Sony unveiled the PlayStation 4 today at an event in New York. The new machine will be extremely powerful, building upon a foundation of PC technology rather than the legacy of the PlayStation 3.
Sony didn’t explicitly say all of the details about the PS 4′s capabilities, but it disclosed a lot of information that gives technologists a lot to discuss. After all, modern-day game consoles pack the technology of recent supercomputers into an affordable and attractive box for the living room. As such, they are among the highest technological achievements that engineers can create. We’ve pinged Kevin Krewell, a longtime chip analyst at the Linley Group, for his reaction.
Advanced Micro Devices is most likely the provider of the processor. Sony didn’t say so, but it referred to an “accelerated processing unit,” which is AMD’s terminology for a chip that combines a microprocessor and graphics processing on a single piece of silicon. As two-chips-in-one, it’s more economical for a game console.
“This is an AMD chip with Radeon graphics,” Krewell said.
Sony’s Mark Cerny, (pictured above), the system architect for the PlayStation 4, said that the machine will have a local hard disk drive, eight gigabytes of main memory (using the DDR5 flavor from fast PCs), an enhanced PC GPU, and x86 processing. Again, AMD fits the bill for all of that.
The advantage of using PC technology — a move that matches that made by Microsoft on the original Xbox — is that Sony can take advantage of all of the volume economics of the PC industry to lower the costs of its machine. The drawback, however, is that Sony may have a harder time distinguishing the performance of its machine from high-end gaming PCs at some point in the PS 4′s life cycle.
The machine uses its GPU for particle physics processing, which indicates GPU Compute, Krewell said. That likely means it will support OpenCL computing, or using the graphics cores to do non-graphics tasks.
Krewell also said that the 176GB/s of memory bandwidth is 3.7 times the bandwidth of the Cell processor (made by IBM for the PlayStation 3, with Rambus XDR main memory) and the Nvidia RSX (GDDR3) graphics chip memory bandwidth combined. The eight gigabytes of DDR5 memory is 16 times the memory of the PS 3.
Sony didn’t give any other speeds and feeds, so it’s hard to estimate clock speeds and which AMD processor and graphics cores are in it, Krewell said.
Sony will use the cloud gaming technology from Gaikai to run “past, present and future” PlayStation titles, according to Gaikai chief executive David Perry, whose company was acquired last year by Sony for $380 million. That means, most likely, that Gaikai cloud gaming will provide the ability to play older PS 3 titles on the PS 4 through cloud-based infrastructure. That obviates the need to put backward-compatibility hardware in the PS 4, cutting its overall costs.
Cerny acknowledged that the “radical technology” of the past Cell processor got in the way of simplicity for game developers. So the new machine is making the leap to PC technology, which can easily run game engines such as Epic Games’ Unreal Engine 4. Sony showed a demo of Unreal Engine 4, reportedly running in real-time on the PS 4.
“It’s very comparable to a mid-to high end gaming PC,” Krewell said. “Even though the CPU is combined with the GPU on one die (like most PC processors today), the higher bandwidth memory and custom design, indicates it should have graphics performance more like a discrete GPU found in a gaming PC.”
Cerny mentioned that the CPU has eight cores and a state-of-the-art GPU, with a combined two teraflops of computational performance. He showed the Havok physics engine running on the PS 4 while consuming very little of the system’s overall computing power.
Will history matter?
In the last technology arms race, the technical specifications didn’t really matter.
By the end of 2012, Nintendo sold more than 100.2 million Wiis, while Microsoft sold 79.2 million Xbox 360s, and Sony sold 75.2 million PlayStation 3s, according to research firm Strategy Analytics.
Microsoft came out first in 2005 at $300 and $400, while Sony and Nintendo followed in the fall of 2006. Sony’s machine packed so much technology in it, including a state-of-the-art Blu-ray player, that it cost $500 and $600 at the outset. But Microsoft used one of the same core designs that Sony did. The three-core Xbox 360 processor proved a match in terms of the visuals it could deliver compared to the PlayStation 3, and Microsoft made improvements along the way.
Both Microsoft and Sony used the traditional console chip strategy. They developed high-end chips using some of the very best — but also cost-friendly — chip designs available. They sold their initial hardware at a loss and then reaped the benefits in terms of software revenues — the razor and razor blades model.
Over time, they rode the cost curve down. Every year, they redesigned the system boards and other components with the goal of cost reduction in mind. And every couple of years, they switched to a more advanced manufacturing process. This switch was enabled by the continuous innovation of the semiconductor industry that has been codified as Moore’s Law, which holds that the number of transistors on a chip doubles every two years. Chip makers did this by shrinking the width between circuits. With shorter circuits, electrical signals could move from one part of the chip to another faster. The power consumed declined, and, since the distance between circuits was smaller, chips could be made smaller and still perform the same functions as earlier generation chips. Since the chips were smaller, they consumed less material, the yield (or ratio of functioning chips out of the total made) improved, and the costs went down.
In the case of console chips, the system makers reduced the cost of the chips while keeping the performance the same. They could reduce the cost of making the consoles dramatically every couple of years, and over the course of six or seven years, they have been able to bring the prices down.
An Xbox 360 now sells for $200 to $300, but with much larger hard disk capacities. The PS 3 sells for the same.
But Nintendo’s Wii turned the technology race in a new direction, unexpectedly adding motion-sensing controls with a remote-control-like controller. It cost only $250, and Nintendo claimed from the outset that it wasn’t selling the hardware at a loss. Nintendo used older chip processes and started out with much lower costs. It now sells a Wii for $130. That enabled it to reach a much wider mass market.
Yet Nintendo’s console ran out of steam more quickly than its rivals did. While the graphics of PS 3 and Xbox 360 games raced ahead, the cartoon graphics of the Wii made it seem like a poor substitute. For the past couple of years, the Xbox 360 has sold more units than the Wii, erasing some of Nintendo’s previous overwhelming lead in this generation.
Microsoft’s fortunes turned around as it continuously upgraded its software-driven Xbox Live service, and then it introduced the motion-sensing Kinect add-on system, which enabled the console to recognize gestures and voice commands. The Xbox 360 had just enough horsepower to pull off Kinect — something the Wii clearly couldn’t do with its aging technology.
This time around, Nintendo’s low-cost strategy appears to have failed Nintendo. The new Wii U ($300 to $350) upgraded the experience from the old Wii to match what the PS 3 and Xbox 360 could do, and it enabled Nintendo games to use high-definition graphics. But the Wii U isn’t even outselling the Xbox 360, and Nintendo has shaved back its estimates for sales ending March 31 to 4 million, down from 5.5 million earlier. Nintendo evidently make enough of a technological leap, even though it tried to incorporate the tablet phenomenon into its new game system.
Meanwhile, all three of the console makers are being buffeted by competition from Apple and Google-based smartphones and tablets.