Actions
Forum Thread Details
  • Replies 15 replies
  • Subscribers 531 subscribers
  • Views 1890 views
  • Users 0 members are here
Related

Affordable Versal parts and their penetration in hobbyist circles

unixux

Good day all,

It's been a while since relatively affordable Versal boards came around - at least two sub-$1000 parts from reputable suppliers are available, as well as very interesting board that combines AMD Ryzen with Versal Edge device (that's a tad more expensive though- but still a far cry from the full sticker for the big devkit boards). All are based on the same xcve2302 part - SOM actually. And considering the amount of innovation that Versal ACAP brings to the table - and the AI revolution we're living through, one could naively expect an explosion in both content and support. But it's been relatively lackluster thus far. To an untrained eye there is very little cross-pollination from non-ACAP projects into Versal and those who do target it tend to just support the big expensive parts. (Not to take away from whatever content that _does exist - it's very informative and interesting to those who seek)

What's missing in this scenario - is it vendor support, or hobbyist interest, or is the added value from new technology still not perceived sufficient to justify the cost ?

  • Although this is a very flexible platform, and there is an impressive tool chain, the learning curve is pretty steep if you only have one application in mind. To make it more widely adopted, there probably needs to be more complete low-cost reference designs available. This would lower the barriers to adoption.

  • Many developers might not see the immediate value of investing in Versal boards, especially when alternative, cheaper platforms already provide decent AI acceleration and FPGA programmability for many use cases.

  • Momentum on any project/product kicks in when it is available, affordable and something novel (that can address an existing problem) is being done. Even missing one of the ingredient, the project might not gain traction (immediately). I feel over the last couple of years there has been a lot of SoC's rolled out (Zynq, Zynq MPSoC, K26) with DPU as the main accelerator. Most of the FPGA engineers are transitioning from Zynq/ZynqMPSoC to K26 primarily because of cost advantages and its availability. Still there is not a major differentiating factor/application that shows Versal is better and easier to program than Jetson(or Nvidia's boards). Moreover competing with Nvidia who is a clear market leader (cost and user base) in terms of AI hardware will take time and effort. So we will have to wait and watch how this unfolds. 

  • $7000 for the  AMD VersalTm AI Edge Series VEK280 Evaluation Kit, and the other Versal kits are above $13,000

    That's a lot for a hobby project.  What kits are under $1000 ?

  • I suspect that the lack of hobbyist interest is for the same reasons that I am completely un-interested in the Versal series of parts from a one person design company point of view.

    They are simply too complex - no useful project using Versal parts is likely to require less than several person-years of engineering effort, lets say approximately 10000 hours minimum. But these chips are aimed at things like fully autonomous vehicles - projects which have so far guzzled up billions of £$ with precious little to show for it (in terms of product or profit).

    10000 hours isn't quite out of the range of a hobby project (build a decent telescope might take the same kind of effort) but there won't be many takers.

    Open source projects can generate that level of effort but leading edge FPGA and Open Source don't sit easily together.

    So I'm interested in sub $20 FPGAs that I can use in non BGA packages, but a 208pin TQFP Versal for £20 wouldn't interest me because it would take me forever to get it do anything.

    MK

  • in reply to rsc

    There are ALINX VD100 and Trenz TE0950, both under $1k. There's also Sapphire VPR4616 which is more but not much more. ~$800-$900 is still not chump change for someone doing just DE0-Nano or Arduino-level projects, but in context of modern FPGAs and even GPU boards like Orin Nano, IMHO it fits the bracket. 

  • in reply to manihatn

    Is that a QSFP cage or a QSFP-DD cage?  I might be interested is it's the non-DD variant.

  • in reply to manihatn

    The chip isn't all that impressive in terms of logic cell count or is it? What key component do you think would be worth it to pay extra and go with the added complications of a versal device?

  • in reply to erik__18

    My client is using a very large Versal part in a project.  I'm responsible for the overall platform management code.  I need the Versal chip to get the complexities of startup and external management correct.  This board would allow me to do that without spending anotger $15k for a second dev board.