The demand for data capture at the edge is increasing faster than engineers can develop solutions to support it. Rather than developing solutions from scratch that enable the capture, analysis, and transmission of data from edge to cloud and back, Intel® Partner Alliance members like Advantech, a leader in embedded and automation solutions, are supplying building block single-board computers, computer-on-modules (COMs), and plug-and-play systems complete with the user’s operating system of choice—all based on Intel embedded processors.
For vision applications, many of the Intel chipsets at the heart of these solutions include integrated Xe-graphics technology for vision processing at the edge. But for use cases that require leading edge-graphics performance, Advantech has also designed discrete PCIe graphics cards and MXM modules around the new Intel® Arc™ GPU chipsets. As Thomas Kaminski, Director of Product Sales Management, Marketing, and Technical Support at Advantech, explains at embedded world 2023, these technologies can combine seamlessly with modular computing standards like COM-HPC to bring more intelligence to the edge than ever before.
Thomas Kaminski: My name is Thomas Kaminski. I’m the Director for Product Sales Management, Marketing, and Technical Support in Advantech Europe.
So, the demand for data capturing at the edge is getting bigger and bigger to a higher demand and requirements from different applications. So, first of all, you have to detect and collect a lot of information at the edge—about objects, about persons, about interactions, and a lot of other customer demands. This is driving the processing power at the edge, and at the same time the vision power at the edge. So you need to get the picture inside of the processing unit. You need to analyze the picture. You need to get the important data out of the picture and transmit them finally from the edge into the cloud to the server.
The challenge for the customer is, first of all, they need to develop their own platform from scratch. So what we can offer first is—like we are showing here a lot of different platforms. We can use a simple, single-board computer where it can say, you need to include it into your application, into your chassis, into your housing, and build the final device out of it. That the first one.
The second option is if the customer wants to be more flexible and have a scalable solution. The other customers are going with the computer or module solution so they can build their own carrier board with specific interfaces, with specific functionalities—depends on the different applications.
You can have different applications in a medical device, others adding special requirements on the connectivity, or whatever it is. Or IO, simple IOs to collect the necessary data in the different applications. If the customer doesn’t want to build the final device by himself, we can offer same solutions already integrated into the chassis, simply saying a plug-and-play device with an applied operating system. It could be Windows, it could be Ubuntu, or we can support a customer with Yocto Linux to build his own BSP and application in the end. So this is where we can support customers for applications in the edge to develop the final application. Could be hardware driven, could be only software driven with finished devices. Since Intel has now launched the Xe graphics inside of the CPU, especially the embedded SKUs themselves were getting much more stronger in terms of the GPU performance than in the past. So I would say, for a lot of applications this performance is good enough without an additional GPU card to perform at the edge in the different applications.
Talking about, for example, medical applications or other applications that are requiring more performance, GPU performance at the edge, Intel just released a new graph discrete GPU chipset, the Arc Series, where we have just launched or are here launching on the embedded. So, an additional GPU card—it is a PCI Express card, or in optional, we can offer the same one as an MXM board. And you will find, again, MXM slots on our motherboards. You will find the PCIe 4.0 slots on the boards. Together with that combination, I think you can cover most of the market requirements from the performance side.
I think the COM-HPC standard, that is an interesting standard because this is developed to collect a lot of data at the edge. So the COM models themself, talking about single boards, talking about also the traditional computer modules, are okay for standard applications. But now looking on the additional demand of data capturing at the edge, you need to process a lot more data. And to deal with such kind of high data volume, the COM-HPC standard, which is shown over here, was developed on the server or client side, because compared with the current computer modules, you can deal with even higher power. You can deal with much, much more data transmission and more interfaces, which is in the end necessary to get the data in, to process the data at the edge and transmit this necessary data back to the cloud, to the server.
So this is our computer, our COM-HPC modules which we are offering to the market. I think we have several options to get more information on several products. First of all, of course, is our webpage. You can find all available and the latest upcoming products on that website with the first data sheet. But I think because of the complexity of all the different applications, I think it’s always good to talk to our local product-sales managers. They have a lot of knowledge about the different products, know-how and technical requirements and possibilities also in combination with operating systems, with software services. So they can help you out to find the right and best-fitting solution in your application. So I would suggest both options. First of all, general check on the website. But then start to talk to our product-sales managers, and get their knowledge to select really the fine and right platform.