New World interview — How Amazon Games brought an MMO (close) to the finish line

Amazon Games hasn’t had the best track record with shipping games, as the tech giant tries to move beyond being a seller to a maker. Previous titles like Breakaway and Crucible were canceled after poor receptions in testing and rollout. But now, Amazon is nearing the launch of New World, a massively multiplayer online game that fulfills the mission that Jeff Bezos gave to the developers: make games with ridiculous computation.

New World is a big MMO set in a fictional age of conquest where three factions battle each other for control of an island continent. The players can fight in huge 100-person fortress battles or craft the weaponry and clothing that others need. It’s a pretty deep game, and a lot can either go wrong or right. Amazon has a lot at stake in the launch, as the company can’t really afford another failure with New World if it wants to be taken seriously in the game business.

I interviewed Richard Lawrence, the studio director at Amazon Games, and Eric Morales, the managing director at AWS Game Tech, about what it takes to bring home a game like New World and how Amazon’s game technology helps. We talked about everything from the delays in getting the game done to the open-sourcing of the Lumberyard game engine amid the project.

The MMO is finally scheduled for its September 28 launch. Hopefully, this deadline will stick, as some fans have been waiting for this game for a long time.

This is an edited transcript of our interview.

Wrapping up

Above: Crafting is in New World. I doubt you could launch an MMO without it these days.

Image Credit: Steam

GamesBeat: It must feel good to be getting to the finish line.

Richard Lawrence: For New World, yeah. It’s been quite a while. As many MMOs do, we kind of grew along the way. Always in relation to what people wanted in terms of gameplay. It’s rewarding to get to the point where we can give them the tech that we’ve been testing with them for quite a while. I’m excited to see how it goes.

GamesBeat: Eric, what sort of work are you doing on your end?

Eric Morales: First off, we’ll be playing a lot of New World. New World is this perfect embodiment of a lot of the technological innovation that underpins AWS, coming together in one place and doing it at pretty impressive scale. It also covers the full gamut. It’s AWS, the AWS game tech service side, but it’s also some of the bigger, longer-term bets we’ve made, like with what was Lumberyard. I’m excited to see what happens next. We’re all excited about the launch.

MMO tech

GamesBeat: A lot of this is behind the scenes, but I wonder if you can describe what they should notice about the technological innovation here. What are some things that they might not even see?

Lawrence: You hit it on the last part there. Ideally, players won’t notice anything specifically. Things will just work the way we intend. They’ll be surprised, hopefully, by what they’re able to do or what they’re able to experience in the game. From a technology standpoint, I firmly believe that technology should be largely invisible, other than enabling a game. It should be there to acknowledge the gameplay. In a couple places we’ve done that with Lumberyard.

We have an interesting graphical system for an MMO. I’m proud to say that many players have recognized that it’s a graphically rich MMO. Typically MMOs have compromises they have to make there — how much you can see on the screen, how much is going on. We have very large battles. Not the largest in the industry, but high fidelity for what they are. We have 100 people in a very tense, spells-and-swords fight. That’s quite an experience. I think people will see that and be surprised by how much is going on in the game. But hopefully they won’t be distracted by it. Hopefully that’ll just be the game, what’s appropriate for the game.

GamesBeat: What kind of technology enables that, where you have so much happening on screen?

Lawrence: It’s a combination of the backend servers, which are on AWS. That serves all our computational needs. New World is a very computational game on the backend. But we don’t want the players to perceive that there are a bunch of servers working hard to make this happen. We just want them to see a huge battle with a ton of stuff going on. That’s the preferable experience.

From a client technology standpoint, there’s a lot going on graphically. We have extremely far view horizons, for instance, up to four kilometers in the game. That’s pretty unusual for the fidelity we’re delivering. Simultaneously with that, we’ll have these giant battles and great experiences in expeditions, the things you would traditionally expect to do in MMOs. When you’re fighting another player or fighting with a bunch of other players, everything should be consistent, high-grade, easy to understand, and not disrupted in any way by technology issues.

Big decisions

Above: Amazon Games’ New World debuts on September 28.

Image Credit: Amazon

GamesBeat: Going back in time, when you needed or wanted to implement these innovations, what kind of decision did you have to make earlier in the project?

Lawrence: There are some very large cycles — when you take a standard game engine setup that will work for pretty much 99 percent of the games on the market and you say, “I want to add 4-kilometer horizon, and I want a continuous open world that’s always streaming, and I want to seamlessly transit to instanced areas of the world as well,” those are not typical features in a game. You have to add those, or at least modify them to suit an MMO. We made some decisions there, and I was happy with the support that we got from Amazon. They said, “Do the right thing for the gameplay. If this is the right thing for the game, let’s invest in the specialized code we need.” For instance, for a very scalable backend. That’s considerable work. It takes a fair amount of time, and not a trivial engagement in any regard.

Amazon was always supportive. Or I should say because I work there, we were always supportive of the idea that this is what the customers want. Do they want to experience a big battle with a bunch of players? We think they do want that. Even though it’s a pretty significant technology undertaking, we were always clear that if we were headed in the direction of what the customer wants to play, let’s do that.

GamesBeat: What sort of tradeoffs are there with that number, 100 players in a battle? Why did you set that limit?

Lawrence: The No. 1 reason, interesting enough, is because it might not be fun. The reason we have the battle size we have is that we tested it until we felt like having more would be more of a distraction than a benefit. We created tactical roles for different groups of players within the combat settings, and eventually, it got to a point where just organizing this and staying focused was more of a burden than it was actually fun. In our particular game modes, if you had 200 or 300 people, which we can arguably do, it just wouldn’t be as fun.

There are still tradeoffs. Characters in MMOs tend to be fairly high fidelity, with a lot of graphic richness to that presentation. When you have 100 characters with a lot of different equipment and spells going on, a lot of things happening in terms of animation space, which can be potentially a really low framerate situation on a lot of machines, even most machines in the market. One thing that you have to do, and it was very easy to do with our artists, is come up with a variety of systems for reducing detail when it’s not necessary when it doesn’t get in the way of players understanding what’s going on.

Hopefully, if we’ve done our job right, and as far as I know we have, that’s not even perceptible. It’s not visible to players. There are some things being drawn at a lower resolution or have changed as far as the depth we draw them to, but in combat that doesn’t become an issue at all. You don’t see that happening.

Server gaming

GamesBeat: Is there any difference compared to a traditional game here, where a lot of computing still happens on the client and less is happening over on the servers? How is that distributed?

Lawrence: It is very unusual for us in that we have an almost entirely server-actuated state system. In many games, even other MMOs, there’s a hybridization where the server does some things and the client does many things, making the decisions easier to process computationally. In our system, almost everything in terms of simulation is done on the server. All of the physics. When a player moves we’re asking the server instead of just telling the server, “Hey, I’m moving here.”

It’s not necessarily better. It’s just different. It enables some different gameplay. We felt like it was necessary for our physical gameplay style, the combat model being a very physical model where space and direction, and facing matter a lot. Again, the goal is that the player doesn’t see a difference here. It turns out there’s a whole lot going on at the server, but the player doesn’t have to worry about that too much.

GamesBeat: Is there a very different tech stack here to describe for people who might be interested in what’s going on behind the game?

Lawrence: Like many MMOs, it’s a fairly proprietary backend. We built it to suit the game we were making. That was a deliberate decision a couple of years back in the project. We could piece together components that are relatively available on the market or within Amazon and build something that did a particular set of behaviors. But we had an interest in this particular game style and simulation concept, and to make that work we knew we needed the servers we were building had to process state a little differently than they would if we had assembled a set of components that were available on hand.

That was, again, a deliberate decision. This was something we wanted to invest in. We liked this combat style. We thought it would work. At that time we didn’t know it would work with 100 players or more. We were given an opportunity to spend the time and do that right, getting to a point where we could test it, and when we tested it we really liked how it played. It’s different from a lot of games. It delivers the fun in that respect.

Lumberyard

Above: Archers in Amazon’s New World

Image Credit: Amazon

GamesBeat: How does Lumberyard and its evolution factor into this?

Lawrence: Lumberyard was the basis for everything that you see in New World. Graphically, for sure. There was, of course — the backend was something we were building on our own at the time. At one point Lumberyard was advancing really quickly, adding features at a very high rate, and then they were moving more toward the O3DE concept. But for us, because we had a built feature set, there was a point where we had to slow down a bit with the technology adoption. We said, “We’ll take the Lumberyard we have now and stick with it.” That’s very common in-game engine development versus product development. Even with other engines available on the market, you often decide, “This is the version I have to retain and use for launch. I’ll keep working with it.”

Since we’ve done that, Lumberyard has seen a tremendous number of features added, some of which we would have liked to have had. But that’s the nature of the beast. You have to be able to make the call. This is the feature set that I actually need, compared to the ones I want that are just around the corner. If I were doing it again now I’d take O3DE and start with that. I’d have DX12 and a couple of other goodies that we’ll have to add ourselves. But it’ll still get done.

GamesBeat: As far as updates go, can those features be added over time, then?

Lawrence: Oh, yeah. We’ve changed — again, because of the nature of MMOs, we’ve changed some things about the render path, how we visualize. It requires some adaptation. But the beauty of having an easily available source is that we can do that in our own time. We’ll definitely have DX12. It’s not necessary for our launch, but it’s something we think we’ll get done fairly quickly after launch.

GamesBeat: What does that deliver for you?

Lawrence: The first-order changes are pretty imperceptible to players. Some drivers will work a little more efficiently with DX12. This is a little esoteric, but it allows for large frame buffer rendering. It has some performance improvements when you go to DX12, because you can split the render set. And then a few other special effects. Some shader effects are much easier to establish under DX12 than other architectures.

GamesBeat: Did the open-sourcing of Lumberyard affect anything for you?

Lawrence: Other than being a proponent of it — I like the idea of having an available choice on the market that academics and other people can pursue rather than getting into any sort of closed commercial license. There are excellent engines on the market, but having the choice of another one that’s relatively unburdened–the Apache license is very permissive. It allows for a lot of choices for players and developers. I’ve been a supporter of that since we started kicking around the idea.

Morales: With O3DE, the intention was to lean into the modularity and the specialization of that engine for other industries and let other people make it their own. If you look at the page for O3DE, it’s not just Niantic and Amazon Game Studios. It’s also Adobe, Intel, Huawei, a few others. The DNA, the work of Rich’s teams, is very much ingrained in O3DE moving forward. That’s one of the core reasons why we’re so excited to see New World launch and be successful and act as a centerpiece, a perfect example of how you can tie together all these AWS and Amazon pieces while also being adjacent to open source.

Selfishly, for us, it helps us tell a great story to all of our other games customers and AWS customers. They’re playing New World and having this amazing high-fidelity experience. At the same time, they’re streaming something on Netflix and checking up on Clash of Clans. Maybe playing some Fortnite in the evening. That’s all coming from the same place.

Overcoming delays, addressing feedback

Above: The Corrupted in New World.

Image Credit: Amazon

GamesBeat: What kind of feedback enabled you to go forward? There were a couple of big games that were canceled before you got to this stage. Were certain things really going right for New World as far as the feedback that came in?

Lawrence: We’ve had the privilege on New World of playing with the customers for quite some time. Luckily we had some customers that were understanding and supportive when we were in a very alpha state. That was two years ago. We were determined to try to keep that live and continuous as a service as much as possible. I believe we did a pretty good job of that. It’s very rare that, even in an alpha state, you never have to take the servers down.

The decisions in terms of investment in features and how long we should work on different aspects of the game were very much driven by that player feedback. We tried to cycle through the audience so we would have a very expressive audience with different play styles. We can’t build a game that’s right for everybody, of course, because I don’t think that’s possible, but there was a large enough consensus. “This is cool, that’s cool, that’s cool.” You put those features together and you create some strategy around how they serve as a greater framework for the game. What are the play styles that we’ll build around? What are the pillar features we’ll deliver to satisfy these players?

Our meter on what we invested in or didn’t invest in was always driven by that. There were other product decisions made, but they didn’t affect us in that way. I’m always sad to see when a project doesn’t work out, when customers aren’t 100 percent receptive to it. But that doesn’t affect our focus on our customers. There are a lot of different styles of gamers. We’re pointed at a style that we have something for and we’ll continue to listen to them and refine the game for them.

GamesBeat: What were some of your thoughts on the balance between what you designed and the control players have through things like user-generated content?

Lawrence: It’s a bit of a paradox. You can’t really build great design-by-committee games, in my opinion. It’s a completely valid approach to take customer input, and we do that all the time. But if you ask 10 people what’s the most important thing in the game, you may well get 10 different answers. You can’t achieve focus and a strategic delivery when you in that many different directions at once.

We picked the focus. We picked the strategic delivery. We’re game players too. We know the games we like, the games we’ve played in the past, the ones we aspire to see on the screen. That creates the core identity of the game. We want this very physical combat style. We want a lot of crafting. We want a game that has a very open world where you can just wander around. You don’t always have to follow a specific quest guideline if you don’t want to. Those were things that we wanted.

Sometimes that’s a risk. We put those in the game, put them in front of the players as quickly as possible. That’s a key thing. Limit your expression of the feature to a point where you can get players playing it, and then ask them. Do you really like that? You have to be prepared for the things that you really adhere to — the players might wholly reject them. A few of the things you didn’t think were that important, they might highlight them as the loop they really want to see a lot of. But it’s a bit of a mutual selection process. You select many of the core things, and then the players identify the things they’ll pick up or they won’t.

Even if you’re really devoted to the idea of a particular feature, there’s no sense in going heavy into investment on it when the players are telling you that it’s not very fun for them, that it’s not something they want to do. There are guideposts in terms of how much you spend on a feature or not. But they’re not a complete description of a product. You don’t get a description of a product from a feature set. You get that by setting down pillars. These are the things we’ll try to achieve in the user experience. Then you pick features that guide that user experience and make sure players enjoy that user experience, the features that come together to deliver it.

Above: The place is dark, but it’s still pretty.

Image Credit: Amazon

GamesBeat: When you think about the balance of power in the game, do you think that players will be more excited about a stalemate that lasts forever, or could some faction actually achieve victory?

Lawrence: Achieving victory is cool. When you do it should be a monumental task that’s well-recognized, even by the people that you’re defeating. “That was an amazing job.” It should be hard to achieve. I don’t think it should be very persistent, because that’s not fun. Whoever takes over, they can’t rule forever. Mostly players select out of that. People like the idea of combat that has meaning, and so while it’s tempting to say they’ll just stay in power forever, there will be people who adjust that in the game. But if not, we’ll create conditions that help that.

One of the things we did right up front was that we definitely wanted three factions. We didn’t want two. There are a lot of logistical problems with two factions and getting players to represent evenly between the two. When one side starts winning, many players will run to that side. But if you have three, it’s not a win or lose. It’s win, lose, or you’re still in the fight most of the time. That tends to rotate. It was an early decision.

We have a bunch of players who may not want to participate in any of those systems. The key to that was building a feature set where a PvE player could participate, but sort of obliquely. They’re not in the fight directly because that’s not their thing. That’s cool. You can still contribute by helping your faction or your township. There are ways for a PvE player to assist in the fight that goes on in the world without actually fighting other players.

Ridiculous computation

GamesBeat: For AWS, what does that achieve for the game? I imagine one thing would be that if you log in from anywhere in the world, you’re still going to have a good connection and experience. You don’t have to wait for servers in Seattle or somewhere else.

Lawrence: There are always finite computational resources. We certainly hope that if we do our job right, the players won’t perceive any of that. It’s always jamming along. But I don’t want to be presumptuous and say we’ll never have any problems at all because, statistically, that kind of stuff could happen to us as a game. We do, at Amazon, specialize in trying to build scale and make that seamless and invisible to the user. To the degree we’ll succeed or not, our application, which is New World, is based on some great underpinnings at AWS. I can do some fascinating things from a game operation standpoint that are probably pretty dry and boring to a lot of players, but just the idea that I can hit a button and a game server appears — that’s not easily achievable all the time. Here it’s second nature.

In the case that, for instance, players need another server, we should be very responsive to that. Amazon has a great network. We should have some pretty good network responses. There are of course always people — if you’re out in the middle of nowhere, there may be some network issues, but generally, you should have a very solid experience. We know that our latency and our packet loss are quite good on the Amazon side. The internet these days is pretty dependable. I think most people will be able to have a good experience.

Morales: On the AWS side we have a track record of this. We mentioned Fortnite. Being able to have a session-based multiplayer game with lots of computationally challenging, ridiculously intense battles is something we’ve done before in partnership with Epic. Being able to do that on just about any continent, 25-plus regions scattered all over the world is pretty impressive. To Richard’s point, what this also helps underscore — that infrastructure is available with an API call. Not a phone call. Being able to automate scaling or launching a new game into a new region, a new country, that’s something we’ve thought long and hard about on the game tech side. We’re helping customers like Amazon Games make their games successful.

In the next wave of new markets, we’re about to launch a new region in Spain. We’re about to launch a new region in the Asia Pacific, in Jakarta, Indonesia. This is a great beginning. We’re excited to help New World scale and launch globally when they’re ready. We’re excited to use this as a proof case for any other game developer in the world to do the same thing.

Above: A town in New World.

Image Credit: Steam

GamesBeat: Folks are talking about the metaverse right now. You’re launching an MMO. What does that make you think about?

Lawrence: Everybody, it seems, has read the books and really wants to build that thing where everybody gets together and it’s your second life that has all these cool things happening in it. Not to be confused with the actual game Second Life. I find those ideas interesting. There are, of course, a few people who’ve attempted that. So far I think the folks that have done so–it’s very perceivable where the technology limits us. I don’t think we’re quite there yet. I mean no offense to anybody. Anybody who ships a game has accomplished something fantastic. I just wonder if we’re at a point where we can make Ready Player One a reality right now.

It’s not impossible. But the question is always, would it be the right level of fun? It’s great when you’re the protagonist in a novel and everything comes together as a storyline that’s a fascinating exercise in ‘80s nostalgia combined with an action movie. But you need to be able to do that for every player. Every player deserves their fun. Metaverses tend to have, descriptively, a lot of feature set, but I’m always looking for–what’s the core gameplay? Are there multiple core gameplay loops that I’d settle into?

It needs a little more focus and understanding for me as a player. I want to understand what I’m doing. Just existing in it is cool when you read about it. Many games do a great job of just existing today. I can show off my fashion sense and hang out with my friends. They tend to be very social organization games. That’s cool. That’s a loop. I’m going to get online and share with my friends using this cool avatar abstraction. But how do you combine that with action gameplay or strategy gameplay or any of the other aspects that are typically described in a metaverse concept?

I will admit to having my own metaverse game idea I’ve been kicking around for a while. I’m going on about how I’ve refined it over time. The first version, when I was a young game designer, just went, “Hey, we’ll have Neuromancer. That’ll be really cool, right?” Well, wait a minute. We actually have to do things. We have to have goals. People like to have goals and achieve things. But I’ve been kicking that around and refining it for quite a while. If I get a chance, I’ll let you know when I start working on it.