GDFireFrog

Yes, it should be no problem to implement MSAA. I don't want to 100% promise you that off the cuff because I need to test it, there are certain limitations with multi-sampled buffers and I may be doing something in some part of the engine that may be an issue with that that I may be forgetting about. I'll update in this thread when I've tested.

It is counter-intuitive but this is true in a sense. To understand when things became worse in this area check the Wikipedia article for "deferred shading", it's pretty good.

A high level philosophical summary to explain the technological regression would be that as sophistication of the engines increased, the inner workings of the engine became opaque to the GPU, that couldn't discern the information that the engine was working with anymore. Information was lost to the GPU because the engine assumed some of the responsibility and started sending the GPU pre-rendered snapshots that the GPU just needs to put together.

It's an analogous phenomenon to game streaming. New innovation allows a controller to assume some responsibility previously belonging to a terminal component. The terminal component now has less information. This allows better performance in some regards but results in loss of functionality and some people wondering why we are moving backwards.

Yes, that's right. I would like to avoid temporal techniques (which is also probably what's used by CoD Warzone which is cited as example in OP judging by the description) since it adds a bit of visual "half-latency" that doesn't really fit the genre, because what you are doing is mixing the two last frames while jiggling the camera to detect edges. But it's not that it's complicated to implement property, it just has that blurring effect. It is used a lot nowadays because it's actually very neat: it gives you good performance and good static quality, and the downside that it has which is the blur may even turn into an upside for your game if you like the effect. DB requires a more snappy old-school 1:1 frame to presentation temporal ratio, IMO.

Now, regarding MSAA proper, it should be no problem to do that. I can't think off the top of my head of any feature that we've added over the years that would be a problem with that, but maybe I'm forgetting about something, I'll co...

Yeah, sorry. It sucks that we had to take out some features. We'll bring that back as soon as our in-house hud engine is more feature-complete.

No, I don't think that's what I was mentioning. Going by your description that things "go into slow motion for a second during a fight", I think that these are specific performance issues related to specific gameplay events. When we finish some stuff we need to do for launch I'll analyse some demos of similar fights and try to find the issue. As an aside, when we start storing server side demos it'll be easier for me to find specific fights that people show me and look at what's going on in specific moments and this kind of stuff should improve quicker.

Alright. Thanks for the update.

Thank you for the help, this may have gone unnoticed for quite a bit, actually, if it wasn't for some details in your thread.

Neither of you elaborated on what you don't like about this netcode, but v294, published a few hours ago, fixes a congestion issue that improves the server-side netcode significantly for some people. This may make you like the netcode a bit more if you are below the backwards-reconciliation ping threshold of 120ms that we have set. If you are over that ping, then, unfortunately you'll now have to do some estimation for placing your shots equivalent to the difference between your ping and said threshold.

But if your ping is less than 120ms and you have tried v294 and you still have issues that you didn't have last week please DM me on Discord if you don't mind, I'd like to hear about that.

Note that there will still be a small difference of accuracy of a few percentage points even if everything is working as expected. We are now testing a conservative backwards reconciliation implementation before we add more tech on top to compensate for the shortcomings of BR. This...

Added this for 295.

Could you DM me on Discord?

To update other people having similar issues. I had a chat with VegetableVampire and I suspect this is a congestion issue from a less than ideal connection to the server and not an issue with our backwards reconciliation implementation. The main thing I want to focus on right now is that he has good aim with pncr but poor aim with shaft. Shaft is a lot of packages very fast so if the connection is not great and our transport has some issues that's gonna cause some congestion and momentarily increase your latency. With the latency changing so rapidly the backwards reconciliation algorithm starts failing.

We have a tentative fix for this and it's being tested. For anybody having a especially bad experience with LG compared to other weapons, I'll update when we think we have a fix.

UPDATE: Got good feedback on the fix so going ahead and pushing this as v294.

Ah good to know, thanks for confirming. Unfortunately Diabotical may be more likely than other games to expose issues like this as it leans very heavily on the GPU and tries to keep CPU usage lightweight (1,2). This is desired for stability/smoothness because, to oversimplify, GPU workloads are more predictable (less variability between frames). It has the inconvenient that cards get a bit hot and people freak out a little.

(1) This is not to say that our current GPU power usage is entirely justified in relation to the frame-rates we are achieving, this can and will be improved over time, and things won't have to get quite so hot to get the same FPS.

(2) This is also not to say that we have been completely successful in making CPU usage super light either, we are getting there.

So far I haven't seen a delta so high, people report a few % difference, and we are going to improve that, right now we are trying a pure conservative approach to make sure it's completely solid. But it shouldn't be that bad right now. Could you DM me on Discord? Thanks!

Ah, sorry, actually I meant "report" in the general sense, as in just saying so on this thread for example. We don't have a bug report form in game yet.

When you say that the screen gets frozen and yet you "can still play", I'm going to assume that you mean that the visuals become static, nothing changes, but you still hear sounds as if the game is still going on.

Your settings folder is C:\Users\AppData\Roaming\Diabotical (You can there by going to %appdata%/Diabotical in Windows explorer or in Windows+Run)

After closing the game, rename "settings.txt" in that folder to "backup-settings.txt" to reset your settings without losing them.

If that fixes it, send us the offending "backup-settings.txt" for analysis. You can send me this on Discord or you can put the context in a pastebin and DM me the link on Reddit.

If that doesn't fix it and you have the Nvidia overlay enabled, try disabling it.

If that still doesn't fix it:

• Can you tell me if you are on a laptop?

• Can you tell me if you on a setup where you are using multiple graphics cards simultaneously?

...

Please report if this happens again after reading this message. We pushed a tentative fix for this to the servers 20 minutes ago, though the bug had more to do with picking weapons than swapping to them. Could it be that's what you both experienced, were you actually getting the weapons in the weapons list? Thanks!

For the record, you shouldn't feel bad about that :) We work with both objective and subjective feedback, and sometimes subjective feedback is very important where hard data is incomplete or hard to analyse. In fact I do often ask specifically for "subjective feedback" on changes, like having someone jump around shooting at things for a while back and forth between two versions and just see what feels better. You'd think this is error-prone but it is actually exceedingly rare that a player gets it wrong when it comes to telling which is better between A and B (at least in my experience). The same goes with "feels like X" statements as you mention. There is more commonality with the way we choose to describe sensations than one might expect and it is rare that things are described in a way that is so abstract that you can't make sense of it.

OK, we might need to subdivide this issue yet again because there's also two things in here that we could be talking about.

The issue that affected mouse input, the one that we think we have improved/solved, also affected keyboard input. That would be conceivably consistent with what you are describing, if so this may also be solved.

To know if the issue is the other thing that I was mentioning in my post (i.e, you find friction is just too low), during the next test, start a custom game pasting this in the Commands field:

phy_surface_friction: 20;

phy_accel_ground: 20;

This is going to make you super grounded, pinning you to the ground as soon as you touch it, kinda like Super Meat Boy. It's an extreme so you can see if that's at the root of your feedback. The ground acceleration is increased in this example to compensate for the increased friction, the value is not exact to keep things as they were, just ...

Regarding this issue with multi-threaded input:

https://www.reddit.com/r/Diabotical/comments/i31g0s/game_started_feeling_slow_or_floaty_is_it_just_me/

We made several improvements and fixes to multi-threaded input yesterday. We have had good results in our internal testing group.

• People who previously felt ON was better than OFF, now think the the difference is more pronounced and multi-threaded is even better. The qualifier "crisp" has come up a few times.

• On the si...

When we change a map we are supposed to bump the version of the game and force people to update after they finish their current game, otherwise the game server is going to see a different version of the map as if your version is outdated and you won't be able to start the game.

We updated Skybreak last night (morning?) and the client version failed to be bumped. Raven stayed around removing penalties manually for a while to people affected. (Hopefully your penalty for this would have been removed too, if it wasn't let me know). We'll make changes to the system today so this doesn't happen again.

There is still an edge case that we need to solve at a technical level, and it's to notify server instances running continuous games.

As an aside, most of these map versioning issues will be solved when we introduce the final content delivery system for map delivery that we are gonna use that allows the community to contribute maps, in this system everything will hav...