DANI RABIN: This is Dani Rabin from Marbin, and you're listening to EETimes On Air.
DANI RABIN：这是来自Marbin的Dani Rabin，你正在收听的是EETimes On Air。
DAVID FINCH: This is your EETimes Weekly Briefing. Today is Friday, April 5th.
Among the top stories this week: Intel’s new memory architecture; an exclusive interview with Greg Travis – a veteran software engineer and instrument-rated pilot. We ask him if Boeing 737 Max MCAS software can be fixed. We also look at STMicroelectronics’ strategic priority on silicon carbide; and how the risk-averse business mentality is making Japan’s car OEMs slow to embrace the global auto market’s shifting focus on mobility services.
本周头条新闻故事包括：英特尔的新内存架构; 对Greg Travis的独家采访--一位经验丰富的软件工程师和经过专业训练的飞行员。我们问他波音737 Max MCAS软件是否可以修复;我们还要看看意法半导体在碳化硅方面的战略重点;以及规避风险的商业心态如何使日本的汽车制造商在全球汽车市场向移动服务的转变中步履缓慢。
JUNKO YOSHIDA: Japanese corporations are willing to flock together into a big group, but not because they are willing to work hard together. They do so because they don't want to take big risks.
DAVID FINCH: All that to come. But first, we begin with a story from our EETimes Silicon Valley Bureau.
Junko Yoshida, global editor-in-chief of AspenCore Media, asks Silicon Valley Bureau Chief Rick Merritt about his recent visit to Hillsboro, Oregon for Intel’s briefings.
JUNKO YOSHIDA: Hi, Rick. So you posted two stories this week about Intel, which I assume came out of your recent visit to Hillsboro, Oregon. Give us the rundown of new stuff they unveiled.
RICK MERRITT: Hi, Junko. So what Intel is doing today, or this week, is rolling out a whole slue of its data center chips. 40+ new Xeons, its first 10 nanometer FPGA, which I think is the most interesting part of the deal. And Optane memories on DIMMs for the first time. This is all chips that go into Intel's strongest and most profitable growth market. So it's really important for the company.
JUNKO YOSHIDA: What does Intel’s new memory architecture entail? Why does the industry need it, and what are the challenges?
RICK MERRITT: The memory part of this announcement was really interesting, because we've been following this Optane-- used to be called 3D Xpoint memory for going on five years now, and there's been a lot of interest in it because Intel billed is as the first new memory architecture next to DRAM and NAND in decades. Unfortunately, as it turns out, you gotta do a lot of work to bring that into a DIMM, and Intel did that work, including a lot of the software work. But it's still going to only be available on select applications where the software is in place. And frankly, the performance benefits weren't overwhelming to anyone. They were interesting but relatively modest in the sort of range of 30, 40% kind of benefits, although it's a complex performance story because it varies a lot with every different kind of application. And there's maybe a half a dozen that Intel is still kind of sorting through right now.
And unfortunately, the chips come at a time when the DRAM market is tanking. The prices are going down, so the price difference that Intel hoped to make a difference with here with Optane memory just isn't going to be there for them.
So one of the analysts I talked to said, back in 2015 when this was first announced, he projected that if all goes really well for Intel, they could make $2 billion in the first two years of rolling this stuff out. Optane DIMMS. Now he says, given the changing dynamics and the new performance metrics, he says it's probably more like 200 million over the next two years. So that's a big difference.
JUNKO YOSHIDA: You laid out in your stories. Rick, that some advances in Intel’s new CPUs, FPGAs and memory modules have come at a cost of Intel proprietary lock-ins. What concerns are you hearing from the engineering community on this?
RICK MERRITT: So, strategically here, Intel is trying to do a really interesting thing that I think a lot of other companies will do too, which is that while Moore's Law is slowing and so the performance gains in any one chip aren't as great as they had been in the past, they're trying to tie many chips together. In this case, Xeon processors and FPGAs over a cache-coherent processor bus with Optane DIMMS using Intel's proprietary DDRT protocol. And then together they will all give bigger boosts than any one chip. It's sort of the dogsled metaphor is the way I think about this. And it's smart. It's what companies are going to increasingly do. But Intel is adding this kicker of having proprietary interconnects between the Xeon, the FPGA, and the Optane.
I talked to two top data center engineers and a couple analysts, and they both said they really think this is going to be a non-starter. Companies don't want to get locked in. It's bad enough for most of them that they depend upon Intel for their processors. They don't want to be locked in and dependent upon some of their memories as well.
So strategically it makes sense. Practically, we don't think it's going to go over. And that was kind of the story that we came away from in Hillsboro.
DAVID FINCH: That was Junko Yoshida and Rick Merritt reporting.
And now, Science and Technology writer George Leopold, who recently wrote an in-depth analysis entitled “Software Won’t Fix Boeing’s ‘Faulty’ Airframe,” sat down with Greg Travis for an exclusive interview with EETimes On Air. Mr. Travis, who was quoted in George’s story, recently posted a damning critique of the 737 MAX fiasco. As Boeing hurriedly updates its MCAS software, Mr. Travis explains if the updates can save the Boeing 737 Max.
GREG TRAVIS: The causative frame that's being promoted by the industry, the media, the FAA, etc. is that there was a problem. And what do you do with problems? You fix them. But let's step back from the problem mindset for a bit. I rest my piece, I argue not so much that MCAS couldn't be fixed, I argue that it SHOULDN'T be fixed. Specifically, what I wanted to convey was the concept of normal failure. That as systems become more complicated, failure of those systems becomes a normal aspect of their nature.
In his book, "Normal Accidents," Charles Perrow points out that a characteristic of normal failure is that the interactions between components of a complex system become incomprehensible. Thus we cannot predict in advance how all of the pieces of a system will couple, how a failure in one area can cascade into other areas, how failsafes nonetheless fail. We don't know just how bad the pitch-up problem in the 737 Max is. But I'm hoping that there is something short of MCAS that can be brought to address it.
Again, I don't want to see MCAS fixed. I want to see it gone.
DAVID FINCH: Mr. Travis also argues that the 737 Max airframe design is fundamentally flawed. But is it too late to fix?
GREG TRAVIS: I think to address that, we need to look at the 737's history. Specifically, how it has evolved over the years. And almost all the story there is the story of the 737's engines.
For access reasons, the 737 was designed to sit low to the ground. That was okay when it had its original low-bypass JT8 engines. But then the A320 came along, twenty years later, and the A320 started life with the CFM56 engine. Boeing had to follow suit and squeeze a smaller version of the CFM56 into the classic line of 737s. And when the A320 got the LEAP engine, Boeing once again scrambled to make it fit.
That hasn’t gone so well. I think this is nature’s way of telling us that the 737 has run out of steam. They tried to kludge the LEAP engine in, using MCAS as the bandaid, with disastrous results.
DAVID FINCH: And finally, as George Leopold points out, the Boeing 737 Max saga raises many engineering ethics issues. Among them, Mr. Travis has cited “cultural laziness” within the software community. Here, Mr. Travis offers his perspective on we got to this point and the lessons we are all left to learn.
GREG TRAVIS: Something unspeakably terrible happened with regard to the development of MCAS. We had a flight-critical system that could configure the aircraft, that had overwhelming control authority, that took as its only input a single angle of attack sensor. You never do that in aviation, despite what Boeing is saying. And the reason it did so was because if it took inputs from more than one sensor, then there would need to be some kind of disagreement detection and failsafe. All of which would require pilot training, which Boeing took off the table.
As the saying goes, a man with one watch always knows what time it is, and a man with two watches is never sure.
For money reasons, Boeing wanted the surety that only one sensor can give, even if it’s deadly wrong. What I don’t understand is how it happened. How could the company leadership imperil the company’s reputation so recklessly? How could the hundreds of engineers look the other way when confronted with something so wrong? This is the organizational story, this is the bureaucratic story, this is the political and engineering story of a lifetime.
DAVID FINCH: We turn now to Nitin Dahad, EETimes' London Bureau Chief Correspondent, who recently visited Catania, Italy, to explore ST Microelectronics’ big plans for Silicon Carbide. Nitin also spoke with ST’s CEO Jean-Marc Chery. Here’s Junko asking Nitin about ST’s silicon carbide strategy.
JUNKO YOSHIDA: Hi, Nitin. You visited STMicroelectronics in Catania, Italy, last week. What was the big news?
NITIN DAHAD: Thanks, Junko. The context is that the big industry driver (excuse the pun) in electronics right now seems to be electric and hybrid vehicles, which is all part of the holy grail of autonomous vehicles that everyone is talking about.
And for that energy and power management, electronics will be key, and ST was keen to emphasize its core competence in this area, particularly with its silicon carbide plant in Catania in Italy.
If you’ve followed ST’s financial earnings over the last year, you’ll have heard how CEO Jean-Marc Chery consistently tells the industry that silicon carbide is a strategic priority and ST wants to capture 30% of a 3.7 billion US dollar silicon carbide market by 2025.
So, ST was essentially saying: look at where we are already in silicon carbide, why this technology is important for the future, and how we’re going to get there.
JUNKO YOSHIDA: So what are the challenges, and how is ST addressing these?
NITIN DAHAD: Good question. I actually put that question to Jean-Marc. This is what he said.
JEAN-MARC CHERY: The short-term challenge is definitely the supply chain. So both the material maker and the device maker/seller, all these guys, they have to align the supply chain in terms of volume versus the fast adoption of let's say the electrical car. Also, sooner or later you will have a strong push to demonstrate that the electrical cars are power efficient. So silicon carbide, okay, will have to go, because the only way to be very power efficient with an electrical car is to use much better silicon carbide. This is our first challenge.
Then the second challenge would be as usual for new devices, we have to decrease the cost. So we would have to (?) the device, we would have to increase the (?) size, and we would have to decrease the cost of material, and we would have to optimize the design of the (?).
So first challenge as a takeaway is the supply chain volumes. Second challenge is business as usual new device, is cost reduction.
NITIN DAHAD: As he says, the big challenges are supply chain and device cost. Just as an example, in electric vehicles, silicon carbide might add an additional upfront cost of $300. But overall, there could be a $2,000 saving as a result of savings in battery cost, EV space and cooling. On the supply chain and ecosystem, ST has been making moves to address this, first with its $250 million silicon carbide wafer supply agreement with Cree, and then with its acquisition of a 55% stake in Norstel, which it ultimately intends to fully acquire.
And we can expect more, as ST executives suggested they want to control more of the supply chain. So I would expect more news on this front coming as a result of the Norstel acquisition.
The company is also developing its third-generation planar technology for silicon carbide MOSFETs, as well as developing the next generation in parallel using trench technology. On top of this, it plans to move from 6-inch wafers to 8-inch wafers by 2025.
JUNKO YOSHIDA: OK. I understand that’s ST’s pitch, but what’s your take on all of this?
NITIN DAHAD: While the big pitch was on silicon carbide, ST’s broader offer to the power electronics needs of industry features gallium nitride too, for which ST also announced it is working with Macom, and we reported on it earlier this year.
JUNKO YOSHIDA: Wait, Nitin. Does ST plan to bring both SiC and GaN to electric vehicles?
NITIN DAHAD: Actually, both silicon carbide and gallium nitride are part of ST’s power strategy, the former targeting electric vehicles due to its higher voltage operation, and the latter targeting RF and especially 5G due to its higher operating frequency. But ST says future developments of GaN products will target automotive, particularly on-board chargers for EV and mild-hybrid powertrains.
Coming back to the big issue in silicon carbide, it is indeed the availability of raw materials. So ST’s moves to address this with Norstel, in particular, will be one to keep a close watch on, especially as ST said it wants to be more vertically integrated.
My personal take is that, while the electronics industry, in general, is betting big on EVs, which is what is driving the demand for silicon carbide, I don’t yet see that translated into consumer uptake of EVs, particularly as many countries still lack essentials like charging infrastructure.
So while the electronics industry is busy readying itself for that consumer-led demand and sorting out supply-side issues, the vehicle manufacturers themselves still need to do a lot of work in enabling more consumer demand for EVs, both in terms of cost and infrastructure.
DAVID FINCH: That was Nitin Dahad and Junko Yoshida reporting from Catania.
Junko is also in Tokyo this week and shares her perspective on Japan’s automotive industry. She draws a parallel between Japan’s automotive industry and the Japanese consumer electronics industry whose complacency led them to go into a downward spiral. Here's Junko with more.
JUNKO YOSHIDA: I’m worried about the future of Japan’s automotive industry. I have this bad feeling that Japanese automakers might be following the same course that led the Japanese consumer electronics industry into a downward spiral.
Usually, two factors contribute to the decline of successful corporations: arrogance and complacency. The combination is fatal, especially when revolutionary changes are afoot in the industry and in the market.
Take Sony, Panasonic, and Sharp. Although they knew applications-intensive digital consumer devices were about to alter the very nature of the CE landscape, they chose to stubbornly stick to what they knew: hardware design. They convinced themselves that so long as they developed high-quality, high-resolution, high-fidelity audio and video products, they could somehow weather the building storm.
Similarly, Japanese automakers today are sailing into troubled waters. On one hand, their vehicle sales are due for a stall, as people worldwide opt for pay-per-use services over car ownership. Meanwhile, new entrants to the industry – such as Google’s Waymo and startups like Uber and Lyft – are shifting the automotive business from building cars to selling transportation services.
So, what do Japanese car OEMs do? They’re keeping their heads down, waiting for a low-risk, high-return option to emerge.
Against this background, Honda and Japan’s truck maker Hino Motors announced last week that they’re investing in a venture called Monet Technologies, founded by Toyota and SoftBank to develop self-driving car services in Japan.
Monet is advocating commercial mobility business based on a platform called e-Palette. Proposed applications range from on-demand stores to distribution and food trucks, as well as on-demand shuttle buses and mobile emergency rooms.
Such commercial mobility services might sound unique. But here’s the $64 billion question: Can Monet Technologies save Japanese automakers? Car OEMs in Japan, generally speaking, have been late to the autonomous driving party and slow to build their own Mobility as a Service business.
I have doubts. Here’s why. First, Monet Technologies is bragging about the formation of a big consortium of 88 Japanese companies. On the list are companies such as Coca-Cola Bottlers Japan, beverage maker Suntory, Yamato Holdings (a Japanese delivery company) and Yahoo Japan.
Here’s what everybody should know: Consortiums in Japan aren’t “ecosystems.” Japanese corporations are willing to flock together into a big group, but not because they are willing to work hard together. They do so because they don’t want to take big risks all by themselves.
Consortiums often founded in Japan do not necessarily guarantee concerted success.
Second, Monet Technologies’ CEO Junichi Miyakawa made a bold promise in an interview with one of the Japanese newspapers by claiming, “We won’t let foreign corporations take over Japan’s Mobility as a Service market. We won’t repeat the same mistake we made when Google, Facebook, and Amazon took over our Internet market.”
To me, this was a clear red flag. Never trust a Japanese executive who resorts to flag-waving. Success in business has nothing to do with allegiance to the domestic market. In this day and age, no company succeeds without a global vision.
This is Junko Yoshida, EE Times, reporting from Tokyo.
DAVID FINCH: This has been your weekly briefing from EETimes and the AspenCore Global Service. You can read all of these stories and more at EETimes.com. Thanks for listening.