It’s nearly Thanksgiving and we’re here to give thanks for all the great stuff in technology, like new ARM64 server chips, ARM's new security platform, the forthcoming 5G standard, and new reports (like the one listed below).
TIRIAS Research announces the publication of two research briefs on technologies intersecting AR/VR as part of the new AR/VR research service. In The Cloud is the New Training Ground for Virtual Reality, we look at the potential for VR to move to the cloud to enable mobile clients to deliver high-performance experiences otherwise inaccessible for the next decade. In Virtual Reality is the New Training Ground for Artificial Intelligence, we examine the intersection of VR and AI in the cloud to advance AI training through advanced visual and sensor simulation.
The TIRIAS Research AR/VR Practice employs a forward-looking framework for forecasting the evolution of AR/VR systems and experiences. We segment the field of emerging systems into tiers, tracking performance, capability, and user experiences by tier. We also look at intersections with adjacent technologies including machine learning, computer graphics, and computer vision to predict major inflection points that shape market transformation. Expert in understanding the market and technical challenges, TIRIAS Research provides direct support to teams seeking strategy guidance for technology investment, product development, and go-to-market.
The TIRIAS Research AR/VR service is led by Simon Solotko, an accomplished VR and AR pioneer who serves as a mentor at the German Accelerator. With two decades of experience spanning HPC, semiconductors, and recent work with dozens of AR and VR companies, Simon is regarded as an AR and VR leader and subject matter expert. Recent articles in the media published by Simon include The Social Matrix: How The Cloud Can Activate A Billion Virtual Reality Users on UploadVR and Virtual Reality is the Next Training Ground for Artificial Intelligence on Forbes.
For more information on the Tirias Research AR/VR service, please contact Simon Solotko at email@example.com or by phone at 512-508-1054.
Broadcom Makes a Bid For Qualcomm
The Broadcom CEO Hock Tan is ready to make the bid of a lifetime. As CEO of Avago he made many deals to acquire companies, including such semiconductor industry stalwarts as LSI and Broadcom (the renamed the company to the latter).
Broadcom has launched an unsolicited $100+ billion bid to acquire Qualcomm in what would create a networking and communications powerhouse with a $200 billion market cap. The proposed acquisition pits Broadcom’s strategy to grow with through acquisitions and operational efficiency against Qualcomm’s long-term R&D investment in future communications technologies. While there is little stand in the way of the acquisition, the difference in business strategies could turn the proposed acquisition hostile. Qualcomm’s board has unanimously rejected the offer at $70 per share setting up a battle for five board seats up for re-election. TIRIAS Research has analyzed the potential impact of the deal and concluded that it would have wide ranging negative impacts on the industry and that there are likely other parties involved in the offer that are not yet visible. For more information please review the full articles on Forbes at:
The Real Deal Behind Broadcom's Hostile Bid For Qualcomm (Part 1) -- The Impact
The Real Deal Behind Broadcom's Hostile Bid For Qualcomm (Part 2) -- The Players
AMD and Intel Team Up To Build a New Mobile Processor
The idea that AMD and Intel would team up on an integrated part that used Intel’s CPU and AMD’s GPU was first rumored last year. Kevin Krewell wrote up his take on the rumor in Forbes 11 months ago.
And now both AMD and Intel have revealed that there is indeed a plan for a cooperative venture. The difference is that there is no IP transaction involved - AMD will provide semicustom GPU silicon that Intel will integrate into a multichip package. This is not a transfer of AMD’s GPU technology to Intel, nor a licensing agreement, this is a custom GPU designed for Intel and supplied as die for Intel to integrate.
But that is not to understate that this is still a historic collaboration. It also points out advantages of AMD forming the Radeon Technology Group (RTG) and the semicustom group, which gives AMD the flexibility to do deals like this.
In addition, this deal could have been brokered by Apple. Apple wants a slim, but powerful MacBook. Intel graphics have never matched the performance or capability of AMD’s Radeon nor NVIDIA’s GeForce graphics. And as AMD has become Apple's preferred graphics supplier, putting Intel’s CPU and AMD’s GPU together in one package is a win for the Apple and consumers.
While AMD is about to launch its new Ryzen Mobile processor to compete with Intel’s mobile processors, this Intel multichip solution, with second generation high-bandwidth memory (HBM2) and discrete GPU die, will be more powerful. And it will not come cheap, but the Intel Core with Radeon Graphics processor module will offer premium graphics performance in a thinner laptop. And even if AMD's new Ryzen Mobile chip is wildly successful, Intel will still have the largest market share and this new module is an opportunity for AMD to address the ultra-premium slim notebook segment.
We still need to see more details of the system in package (SiP) module specification to know where it will be positioned. We don’t even know which graphics core is being used in the semicustom GPU-the new Vega graphics or last generation Polaris graphics. While Vega already supports HBM2, it could be that this design predates the shipment of Vega, so perhaps AMD adapted HBM2 to the older Polaris generation. Even if the Intel SiP is using the older Polaris generation, it will still be far superior to Intel’s latest Iris graphics.
In addition, AMD will be providing the graphics drivers for the Radeon GPU that Intel will integrate with its graphics driver. Intel may provide addition value with validation and stability.
In the announcement, Intel is touting its embedded multi-die interconnect bridge (EMIB) technology which is used to manufacture this module without the use of more expensive 2 ½ D silicon interposer packaging technology. Based on the diagram picture of the module, it appears that the EMIB is being used between the GPU and the HBM2 die. It would appear that the CPU and GPU are too far apart for EMIB to be useful and as a result, it appears that CPU and GPU die are connected using more traditional PCI express lanes routed in the substrate.
This may well be a one-off experiment by Intel to see if higher performing graphics in a thin and light notebook will sell. There’s no indication that this is an ongoing relationship between AMD and Intel beyond this single point project. Still, this is unprecedented in the history of the two companies since the acquisition of ATI by AMD back in 2006. And while many in the press have characterized this as AMD and Intel ganging up on NVIDIA, this is really more about AMD and Intel cooperating to build a better product. That said, it’s premature to say that this is a strategic relationship between Intel and AMD, and rather a partnership of convenience. We look forward to seeing the results of this experiment and to see if Intel and AMD can overcome decades of hostility to make this product successful.
And after the Intel-AMD deal was announced, another bombshell dropped. Shortly after resigning from AMD, Raja Kaduri, the former head of RTG, Intel announced that he is joining to lead a new group that will develop discrete graphics. This move was unexpected – Intel had never really taken graphics seriously. Integrated Intel graphics was always of the “good enough for business use” caliber. The company had a chance to buy 3Dfx in 1999 and passed on it; Intel tried to build a graphics chip called Larrabee using Intel technology that never made it to market. Intel’s internal graphics was never able to keep pace with AMD (ATI) or NVIDIA. Now, suddenly Intel has decided to try to build a new GPU from scratch. Well, good luck with that. We remain skeptical.
ARM Techcon - Not All Electronic Device Are Secure, But ARM's PSA May Change That
ARM Security is one of those never-ending challenges for the electronics industry.
As an example, security always ranks near the top of consumer concerns in surveys. However, few consumers ever ask how secure a product, such as a smartphone, smart TV, car, or any other product is before purchase. Consumers simply expect security to be included, and often don’t look for it or cannot tell when it is not included.
For the past several years, Arm, the leading provider of intellectual property (IP) cores for semiconductors, has been promoting the concept of building security into every device. While the company has offered many technical solutions for security, it is now offering a complete suite of hardware and software solutions for every device connected to a network. Arm calls their new security suite Platform Security Architecture (PSA).
The PSA framework includes the Armv8-M architecture with new firmware (Trusted Firmware-M), and two new semiconductor IP blocks – a separate security core called CryptoIsland-300 and a secure debug channel called SDC-600. Trusted Firmware-M initially supports only Cortex-M (MCU) devices and will be open sourced to allow for industry-wide collaboration.
CryptoIsland-300 is a completely separate IP core that sits on an SoC die, but has its own boot and runtime resources. Most solutions today rely on trusted execution environments (TEEs) that cordon off resources for security. The CryptoIsland takes this one step further by creating a separate security engine. Arm has yet not provided an estimate of the potential SoC die size impact. Given that CryptoIsland-300 is aimed at MCUs, the impact must be nominal or SoC costs will increase too much for target markets. Additionally, the SDC-600 provides a secure channel for debugging, closing another potential point of attack.
Qualcomm And Mercedes Race To Download Data From F1 Cars
Kevin Krewell and Paul Teich had a chance to meet with Qualcomm and the Mercedes-AMG Petronas Motorsport team at the Formula 1 (F1) United States Grand Prix in October, in Austin, Texas. We talked with Steve Pazol, VP & GM of Wireless Charging at Qualcomm and Geoff Willis, Director of Digital Transformation for Mercedes-AMG Petronas Motorsport.
Increasingly, high-tech companies like Qualcomm look to high-end motorsports to hone their research and development (R&D) edge. A race weekend may generate up to 500 gigabytes of data. How to get data from a car in a timely fashion during a race is a challenge, and it is an opportunity for Qualcomm to bring connectivity technology to the race program.
There are hundreds of sensors embedded throughout an F1 car. Perhaps a dozen of them continually assess the condition of each tire, monitoring pressure, temperature and vibration.
There are substantial trade-offs in mounting sensors in a car where weight and aerodynamics matter so much. Not only does each sensor add weight, but so do the power and signal wires that must be run to each sensor. This why Qualcomm and the race team are using wireless power connections to some sensors.
The top priority of the Qualcomm R&D relationship with Mercedes-AMG Petronas Motorsport is Wi-Fi, both in-car and car-to-pit. Car-to-pit Wi-Fi connectivity goes way beyond designing the next consumer or commercial Wi-Fi access point. Each car aggregates gigabytes of sensor data during a race, and the race team in the paddock needs this data in real-time to make adjustments to the car. As a car approaches its trackside command post and paddock, 5 GHz Wi-Fi engages between the car and pit crew.
Read the whole story on Forbes.
TIRIAS Research additional readings:
Qualcomm Centriq Aims At Intel Xeon With Competitive Performance And Low Power.
Qualcomm Takes Aim at Intel with ARM Server SoC
Imagination Technologies: Enabling Efficient Implementation Of Neural Networks In Mobile Devices.
Microsoft And Eastern Washington University Collaborate on Bachelor of Science in Data Analytics.
NXP Bridges the Gap In Processors.
Quantum Computing Will Not Break Your Encryption, Yet.
TIRIAS Research can be found in public!
You’ll also spot TIRIAS Research’s Principal Analysts attending these events over the next month:
- Supercomputer (SC) 17 (Denver, CO)
- HPE Discover (Madrid)
- RISC-V Workshop (Milpitas, CA)
- NIPS (Long Beach, CA)
- CloudNativeCon + KubeCon (Austin)
Say Hi when you see us!
As always, we encourage your feedback
Kevin Krewell, Jim McGregor, Paul Teich
TIRIAS Research is a high-tech research and advisory firm, an independent third-party resource to high-tech companies. We provide custom research and advisory services on technologies, markets and ecosystems to a select group of technology industry leaders. Our Principal Analysts have decades of in-depth expertise in silicon, software, and systems specification, design and deployment.