In Theory: Could Sony release PlayStation 5 in 2018?
Digital Foundry on the hardware options available for a next-gen PlayStation and how powerful it could be.
Could Sony be releasing its next-gen PlayStation in 2018? That's the notion put forward by Maquarie Capital Securities analyst Damian Thong, cited in a recent edition of The Wall Street Journal. But is the time right to replace PlayStation 4? And what kind of hardware could Sony conceivably deliver next year?
First up, it's worth pointing out that Thong is credited with predicting PS4 Slim and PS4 Pro, making him a creditable source. However, a comprehensive look through Google's news archive doesn't appear to show any prediction for the Pro that predates Kotaku's initial PlayStation 4.5 exclusive. Also worth pointing out is that Thong's prediction is nothing new - he made the same claims about a prospective PlayStation 5 launch last September. Did he predict the PS4 Slim? It certainly looks like it, but then again, Sony has made physically smaller iterations of every one of its previous consoles. That said, his CV certainly looks first-class and clearly, he knows his onions.
We can add some additional flavour to the speculation as we spoke to PlayStation system architect Mark Cerny last year, where he was very specific about how PS4 Pro was not a next-gen console, sharing with us the criteria he has for what constitutes a new console generation. Games hardware is years in the making - work on PS4 began in 2008 - so we can safely assume that his words do have some bearing on what to expect from the next PlayStation, which is almost certainly in the works now, even if a 2018 launch does seem unlikely.
"Each new generation brings with it a new set of capabilities: CPUs, GPUs and the like but also controllers and new types of display devices. If you go back to the 1970s, it was colour TV. That was the new display device," Cerny told us. "These capabilities unlock new potential for the type of games that can be created. For example, increased CPU power might not seem like a game-changer but it actually allows for much better enemy AI, more enemy characters, better world simulation and a whole host of other evolutions in the game experience."
But this kind of innovation doesn't come for free. Switching to a new console generation effectively pushes a giant reset button on game development, not to mention the prospective financial fortunes of the industry, which has to deal with a challenging period of cross-generational development.
"The flipside of that new potential is that the game creators, who we depend on to create the experiences that'll meet the ever-increasing expectations of the gamers," he continued. Those game creators have to adapt to the new tools and technologies that we provide. That is certainly a disruption of sorts. It's a good one because it allows for those important breakthrough experiences but it definitely has an associated cost."
Looking back over our conversation, Cerny laid out what he considered necessary for a new console generation. He discussed a move to a new CPU architecture (retaining x86 CPU architecture as a possibility, albeit one that still poses compatibility issues with existing PS4 games), increased graphics capability, significantly more memory, the IO required to feed it, along with the sheer mass storage required to house these advanced new titles.
But to cut to the chase - by the criteria he laid down, is a new PlayStation in 2018 technically feasible? In theory, yes. But would the resulting console deliver the generational leap he describes above? That's less clear. Elsewhere, Cerny has discussed the notion of an eight teraflop console required for native 4K gaming, and that's an interesting figure to put out there, because such a machine could conceivably be built for a late 2018 launch. But the question remains as to whether this is enough to kickstart a new console generation. Higher resolution versions of existing games doesn't really constitute the "breakthrough experiences" Cerny associates with next-gen hardware.
But let's look at the evidence for a viable console on a per-component basis. First up, the CPU architecture issue can be addressed easily enough - AMD has its new Ryzen processor line out now and a version of this architecture will be repurposed for its all-in-one APUs, meaning that the core technology (a version of Ryzen integrated with Radeon graphics technology in a single chip) almost certainly exists in the here and now.
A single AMD Ryzen CCX module featuring four physical cores and eight threads would be a good match for a console and clock-for-clock would represent a generational leap from the Jaguar-based cores found in the current consoles. Ryzen APUs in the PC desktop space are expected later this year, so the notion of a semi-custom chip based on the same architecture arriving a year on isn't beyond the realms of possibility.
The next point concerns the configuration of the GPU, which ties in very closely with the process technology that would be used for the next PlayStation processor. Realistically, a 2018 console would still be using TSMC's 16nm FinFET technology - as used in PS4 Pro and Project Scorpio. The larger the graphics component of the chip, the more powerful it will be, but this makes the chip larger, more difficult to produce and more expensive. PS4 Pro achieves 4.2 teraflops with a processor using a chip in the region of 310-320mm2. Microsoft's Scorpio chip is 360mm2 but uses innovative technology to increase clock-speeds to hit its six teraflops.
A PlayStation 5 processor can afford to be larger, but not that much larger than Scorpio's, if we're remaining on the 16nmFF node. Mark Cerny's notional eight teraflop GPU seems like the realistic limit for a console processor under these conditions. The issue here is that this would only represent a 4.2x improvement over the base PlayStation 4 and a 1.9x boost over PS4 Pro.
A move to AMD's more efficient, more innovative Vega technology with Sony's bespoke customisations, along with continued, improved support for checkerboarding would allow an 8TF GPU to punch well above its weight, but fundamentally, would we be looking at a game-changing piece of technology worthy of a new console generation? Historically, a generational leap in console power has been defined as a 6x to 8x improvement in performance in addition to the enhancements brought by technological innovation. The reason we are getting mid-generation refresh consoles is exactly because this level of advancement - not to mention the frequency of new processes - is slowing down.
In our briefing, Cerny only specified 'significantly more' memory as a requirement for the next-gen PlayStation, but the question is how much RAM is required to service a machine that must notionally last around five years. Microsoft has targeted 12GB simply to embellish the current generation of games running at 4K resolution. A true generational leap will require not just more memory - it must be a lot faster too. HBM memory - which stacks modules vertically for huge improvements in bandwidth - is the obvious route forward, but this will remain cutting-edge, expensive technology well into 2018 and most likely beyond. In this case, providing much higher levels of memory and bandwidth seems like an innovation that we really need, but can't be delivered in both cases by next year at a console-friendly cost.
There's also the question of storage. 1TB hard drives in Pro and Scorpio are already looking underwhelming - but more than that, once again, there is the question of bandwidth. Loading times for this console generation are already too long in many cases, and there's a strong case that the existing 2.5-inch laptop hard drive has been taken as far as it can. Microsoft didn't want to tell us how it is delivering an extra 50 per cent of bandwidth from its Scorpio mass storage, but we did note that Seagate drives were in place in the prototype devices we saw, meaning either a hand-picked drive or maybe even an HDD/SSD hybrid, along the lines of the Seagate Firecuda drive we recently reviewed. The true generational leap will come from solid state media, but again, the combination of mass storage (1TB+) and super-fast access is just too expensive right now.
Key to delivering a console that represents a huge upgrade is the availability of new microprocessor production technology, or to put it more simply, a new process that allows more transistors to be crammed onto a piece of silicon. Chip-producer TSMC (who produce all current console APUs) has 10nm and 7nm technology en route, to replace the current 16nm process. 10nm is shipping now in the mobile space, but there's still no indication of whether this is a viable process for a console - after all, 20nm proved unsuitable for anything other than mobile devices. We should look to components in the PC space and where they move on from the current 16nm and 14nm processes (this Anandtech article shows us the state of play with new fabrication technology right now). This week, Nvidia revealed the existence of a 12nm TSMC node for its next-gen Volta GPUs, but don't be fooled - despite the name, it is just an evolved version of the current 16nm process.
There are other options for Sony of course, but it would involve cutting all ties with the PlayStation 4 generation - making backwards compatibility extremely challening. Nvidia could doubtless provide its own processor on the existing 16nm/'12nm' process, swapping out x86 CPU cores for ARM alternatives. AMD's arch-rival can achieve better power efficiency and higher utilisation of area on a processor (GTX 1060's GP106 processor is 14 per cent smaller than Radeon RX 580's Polaris 10) and we're already seeing some phenomenal gains of 'to the metal' access to Nvidia graphics technology via Nintendo Switch.
However, Nvidia doesn't license its graphics technology as flexibly as AMD, and x86 support has gone down extremely well with developers - and this is something that Nvidia can't supply. Returning to the dual chip CPU/GPU set-up as seen in PlayStation 3, with individual vendors supplying each component, is incredibly unlikely. Increased performance via a more efficient Nvidia chip is conceivable, but it's still hard to imagine that this would power "groundbreaking experiences" worthy of a next-gen console. As things stand, AMD does seem like the most probable partner for a prospective PlayStation 5 right now.
Ultimately, a 2018 console is viable, and it would represent a considerable improvement over the base PlayStation 4, but comparisons against Pro or Scorpio will be less impressive - especially when 4K displays will be the likely target, vacuuming up most of the extra power such a machine could provide. A 2018 PlayStation 5 would enjoy the advantages of AMD's latest CPU architecture, and a GPU that benefits from five years of improvements to Radeon technology, not to mention a 4x boost to pure compute power. It could provide more memory, but bandwidth would remain an issue. Bigger games could be provided via UHD Blu-ray media, while a hybrid HDD/SSD would speed up storage.
It would be a good machine, but it's hard to believe that it would reset the paradigm - and with PS4 still having so much to offer and with so much invested in it from consumers, developers and publishers, the notion of it arriving within 18 months does seem unlikely, bordering on the unwelcome. The one big unknown in such a design would be the level of customisation Sony introduces to the silicon - we saw big, big gains in PS4 Pro from this approach, allowing for experiences like Horizon Zero Dawn, a game that looks like it's running on hardware that's twice as powerful. Meanwhile, Microsoft's claims for Project Scorpio exceed what we would expect from a vanilla six teraflop AMD graphics core. How that will translate into final software obviously remains to be seen.
In conclusion, the potential technological underpinnings for a machine that is a worthy generational leap over the base PlayStation are out there - but it'll take time for them to become financially viable for a mass market box, and in the here and now, there isn't the sense that developers are hitting a brick wall in terms of the technology available. Games continue to improve technologically year on year, and the appetite for current-gen titles is still strong.
However, we do like what Mark Cerny told us - that Sony believes that the traditional console generation model isn't going away, that it's healthy, and that it acts as a catalyst for a renewed push to breaking technological boundaries in the games we play. It's in stark contrast to Microsoft's vision for a more evolutionary approach, but at the same time, the Xbox team's outlook may be more realistic bearing in mind how technological progression is slowing down. This year, it's all going to be about PS4 Pro vs Scorpio - but what happens when the next wave of consoles arrives is going to be fascinating.