The clues that Huawei has been leaving during these last days anticipated that today it would present its new SoC for its high-end smartphones. And so it has been. The new Kirin 990 has arrived, and, on paper, looks really good. In its manufacture, the Chinese company uses the advanced 7 nm FinFET Plus integration technology , which has allowed them to integrate an overwhelming number of transistors: no less than 10.3 billion.
However, this is probably not your most impressive asset. During the presentation Huawei has emphasized that this new SoC incorporates a 5G modem designed to access both SA (Standalone) and NSA (Non-Standalone) networks , a feature that, according to the Chinese brand, puts a solution through its solution in front of its competitors, which at the moment only contemplate the possibility of accessing 5G NSA networks. In any case, this is just “the tip of the iceberg.”
Kirin 990 5G and Kirin 990: technical specifications
|KIRIN 990 5G||KIRIN 990|
|LITHOGRAPHY||7 nm + EUV||7 nm|
|CPU||2 x Cortex-A76 @ 2.86 GHz
2 x Cortex-A76 @ 2.36 GHz
4 x Cortex-A55 @ 1.95 GHz
|2 x Cortex-A76 @ 2.86 GHz
2 x Cortex-A76 @ 2.09 GHz
4 x Cortex-A55 @ 1.86 GHz
|GPU||16-core Mali-G76||16-core Mali-G76|
|NPU||2 Big-Core + 1 Tiny-Core||1 Big-Core + 1 Tiny-Core|
|UFS||UFS 3.0 and UFS 2.1||UFS 3.0 and UFS 2.1|
|MODEM||2G, 3G, 4G and 5G||2G, 3G and 4G|
Redes 5G SA vs NSA
To understand what is undoubtedly one of the most important assets of the Kirin 990 SoC, it is good for us to briefly review what the SA and NSA networks are and to what extent our experience may determine the fact that the SoC of our mobile phone is not compatible with both. The complexity involved in the deployment of the 5G communications infrastructure has caused the companies that are carrying it out to choose to divide it into two large phases known as NSA, which we can translate as’ non-autonomous’, and SA, which means’ autonomous ‘. The NSA phase precedes the SA stage for a compelling reason: it is easier to implement and less expensive. And it is that during the NSA phase 5G networks settle on the existing 4G infrastructure.
The complexity involved in the deployment of the 5G communications infrastructure has caused the companies that are carrying it out to choose to divide it into two large phases known as NSA and SA
The 5G devices that intervene in the communication during the NSA phase, like any of the 5G smartphones that have already been presented, establish the link using the frequency bands characteristic of 5G communications because, otherwise, they would not be able to reach the speeds of transfer that this technology has promised us (average rates higher than 100 Mbps and theoretical peaks of up to 20 Gbps). However, the communication between these terminals and the base stations is carried out through 4G links such as those we use now, hence the NSA phase does not define a 5G infrastructure that takes advantage of the benefits of this technology at all levels.
This limitation is resolved in 5G SA networks. And it is that the devices involved in the communication use the frequency bands that require 5G links both during data transfer operations and when conducting the control dialogue with the base stations. This improvement is possible because the SA phase defines 5G networks independent of the 4G infrastructure that we currently have.
7 nm FinFET Plus EUV lithography
As I anticipated a few lines above, the integration technology used by Huawei in the manufacture of this new SoC has made it possible to include a huge amount of transistors in the chip core. And this, in turn, has facilitated the integration of the 5G modem we have just talked about and which, according to the Chinese company, makes it easier for us to access both NSA and SA networks. Leaving aside “the nanometer war” of which we have been talking for several years due, above all, because Intel is having a hard time following the pace imposed by some of its competitors, it is clear that lithography seriously conditions development of new microprocessors, hence Huawei has given so much importance to the integration technology that supports its new Kirin 990 SoC.
In addition, the development of integration technology can have a direct impact on the size of the chip because it not only affects the size of the transistors; It also allows you to place them closer together. Huawei has confirmed during its presentation of the Kirin 990 SoC that it is at least 36% smaller than its competitors. Should this matter to users? In an indirect way, yes, especially due to the reduced space available within the chassis of a smartphone and what this entails in terms of heat dissipation capacity in the form of heat.
GPU Mali-G76 with 16 cores
The improvements introduced by Huawei in its new SoC also involve graphic logic, which in the Kirin 990 chip is a Mali-G76 with 16 cores . The Chinese company has insisted during the presentation that it is the first SoC in the market that has these characteristics, and yes, on the wallpaper really well. But what interests us is to know what impact this new graphic logic can have on our experience. We will not go out of doubt until we can thoroughly test the first Huawei mobile phones that will incorporate this SoC, which will, in all likelihood, be the impending Mate 30 and Mate 30 Pro .
Of course, we can intuit that the new GPU will represent a remarkable step forward when executing multimedia games and applications that require an important effort to graphic logic. According to Huawei, the new microarchitecture also benefits from a cache memory linked to the graphic subsystem that has allowed them to reduce the need for bandwidth imposed by the link between the GPU and DDR memory by 15%, and has also reduced 12% DDR memory consumption. These figures are not but not bad, yes, as long as they faithfully reflect reality.
Kirin 990 and artificial intelligence 2.0
Much of the presentation of Huawei’s new microprocessor has revolved around the alliance that establishes 5G connectivity and artificial intelligence in the Cloud. Two of the most notable improvements introduced by 5G communications (although not the only ones) are a very substantial increase in transfer speed and an equally noticeable reduction in latency . And, as we can foresee, the artificial intelligence tools that we access from our smartphones can benefit greatly from these improvements.
According to Huawei, the combination of the double NPU ( Neural Processing Unit ) or Neuronal Processing Unit integrated in the Kirin 990 and the 5G modem allows real-time inference processes to be carried out not only on the mobile phone itself, but also on the Cloud. This means, in simpler words, that the greater power of the section of the Kirin 990 chip that is responsible for the execution of artificial intelligence algorithms, and, above all, the presence of the 5G modem, facilitates a much faster and faster communication. with less latency towards the Internet servers where the artificial intelligence applications we use reside.
This is the reason why the Chinese company talks about Artificial Intelligence 2.0 , stressing in some way that its new microprocessor with 5G connectivity will put at our disposal AI tools that require a very important calculation effort, and therefore, only They can be processed on Internet server farms. 4G communications do not offer us the immediacy that these applications require, which must send the result of the inference in real time to our smartphone. But 5G links do. The impact that this “new” artificial intelligence will have on our experience is not clear yet, but we can assess it when the first Huawei phones with Kirin 990 SoC are available and the 5G deployment reaches maturity.