All hardware specs: PlayStation 5 GPU breaks the 2 GHz mark

Sony has unveiled the specifications of the upcoming PlayStation 5 game console. The focus is on a system-on-chip (SoC) designed with AMD with eight Zen-2 cores with speeds of up to 3.5 GHz, 2304 shader units and its own SSD controller, coupled to 16 GB GDDR6 RAM. Sony also relies on an internal 825 GB NVMe SSD with PCI Express 4.0 to reduce loading times and RAM requirements.

2304 shader units may seem familiar to owners of a PlayStation 4 Pro: Sony's previous game console had just as many. The GPU in the PlayStation 5 should nevertheless calculate by several factors faster thanks to the higher clock frequency and modern RDNA-2 architecture. The GPU clock increases from 911 MHz to 2230 MHz, which results in a theoretical FP32 computing power of 10.3 instead of 4.2 TeraFlops (+ 145 percent). AMD's RDNA-2 architecture makes it easier for developers to utilize the shader cores than on the PlayStation 4 with GCN graphics unit. Just like the Xbox Series X, ray tracing cores are on board to create pretty, realistic graphic effects.

By continuing to use 2304 shader units, there is synergy in backward compatibility – Sony can more easily estimate how previous PS4 games are running on PlayStation 5 and, if necessary, adjust the clock.

Dynamic boost instead of fixed clock frequencies

The manufacturer does not guarantee fixed clock frequencies for the PlayStation 5, but relies on a boost for the processor and graphics unit. Unlike desktop graphics cards such as the Radeon RX 5700 XT, this boost is not primarily based on the chip temperature, but on the computing load and power consumption. The PlayStation 5 should maintain a specified (not mentioned) power consumption at all times. In order not to exceed this threshold, the CPU and GPU can throttle their clock. According to Sony's chief architect Mark Cerny, a few percent loss of clock speed would be enough to reduce power consumption by 10 percent – players would not notice anything.

At this point, Sony takes a swipe at Microsoft's Xbox Series X, whose 3328 shader cores with 1825 MHz reach 12 TeraFlops: 1 TFlops is not equal to 1 TFlops, explained Cerny. A smaller GPU with a high clock frequency enables a higher frame rate than a larger GPU with a low clock frequency – despite nominally the same raw power. There are two reasons for this: Firstly, each additional megahertz not only speeds up the shader cores, but also the backend and the caches; the GPU feeds the computing units more or less faster. On the other hand, a few shader cores can be used more easily.

16 GB GDDR6, 825 GB Flash

On the memory side, Sony meanwhile uses 8 × 2 GB GDDR6 on 256 data lines with 7000 MHz (14 Gbps). That gives 16 GB of RAM with a continuous 448 GB / s. Sony, like Microsoft, argues that the internal 825 GB SSD reduces the RAM requirement because the GPU has to store less data locally that would ultimately not be used. The AMD combination processor has its own SSD controller, which takes care of data processing and its decompression. This relieves the CPU cores.

Users of a PlayStation 5 can expand the memory with conventional PCIe SSDs from the trade, but with restrictions: Sony requires high-end models with PCIe 4.0, which reach the 5.5 GByte / s of the internal SSD. The size does not matter – even 4 TB models should work. A list of compatible models follows for the market launch. PS4 titles can still be started from an external USB HDD. If desired, users can also copy them to the internal SSD.

(mma)