Could Low-Cost, Low-Power FPGAs Benefit Your Application?

Could Low-Cost, Low-Power FPGAs Benefit Your Application?: Audio automatically transcribed by Sonix

Could Low-Cost, Low-Power FPGAs Benefit Your Application?: this mp4 audio file was automatically transcribed by Sonix with the best speech-to-text algorithms. This transcript may contain errors.

Ray Hoare:
Ray Hoare here with Concurrent EDA. Quick take on what's new in FPGAs. We are an AMD, which bought Xilinx a number of years [ago], embedded partner, and we are "elite certified." So we're here to help you.

Ray Hoare:
So what's new? In the cost optimized portfolio, which is not their high-end networking data center chips, but in the embedded space where you're going to put this inside of your product or put it on a system, on a module. We have Zynq, which has ARM core processors in it —— we talked about that previously —— and logic. So, it's hardware and software. You have the Artix, which has lots of compute transceivers, lots of bandwidth. And then Spartan. Spartan has been around forever. And it, you know, put logic in there and you've got a lot of I/O. It started with the Spartan-3 and then moved up to Spartan-6 and the 7, and now we're into UltraScale. And UltraScale and UltraScale+ is the 16 nanometer, very high density chip. So this is something that's the same fabric that's in your UltraScale, your Zynq UltraScale+ or your Artix of the same exact fabric, but a little different ratio of I/O to the fabric. So I want to talk about that today. And the good thing here is they're going to be in production through 2045 —— and, I think 2045... Where will I be? Oh wow. That's a long time from now. It's 20 years from now, they'll still in be production. Still be IN production, pardon me —— and I might not be talking so well then, so who knows? But anyway... So where do we see these? Why do we care? If you think about a PC and you think about the system on a chip, you've got your processor plus your logic. That's what the Zynq is. And now on your motherboard, you've got this Southbridge, and it's got all this I/O that you hook everything to. That's where the Spartan is going to come in —— and maybe you use a Spartan with your Zynq. Or maybe you just use the Spartan by itself. Let's see what Spartan has to offer.

Ray Hoare:
In very small packaging —— let's just talk about what's in production now, because that's what you can actually apply to your product. Let's talk about this. This is in a 10x10 millimeter package. And that's very abstract to me. So here's a little inch. And so it's less than a half-inch by a half-inch. So we're talking a little, teeny, teeny, little chip with a lot of compute for that little bit of I/O. So you can look at —— we have 10,000 logic cells up to 36,000 logic cells. So that it's not a lot compared to up here on the high end, which is hundreds, and you get even bigger into millions. So from a compute scale, it's pretty small compute, but in that really, really tiny package. So you can do a lot of that I/O processing in there. And then let's look at what we have to offer in there on the small end. And we'll look at the data sheet. Sometimes it's a little confusing, but let's go through it.

Ray Hoare:
All right. I know, eye chart, sorry about that. If we see ... I highlighted in blue here, like the 10x10 millimeter package, which is really, really small. And let's ... The major benefits here are, it's got a lot of memory inside of it, and it's got lots of I/O. And why do you need I/O? Well, you're going to talk to UR, you're going to talk to SPI. You're going to talk to Ethernet. You're going to talk to cameras or displays or you're going to control LEDs or receive things in that are digital, jumpers, etc. This is where all of that I/O comes into play, and we have different kinds of I/O that I'll talk about, but we have single-ended, which is, they call this high density, and then high performance. So we have 252 pins of high density I/O in something that's the size of your fingernail. So we can put that on your board. It works. Got your I/O. You attach it to your Zynq or your CPU or your embedded micro, whatever you need to hook it to. You can do that. If we take a jump up —— when I say jump up, I mean go from 10 millimeters all the way up to 12, almost a half-an-inch. We can actually now add in something called XP5IO. And that's this third column here in this chart. And when you see that ... So when you look at this chart, it's got four positions: high density, high performance, and then this extreme performance, this XPIO, and then your transceivers —— your GTH, your gigabit transceivers. So we're going to look at these XP5IO. In the XP5IO you... When you go to 12x12 millimeter —— not in production yet but it's coming —— you get 132IO, in addition to your high density and your high performance. And so you get all sorts of, a mix of I/O. And then you also get PCI Express in some of the parts. And that's where your GTHs come into play. So Gen 4x4. Which means you can hook it to a real motherboard and get phenomenal performance. The memory also goes up. So now we're up into 5 megabits of memory on this little, teeny little thumbnail of a chip. All right. Let's dig deep into like, what does that mean? What are these three different kinds of I/O that you can use? And why would you want to care? Why would you want to use them?

Ray Hoare:
So here I just took this one example. Pardon the poor resolution. I wanted to zoom in and show you, on this particular part ——.

Ray Hoare:
And that's this one right here. So if you go to these two different pin packages, this, the bottom one is a 19x19. Okay. So now we're up to almost three-fourths of an inch by three-fourths of an inch, a monster part —— no, it's really, really small. So you can get more I/O in there.

Ray Hoare:
So, in that case, you got more high density, as opposed to this top one, you have transceivers. So on the high density side, you can go all the way up to 3.3 volts. If you need to go to 5 volts, you're going to need level converters. Once you get down to the 16 nanometers, you're not doing 5 volts. So you need a level converter on those, which are ... They're easy to put on. As you see on the high density, you don't get some ... You don't get the 5 supports or the memories or camera supports. You know, they don't have termination built into it. So you got to pay attention to that yourself, and we can help with that. But then when we go to the high performance, we don't go to 3.3, we drop down to 1.8 volts. But, the things that you care about are all running at 1.8 volts that are higher speed, because the more voltage, the more ... The longer it takes to transition. And so higher performance things don't go as high on the voltage. So you have to worry about this. So you have input serdes and output serdes, which are serializers. And then you also have memory like DDR4 and then a DFI.

Ray Hoare:
Now, what's a DFI? A DFI is like... A MIPI camera uses a DFI or a DisplayPort uses a DFI. That's the physical interface specification. And then we have differential —— LVDS is what people are familiar with. Interfaces, where you're sending two pairs of wires to send a data signal. And what that means is when you send two wires, if you have some electromagnetic interference, they both go, both of those signals go up and down. And that's okay, because you're looking at the difference between the two of them. And that enables you to go really, really fast. And all your high-speed peripherals use LVDS or some sort of differential. Now when we go to the Extreme Performance 5, the XP5 I/O, we drop down a little bit in the voltage to 1.5. But now we bring in support for LPDDR5. So that's crazy fast. And then you have also D5. And these D5s go up to 3.2 gigabits per second. So the reason we care, is on a camera —— if you have a MIPI camera —— well, you've got, you know, usually two lanes of MIPI or four lanes of MIPI. Well, if you can only go a gigahertz off of four lanes, you get your 4 gigabits. But if I go to 3 gigabits per second, I get 12 gigabits over four lanes. And so I get much, much higher throughput. So your higher resolution cameras need that performance. Okay.

Ray Hoare:
So that's just a quick summary of what the Spartan UltraScale+ can do for you. You can use that with your Zynq UltraScale MPSoC. Right? Which has your ARM processors. You can use that on a different board. You're talking to an embedded micro that you need to put some I/O on there. So... Good part. Really exciting. This has been a long time coming. It's now in production. At least the first four devices are. If you need help with part selection, we're here to help. We've been doing this for 19 years. If you have a design and then it's in a Zynq 7000 or an older Spartan, we can help you with that migration. If you have an algorithm that you want to say, "hey, I've got this algorithm, I need to put it in this little part. How do we do that?" We can do that for you. We take Matlab, we take Python, and we take you through a rigorous process that we've been developing for almost 20 years and getting it down into the VHDL and Verilog. We also do VHDL Verilog interfacing, SPI, Square, UArts, you name it, we've probably seen it. And then High-Level Synthesis —— that's where we can take C-code that's been optimized and get it down to gates and do verification along the way. If you need any help, send us an email. I'm Ray, This email address is being protected from spambots. You need JavaScript enabled to view it.. Thank you very much. Have a great day.

Sonix is the world’s most advanced automated transcription, translation, and subtitling platform. Fast, accurate, and affordable.

Automatically convert your mp4 files to text (txt file), Microsoft Word (docx file), and SubRip Subtitle (srt file) in minutes.

Sonix has many features that you'd love including enterprise-grade admin tools, collaboration tools, advanced search, transcribe multiple languages, and easily transcribe your Zoom meetings. Try Sonix for free today.