Under the hood of the PlayStation 5: lead system architect Mark Zerny spoke in more detail about the technical decisions made

March 18, Sony introduced the technical specifications of the PlayStation 5. Leading system architect Mark Zerny made a presentation during which he spoke about the basic principles of the next generation console. You can read more about this in the Eurogamer article.

Two weeks later, the engineer again met with Digital Foundry to supplement the available information. There were no new details about the console this time, however, their conversation deserves some attention.

The main thing from the interview with Mark Zerni technical division of the site Eurogamer read below.

Overclocking innovation: how it works on the PlayStation 5

One of the main features of the new system is certainly the dynamic frequency of the central and graphic processors. The innovative technology used in the PlayStation 5, in fact, provides the system with the ability to choose the frequency of the chip based on its heat dissipation.

Zerni recalled that all consoles will work the same regardless of the situation and external factors. Your PlayStation 5 cannot work slower than that of a neighbor, since you have it in a closed rack and it has it in the refrigerator.

“One of the main differences is the ambient temperature. The console can be located in a hotter or cooler place in your game room, ”says Zerni. “Another is the presence in the console of a non-standardized chip. Some chips heat up more, others less. To “align” the values, the creators used an algorithm in which the frequency depends on the activity of the central and graphic processors. This maintains consistency between PS5 components. ”

Inside the processor is a power control unit that constantly measures the activity of the central processor, graphics processor and memory interface, evaluating the nature of the tasks they perform.

As for the developers, they will have to get used to the new conditions. However, they will understand perfectly how much energy is used by the processor and the graphics subsystem.

The idea is that the creators of the games will learn how to optimize their creations in a different way, achieving identical results from the GPU, but doing it faster by increasing the clock speeds obtained by optimizing power consumption.

“Each CPU and GPU has a common energy budget,” says Zerni. “And of course, the GPU power budget is a priority. If the central processor does not spend its entire energy consumption budget, then unused power goes to the schedule. This is what AMD calls SmartShift. There is enough energy in the system so that CPUs with GPUs can operate at maximum frequencies – 3.5 GHz and 2.23 GHz. Developers will not have to choose which one to slow down. ”

Zerni emphasizes that power consumption and clock speed are not linearly related. For instance, if you reduce the frequency by 10%, then the power consumption will drop by about 27%.

Probably, many will be interested to know how “overclocking” will affect game design. Several developers told Digital Foundry that they are forced to reduce the frequency of the PS5 graphics core to 2.23 GHz, since most game engines are designed with low Jaguar performance in mind (current hardware).

Eurogamer specialists compare this to the performance profiles that they saw on the Nintendo Switch.

“As for the profiles with a locked frequency, we supply them with all the devkits. Developers may not need the “floating” frequencies during optimization. Games released for the PS5 will always use higher frequencies to unlock extra power, ”says Zerni.

But what if developers are not going to use the entire power ceiling of the PlayStation 5? Is there some kind of “worst case scenario” equivalent to the base frequencies, like on a PC?

“Developers do not need to optimize their project in any way; if necessary, the frequency will be independently adjusted depending on the actions of the central and graphic processor, ”notes Zerni.

An interesting fact: if you write code that causes the console to overheat and perform a thermal shutdown, then it will work only on PS4. “The PS5 will handle such an unrealistic piece of code more elegantly,” Cerny smiles.

What was required from the CPU for backward compatibility

Regarding backward compatibility, Cerny confirmed that his words were misinterpreted. Speaking about launching the 100 best PlayStation 4 games on the PS5, he meant something completely different.

Additional logic has been added to the PS5 architecture to ensure full compatibility with PS4 as well as PS4 Pro.

“All of the game logic created for Jaguar processors works correctly on Zen 2 processors, but the execution time for instructions can vary significantly,” says Zerni. “We collaborated with AMD to create specific Zen 2 cores; they have modes where frequencies are close to Jaguar values. We keep this a secret, as the work on setting up backward compatibility continues. ”

The main advantages of SSD and what they give

Developers will be provided with a new I / O API – low-level and high-level. The first simplifies the work, and the second opens up complete freedom, allowing you to configure everything and everything.

The file system has also been revised. Instead of the traditional access to files by name, identifiers were used. Now developers only need to specify the identifier, the starting location and the ending, and after a few milliseconds the data will be downloaded.

The new system allowed to significantly accelerate the loading of games, amounting to a few milliseconds. For comparison, on the hard drive the same process usually took up to 250 ms.

These 250 milliseconds were a big problem for developers, forcing to keep a lot of data in RAM. Now, to get instant access to urgent data, you no longer need to store them in RAM. This opens the door to huge savings in next-generation efficiency.

Now there is no need to cache a lot of information that the console might need … or not. The double system is also a thing of the past. Since the HDD head is moving rather slowly and data searching can take as long as reading, the same data is often duplicated hundreds of times.

If you refuse duplication, then disk performance will drop to a minimum and data reading speed will drop from 50 – 100 MB / s to 8.

The developers of Marvel’s Spider-Man had to create 600 duplicate trash bags located in each quarter of the city. And the size of one object, for a minute, is 1.6 MB.

“The Insomniac rule states that any object used in the game more than four hundred times must be in RAM. Packages were one of those facilities. ”

This is another example of how SSD Helps Reduce Next-Generation Games Distribution Size. The installation package will be optimized because duplication is not required.

How 3D sound works and how powerful the Tempest Engine is

Sony technology is ahead of Dolby Atmos and PlayStation VR, processing hundreds of sources in three-dimensional space (32 and 50, respectively). Sound transmission does not require special equipment, and solutions such as ILD (interaural level difference – the difference in sound levels between the ears) and ITD (interaural time delay – sound delay between the ears) will be taken as the basis.

Sound is processed at a frequency of 48000 Hz with 256 samples – 187.5 audio ticks per second. That is, a new sound reaches the ear every 5.3 ms.

Tempest Engine itself is a computing unit that runs on a GPU frequency and provides 64 flops per clock cycle. Thus, its peak performance is about 100 gigaflops, which corresponds to the eight-core Jaguar CPU cluster used in the PlayStation 4. Despite the fact that the unit is based on the GPU architecture, its use is very different.

“The GPUs run on hundreds or even thousands of fronts, and the Tempest Engine only two,” explains Mark Zerni. “One is for 3D sound and other system features, and the other is for playing. The bandwidth of the engine is more than 20 GB / s, but you should be careful, because the sound can overtake on itself the part required for graphics processing. “

Tempest Engine is compatible with the Ambisonics format, which is a virtual speaker system combined with physical speakers. When using it, an expanded sense of presence is created, since any sound can be reproduced at one of 36 levels. Discrete sound tends to “block” physical speakers and may not appear at all on some of them. Ambisonics is already available on PlayStation 4 and PSVR, but with much fewer virtual speakers.

What opportunities will the new 3D-audio system provide owners of speakers 5.1 / 7.1

Deploying 3D audio on the PlayStation 5 may take some time. Currently, developers are using basic technology, so the transfer of revolutionary sound to users using speakers is still under development. At launch, the Tempest Engine will be available exclusively to owners of standard headphones.

“Players can turn on the TV Virtual Surround function on the stereo speakers of their TVs and get the ability to play sound that has no analogues,” says Zerni.

As the architect mentioned, the basic implementation of 3D sound for TV and stereo speakers is currently under development and the PlayStation 5 team continues to optimize it.

“As soon as we are satisfied with the result for two-channel systems, we will move on to 5.1 and 7.1,” he added. “At launch, 5.1 and 7.1-channel systems will get a solution close to what we have on the PS4.”

Read also: Designing the Xbox Series X: How Microsoft Rethinks the Console Form Factor

Add to our Telegram channel via the link or search for it manually in the search by name gmradost. There we publish, including that which does not fall into the news feed.. Also subscribe to us in Yandex.Zen, Twitter and VK. And do not forget that we have a dark theme and a ribbon instead of tiles.