Mark Cerny, speaking at Gamelab in Barcelona, has spoken out about the road to the PlayStation 4. In a live presentation the PS4′s system architect talked about the complexity of developing for the PlayStation 2, saying there was a “loss of flexibility” in game design because of the “low level graphics programming” required to get the most out of the console.
Cerny worked on graphics engines during that period, one of which resulted in Jax and Daxter.
In 2003, Yoshida talked to Cerny about his “concerns”, many of which were about costs of development – the “typical team had tripled”. Yoshida was talking about the next platform transistion, and would SCE be “up to the challenge of PS3″ – could they use a “common engine” between development teams, and grow the tech team (known as the ICE Team) to share knowledge and expertise.
Yoshida planned for “inter-team collaboration”, and even though at the time Cerny knew nothing about the PS3, the team studied hard on new technologies, like shaders. In the early days the hardware teams didn’t communicate with the game teams, but with the PS3 the ICE team would be integrated with the hardware teams, and Cerny would be central to this collaboration, not least because of his Japanese language skills.
Kutaragi’s Cell chip has already developed at this stage, and was passed to Cerny to see “what he could do with it”. He explains that developing for the new CPU “required huge effort” and was highly complicated. Cerny worked on a formal presentation, making “quite good process” on the Cell’s SPUs, but explains he was “focused on the task at hand” rather than thinking of practical applications for the chip – like games.
“It was hard to do the most basic tasks on the hardware,” he admits. It took a year or so to “crack the puzzle” and “there was now potential” with some great first party titles. Sony were happy they’d got “a tremendous lead over all the third parties on the system” – yes, really. “This was completely the wrong attitude, but we just didn’t know any better” he says.
Less than two years before the PS3′s launch, Cerny shifted to working on launch titles. At that time focus was “99% hardware, 1% software”. There was no debugging, no profiling, no graphics driver – it was all “in a primitive state” – and the third parties like EA and Ubisoft were “having an easily more difficult time.”
Engine development on PS1 took 1 month, on PS2 3 months, and on PS3 up to a year, Cerny added.
Of course, that all changed, and after the machine shipped Sony looked at “what worked and what hadn’t”, in order to look at what to change for PlayStation 4. This process was “inclusive and collaborative” for the first time, Cerny said. PS4 was originally going to use Cell again, at least it was an option to enhance it, but there were “other options”.
“Choosing a CPU would be a very big deal”, he admitted. “Timeline, business structure, development cost”. Cerny mentioned PowerPC and X86, the latter was apparently claimed to be “unusable in games consoles”. Cerny disagreed, saying he “spent his holiday in 2007 researching the X86″ and added that progressive enhancements changed conventional thinking.
Cerny went to Yoshida, pitching his skills as a lead architect on the PS4. He’d have to “leave WorldWide Studios” and work for SCE, although he wasn’t actually a company employee. In 2008 the development of PS4 began “in ernest” and began conversations with game developers, and started on the hardware.
“We wanted their input on the design” he said, referring to third parties, but said their solution was to make a questionnaire and present that to the developers and studios. Some questions were about GPU types, bandwidths, and so on, but more importantly Sony asked third parties about “the flavours of next-generation consoles”.
“Everyone we sat down with knew we were asking for feedback of PS4″, he admitted. “The number one piece of feedback was that the system had to be unified, one pool, and that if we had the money to spend we should invest it in an expensive GPU.”
They didn’t want an “exotic” GPU, though. Ray tracing, for example, would be a waste.
Cerny concentrated on some “rich features” for year one, and more “speculative” features for further down the line. Sony concentrated on ensuring the architecture was easy to use, but with powerful tech – GDDR5 memory, for example. The “Time To Triangle” for PS4 is back to 1-2 months, a huge contract to the 6-12 months that it was on PS3.
“Titles are much easier to bring to PlayStation 4″, Cerny claims. The GPU can be used for raycasting for audio, physics, collision detection, decompression, and so on, for example. “As developers learn to use these techniques later on”, he adds, “we’ll see richer and even more interactive worlds.”
It seems like lots of things were run by first and third parties during the development of the PS4 console. Cerny mentions Rockstar, Ubisoft, EA and Activision, and large-scale technical presentations that went well, and some that “let them know when they’re on the wrong track”.