hands have 1,200 triangles each” however, there is 278 per hand. The fingers are already squares and the joints are made in the standard way to animate flexion.

@Canessa, I am not 100% in favor of putting limitations since I am aware that too many rules or too many restrictions create an entry barrier for new users, however I think that if we do not get Decentraland to go faster it will be very difficult to attract new visitors .

I have raised this proposal because I believe that at this time it is a priority to speed up the loading speed, another thing is that this proposal achieves a negligible result (which I do not know about), so @HPrivakos’s contribution is very valuable in this thread, since who has given a very appropriate example with WebGL which helps us to evaluate this idea more objectively.

At the same time I believe that big changes are made up of smaller changes, sometimes insignificant, such as removing the colliders from all the rocks that make you trip every 10 seconds.

Whether removing colliders is a substantial improvement in terms of user scrolling quality and loading time is something I trust will be debated here.

Thanks PinkPunk! I’m really glad you’re asking these kinds of questions! I’m voting no for now, but I really would love to keep this conversation going!

1 Like

I dont see any reasons for reducing triangles. For example, Spatial scenes have thousands and millions polygons and works very smooth in Web. Decentraland requires code and server optimizations to reduce freezes, crashes, and FPS dropdowns. Reducing and polishing models will not affect on FPS that much.

3 Likes

Reducing tris won’t make a significant difference in the long run (especially by a couple thousand, I’d understand more so millions), optimizing the game engine however would. New updates get bought out but the game still runs just as bad even with some higher end PCs and we have designated videos just to make sure the game can run on mid to low end PCs.

2 Likes

Is it worthwhile to reduce avatar tris and colliders to improve loading speed?

This proposal is now in status: REJECTED.

Voting Results:

  • Yes 26% 1,112,337 VP (36 votes)
  • No 54% 2,171,332 VP (40 votes)
  • Abstain 20% 872,254 VP (14 votes)