The 70,000 polygon count for a Very Poor rating is perfectly fine. I’m not sure why so many users think that the performance ranking system needs an “overhaul.” The problem has always been that there is no limit beyond that. So an avatar with 3,000,000 polygons is the same performance ranking as an avatar with 70,001 polygons. Insanity.
4 Likes
While we encourage discussion and feedback in the replies always remember to remain civil and respectful. Anything else will be deleted.
5 Likes
Are you all aware of issues involving AM5 stability & crashes with UnityPlayer.dll exception code 0xc0000005/GPU TDR crashes? Had multiple friends with same GPU/CPU crash the same way (9800x3D + 5070Ti) using VD on a QPro.
Any idea what causes these crashes?
1 Like
Edit to be clear i think this is a really good addition! i realized this post may make it seem like im against performanfe gating, but i really think its a great idea.
Its just really hard to hit with some bases. When 60k polygons are taken up by the avatar base alone, and accessories for it are 40k, a ton of manual work is needed to make it functional. When you gotta shave off that many polygons, merely decimating the mesh isn’t enough and would likely just break the weight painting. The level of skill and blender knowledge needed to significantly reduce polycount is far higher than the amount of technical know-how needed to reduce the number of textures or memory used.
And like… switching bases isn’t really a solution, cuz at that point its no longer your avatar.
Hitting every other performance limit is comparatively trivial. Like, its two button presses to lower texture size and drastically reduce Vram usage. Thats why i think the poly limit keeps getting talked about: its hard to fix! And fwiw, i would also like to see it raised slightly for medium and poor avatars, provided the performance impact isn’t significant.
4 Likes
honestly I would say that’s kind of a failing on the Avatar base maker part to make it so high polygons that there’s no budget left for anything else
3 Likes
Maybe for the popup add in a button to it that let’s the user generate an imposter version if one hasn’t been generated yet?
That way people will get at least a form of their current avatar generated from within the game without the need to go tbrough the website.
2 Likes
Will there be an icon indication when a group instance is Avatar Performance Gated, similar to the icon for Age Gated (Verified 18+) instances? The question is if users can check the current avatar performance gating from within an instance and before attempting to join one.
3 Likes
Are there any plans for actually adding a “poor” range for the poly count of avatars? Too many creators put their avis at just shy of the 70k that immediately take you from medium to very poor and it leaves no space for adding accessories
Also, you should consider adding another option to the performance gating to let avatars be gated by download size or vram usage. A lot of avatars end up “very poor” despite being comparably light, and a lot of the performance issues I get in larger instances are due to my VRAM getting capped out by people with too many outfits and 4k+ resolution textures
2 Likes
You arent wrong! But also… pointing that out doesnt actually fix anything for anyone lol. Like i said, switching bases isnt a solution because then its no longer your avatar.
There may not be a solution, tbh. It just kinda sucks in here and now.
Welp, time to learn how to retopo lol
do agree, the rankings do need to get adjusted slightly as the average avatar base usually falls under poor already with maybe some decimation pushing it to medium , add in a marker or prop and you are back in poor/very poor.
2 Likes
This is an incentive for avatar base creators to make more optimized avatars. It could even be as simple as offering the primary base, and then offering an “optimized” base which has reduced poly counts other other reductions.
This is my entire avatar for reference, which has two pants and two shirts, custom hair, complex tail, a 4K body texture, and repeating textures:
This is an incentive for people to spend money on people who do know how to optimize avatars.
I guess it depends on the base. A lot of users manage to transfer their identity across avatar bases pretty well. If you are certain on making a particular base work, then it would be worth it to find ways to optimize it.
As mentioned before, the performance impact would likely be significant. Every polygon has a theoretically infinite number of cascading effects done on them. Each render pass may have multiple steps which have a cost scaling with poly count.
PolyQuilt! PolyQuilt Retopology Tool — Blender Extensions
A video I really like on how to do retopology: https://www.youtube.com/watch?v=OLwtul3ngB8
Also, it’s a bit outdated, but a video I made on the basics of Blender, to hopefully encourage people to feel more comfortable in the viewport and the program. I think this is critical since it provides a foundation for navigating Blender as a whole. https://www.youtube.com/watch?v=rEfCs_9LzeQ
2 Likes
That said, you are not allowed to override the instance level restrictions to make other’s appear, even if you force-show their avatar or are friends. This is to prevent the slightly awkward dynamic of players asking others to show their avatar, thus defeating the point of the feature.
I’m not even going to get into the ‘experience’ discussion. Just make sure there’s an option for instance creators to allow a bypass for users, so they can control their experience, what they see, based on their systems - if the goal here is genuinely to support performance.
4 Likes
My guess is that a full block on joining, and preventing users from switching avatar is way harder to implement.
There is already underlying code to handle hiding avatars/showing imposters in various circumstances, but you can change avatar through the website, or via a API call. It would be really difficult to prevent that from working.
1 Like
Will groups have the ability to control with role permissions who can Avatar Performance Gate group instances, similar to the separate Create Age Gated Instances permission? Or will avatar performance gating be available by default to everyone who has Create Group Public Instances / Create Group+ Instances / Create Members-Only Group Instances permissions?
Note for clarity, Create Age Gated Instances doesn’t grant instance creation permissions by itself, only the permission to control who can enable gating (afaik).
Can the Avatar Performance gating be edited/changed retroactively (with appropriate permissions, if any) for an existing instance after a group instance has been created, or will it be only configurable at instance creation time?
Some group events may start out without avatar performance restrictions, but ask later to use more performant avatars. e.g. as more than expected people show up or performance suffers.
1 Like
This is an incentive for avatar base creators to make more optimized avatars. It could even be as simple as offering the primary base, and then offering an “optimized” base which has reduced poly counts other other reductions.
Ita really… not, though. Avatars are still coming out where the base eats most of the 70k polygons. And tbh? i think its because theyve given up on the performance ranking system.
70k genuinely is a hard cap to hit and if every avatar above 70k gets forced into the same very poor box, at that point why try? Itll just get blocked by most people anyways. I worry the unified poly limit is having the opposite effect of what they hope. the monkeys paw could always curl and make the limit even tighter at higher performance levels, but tbh? i kinda think thatd still be better than what we have.
This is my entire avatar for reference, which has two pants and two shirts, custom hair, complex tail, a 4K body texture, and repeating textures:
Thats super cool! But also, the base i like using is 60k polygons out of the box. Ive made medium versions of my avis, but it took days of blender work and a lots of custom modeling. Its a lot of work!
I guess it depends on the base. A lot of users manage to transfer their identity across avatar bases pretty well. If you are certain on making a particular base work, then it would be worth it to find ways to optimize it.
I feel like we have wildly different philosophies on what an avatar actually means, ngl. If i switch bases, thats a different version of me, kinda? its not bad, but its not necessarily how i want to present. Switchicjg away from how i want to actually look for events sucks (which is why ive made optimized versions of my avis lol)
This is an incentive for people to spend money on people who do know how to optimize avatars.
Thats… actually super fair. Im just broke lol, so i typically do everything myself.
As mentioned before, the performance impact would likely be significant.
Ive heard the opposite regarding performance impact, fwiw (Edit: This, as it turns out, isn’t quite right! See Mystical’s comment). It makes a difference on much older hardware but has a much smaller impact on newer hardware. But to be clear, thats just hearsay, and i could totally be wrong.
Thanks for the polyquilt recommendation, though, ill give it a look! would probably beat my current method of “selective decimation and a lot of prayer”
3 Likes
There’s a lot of (almost all unintentional!) misinformation going around about avatar performance in this thread, it’s hard to address each of the times it comes up – trust me, these misconceptions are my pet peeve and my eyebrow twitches every time I can’t speak up about them. 
That being said, if you would like to talk to us during our Dev Stream tomorrow on Twitch, we’d be glad to chat a bit more deeply about it.
5 Likes
Unfortunately a linear increase in the amount of polygons does not result in just a linear increase in the time to rasterize a frame. I believe it’s closer to n x log(n)
The unity build in render pipeline does not do great with millions of polygons with the sort of GPUs most consumers have. Modern gpus are also no longer focusing on traditional raster performance, so this is not something we can just wait on a few gpu generations to brute force.
Asking for 30k more polygons per avatar does not sound like much, but that’s a max of 2.4 million more polygons in a 80 player instance.
We need to continue to encourage avatar makers to put in the effort to optimize their meshes. It’s not an easy task but it’s an essential one.
3 Likes
Is there anywhere its been addressed before that could just be linked back to? Im super interested in learning more, but dont really have time for a full stream.
All I can really say is: It seems people want both good performance and the sky being the limit for avatar customization, but that’s not how it works. If we did have infinite performance, then sure! But as-is, even just Poor and Medium can still be intensive at times due to certain factors.
I no longer will accept “can’t” as a reason for not optimizing avatars. People either have to learn or find/pay someone to do it for them. This is a social platform whose purpose is to allow people to communicate, hangout, and express themselves. There have to be sacrifices in some way, and in this case, it’s a relatively standard limit targeting average hardware.
The higher the limits go, we exclude people with worse hardware, which is especially hard to justify in this current tech market.
If you want your avatar to fit within the performance limits while also expressing yourself as you wish, you have to put in the work. There is no way around this.
I started learning this stuff after spending a decade not learning it because I finally wanted to fully express myself. Learning Blender and Unity better has provided me the opportunity to not only optimize my stuff, but express myself in whatever way I want.
There are three choices: Do it yourself, pay someone to do it, or don’t get what you want.
Edit: I should add: I have in the past tried to make resources to help people get a full picture of the content creation pipeline for VRChat. My goal has been to provide all the steps needed and context to the entire pipeline, to reduce friction for this process and give people the tools to do so.
I just haven’t yet due to ADHD and other life priorities. Plus there are people who are much more talented than me that have been putting out resources. If there is a need (especially now) to have a centralized place to learn about this stuff, then I’d be happy about setting up a website.
There is already VRC School, the VRC docs, and other websites/accounts that post stuff for now.
The Avatar Optimization Tips part of the docs I think provides a good overview of what contributes to poor performance. It may be a little outdated but I think it should be mostly good.
3 Likes