1. 8nm process! Xiaolong 730 exposure, integrated NPU
According to information gathered from the firm's firmware by well-known informant Roland and XDA, the previous Xiaolong 670 actually corresponds to Xiaolong 710. In addition, Qualcomm also prepared a Xiaolong 730.
Indian media suggestphone disclosed the specifications of Xiaolong 710 and Xiaolong 730 exclusively. It seems that Xiaolong 730 further bridges the gap between Xiaolong 600 and Xiaolong 800, which is very powerful.
Specifically, the Snapdragon 730 is built on Samsung's 8nm LPP process (theory is 10% more energy efficient than 10nm LPP). The CPU is designed with a '2 Big+6 Little' size core design. The big core is the Kryo 4xx series, clocked at 2.3GHz. , 256KB slow, small core clocked at 1.8GHz, 128KB slow, and share 1MB three slow.
The GPU is an Adreno 615, clocked at 750MHz and up to 60FPS@2K.
ISP upgrade to Spectre 350, up to 32 million pixel camera, native three camera support. A relatively big highlight is the independent NPU 120, with the HVX vector unit to enhance the overall computing efficiency and battery life.
In contrast, the Snapdragon 710 is based on Samsung's 10nm LPE process (with Snapdragon 835, a 27% increase in performance over 14nm, and a 40% reduction in power consumption). It is also a 2+6 8 core design and Kryo 3xx has a maximum frequency of 2.2GHz. The ISP is the Spectre 250. The display supports 19:9 ratio of up to 3040x1440 resolution, 10bits depth and HDR10, no separate NPU unit.
The Xiaolong 730 and Xiaolong 710 both have SDR660 RF chips, but the former support Bluetooth 5.1 technology, which is better than the Xiaolong 710 Bluetooth 5.0. In terms of supported memory specifications, there is no difference between the two processors, but the display The screen processing unit (DPU) model is higher, so that even in the case of the same support 3040 × 1440 resolution display, including multi-tasking switching effect should be more smooth.
As for the ISP configuration of the two processors, the Spectra350 equipped with the Xiaolong 730 has increased the support for the FS2 sensor compared to the Spectra250 equipped with the Xiaolong 710. However, the highest supported pixel is also 32MP. However, it is worth noting that At this time, Xiaolong 730 also seems to integrate a dedicated NPU artificial intelligence chip.
The media expects that the Xiaolong 710 will be launched in the first half of 2019. Since Samsung's 8nm did not complete verification in October last year, the Xiaolong 730 may have to wait until late 2019.
It is understood that Samsung's 8nm LPP process completed its verification work in October 2017. Samsung said that compared to 10nm process technology, the new 8nm process will increase the chip's energy efficiency by 10%, and the chip area will be reduced by 10%.
Samsung also revealed that the 8nm LPP process process is one of the most advanced process processes before 7nm production, and vice president of Qualcomm also stated that '8nm LPP will use proven 10nm process technology while providing better performance and scalability than current 10nm products. Sex, so the speed will grow quickly. '
Analysts said that Samsung's 8nm LPP process process is mainly for the preparation of 7nm process, Samsung and Qualcomm's ARM processor may use the latest 8nm LPP manufacturing process.
However, the first exposure of the Snapdragon 730 processor has not yet been disclosed, and when the market and shipments have no information, but according to Qualcomm’s previously announced information, the first batch of Snapdragon 700 series processors will be In the first half of 2018, commercial samples will be provided to customers. This means that models equipped with Xiaolong 730 processors may have to wait until the second half of this year.
2.Google releases artificial intelligence chip TPU 3
TechCrunch reported that Google CEO Sundar Pichai released the third generation TPU (Tensor Processor Unit, TPU3) at Google I/O in 2018. He said that the TPU 3 rack cluster (Pod) is more efficient than last year's TPU 2 Pod. 8 times, up to 100petaFLOPS.
Venture Beat reported that Pichai also mentioned that TPU 3 is too powerful to install liquid cooling equipment for its data center.
CNBC reported that the new TPU can help Google improve artificial intelligence (AI) speech recognition, find specific objects in photos and videos, and understand the emotions hidden behind the text. Google announced the first generation of TPU in 2016. The test results show that TPU 2 outperforms existing GPU options in specific scenarios.
Depth learning (Deep learning) promoted by tech companies such as Google usually consists of two phases. First, researchers train built-in GPU computers to identify cars in photos by giving them large amounts of data. After completing training, researchers In the second phase (referred to as inference Inference) requires the computer to make predictions based on new data. TPU 1 can only be applied in the inference phase of deep learning, and TPU 2 released last year is able to handle training (in the first phase) and inference.
Amazon founder and CEO Jeff Bezos mentioned in an annual letter to shareholders last month that in the United States, the United Kingdom, and Germany, Amazon promoted machine learning components and used semi-supervised learning. , SSL), increased the verbal comprehension of Alexa's digital virtual assistants by 25% in the past 12 months.
Salesforce.com Inc., a customer relationship management (CRM) vendor, announced in November last year that it will work with Google to launch new strategic partnerships. Google Cloud Platform will become the preferred public cloud provider for Salesforce’s international infrastructure expansion.
3. Giants are frequently exposed to chip vulnerability How can both performance and information security be achieved?
Frequent chip vulnerability problems, this time for Nvidia. The ReSwitched team, which specializes in cracking Nintendo Switch game consoles to execute homebrew software, has recently disclosed the vulnerability details called Fusée Gelée, which is hidden in the Switch host core Nvidia Tegra X1 chip. The security loopholes in the system will allow hacker to execute arbitrary programs, and claim that they can't be repaired after the host leaves the factory. Since 2018, Intel and AMD have been pushed to the forefront by chip vulnerabilities, and the chip vulnerabilities have been continuously exposed. The issue of cyber security has become a target of public opinion. To ensure performance or security, it is a proposition that needs to be balanced among chip vendors.
Chip Vulnera Causes Industry Concerns
The chip market in 2018 is doomed to be an extraordinary year. The chip off-duty incident on ZTE's side has continued to ferment. The chip giants there have frequently exposed chip vulnerability.
Recently, Nvidia Tegra chip was leaked. In the first year of its release, Nintendo's new host Switch was powered by the NVIDIA Tegra X1 chip. According to media reports, Katherine Temkin of the ReSwitch hacker team announced the use of NVIDIA Tegra X1. A bug in the chip cracks the Switch host. This vulnerability allows anyone to run arbitrary code on it. This means that home-made systems and pirated software can run at will. According to technical documentation, this vulnerability exists in the chip's Bootrom, after the device leaves the factory. Can not be patched by the patch, that Nintendo can not patch it through software updates, the only repair method is to return to replace the processor. Simply put, the discovery of the loophole is the use of a defect in the Tegra X1 chip. Not just Nintendo, all devices that use the chip will be affected.
Since this vulnerability exists in the boot area, it cannot be fixed by a simple patch. This means that, theoretically, all existing Nintendo Switch can be perfectly hacked.
In addition to Nvidia, in January 2018, Intel chips were exposed to technical flaws that caused major security holes. Hackers could use this vulnerability to read device memory and obtain sensitive information such as passwords, keys, etc. Two design flaws were called 'fuse' ( Meltdown) and Spectre. The former allows low-privilege, user-level applications to "cross-border" access to system-level memory, while the latter can fool security checkers to make applications access memory anywhere. 'Specter' affects AMD, ARM, and Nvidia's chip products, and it affects almost the entire computer processor world. The 'Specter' loopholes are believed to not be completely resolved within a few years. Apple responded that all Mac and iOS systems are Will be affected. In response, Apple has responded that all Mac and iOS systems will be affected, because Apple's Mac computers and some iPhones use Intel chips.
After a wave of waves and a wave of waves, a German computer magazine recently revealed that researchers have found eight new loopholes in the computer's CPU chip. These loopholes are related to the well-known Intel chip flaws 'Meltdown' and 'Specter'. Very similar, the researchers said, Intel is ready to issue patches to fix these loopholes. Of course, ARM has not been spared this time, and some of its chips have also been affected.
In fact, Intel mentioned in a statement earlier this year that the loopholes are not limited to Intel, and that AMD, ARM and other friendly companies have similar problems. Sure enough, apart from Intel being exposed to major loopholes, an Israeli security company, CTS Labs, also recently. Published a security white paper, which pointed out that there are 13 security vulnerabilities in AMD's chips for computer chip manufacturing. According to the report, the vulnerability disclosed by CTS this time requires administrator privileges to be discovered. These vulnerabilities involve AMD Ryzen desktop. Processor, Ryzen Pro enterprise processor, Ryzen mobile processor, EPYC data center processor, different vulnerabilities corresponding to different platforms, of which 21 kinds of environments have been successfully used, and there are 11 possible use.
The chip hole is endless
With the fermentation and upgrading of the 'chip hole door', its influence and coverage are also difficult to measure. In the 'vulnerability door' incident at the beginning of the year, although it was Intel-based, but this cross-vendor, cross-border, cross-border Cross-OS major vulnerabilities have engulfed the entire IT industry. Most mainstream processor chips such as ARM, Qualcomm, and AMD have also been affected by vulnerabilities. Processors with IBM POWER details have also been affected. Windows, Linux, and MacOS, Android and other mainstream operating systems and computer, tablet computers, mobile phones, cloud servers and other terminal equipment are subject to the above loopholes.
Today, we have a large amount of information in the form of digital computers and the cloud. Design defects from the processor will cause the data to be stolen and misused, causing users to lose personal property. Cybersecurity experts warned that the use of cloud computing services Businesses may be particularly vulnerable to attacks. If an attacker pays to gain access to a particular area of the service, they may gain information from other customers, such as Bitcoin exchanges, social chat applications, or even a large number of government agencies. Store passwords or data on the cloud server. Hackers may attack the object of the cloud service, bitcoin wallets may be stolen, and important data from government agencies may also flow to hackers.
Companies such as Amazon, Microsoft and Google have been affected. In Amazon's AWS or Microsoft's, Google's cloud service, in the same server connected user group, if a hacker uses a device, they may be able to read it. The data of another device of the same chip. Future high-precision industry chains such as the Internet of Things, cloud computing, and artificial intelligence all have chips. If there is a hardware Trojan in these chips, once triggered, it will lead to data loss. , Functional failure or even system failure.
Blocking loopholes requires the entire industry to work hard
While vendors are pursuing chip speed, they can't be taken lightly about information security. However, how to balance the efficiency and safety needs to be handled with caution. Paul Leukec, one of the 'Spectre' discoverers, points out, 'Ghosts' Affecting almost all high-speed microprocessors, the high speed requirements when designing new chips cause them to be prone to safety problems. 'The entire industry has always wanted to be as fast as possible while achieving safety. But the 'ghost' shows that fish and Bear paws can not have both... This is a problem that will continue to deteriorate as the hardware life cycle continues.
Industry analysts also believe that related fixes will affect the processor's computing speed. Linux system developers said that the repair of chip security vulnerabilities will affect the speed of the operating system. The typical performance degradation rate is 5%. The US Department of Homeland Security at a The statement stated that the security patch is the best protection method to solve the chip vulnerability, but the performance of the computer may be reduced by as much as 30% after patching, and the chip vulnerability is caused by the CPU architecture rather than the software, so patching cannot be complete. Solve the chip vulnerability. If the security problem solves the result but the CPU performance is worse, isn't this outweighed?
In response, some experts also stated that currently all major suppliers have provided corresponding mitigation measures. Although the risk of chips being attacked is reduced, it is only a matter of “treating the symptoms.” They need to consider whether the next-generation products need to be Replace the new architecture to fundamentally solve this type of problem.
Chip vulnerabilities have become 'all-in-one-loss' all-industry incidents and require the close cooperation of the entire industry chain to solve them together. Well resolved, everyone is saving their lives. In the event of a security attack, it is not only Intel or which company is harmed. Manufacturers, but the entire industry. Communications Newsletter
4. Strong demand for new mobile phone parts stocks. Estimated Q2 revenue increase 19-23%
Thanks to the strong demand for new handsets, the leading chipmaker of the driver chip is expected to increase its revenue in the second quarter by 19% to 23% in the second quarter. The increase in revenue will exceed market expectations. In addition, Wang Suiren, general manager of Lianhe, also disclosed that The second half of the year will be even better than the first half of the year. The operating target of the second quarter is quite optimistic. It is estimated that the combined revenue in the single quarter is expected to be between RMB 12.5 billion (NTD, the same below) and RMB 12.9 billion, and the quarterly increase will be 19%. To 23%, the second quarter gross margin is expected to be between 28.0% and 29.5%, while the single quarter operating profit rate is expected to fall between 12.0% and 13.5%.
The joint venture announced the first quarter of this year's transcript, single-quarter consolidated revenue of 10.466 billion yuan, a decrease of 12.43% from the fourth quarter of last year and a decrease of 4.15% from the first quarter of last year. The company explained that the first quarter revenue decreased compared with the previous quarter, mainly Due to the impact of the traditional off-season and the reduction of working days during the Spring Festival, the gross profit rate in the first quarter was 29.44%, which was 29.37% higher than the fourth quarter of last year, an increase of 0.07 percentage points, mainly due to the difference in product mix. The net profit was 1.156 billion yuan. After the single-season tax profit was 917 million yuan, it was reduced by 30% in the quarter and 1.32% in the year.
In view of the second quarter operational outlook, Wang Shouren said that the main reason is that the demand for spare parts for new mobile phones is very strong, and the performance of single-chip panel-driven chips and system-on-chip products will all grow in tandem, but in turn, demand for panel-driven chips will increase. The most powerful.
Wang Shouren further mentioned that in driving the production line of chips, TDDI chips with integrated driving and touch functions are being promoted faster than expected. More than 10 million units will be shipped in the first quarter of this year, and will be more in the second quarter. Annual total shipments do not rule out the opportunity to reach 100 million sets.
In addition to the overall market, Wang Shouren also expects that not only will the outlook for the second quarter be optimistic, but also the performance in the second half is expected to be better than in the first half of the year. As for the performance of the joint venture in April, it is equally eye-catching, with revenue rising by 10.34%. Increase 6.79%.
In addition, due to rising prices of silicon wafers, which led to higher costs for designers, Wang Shouren mentioned that in response to the increase in the cost of memory and passive components due to silicon wafers, it is currently communicating with customers to adjust the selling price of products.
5.Qualcom plans to launch new chip platform for Wear OS smart watch
In February 2016, Qualcomm released the Xiaolong Wear 2100 SoC for wearable devices. However, two years later, the Android Wear (later named Wear OS) market performance has been unsatisfactory. Fortunately, in the near future, In an interview with Wearable of foreign media, Pankaj Kedia, Qualcomm’s wearable director, revealed: “The company is aware of this issue. It will release an upgraded chip for flagship smart watches this fall. Many partners will also launch during the holiday season. Wear OS Wearable Device Based on New Chip'.
Kedia said that the third-generation processor will 'create an uncompromising smartwatch experience' from scratch, which is said to give the watch a better battery life.
In addition, Qualcomm is clearly designing chips that will make the alway-on display even more energy-efficient. He said: A smart watch, you must first do your watch's job. Look at it, or not watch it, It must be mellow and seductive. When I do not look at it, it should not be black and white or static.
We have had conversations with consumers of Fossil and Michael Kors who do not wish to compromise on smart watches.
Although Qualcomm is not yet ready to disclose its specifications, Kedia claims that the new hardware platform will 'significantly change the Wear OS ecosystem'. cnBeta