Caffeinated_weeb:

I don't really know who these "many people" you speak of who think optimizing makes an avatar ugly are, as I do not have access to your memories or your experiences. But based on my own, I have not really seen people in public instances of The Great Pug, or Japan Shrine, or The Box (worlds I frequent almost every night with my friends) who have very poor avatars that are "5 or even 10 times" the suggested values, and I have literally NEVER met anyone say that optimizing their avatar makes it ugly. I have been playing VRChat since November of 2017, became involved in several streamer communities and spoke to a great deal of users from Xigneon's community, Naggz' community, Ikrium, Zentreya, Nurfee, Pokelawls, and a few others. I've been in contact with some of the friends I made from those communities since then, and never heard any of them make the claim you just did.

I will tell you exactly what I DO see. I see people using models created by an avatar creator named, Yuumi. These models typically are rated "very poor", however, they do not exceed the recommendations by very much. Most of them are already quite optimized. I think these models are extremely creative and fun, and optimized just fine for VR.

I also used to see models created by Mitch Murder. Mitch Murder's models were also "very poor" rank, and yet, again, they were not "5 or even 10 times" the recommended values in any category. However, I no longer see models from this user anymore because their world was taken down. It was taken down after this new avatar ranking system was implemented, and furthermore, my understanding is that it was taken down BECAUSE of this ranking system.

Which is really a shame, because those were very nice models and I enjoyed seeing people using them. I did not experience any serious frame drops around people using them either. My specs used to be an i7 4790k, with 16gb of ram, and a GTX 980 Ti. A rather low-end "VR ready" machine, compared to what the game really operates best on, or what most streamers run, or several of my friends. And still, I kept a stable 30-45 FPS most of the time, using my Vive and full-body-tracking, even with up to 20 people in the same room, with real-time lighting, and a mess of dynamic bones, colliders, and mountains of materials.

Now, my system specs are much better than they were before, following some upgrades this Christmas, and I hardly ever drop below 45 FPS, no matter WHAT is happening around me. While I realize not everyone can meet the recommended specifications to run this game, and not everyone can afford to upgrade their machine to something more suitable, I also realize VRChat is not a very well optimized program, and sympathize with the complaints about that to an extent. However, I do not sympathize with any of the people who are trying to run this game below the minimum REQUIRED specifications. If you can't run this game, either you stop acting entitled and accept that it won't run well, or don't play it. It's as simple as that.

冬寂Wintermute: 45 fps is lower than intended. The game should ideally be running at 60fps.

冬寂Wintermute: One of the arguments I've heard against optimized avatars is "I don't want my avatar to look like a PS2 character!" which only stems from misinformation. I'm not sure why you haven't heard it yourself, maybe you're just around experienced users in general.

But in that case, I'm curious, why do you think people make "very poor" models?

Finding avatars which go multiple times above the limits is easy. I frequently meet people with 250K polygons, 50K+ particles, 50-100 materials, 50+ meshes, 2000+ dynamic bone collision checks(!), etc. Not all on the same avatar of course, but still.

Some of these were from public worlds (Not sure how they've managed) but a lot are also uploaded by people themselves.

And even if avatars are just barely above "very poor", they're still that, very poor. I remade a friends "very poor" avatar as "excellent" to show him how easy it was. It only took me a few minutes, I'm not exaggerating.

And these are not just excessive, they're maliciously bad.

They could be running 10 times faster with no real effort and no visual differences. Their laziness is preventing other users from enjoying the game, and it's only right that they get a red star and "very poor" title.

A 980TI and a i7 4790K is not low end. Most of VRChat is running in 1 thread, and that CPU is only 15% performance away from being the best in the world at that (Google cpubenchmark). And the 980TI is much faster than the recommended 970 (which I'm curretly using), so if you're getting 30 FPS (which by the way, is objectively bad performance in a video game) somewhere, then I'm likely getting 10.

↻ Protagonist-kun: In general all VR applications should run at a solid 90 FPS, 60 FPS is only acceptable for desktop mode.

TCL: Ah you're right. This makes 45 completely unfeasible.

Demirramon: Did you read the original post? This isn't about being against the system entirely, it's about how broken and misleading it is. Your avatar is immediately labeled poor for having a single dynamic bone collider on it, and the bone component itself drops the perf rating a ridiculous amount. Anyone that actually plays this game can tell you reasonable amounts of dynamic bones don't drop frames by a significant amount. If they say otherwise, or even suggest cloth physics instead, they're terrible trolls and definitely don't know anything.

Polys don't matter and material count is equally misleading since particle systems factor into that, but it's really shader draw calls that impact performance enough to warrant any action to optimize the avatar.

We want the system to give a proper weighted average, and I've heard it's in the works.

CurtisVL: " try joining a Great Pug 20+ "

Right.

It has nothing to do with how awful the PUG performs without anyone in it as it is.

So all avatar creators should just cater to the worst offenders of world frame loss.

No.

Zennshi: You do realize the Pug is pretty well optimized, right? The biggest issue avatars in that case.

You can only optimize something so much, it'll never be a 'lag free box'. :p

CurtisVL: How many interactable world items are in the pug? Ever joined one where someone kicked them all around? It almost crashes people on entry.

Zennshi: My point isn't that it's a high performing world, it's that it's as optimized as it can be for being as high detail and intractable as it is.

the pickupable items should always be rendered, as dynamic objects can't have object culling AFAIK, however, whole other floors of the Pug will 'despawn' when you're not on them.

Check out object culling for yourself and realize that the Pug isn't as bad performing as you think before you criticize it so heavily. Your personal experience may differ to others. :p

CurtisVL: ohhh, so you do understand nuance. You just have a gross favoritism towards map makers vs avatar makers. Okay. Now I get the kind of person I am dealing with.

Everyone I know thinks the pug is a shit map in terms of optimization. Period. I do too. Period.

Zennshi: There's a difference between opinion and fact, the Pug has had a lot of optimization done, as much as you can really do, that's a fact. Developing a map is no different to developing an avatar, both are just content creation, same types of optimization are done for props in a world. :p

If a world developer is lazy with their optimization it's equally as bad as an avatar creator doing it. It impacts users in the same way.

Zennshi: What I meant was that even though the amount of dynamic items in the Pug can be pretty performance impacting, it's as good as it can really be in terms of performance before removing content.

I just don't really understand what you're getting at, or why you're so mad about it. I don't see why it's not obvious to you that the biggest cause of frame-loss is yourself, why be so fast to blame everything else first? Do you realize that almost every other game with custom content has hard limits to avoid the laggy experience VRChat has become?

As soon as I realized the impact my avatar was having on other's performance, I was quick to learn how to optimize and tips and tricks to keep everything looking good whilst doing it. I didn't want to destroy other people's experiences with my own ineptitude.

Evidently some people don't think like that... :/

CurtisVL: Okay... so you do understand nuance.

You also don't know what avatars I run, so no the biggest cause of frame loss isn't me. My performance impact relies heavily on other people's specs and if their computer is bad enough being in a green avatar alone will cause them frame loss.

Everything causes frame loss to some degree. Which is what people like you don't seem to want to get. The baseline is the way and efficiency that the game uses resources and processes data, that alone determines how much anything will impact your performance. First impact is the world itself, the the number of avatars is next, then IK, then nametags, then the actual complexity of the avatars themselves.

You think performance should come at the cost of avatar creators.

I think the game itself needs to do the most it can to be the most effective at using a PCs resources to allow everyone to make feature rich worlds and avatars.

Optimizing is easy. It has nothing to do with my inability to optimize but rather their unwillingness to make their client more potent.

Are there people whose avatars are draw call heavy for no reason other than sheer laziness? Yes. Malicious? Yes.

Is that my issue? No. It isn't. So please stop talking about goofy extremes, I am not addressing that aspect of it.

The current system is based on a simple tally next to basically zeros. It is shitty and lacks an actual understanding of how draw calls work.

If you are one of those people with 50+ materials and 5+ meshes of over 250k polys and 200 dynamic bones. Then I can understand why you seem to think other people are like that.

I don't make goofy stuff like that. If I do have lots of materials it is more so I can keep the option for changing the materials in the future. But I don't believe the system takes into account same materials vs unique ones.

Seriously. I'm done talking to you because you honestly don't seem to get a lot of things and you seem to want to believe I'm stupid or malicious or lazy or something. You honestly have no earthly idea the amount of effort I put into things I upload.

While you are concerned about polycount I am concerned about proper mesh deformation and proper weight painting. Maybe you don't give a damn about how bad decimation ruins weight painting but I do.

While you are worried about material count I am worried about a proper and attractive shader application. Maybe one without boatloads of sliders that you don't need in reality.

You worry about dynamic bones or cloth performance, I worry that they look awful. Or moving parts clip.

You seem to get that in order to achieve certain things the performance hit is inevitable - when it comes to maps. So I'm just gonna say this - its the same for avatars. Maybe your standards are lower for what qualifies as a good avatar - which is fine. But when I set out to actually make a quality avatar - I only do what I need to for the effect I want.

To sit there and tell me I'm just lazy and need to compromise -

No.

VRchat will have to ban me first.

Layfe_Kuro: thats the thing it really depends on how many people are in the room and even if you would block avatars it would still lag its not really the avatars which kill your frames more the amount of people that are in a room

WandDschungel: indeed... the amount of avatars per amount of screenspace should be calculated,it is exactly how Horizon Zero Dawn can render 1500+ full game object with almost no lag in real time, the area visible to the player is rendered,so a 100x100 Block map for instance could handle 800 full detail world objects and maybe 12 70K poly avatars with 5 dyn bones and 3 colliders, maps like the great pub are pretty small(20x20 blocks i think) but are still set to 25+ avatars while also having 120 world objects etc and has alot of lag that gets blamed to unoptimized avatars

Bleu: That I agree wtih, another good example is that an avatar could have some Dynamic Bones with no colliders, but be ranked Medium, whereas an avatar with 15 materials that causes frame loss could also be ranked Medium.

I'm sure they'll try to iron this out over time, it is the first iteration of it after all. :p

CurtisVL: They had ample time to test, no actual need for it to exist, and plenty of feedback, and pushed it through anyway.

That excuse won't fly.

Zennshi: My point isn't that they didn't spend enough time testing it, which I totally agree with, my point is that it'll get small changes over time as they're needed, like what has happened with the trust system.

More than likely it'll all be behind the scenes or small changes to the current recommendations, which in some places are a little strange, but others perfectly reasonable.

I don't understand the big issue people have with this, have you seen how nice of an avatar you can create and still be ranked Good/Medium?

It's not a case of 'the limits are too low', it's a case of we're all used to laggy dynamic bones and lack the knowledge to optimize our avatars.

(Though, I do think some of the recommended limits will be upped slightly)

CurtisVL: Why on earth would that need to be done in secret?

If they knew this information shouldn't they make that sort of thing public? You know, so people who actually do know can cross check their information?

That is what their problem is. They are ignorant.

Zennshi: I don't think I said anything is secret; Some stuff just doesn't make it to changelogs, with all games, especially something minor.

Maybe they'd list it as "Minor changes to performance stats system"? Who knows.

Almost every game developer will do this, not just VRChat. Why confuse a none-technical audience with technical information? Refer to SteamVR changelogs for comparison, they give a 'in a nutshell' of the changes, as does VRChat.

CurtisVL: "More than likely it'll all be behind the scenes or small changes to the current recommendations, which in some places are a little strange, but others perfectly reasonable."

That part. Behind the scenes. AKA in secret, without any sort of outside input besides their echo chambers. They have people willing to tell them why they are being dumb. Why not use it?

"none-technical audience with technical information?"

... You can find out draw calls by turning one gizmo on. They literally just have to test shit to get the results. Draw calls are measured in numbers... I don't really see what is so technical about saying "dynamic bone = 4 draw calls per bone"

"I don't understand the big issue people have with this, have you seen how nice of an avatar you can create and still be ranked Good/Medium?"

'how nice'

You mean how featureless?

A basic avatar with clipping hair, angular knees, no particles, no world particles, no frill of any sort, no props, no audio... Those sorts of 'nice' models?

I'm starting to think you just can't make models that do this and that is why this issue is so black and white to you.