BRIAN SANTO: I’m Brian Santo, EE Times Editor in Chief, and you're listening to EETimes on Air. This is your Briefing for the week ending May 31st.
BRIAN SANTO：我是EE Times主编Brian Santo。你正在收听的是EETimes全球联播。本周热点新闻汇总如下：
* This week we’ve got a dispatch from Monte Carlo about the recent Grand Prix ELECTRIC vehicle race.
* 本周我们收到来自蒙特卡洛的报道：最新的Grand Prix电动车大赛消息。
* Also, a separate report on what’s going on with chips for Autonomous Vehicles.
* 另外还有一份关于自动驾驶汽车芯片的单独报告。
But first, Rick Merritt, our Silicon Valley bureau chief, picked up the advance program for Hot Chips, an annual three-day event that’s sort of the… Coachella for fans of microprocessor architectures. This year, the festivities will be in mid-August, on the Stanford campus. Rick has parsed the program, and he thinks the lineup signals a significant shift in microprocessor tastes. Here’s Rick:
但首先，我们驻硅谷记者站主任Rick Merritt来介绍Hot Chips先进项目会议，这是一项为期三天的活动，相当于微处理器架构师们的Coachella音乐节。今年的庆祝活动将于8月中旬在斯坦福大学校园举行。Rick深入剖析了这一项目活动，他认为今年的演讲阵容标志着微处理器研发的重大转变。有请Rick：
以下为英文原稿，Enjoy!
RICK MERRITT: The advance program for the Hot Chips annual conference came out recently, and it’s a real bellwether because it’s a an event put out for microprocessor architects by microprocessor architects.
This year, not too much of a surprise: about half the papers and talks will be about AI in some form. Among the more interesting ones, Cerebras, a much-watched startup, will talk about its chip, which is using wafer-scale integration. We'll see startup Habana will talk about its training chip. Intel will talk about two AI accelerators-- one for inference, one for training. And they will be a talk by UpMem, one of the emerging in-memory compute architectures that could be pretty interesting.
Separately, the tutorials this year are actually going to be amazingly interesting, because Google, Amazon and Microsoft are all going to talk about their AI hardware in the Cloud. It’s pretty unusual for these companies to talk about their AI hardware at all, and you almost never see it with them talking at somebody else's event, let alone together. So that in one day you get a view of all three. So I think that' going to be a well-attended tutorial.
In an interesting sidelight, Huawei is confirmed to speak about its Ascend AI accelerator, but they're still crossing their fingers that they'll be able to get visas and approvals to come, given the US/China trades wars and the way Huawei has become in the midst of all that. So we'll see.
PCs and servers still exist. There’s still a market there. And AMD is going to give a keynote. It's going to talk about its Zen 2 x86 core, and IBM is expected to talk about its Power 10 server processor.
But it’s a different day in the semiconductor industry. And you can also see that by the fact that Intel is doing four papers I think. And two of them are on AI accelerators, one of them's on Optane memory and a third one on its packaging technology. Nothing on x86 chips. It’s a different day in the semiconductor industry; it's a different in Silicon Valley.
And this is Rick Merritt reporting for EE Times from London.
BRIAN SANTO: Junko Yoshida, our Chief International Correspondent, recently interviewed David Fritz from Siemens. Fritz manages Siemens’ technology for autonomous vehicles (AVs) and advanced driver-assistance systems, which you’re going to hear being referred to as ADAS.
Junko called Fritz to discuss the changes in automotive manufacturing, but when she found out Fritz used to work for Qualcomm, she found the one guy who might be able to answer a question that always bugged her, which is: What’s more difficult to design? A processor for smartphones or an IC for Autonomous Vehicles? Listen in to their conversation.
JUNKO YOSHIDA: Many of us have spent many years covering smartphone apps processors. Well, we still do.
But now we encounter a slew of brand new, computationally intensive, big autonomous vehicle processors. Reportedly, they’re designed to function as the brains in next-generation cars.
Boy oh boy!
But I start to wonder: Which is tougher to design? A smartphone apps processor or an autonomous vehicle system on a chip? There’s no contest, according to companies like NXP or Renesas, given the stringent functional safety requirements their chips must meet.
The argument is valid. After all, autonomous vehicles could kill a pedestrian, but a few bugs in apps processors? Well, not so much.
But look closely. The processing demand for two big chips inside the systems-- one for smartphones and another for autonomous vehicles-- are pretty similar. Both processors take in a huge flow of sensory data from the world. They digest it and they act on it.
Where they diverge is the design and validation of these complex processors.
The smartphone industry has already gone several steps ahead. Companies practice rigorous pre-silicon validation, way before they start designing a system on a chip. In contrast, the automotive industry is still trapped in an arduous cycle of going back and re-spinning the chip to get it right after the chip came out.
What’s up with that?
I sat down with David Fritz, global technology manager for Autonomous and ADAS systems at Siemens, and asked him: Explain the differences in the two worlds.
Hi, David. Thanks for coming to the show. I understand that you used to work at Qualcomm before joining Mentor/Siemens. What initially surprised you about the way automotive chip designers develop their complex autonomous vehicle SoCs? Walk us through their process and design challenges.
DAVID FRITZ: Let me start by saying that I really can't speak to how any one company in particular develops their autonomous SoCs. Only in general terms.
It's important to understand that in the primarily electro-mechanical or mechatronic world of automotive, methodologies geared towards addressing this class of solution, model-based system engineering for example (MBSE), have been pretty effective. Continuous integration processes-- including model in the loop, software in the loop, hardware in the loop-- are commonly used. The issues arise when you look into shortening design cycles to gain a shift-left benefit while also trying to managing the complexity of consolidated ADAS or complex AV chips. Here, the Mill-Sill-Hill models start to break down and require more sophisticated approaches.
JUNKO YOSHIDA: In contrast, though, how do smartphone application processor designers develop their own SoCs?
DAVID FRITZ: Well, Junko, when you take a close look at how smartphone chipsets are developed today, those that are required to meet stringent power and performance and scheduling constraints while simultaneously running a full software stack-- all of this pre-silicon, mind you-- it's clear that automotive ADAS and AV SoCs could be designed and tested in very similar ways. The result would bring significant value to automotive chip makers and their customers.
The concept of delivering use-case proof points and models to customers in advance of silicon is well understood and common practice for smartphone chipset makers. But the idea seems to be foreign to many automotive suppliers.
JUNKO YOSHIDA: I see. In your opinion, David, what’s missing from today’s autonomous vehicle SoC design process?
DAVID FRITZ: Well, one important point to make here is that the higher level of autonomy that's reached, the more interdependent the systems of the vehicle become. Correct decisions in an AV are highly dependent upon accurate sensing data and vehicle dynamics, as well as AI inferencing and awareness of environmental conditions.
When an SoC is developed and tested independently from the rest of the vehicle, subtle issues arise that can cause accidents. This type of gap issue can be very difficult to diagnose, and AV SoC designs can benefit from the pre-silicon validation methodologies pioneered by their smartphone colleagues.
JUNKO YOSHIDA: Thank you, David. That was very interesting.
This was Junko Yoshida, EE Times, with David Fritz at Siemens.
BRIAN SANTO: London correspondent Sally Ward-Foxton gets all the tough assignments. Recently, she was dispatched to Monte Carlo to attend this year’s E-Prix rally. Given a chance to peek under the hood of one of the electric case, Sally--whose background is in electronic engineering-- just couldn’t resist.
Here’s her report.
SALLY WARD-FOXTON: AUDIO: I had the opportunity a couple of weeks ago to visit Monaco to report on the Formula E electric vehicle championship. Like Formula 1, Formula E is essentially a technology competition, and it basically showcases the latest innovations in powertrain components and software for electric vehicles that will eventually end up in all electric vehicles.
I visited the garage for the Venturi team the day before the race. Venturi are pretty well known in the electric vehicle industry. One of their most famous projects has been to set the land speed record for electric vehicles a couple of years ago. They’ve had a Formula E team for several years.
Now, new to Formula E this season is an entirely new car. About 80% of a formula E car is standardized-- like the aerodynamics, the wheels, the tires-- but the teams can work on the powertrain, so the inverter, the motor, the regenerative braking system, things like that. The second generation car has doubled the size of its battery to 54kWh, so not as big as the 100kWh battery in a Tesla Model S, but crucially, it means that one battery charge can last the whole 45-minute race, so no more changing cars like in previous seasons. There’s no pit stop at all now.
Of course, the race is all about squeezing as much as possible out of one full battery charge. Any efficiency increases from the components are obviously really important, but I was a little surprised to learn that software plays a huge role as well. Venturi’s team principal, Susie Wolff, told me: “the big race in formula E is the software race.” She said teams can update the software quickly in between races to try and finesse aspects of performance, which is especially important for the new brake-by-wire system in the new car, which automatically balances between torque from the mechanical brakes and torque from the regenerative braking system. This was previously done manually by the drivers, and they frequently lost control of the cars as a result.
One of the things that’s unusual about team Venturi is, it’s relatively diverse. There are seven women on the team of 30, working at almost every level, from team principal right down to the mechanics. This kind of ratio is practically unheard of in motorsport.
I was shocked, but unfortunately not surprised, to hear Susie Wolff describe her first press conference when she joined Venturi as team principal. She said the first question she was asked was: What makes you think you can do this job? Remember, she’s had a long, distinguished career as a professional racing driver. She’s hardly inexperienced! Then the second question: Did your husband place you in this job? Her husband is the team principal for Mercedes Formula 1. She was shocked by that question, and rightly so! And the third question was: How are you going to do this job as a mother? She said that was like a slap in the face.
It’s hard to believe women are still being asked these kinds of questions in 2018, or 2019, but the world of motorsport is still very male-dominated at every level. Susie Wolff has set up her own initiative to encourage girls to think about careers in motorsport, whether that’s as drivers, engineers or mechanics, but it’s clear that there’s still a lot more work to do.
On race day, Venturi’s lead driver, Felipe Massa, started in fourth position but managed to overtake due to a small mistake from the driver in front of him. Massa was able to successfully hold him off for the rest of the race, to finish in a well-deserved third place.
This is Sally Ward-Foxton reporting from Monte Carlo for EETimes.
BRIAN SANTO: Next up, with the US-China trade war escalating, it’s time for us to hear the Chinese side of the story. Junko Yoshida got on the horn with one of our colleagues at EE Times China. Here’s Junko.
JUNKO YOSHIDA: It looks like neither the United States nor China is prepared to end the trade war any time soon.
Over here in the United States, we’ve done extensive reporting on what Americans think of this clash. We talked to economists, bureaucrats and representatives of industry associations to figure out what’s really going on.
What we haven’t done, though, is reach out to our colleagues in EE Times China. Shame on us. It’s time for all of us to step back and get curious about how the Chinese perceive this so-called trade war.
During any crisis, each side in a conflict tends to be fixated on defending its own position. Each side, feeling pressed, believes there’s little time, or no good reason, to walk in the other guy’s moccasins. Which poses the question of whether the cause of this whole mess might be a glaring perception gap.
Earlier this week, I got on the phone with Echo Zhao, senior analyst at EE Times China based in Shenzhen. Here’s the broad-brush outlook our colleague Echo offered us.
We've been spending a lot of ink, both on EETimes US and EETimes China, on the ongoing US-China trade war. You know, Echo, I thought it would be prudent of us to ask how you guys are perceiving this whole conflict. To understand the problem, I thought the first step is for us to learn how you see the unfolding situation right now.
So let me ask you this first: What are you hearing from China’s social media like WeChat, Wavo and so forth? I'm sure a lot of people in China are talking about his. What’s the mood of the country like, and what are the people saying?
ECHO ZHAO: There are some we media that inciting national sentiment, call that buy a Huawei phone is “paying patriotic tax.” But I feel that the whole electronics industry, including Ren Zhengfei and Huawei, are still very calm and pragmatic. In fact, Ren thanked all US suppliers publicly for their support in the past. And said he hope that compete in a fair environment and we can't say that using Huawei products is patriotic.
更多...
——听全球专业科技资讯，只在EETimes
用户评论