Actions
Engagement
  • Author Author: rscasny
  • Date Created: Date Created
  • Last Updated Last Updated:
  • Views 166 views
  • Likes 1 like
  • Comments 8 comments
Related
Recommended

Poll: How Will Adaptable Hardware Change How You Work as a Design Engineer?

Tech Spotlight Summer of FPGAs:

Main Agenda | Xilinx | Tech Connection  | FPGA Group | Related Tech Spotlight | Participate in the Discussion

 

This poll is related to the following article: Adaptive Computing for Compute Intensive and Latency Sensitive Workloads

 

The value of FPGAs has long been the ability to change designs in the field. Whether to correct bugs, optimize algorithms, or add completely new features, programmable logic offers unique flexibility compared to all other semiconductor options. The Xilinx Versal ACAP takes this concept a step further, speeding up configuration time by nearly an order of magnitude, enabling dynamic swapping of partial bitstreams in milliseconds. Combining Scalar, Engines, Adaptable Engines, and Intelligent Engines, it achieves performance improvements of up to 20X over today's fastest FPGA implementations and over 100X over today's fastest CPU implementations—for Data Center, wired network, 5G wireless, and automotive driver assist applications.

 

Poll Question: How Will Adaptable Hardware Change How You Work as a Design Engineer?

Anonymous

  • Xilinx Versal ACAP seems like bringing in an entire forest when you want to have one tree. Sure it look amazing, but the learning curve is high. Is any of the Versal products even supported with Web PACK? I doubt I can afford Versal from a hobby standpoint, and even if I could, most project would be done faster with a 555 timer or ardunio.

  • Like bit slice before it, FPGA provides the engineer with a flexible tool for building a custom solution when off the shelf components are just not enough to do the job.

     

    They are also very useful for replacing non-existent legacy parts and boards. You can duplicate the implementation using a custom FPGA to extend the life of a product.

     

    DAB

  • in reply to michaelkellett

    Thanks for the information.The other issue that I have with a lot of the available FPGAs have to do with voltage levels.  With a lot of my designs involved in hobby electronics, the majority of the designs need to drive to 5V levels.  The only programmable logic devices that have real 5V outputs (readily available and not too expensive) are the PSOC parts (not true FPGA, but very capable devices with programmable logic).  This is a line of parts that I see me using in future designs (I have already included it in some quotes that I have submitted to some clients).

  • in reply to michaelkellett

    An interesting use of the higher end ones is to use the fpga to execute algorithms. I'm learning how to dynamically deploy compression, encryption and image processing pipelines in fpga. With DMA and memory shared with ARM/Linux. All on the same silicon die.

     

    A different aspect of FPGA capabilities. It can offload processor heavy tasks from the ARM part. And if you can turn it into a reasonable flat stream, orders of magnitude faster than when executed on the ARM.

     

    edit - example: Learning Xilinx Zynq: Hardware Accelerated Software

  • in reply to genebren

    @

    Altera (Intel) have a better range of parts in TQFP packages than Xilinx.

     

    Lattice have some parts in QFN which are OK for simple assembly processes.

     

    Gowin have some interesting parts in QFN too, but they are hard to get.

     

    (Of course, right now, pretty much any FPGA is hard to get )

     

    MK

  • I answered other because the simple answer is that products at the top and cutting edge of Xilinx's portfolio will make no difference at all to my work.

    (Although I may bump into them built into things.)

     

    I have done  many designs with FPGAs and expect to do more.

    Most of my designs use less than 40k LUTs (and many have used less than 5k) - this is probably typical for a small (in my case one person) EE business.

     

    What would get me excited is Xilinx putting some of the smallest series 7 Artix and Spartan parts in non BGA packages - which is so far from the kind of business that the Versal ACAPs are aimed at that we might as well be on different planets

     

    Another really neat trick would be to put DDRAM in the same (non bga) package as a small series 7 part.

     

    I'm not really expecting it from Xilinx  - Gowin have a small FPGA and SDRAM combo - it looks like the US based companies are already surrendering this space to China.

     

    MK

  • While FPGAs can add options to some designs, I am not sure that they will make it into many of my designs.  I would be more likely to utilize FPGA if they came in packages more suitable to hand soldering, but alas the vast majority of FPGAs are BGAs and other packages that don't work well in my process.

     

    I have designed products with larger CPLDs (100 and 144 TQFP) that acted as programmable glue logic which was a huge success as I was able to correct several design issues and add features without pulling out a soldering iron, proving the system adaptable point, but moving up to and FPGA there are not too many choices in that style of packaging.

  • I am not quite convinced that an FPGA would be selected just because it makes the system adaptable.

    If an application requires an FPGA then it is needed anyway.

    If an application uses a CPU instead, then it can be just as adaptable via software changes.

    I have seen applications where a big FPGA was added to future-proof the design and allow features to be added, but they all ended up with huge schedule delays due to the added complexity, and ultimately new products were completely redesigned, so the FPGA was never used for upgrading.