BeagleV Catches Fire With The BeagleV-Fire

A new BeagleBoard is on the way, full of FPGA hotness: the BeagleV-Fire has been announced. The new $150 Single-Board Computer (SBC) from the pioneering open source BeagleBoard company is built around a RISC-V chip that has FPGA features built in. The BeagleV-Fire is built around the snappily named Microchip PolarFire MPFS025T FCVG484E, a System on a Chip (SoC) that has five Reduced Instruction Set Coding Version 5 (RISC-V) cores and a big chunk of FPGA fabric built in. That means it combines the speed of RISC-V processors with the flexibility of Field Programmable Gate Arrays (FPGA), a big pile of logic gates that can be reprogrammed.

The new BeagleV-Fire includes a sizeable chunk of FPGA to work with: the core chip includes 23 K logic elements and 68 Math blocks, plus 4 Serializer/Deserializer (SerDer) lanes that can throw about 12.7 Gbps of data into and out of the fabric. On the BeagleV-Fire, the main chip is supported by 16 GB of eMMC and 2 GB of LPDDR4 RAM, plus a micro SD slot for extra storage. Gigabit Ethernet is also included, plus USB-C power and a few serial connections for debugging. There is no WiFi built in, but there is an M.2 Key E connection were you could plug in an a wireless adapter if you need it.

Like most other BeagleBoards, the BeagleV-Fire has two headers with 92 pins, which offer access to pretty much every signal on the board, plus lots of analog to digital stuff that works with add-on boards (BeagleBoard refers to them as capes). Also present is the usual 22-pin CSI connector for attaching cameras and other devices.

Want one? They are available for immediate order on BeagleBoard.org or from the usual suspects. It looks like they are already in stock for next-day delivery. If this all sounds familiar, it’s probably because we’ve been posting about this particular board for awhile now, covering both the announcement and first tests.

34 thoughts on “BeagleV Catches Fire With The BeagleV-Fire

        1. “I have no use for it, therefore nobody has a use for it” is not as compelling a statement as you might think.

          It’s got the BB up charge I’d expect, but they’ve also been super reliable, long term easy to acquire, and quite capable. This sits in a nice zone that’d help my team in bridging some prototypical integration testing hardware together, so we’ll be looking into them.

    1. The SOC itself is $68 in quantities of 100, so the $150 price tag on this board doesn’t strike me as unreasonable. Comparing it with pricing and features on other FPGA development boards, $150 is quite reasonable. I just might order one to give it a try.

    2. I wanted to disagree with you, but then I looked up the specs. 667 MHz maximum frequency on those RISC-V cores. And those cores are already at a pretty decent IPC performance disadvantage. That’s going to be a hard pass from me.

      1. 5x 64 bit 667 MHz CPUs (4 Linux capable) is I think the most conventional computing power available in any FPGA. Zynq I think tops out at dual 32 bit 800 MHz Cortex-A9.

        Like the Icicle from late 2020 (same FPGA family but 10x more FPGA resources), this comes with Linux pre-installed, along with FPGA programming that connects the CPU cores to the peripherals. You *could* use it just as a RISC-V Linux computer and ignore the FPGA component, but that would be stupid when you can get a quad core 1.5 GHz JH7110-based SBC (VisionFive 2, Pine64 Star64 and PineTab-V, Milk-V Mars (and Mars CM)) for as little as $34. Heck, the PineTab-V is only $10 more than this board, and that’s a tablet with docking keyboard/trackpad cover.

        And of course there are much cheaper pure FPGA boards which you can load a soft core running at maybe 100 MHz into.

        But if you want pretty decent computing power AND actually want an FPGA then this is pretty good.

  1. Hmm, I do wonder how much does the PolarFire tooling from Microchip/Microsemi cost – there is a free (1 years) license but no idea whether it applies to this part too or not.

    Don’t see much point in getting stuck with locked down proprietary tooling Microchip is known for when there are alternatives that don’t require it. Unless Beagles ensure free tooling support for this board, it is going to be a hard sell. “Open Hardware”, my ass. I wonder what is open on this, apart from perhaps the board iself.

    1. What is a closed FPGA today is an open one tomorrow. It wasn’t so long ago that nextpnr only supported lattice chips, now afaik there are 6 different families supported.

      I wouldn’t be surprised to see interest in porting nextpnr to this chip simply on account of this board existing.

      1. “What is a closed FPGA today is an open one tomorrow.”
        I doubt that. After more than two decades there’s still no official documentation on the inner workings of such simple things like GALs/CPLDs and on how to program them. Anything the average hobbyist can access (and afford) is based on reverse engineering and guesswork. While some 3rd-party tools qualify as “open”, the programmable devices are not and will probably never be.

      2. Good luck with that.

        There is no “porting”, the fabric and the bitstream format have to be reverse engineered from scratch first, before one can even think about porting the tooling.

        There are thousands of FPGA models on the market for decades – and we have free tooling for about 3 of them and only one (ice40) is somewhat usable. And those are really basic low end parts with no fancy features or hard IP to speak of.

    2. I don’t understand why any FPGA company would charge for the tooling. It’s not like someone can use their tools for the competitors product. This is why I’ve avoided Microsemi, I don’t want a useless board when the free year is up.

  2. This is a lot cheaper than I was expecting, at only $150. I’m sure it’s going to be VERY popular with people looking to run RISC-V hardware and maybe even build farms, since all the alternatives there are weird boards with even stranger chinese RISC-V chips.

    The fact that this is reconfigurable is going to be a bit weird, but as long as there’s some “standard” FPGA image as a base, that should get things rolling. Will have to ask around to see what support for this FPGA is like or how hard it would be to support.

    1. The tooling is ok. Hard to install. The free license covers this particular SOC. Their software package also comes with a *lot* of prefab soft IP components.

      Whatever default gateware they have makes the board heat up a lot at idle. It needs a heatsink at the very least.

      If you are intending to use it just for a generic RISC-V CPU, get something else. This is a board to get if you want a tightly integrated CPU cluster and FPGA, there is no other reason to get one.

  3. There is also the Beagle Bone Ahead (RISC-V) board. Thinking about grabbing one of these to play with as I don’t currently have any RISC-V SBCs. Not into FPGA so much, but the BB Fire looks like a neat board for those people that are interested!

  4. It all hinges on the availability of the FPGA tooling. I still have my hands full with Zynq thanks to the cheap EBAZ boards I picked up — and Xilinx has made tooling for Zynq nicely available (thanks!). More than the money for this, the need to climb another FPGA tooling world would be the biggest negative, and if I had to pay to suffer in that fashion, it would be a definite no.

  5. From a RISC-V perspective it is not jaw dropping RISC-V board. I know very little about FPGA boards, it is at least an OKish FPGA board ?

    Like if you wanted to use this hardware as a learning platform for say writing basic DSP in VHDL/Verilog/SystemVerilog would this be an OKish board. I know that the board has no ADC, but surely you could simulate/replay an IQ data stream from one of the 5 RISC-V cores as a source and/or use a core as a sink to simulate TX. I would actually see no direct ability to TX as a major advantage while starting to learn, to sidestep any possibility of spewing interference, generated by bad gateware, from an actual antenna.

    Or can anyone recommend a better board than the one above as a FPGA learning platform from basic to just below advanced DSP algorithms in VHDL/Verilog/SystemVerilog.

    1. It is certainly an OK FPGA for learning, although the fabric is quite small. However, the vendor is uncommon in hobbyist circles, and it’s a SoC combining a CPU and the FPGA. You might be looking at figuring out quite a lot of things in your own before you manage to blink a led.

    2. I can think of several applications. One of them is a reduced version of the very impressive “Panoradio” wide band software defined radio (google “Panoradio” for all the details). That runs on a Zedboard, which goes for about $600. This one isn’t quite as capable but a reduced features Panoradio (say, without the wideband waterfall display) might well fit. Either way you’d have an external A/D.
      Another possible application would be to construct interfaces between software-emulated processors and actual historic peripherals.

  6. See I wasn’t the only one who thought incorporating an fpga into a cpu was a great idea

    Now if teh have flaws wit dat new instruction set logic or gpu logic

    Just flash teh fpga to patch or enhance

  7. My first thought is that this is what a lot of small robotics groups have been waiting for. The barrier to entry on this module is higher than average (CPU Arch + Board platform + FPGA(!)) but for some people trying to do lots of I/O and number crunching this may well be worth it. Adoption will depend on how well the people producing it can create an ecosystem of tools that make it quick and easy to get up and running, as well as FPGA toolchain costs. That will be a big thing. Part of me wished they’d just stuck with a regular processor and used an off-chip Lattice FPGA with an open toolchain.

  8. Too expensive and based on PolarFire that is supposed to be on its way out. Micrichip has been talking about PolarFire2 for several years now. It also had some presentations that were spilling the beans on details (much more potent cores etc).

    PF is far too expensive for general-purpose use and it has quite limited SERDES. It can do 12GBps, but only PCIev2 support.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.