Naughty Dog using PS3 engine on PS4

[anexanhume]

http://www.digitalspy.com/gaming/ne...-uncharted-the-last-of-us-engine-for-ps4.html

What are the implications of architecture changes between generations?

Will the engine stay optimized for their purposes?

Will the engine be easy to add to when new effect types become prominent?

 

[Arwin]

Replied this on gaf:

I remember reading that one major reason that they started from scratch was that they had their engine in the wrong language for sharing with other developers anyway, so all the code would have to be rewritten. Perhaps they didn't have to start from scratch for their engine, but I also think that part of the reason they don't have to rewrite their engine now is because they did so then.

The PS4 was already quite complex, with many CPU cores (8) and close integration with the graphics pipeline already possible. In that sense it was far ahead of the curve, and things haven't really changed drastically since then. Compute Unites have replaced SPEs where SPEs assisted the GPU before as well as did physics, raycasting style AI and physics type calculations, and added are 8 easier to program, out-of-order CPU cores. Shaders and such haven't fundamentally changed either as far as I know. And their engine had very good texture streaming features and such that are also ahead of their time, but can now for instance simply work with bigger RAM buffers.

In other words, their current engine should do pretty well and require only minor adjustments to optimise for PS4. At least, as far as I can tell from the sidelines.

 

[Gradthrawn]

Appears to be the same approach GG is taking, based on their KZ:SF postmortem. Bring their engine over to the PS4, watch the Jag cores choke on SPE code , move off SPE jobs better suited to the GPU, "jobificate" their code even further and fix "high level code".

 

[Rangers]

on the surface i dont like this. and i always said shadowfall looked ps3+ to me, maybe this is why.

it's probably just a time thing.

that and a code gets more and more vast and expensive every gen thing. now they probably dont have a lot of choice. and throwing away ps3 code is throwing away a magnitude more work than throwing ps2 code, etc.

it could hint at greater eventual gains, i guess.

 

[DrJay24]

on the surface i dont like this. and i always said shadowfall looked ps3+ to me, maybe this is why.

Don't worry, Forza 5 is a 100% new engine, it will never look like a 360+ game.

I'm not sure there is a strong correlation between the tools/pipeline/work-flow ("engine") and the image you see on screen.

 

[patsu]

Appears to be the same approach GG is taking, based on their KZ:SF postmortem. Bring their engine over to the PS4, watch the Jag cores choke on SPE code , move off SPE jobs better suited to the GPU, "jobificate" their code even further and fix "high level code".

Yeah, sounds like the "easiest" way forward. They can avoid unnecessary copying in PS4 too. I am more interested to see their workflow and backend tools. e.g., How fast does it take them to change stuff ?

The SPU jobs are one of the main reasons I think GPU-Compute will be used immediately.


Second is the bottlenecks. Where are they in PS4 after jobification ?

Best way to prove this porting exercise is to bring tLoU over to PS4.

 

[Deleted member 11852]

This makes sense. People tend to assume 'graphics' when developers talk about the engine but Naughty Dog are referring to their framework that covers controller input, UI, asset/RAM management, streaming, audio, and the building blocks of games like how objects interface with the world, how the world is defined (think the tricks Naughty Dog pulled in Uncharted 2 to ensure that train journey was exactly the right length for every person who played it) and joined together, the AI, the rules for interaction, special events and dependancies for them.

Given the quality games Naughty Dog turn out, they've obviously parallelized the heck out everything while maintaining coherency and the PS4 is just another highly parallel architecture. They will have to re-write the graphics rendering engine but everything else should be easy to carry across.

Best way to prove this porting exercise is to bring LotU over to PS4.

I would be so, so happy if ND bring The Last of Us over and Rockstar bring GTA V over

 

[Billy Idol]

Hm, sounds dissapointing. I thought PS4 is way different than PS3...a new engine seemed logic to me.

Sony Santa Monica stated the same for GOW3, they said they used their PS2 engine as a base.

The obly good thing I see: if they ported their engine to PS4, the Last of Us PS4 version seems to be realistic...

 

[dagamer]

How long does re-inventing the wheel with a new engine take anyway? A year? And I'm sure there's a lot of good code in that base as well: multi-threading, networking, shaders, etc.. stuff brand new to the PS3 that should in some way work on the PS4.

 

[patsu]

They said it took them 4 years to get the engine and tools to where they are today, with the help of brilliant folks.

I remember they also mentioned that during the early days, they lost about one staff every 2 weeks due to the frustration of developing new engine on PS3.

 

[Lefungus]

As a programmer, I must say that starting from scratch is really a dumb idea. It is much better to replace outdated components and, if needed, refactor the whole thing. After their first ps4 game is out, they'll really customize and optimize pipelines for this new platform.
Who wants to spend 2-3 years without any naughty games on ps4 ?

 

[3dilettante]

A number of relatively high-level decisions for the job-based system seem to work well enough for the PS4 with relatively modest rework, as indicated by the KZ demo.

Why commit significant resources and tie yourself long-term to fundamental design decisions for a platform before you have more full-cycle experience with it?

 

[eastmen]

Most likely games that were pretty far in development getting switch over to the new system.

 

[Cjail]

DICE, Epic Games and Crytek or iD are not making a new engine from scratch either but I didn't see anyone complaining.

IMO it's not a coincidence that the best engines in existence are not getting scrapped for the next gen.

 

[DJ12]

Why would they. As far as I know multithreaded engines using a dx based gpu hardware is exactly what this 'next gen' needs. The only real change is to x64 instead of ppc. The programming methodology surely remains the same.

 

[archie4oz]

Is this even worth discussong? Both systems are fundamentally heterogenous compute environments, better to just get the existing engine (assuming it's not a total stinker) running on the new system, and work out the various issues and optimizations over time. Granted Naughty Dog has a bit of an advantage over 3rd parties since they've got a little more insight as to what's coming and start more solid project plans to make the transition more smoothly.

 

[dagamer]

DICE, Epic Games and Crytek or iD are not making a new engine from scratch either but I didn't see anyone complaining.

IMO it's not a coincidence that the best engines in existence are not getting scrapped for the next gen.

Though you do have to question how much gutting they did to their previous engines when upgrading them to their latest versions such that it might as well be from scratch. I mean, Unreal Engine 4 was worked on for years!

In the end, any engine is really only as solid as its foundation. Assuming engine designers didn't make horrendous mistakes and assumptions when they started from scratch, it should be adaptable. Granted, you shouldn't expect Naughty Dog to be building an MMO or racing game any time soon.

 

[anexanhume]

DICE, Epic Games and Crytek or iD are not making a new engine from scratch either but I didn't see anyone complaining.

IMO it's not a coincidence that the best engines in existence are not getting scrapped for the next gen.

Well, one could argue those are still PC first engines (some of them), which is a sliding window more than a generational jump. Their engines are based on targeted features in the DirectX or openGL release of their choice. They're not targeting specific hardware implementations necessarily.

 

[Shifty Geezer]

ND are suitably progressive in their engine too, I believe. It's not beyond them to start with a PS3 engine and refine it over time to incorporate features enabled by the hardware (DX11, compute). Sony have also been working on a compute language that can run on CPU and GPU, which I expect ND had a hand in. Certainly ND can take their SPU code and supply it to Sony's scheduler and extract compute for little effort; in theory.

At this point I'd like to hear what those who feel it's a idea are worried about? What will the engine be missing out on?

 

[Deleted member 11852]

At this point I'd like to hear what those who feel it's a idea are worried about? What will the engine be missing out on?

My first reaction, before I thought about it rationally, was graphics and PS4 games not making the most of the new hardware - and I think most people probably at least pass through this line of thinking.

Of course, while graphics hardware has become more parallelised and more powerful, the basic primitives aren't such a huge departure from RSX in PlayStation 3 and Xenos in Xbox 360. This generational leap won't be as visually overwhelming like moving from the original PlayStation to PlayStation 2, or PlayStation 2 to PlayStation 3, where the graphical advances were enormous. Like jumping from basic vertex shaders to pixel shaders. Hardware has improved leaps and bounds in performance but not so radically - if you ignore the inclusion of more generalised compute hardware. This generation looks like mopping away strained resources, bumping resolutions to 720p at an absolutely minimum and attaining that elusive 60fps in more games. It's not like we've all jumped to ray-traced full CGI. Yesterdays engines were capable of driving better visuals, it was the hardware that constrained them.