Like the rival Xbox Series X, Sony's PlayStation 5 chipset is based on AMD's Zen 2 and RDNA2 architectures.
The next generation of game consoles will be released this year's Christmas market. Microsoft revealed the technical details of its Xbox Series X console this weekend, and today Sony responded to the challenge with Mark Cerny.
Sony's PlayStation 5 console is based on AMD's semi-custom system chip. Equipped with eight Zen 2 processor cores, 36 Compute Unit RDNA2 graphics controllers and a highly customized I / O controller. Alongside the system chip, the 256-bit memory bus extends to 16 GB of GDDR6 memory, providing a total of 448 GB of bandwidth per second.
The system circuit I / O controller is equipped with a Kraken decompression unit, cache coherence units, a dedicated DMA controller, SRAM memory and two auxiliary processors. Auxiliary processors have memory and SSD traffic management. The use of compression is made transparent for developers and allows them to handle their games as if they were using uncompressed data.
PlayStation 5 focuses on SSD performance in particular. It utilizes the company's proprietary Flash controller chip, which, among other things, supports the newer Kraken compression technology, which offers 10% more compression compared to the ZLib used by PS4. The memory chips that extend the 12-channel controller provide a total of 5.5 GB / s of bandwidth for a total of 825 GB. However, with compression technologies, bandwidth is efficiently 8 to 9 gigabytes per second. The SSD communicates with the system circuit via the PCI Express 4.0 bus.
Additional storage can be added to both external hard drives and SSDs, but they are mainly suitable for PlayStation 4 and older games. In addition, the console will have a dedicated port that will accommodate standard M.2 NVMe drives. However, virtually any NVMe drive will not work, but it must fit into the location not only because of its physical limitations, but also fast enough. The company will announce later which SSDs will be compatible, but one of the requirements will be at least PCI Express 4.0 support. According to Cerny, the list of compatible drives will probably not be released until after the console itself. The console will use a 4K UHD BluRay drive.
The PS5 graphics driver is based on RDNA2 architecture. Equipped with 36 RDNA2 Compute Units, it offers a maximum clock frequency of 2.23 GHz, which means 10.28 TFLOPS performance. The specified clock frequency is the maximum Boost clock rate for the graphics controller. The graphics controller caches are customized with dedicated controllers that manage cache coherence in collaboration with I / O controller units. According to Cerny, despite AMD's push for backwards compatibility, the leap between consoles is so big that at least some of the games will require apparently separate upgrades to ensure compatibility. The company expects most of the 100 most played PS4 games to work on the PlayStation 5 when it is released.
The graphics controller also supports beam tracking acceleration with the RDNA2 architecture. At the same time, Cerny confirmed the radiation tracking of RDNA2 based on a previously disclosed patent where the Intersection accelerator is part of the TMU complex. In practice, the shade units send the Intersection accelerator a command to inspect hits of a particular beam with a single command, then it is available anywhere while the Intersection accelerator is doing its job. From Sony's point of view, support for primitive shades is also new, but it is unclear here whether AMD's primitive shades have already been introduced in the Vega generation, or newer and more advanced Mesh shades.
According to Cerny, the graphics controller's Boost algorithm is based on utilization, which means that players in hotter environments do not have to suffer from performance, but clock speed is determined by the current strain. Different types of tasks exert different pressure on the graphics controller, so the clock frequency drops slightly lower in more power-hungry situations. In addition, AMD's SmartShift technology is used to transfer, as far as possible, the portion of the TDP budget allocated to the processor to the graphics controller. The processor also uses Boost technology and has a maximum clock frequency of 3.5 GHz. According to Cerny, both the processor and the GPU will drop below the advertised clock frequencies in some situations, but they should spend most of their time staying within those numbers.
Sony's new console has invested in audio through its own proprietary Tempest 3D audio technology. This is an interesting solution in that it is based on AMD's unspecified Compute Unit, which has been modified as much as possible to the Cell Processor SPE. According to Cerny, the SPE units were nearly perfect for audio processing and, for example, all caches have been removed from the Tempest Compute Unit because the SPE units also operate without caches. In addition, Sony hopes that in the future, it will be able to customize the HRTF map used individually for each user to ensure the best possible experience.
(Embed) https://www.youtube.com/watch?v=ph8LyNIT9sg (/ embed)
The full presentation of Cerny can be seen in the video above.