343 GDC Presentations on Infinite

For anyone interested, here are the presentations 343 did on Infinite at GDC, now available in the GDC vault:

One Frame in Halo Infinite [This is a focus on CPU simulation of game state, not GPU rendering] - Video: https://www.gdcvault.com/play/1027724/One-Frame-in-Halo-Infinite - Slide Deck: https://www.gdcvault.com/play/1027657/One-Frame-in-Halo-Infinite

Building Zeta Halo - Slide Deck: https://www.gdcvault.com/play/1027569/Technical-Artist-Summit-Building-Zeta - Video is not free

How Infinite’s Bots Make Decisions - Slide Deck: https://www.gdcvault.com/play/1027689/Thinking-Like-Players-How-Halo - Video is not free

7 Likes

Did not know they did any presentations.

Imagine if 343 had a community manager.

3 Likes

Well, these presentations are aimed at an audience of game devs. GDC - Game Developers Conference - isn’t a free event, and only some of the presentations given at the conference are made free after the fact. However, they did put out an article pointing out they’d be at GDC doing presentations that they tweeted about - https://www.343industries.com/news/articles/343atGDC

Most of the community doesn’t care about this stuff, to be honest. It’s more a behind-the-scenes peek at work they’ve done that they think is noteworthy and that they want to share with other devs, who may find it useful. For instance, in the “One Frame in Halo Infinite” video, at the end during the Q&A section other game devs in attendance were asking questions about their work, with one even stating that they’re struggling to implement something similar, but in reverse.

6 Likes

Are any of these the talk one of the technical engineers was going to give about the SlipSpace engine? There have been quotes from some of the design leads claiming really high praise for what SlipSpace is supposedly capable of, so I’ve been keeping my ears to the ground for that one to go public

Yes, the “One Frame in Halo Infinite” details some of the work done going from Halo 5’s engine to SlipSpace. It focuses on the CPU-side architecture, and not the rendering/GPU side, but it is interesting nonetheless, IMO. They detail how they reworked things to support arbitrary framerates on PC and Xbox for calculating the game simulation and how it was set up previously. The presenter does a pretty good job of simplifying the concepts - I’m no coder and could keep up.

2 Likes

I’m 15 minutes in and he’s already explained a high level view of the cause of H5 heavy aim, and given a high level overview of how Blam! had reached a point of crazy inter dependent code systems that were so intertwined it made it hard to onboard new employees (contractors) :nerd_face:

2 Likes

Yep, really seems like SlipSpace has undone a lot of that (at least judging by the presentation). Engine seems like it’s a more solid foundation now than most are giving it credit for these days.

I’m not able to watch the video right now, and I also don’t have a lot of coding or game development knowledge. If the new engine makes it easier for contracted coders to onboard, does it shed any light on what may be causing problems with so much of the game at the moment?

1 Like

The only thing the video may shed some light on is desync. Desync is an issue with differences in the calculated gamestate between the server and client resulting in different results between them, when they should be in sync. It seems like there may be some wiggle room around gamestate calculations and how they try to compensate for them at the server.

However, the rumored/often discussed issues with Infinite - UI being built to only handle a small number of playlists - this is a design decision that ended up just not going the way they thought. The UI was specifically designed to run a small number of playlists, like most other F2P titles, and it just didn’t go down the way they wanted. This is not an engine issue (as in bug), this is a design issue that will require the UI to be rearchitected. (This is not covered in the video, to be clear)

The rumored contractor issue - IMO, non-issue, contractors are common in software dev. Might have been initial communications issues with how technical teams are supposed to set things up for the new systems, but overall seems like the new engine is better for onboarding as they’ve undone a lot of spaghettification and reduced the high level information for any system to be available in the single file for that system - big, important step for making it easier to understand. (Covered only tangentially in the video)

As far as slow content, this video doesn’t really touch on that at all, but basically I think the below timeline sums up my thoughts:

  1. It took a long time to get the engine reworked to where it is now
  2. Due to the above, the game’s development cycle was tight
  3. Once the game hit the public, the public tore apart some of the design decisions
  4. This caused a reprioritization of work that needed to be done to accommodate player feedback
  5. BTB broke for a month and required resources to be reprioritized again, along with important bug fixes being identified
  6. Given the above items causing reprioritization and a re-evaluation of design decisions, it has taken significant engineering time to work on setting up a track to incorporating feedback (scoping how much work, how many resources, how long, which systems are impacted, etc), and has upended previous roadmaps

Overall though, this video simply lays out the results of the work they did, and some of their methodology, for updating the engine to be more adaptable to a wider variety of hardware. They did a good job utilizing the hardware better and not tying it as closely to any specific console architecture. But they do note that they’ll be doing work to update the engine to further embrace this model, which will be interesting to see over time.

6 Likes

Thanks for posting this; always interesting to peak behind the curtain.

3 Likes

Thanks for the links.

I’ve only watched the One Frame in Infinite one.

Gives a great perspective on how much the engine has changed since Halo 5.

Overall though, gives a good indication / impression of their confidence in the process going forward.

1 Like

I am convinced that the people who approve this kind of stuff don’t play Halo (or any video games for that matter). They just look at other successful games and think they can just copy pasta stuff into Halo and it will automatically be great.

1 Like

I think it’s more market research than higher ups being out of touch. A lot of gamers just want to fire up their game, hit play, and off they go. Xbox and 343 do lots of market research, and sometimes the general gaming market doesn’t see eye-to-eye with the market for a particular game. IMO, that’s likely what happened here. A lot of F2P gamers probably got interviewed/sent questionnaires and answered honestly about how they play those games and what appeals to them about it - they’re uncomplicated and finding a match is quick and easy and there’s not a million options to get lost in. And, were I to put myself in 343’s shoes and I’m making a F2P Halo MP, I have to try and strike a balance of F2P norms/expectations and Halo norms/expectations. Unfortunately they erred too far towards F2P for most Halo players’ tastes, but I understand the why behind the decision. Personally, I know the feedback’s been heard and I know it’ll change down the road, so I’m not too hot and bothered by it, nor do I think that 343 or its management don’t understand Halo. They simply tried something and it was a thing we didn’t like.

No problem, I enjoy this kind of stuff - like a Digital Foundry video except it’s direct from the studio, with more details. I found it refreshing to see that there was substance behind the confidence in the game/engine, which is something you really only get from an engineering standpoint, and not from PR/games “journalism” standpoint.

1 Like

Their copy pasta technique is not new with Halo Infinite. They’ve been doing this since Halo 4. And they’ve been bleeding players ever since. I would think they would have learned by now. But alas, 343 continues to think they know what the players of Halo want better than the actual players of Halo.

It’s sad to see 343 listen more to generic FTP game players’ feedback than actual Halo players’ feedback.

1 Like

I think the difference between Halo 4 and Infinite is that with 4 they tried to make Halo more like CoD, whereas with Infinite they’re trying to do a Halo take on F2P. I think there’s an important distinction in the mindset between the two. Personally, I’d be annoyed if they basically just regurgitated out Halo 3 for the 4th time, so I’m glad they’re trying new things, even if short-term there are things I dislike, as I think long-term Infinite will shake out to be a good, if not great, Halo game. It’s got hurdles to overcome, absolutely, but the mindset is right - to do a Halo take on the genre instead of rearranging Halo gameplay to fit the genre, it’s just finding the best balance of exactly what that looks like. End of the day, UI stuff is a support system that can be rearchitected. The gameplay is right for Halo and that they didn’t go out and clone Fortnite, Warzone, or Apex is the most important piece of foundation to have.

Per the GDC presentation, it appears as though the tech is much less of a hurdle than it used to be overall, and will be continued to be iterated on going forward, so that bodes well for improvements in the future.

1 Like

Halo Infinite has the potential to be the best Halo game ever. The pieces are there.

But there are a lot of obstacles that will need to be overcome and 343 has been shooting themselves in the foot with many of their visible responses to feedback. Perhaps there are actual good changes going on behind the scenes but it seems like any substantial improvements will not happen soon.

TBH, I think most (not all, to be clear) of the community’s upset at their responses has been the community looking to be angry, as outrage seems to be a hobby for gamers these days. But I’m also an old guy who doesn’t care about stores and FOMO and just plays the game with friends to hang out and have a good time, so maybe that’s why I’m less cranky. However, I digress - the GDC presentations don’t touch on the design decisions of these systems or the mindset behind them or their responses to feedback.

End of the day, those presentations have laid to rest the majority of concerns I had about the tech being a hurdle - it seems like while they haven’t completely eliminated the tech debt, it has been greatly reduced, and therefore they’re set up well to actually make headway on things from a technical perspective.

2 Likes

Little_Squash isn’t one of those types of people. Just so you know.

I do understand where you’re getting at but that outrage doesn’t come from everyone.

1 Like

Haha… Me too…

My issues with the game have nothing to do with the store, or FOMO, or cosmetics in general.

I’m more concerned with the actual playability of the game.

Like desync… It makes for a frustrating experience to watch rockets do no damage, grenades disappear, shots through walls, getting backsmacked from the front, etc.

Or MMR… Why does how I play in Fiesta or BTB or Swat affect my skill placement in the ranked playlist? Does 343 think there is no difference between social playlists and ranked playlists?

Or the lack of playlists… Hopefully you are correct that changes to the UI are coming, which will allow for better game type selection. There would be a lot fewer quitters if people weren’t being put into games that they don’t want to play.

Or collision… When 90% of players want it and 343’s response is that the feedback is “mixed” so they are not going to change it. That doesn’t make sense to me.
༼ づ ಠ_ಠ ༽づ Summon Collision

1 Like

Not directed at Little_Squash in particular, just a generalized statement about the community, just for clarity.

They did touch on desync in that video a bit (Q&A section), and I think this is the big one that’s impacting the game experience. I think their mechanism to rewind time on the server to process the client-side updates appropriately is where this issue comes from. Being that they’re having to go back and update simulations with data from clients, my suspicion is that there’s something buggy in how the updates are ordered for processing or something screwy with number of updates from clients with unlocked framerates is happening. The presenter was the principal architect, and seemed competent and familiar with the system he/the team designed, but I think this is something that’s going to be hard to debug, based on what they showed. I was also a little concerned that the VMs running the server instances are two thread. It’s a lightweight server instance given the limited computing that needs to be done since there’s no audio or rendering, so it should be fine in theory, but I’d prefer more overhead than that.

Edit: I’m pretty sure, thinking on it, the time rewind thing is common for most server architectures for games - I’m imagining this is a bug related to the jobifying of the game updates, since that’s the delta between the old servers and new ones (at least in theory, given my limited knowledge). Perhaps an odd interaction between the rewind mechanism and the jobification. Either way, I do know they added debugging to the game client to look into this more closely, so hopefully there are improvements.

3 Likes