The scalability and evolution of game engines *spawn*

[Jay]

I'd hazard a guess that it won't be long before Microsoft drop the mandate to run on both gens.

...which is a good thing to my mind.

Edit: we haven't even considered CPU limitations.

Only for 1P, and I think it's linked to things like xcloud more than 'supporting' gamers which is a nice pr spin i guess.

Every 1P game is going into gamepass, xcloud when goes live (this year) will play all ms 1P games that get released. This wouldn't be possible if was next gen exclusive.
Things like that and most of the new studios was probably already halfway through making their game, may as well just wait for everything to align, i suspect is their view and makes messaging easier.

Slipspace must cost a lot to develop, and even if it's for the next half dozen halos still be useful if other internal studios could make use of it once it matures.

 

[cheapchips]

I'd hazard a guess that it won't be long before Microsoft drop the mandate to run on both gens.

That's not their mandate. They have commitment to first party cross gen for this year and next-ish. Booty's call about this in Jan keeps being misconstrued.

 

[Deleted member 86764]

That's not their mandate. They have commitment to first party cross gen for this year and next-ish. Booty's call about this in Jan keeps being misconstrued.

I didn't say they're planning it the whole generation. That'd be daft. Let's see if it goes on for as long as end of 2021.

 

[iroboto]

Realistically that far favours a larger delta as XBSX is a significant advance in hardware efficiency at those flops. If we go with the rule of thumb RDNA versus GCN flops is 3:2 or whatever, you'd be looking at maybe more like XBSX being 15x the GPU power.

I believe this is for non-compute ie. unified shader pipeline.
It comes down to inefficiency in scheduling the unified shader pipeline. It takes 4 cycles to complete an instruction, to have maximum efficiency you need to fill all 4 SIMDs. This is easier to accomplish in compute in which you have control over how the work is submitted to the compute units, but with 3D pipeline you're at the mercy of the command processor.

I believe CDNA will continue to follow this path. RDNA separated to support the 3D pipeline better.

 

[Shifty Geezer]

I'd hazard a guess that it won't be long before Microsoft drop the mandate to run on both gens.

...which is a good thing to my mind.

Edit: we haven't even considered CPU limitations.

This thread has been spawned to discuss engine scalability, so let's keep the business/market discussion in the other Console industry thread.

 

[Ronaldo8]

The engine is designed to cater for 60fps at 4k, my feeling is performance takes more priority than pushing visuals. The first engine demonstration trailer is probably indicative of the general look and feel of what's to come, more or less.

The Index Value Blending method described in the MS patent is a perfect solution for uncompromising fidelity on a high end machine as well as scalability to a low end/entry level device. In fact, that's what motivated my op in the first place.
343i technical leads have been openly bragging about their game engine on linkedin, declaring it the "most advanced" with a straight face. The engine being capable of fully using the IVB method should provide bragging rights aplenty.

 

[Ronaldo8]

I believe this is for non-compute ie. unified shader pipeline.
It comes down to inefficiency in scheduling the unified shader pipeline. It takes 4 cycles to complete an instruction, to have maximum efficiency you need to fill all 4 SIMDs. This is easier to accomplish in compute in which you have control over how the work is submitted to the compute units, but with 3D pipeline you're at the mercy of the command processor.

I believe CDNA will continue to follow this path. RDNA separated to support the 3D pipeline better.

There has been significant modifications added to the command processor of the XSX. Better to wait and see what it amounts to. (I have it on good authority that MS is trying to make strides towards a fully programmable GPU with no fixed functions. In fact this idea was expounded upon in the D3D devblogs a few years back.)

 

[iroboto]

That's probably a relic from previous generations where the engine empowered the artist and there was a clear difference between a 'good' engine and not so good engine. Getting a good looking game on PS2 meant very clever coders who could write an effective engine that got the most from it. These days the hardware is very accessible and any game can use PBR and baked lighting to look good, so there ain't a huge amount to differentiate on overall look, and the difference will come in terms of optimisation and how many FPS a game gets with a particular engine.

Yea I wonder if we're headed back into that direction when you look at what Unreal has been working on.
If you put the engine in charge of all of that, the artists don't need to keep optimizing models and textures down.

That said here's an interesting comparison between Halo 4 and Halo 5.
30fps720p.
they moved to 60fps and dynamic resolution that was probably about 900p~1080p.
They gained some weird version of PBR and changed the lighting model a bit.

I'm going to borrow jroc's post here at resetera
This is the same generation of hardware

Money and production values go a long way. But this is also pretty insane amount of work done here.
Looking back, 3 years was not long enough away from Halo 4 to be ready to ship Halo 5.

But this is the same engine here on display moving from 4 to 5.

 

[function]

If you factor in ray traced shadows and reflections on series X, you're already looking at a type of scalability that can eat up large amounts of performance relative to the Xbox One even at the same resolution.

And it's a type of scalability that will be highly valuable for the PC release too, so it would be a mistake to think this is all about trying to make a game run on X1 and Series X. It pays off in both the PC and console spaces, abundantly. Same for HDD vs SSD - the PC isn't ready to move completely to SSD just yet. Same for CPU to some extent - quad cores will be supported for a while yet.

Unlike Sony, MS have to support a relatively broad spectrum of hardware anyway. Stretching it to support X1 and XSX simultaneously for a year or so makes perfect sense.

 

[cheapchips]

There's a third Microsoft article up at Gameindustry, this one on MS's cross generational approach. More philosophical than technical but not sure where else to put it!

https://www.gamesindustry.biz/artic...ries-x-developers-being-held-back-by-xbox-one

 

[BRiT]

Some snippets from that article ...

"Frankly, held back is a meme that gets created by people who are too caught up in device competition," says Microsoft's executive vice president of gaming, and Xbox chief, Phil Spencer. "I just look at Windows. It's almost certain if the developer is building a Windows version of their game, then the most powerful and highest fidelity version is the PC version. You can even see that with some of our first-party console games going to PC, even from our competitors, that the richest version is the PC version. Yet the PC ecosystem is the most diverse when it comes to hardware, when you think about the CPUs and GPUs from years ago that are there.

"Yes, every developer is going to find a line and say that this is the hardware that I am going to support, but the diversity of hardware choice in PC has not held back the highest fidelity PC games on the market. The highest fidelity PC games rival anything that anybody has ever seen in video games. So this idea that developers don't know how to build games, or game engines, or ecosystems, that work across a set of hardware... there's a proof point in PC that shows that's not the case.

"That said, we're shipping Xbox Series X this year. I'm playing it every day at home, and it is different to playing on an Xbox One X. We should applaud the work that is going on with the SSD, and the work that is going on with audio, to pick some of the areas that Jim [Ryan] and Mark [Cerny] and the stuff that [PlayStation] is focused on. We should applaud load times and fidelity of scenes and framerate and input latency, and all of these things that we've focused on with the next generation. But that should not exclude people from being able to play. That's our point. How do we create an ecosystem where if you want to play an Xbox game, we're going to give you a way to go play it?"

...

"Sorry, I am a bit soapboxy with this one. Gaming is about entertainment and community and diversion and learning new stories and new perspectives, and I find it completely counter to what gaming is about to say that part of that is to lock people away from being able to experience those games. Or to force someone to buy my specific device on the day that I want them to go buy it, in order to partake in what gaming is about.

"Gaming is bigger than any one device, and that is something as an industry that we've embraced all up as we bring more and more players in. I think it's vital to the role that gaming can play on the planet."

 

[Jay]

Not read it yet, but given your extracts then why bother even have next gen exclusives in a year or so time if its is not going to be held back.

If he put it in another context of, we find as 1p that it takes around a year to move the design, AI, scale across there is little reason not to make the commitment we have. I could appreciate that more.

Everytime I see the conversation about holding back it always refers to graphics. When in my eyes it's everything else. Graphics isn't the biggest issue in this regards.

And don't get on soap box and say it doesn't, but yet gonna start to have next gen exclusives in a year. What are you going to be saying then? It just wouldn't be possible on previous gen machines without it being held back.

 

[BRiT]

A good part of it has to do with how long and how much investment game development takes. It's natural for businesses to want to maximize the potential to recoup their investments. It seems natural to not see the real impact of next-gen hardware on games until 2 to 3 years into it. Now if you could cut down the time of game development, you'd see the impact of that much sooner, perhaps to the point of not targeting last-gen.

Yes, load times of current-gen sucks, but you still have hundreds of millions of gamers playing on it and millions of games released and sold on them. Releasing games on those old systems won't entirely break new games. It even pushes game engines to reach amazing feats, look at the COD series and how quickly they load into games.

 

[Jay]

A good part of it has to do with how long and how much investment game development takes. It's natural for businesses to want to maximize the potential to recoup their investments. It seems natural to not see the real impact of next-gen hardware on games until 2 to 3 years into it. Now if you could cut down the time of game development, you'd see the impact of that much sooner, perhaps to the point of not targeting last-gen.

Yea, I'm not even debating the merits or lack of. It's just the framing I have issues with.

Doesn't make sense when they've already said they will be making exclusives, just not yet.
So in a year all this talk of scaling goes out the window for some games.
It's not like the interview are small soundbites either. Could say graphics can scale pretty quickly, game design takes longer to do.

 

[BRiT]

Oh, yeah, I can understand the issues.

However, I think games will have scaling issues for more than just until end of 2021, namely in the PC space since I don't think any but the largest games will have the guts to push for NVME required. Statistically none of them push for SSDs now, and they've been available for over a decade now. Maybe they hadn't pushed it because of consoles using HDD?

As for the jump in baseline hardware, I just don't know that they make much of a difference as far as gameplay is involved. We'll still have the same sort of games as we have now, just without certain aspects limited by the current-gen systems (loading screens or aspects to hide that). I have hopes they will be able to improve the AI and interactivity in the environments, but then again I've had hopes for better AI for decades now and it always falls short. I'm hoping for more, but ready for nothing.

 

[function]

Oh, yeah, I can understand the issues.

However, I think games will have scaling issues for more than just until end of 2021, namely in the PC space since I don't think any but the largest games will have the guts to push for NVME required. Statistically none of them push for SSDs now, and they've been available for over a decade now. Maybe they hadn't pushed it because of consoles using HDD?

Yeah, the PC is in no place to leave the HDD behind just yet, barring the inevitable couple of of outliers. There are lots of PCs with SSDs, but many still use it for the OS and a handful of programs but put games on a much larger HDD. *puts hand up*

Even once you move beyond HDDs, lots of people will rely on SATA III SSDs due to form factor, cost and expansion practicalities. Then beyond that there are the SATA III m2 drives which are still very much in production and in demand for lower end m2 form factors. Beyond that, you have mid range nvme drives with 1.5 ~ 3 GB/s peak sequential which are still far higher latency and far lower effective throughput than the drives in XSX and Lockhart.

When SSD does become the baseline for multiplatforms a couple of years from now it'll be SATA III as the minimum and not fast NVMe stuff. In the next year or two there's not really any reason to skip X1 / PS4 for most PC targetting multiplatforms games. And even if you were prepared to skip the PC. you probably wouldn't want to skip the juggernaut that is the PS4 anyway.

As for the jump in baseline hardware, I just don't know that they make much of a difference as far as gameplay is involved. We'll still have the same sort of games as we have now, just without certain aspects limited by the current-gen systems (loading screens or aspects to hide that). I have hopes they will be able to improve the AI and interactivity in the environments, but then again I've had hopes for better AI for decades now and it always falls short. I'm hoping for more, but ready for nothing.

I think scale will improve quickly once we move beyond Jaguar, but the quality of what's scaling up in quantity will be a much longer slog. So I think I'm in the same boat as you in terms of expectations.

And even once Jaguar is a memory (assuming it's that rather than core game design that's the issue), quad core PCs will be with us for a long time. Multiplatform baseline is going be below 8 core Zen for years to come.

 

[eastmen]

Bethesda annoyed some of their older, smaller fanbase to attract a new larger fanbase. I too would have liked to see an Elder Scrolls on this gen but I also want it to be a genuine leap forward in every way over Skyrim and I don't think it could have been that big a leap if constrained by this generation's consoles. Nextgen though, with decent processors and fantastic I/O to make loading a thing of the past? That is a hardware platform for Elder Scrolls VI.

StarField was always nextgen, this was part of the announcement of Bethesda's E3 2018 conference and was further clarified in an interview with Eurogamer. You can't blame Bethesda because you didn't listen to what they said.

Do you think the Netimmerse engine can be pushed further ? Talk about lipstick on a pen. I know they tried to modernize it and dubbed it gamebryo and then again and named it Creation engine but man there is some really old code and design decisions on there.

I really wish Starfield and Elder scrolls were moved over to new engines. The id tech stuff like in Doom could make a beautiful fallout game

 

[Deleted member 11852]

Do you think the Netimmerse engine can be pushed further ? Talk about lipstick on a pen. I know they tried to modernize it and dubbed it gamebryo and then again and named it Creation engine but man there is some really old code and design decisions on there. I really wish Starfield and Elder scrolls were moved over to new engines. The id tech stuff like in Doom could make a beautiful fallout game

There is really old code in Windows, old code is well-tested code. But what other game engines offer the flexibility of CreationKit?

For me, Bethesda's engine is not about graphics (obviously!) but all about the ability to have a massive, connected open-world in which millions of persistent object and thousands of named and unnamed NPCs can interact with objects, each other, and can initiate (and be part of) quests, have routines, and may move through the world in a believable fashion - like a courier having a schedule of travel and being in certain places at certain times and you being able to track them down travelling from X to Y. There is a lot of detail and tiny moving-parts in Skyrim that aren't really apparent on the surface and much of it is really only exposed through some of the mods.

What other engines have native support for all this, or are Bethesda going to have to rebuild all this tech from the ground up just to be more pretty so they can use idTech? How will idTech's cope with thousands of persistent movable objects with real (not realistic!) physics interactions?

 

[PSman1700]

The engine is supposedly an upgraded versions of PS4's spiderman engine, not a newly written one for the PS5.

 

[Shifty Geezer]

The engine is supposedly an upgraded versions of PS4's spiderman engine, not a newly written one for the PS5.

What is this statement in relation to?