Will Spartan-6 be totally left out migrating to Spartan-7 and further in series? Your thoughts?

For a great many applications the Spartan 7 offers no advantages over the Spartan 6. Or to put it annother way, for applications where, for example, an Altera MAX10 or a Lattice XP2 is doing the business there is no point at all in replacing it with something else.

It is interesting to see how the FPGA business has changes so radically over a few years, in 2019 I think that Altera and Xilinx had 80% of the total FPGA business between them.

Today neither company exists, the pair having been swallowed by Intel and AMD.

I suspect that neither of the parent companies is very interested in low end FPGAs.

There is a huge gulf between the players, Xilinx can ship a 4million LUT FPGA (at, I believe £20k+ per chip) Gowin can sell you a bluetooth enabled 4kLUT FPGA for (I undertand) < $2.  They are not aiming at the same markets !

In the arean with smaller parts there are old players like Lattice and Microsemi (now part of Microchip) and new players like Gowin, Efinix and Renesas. I suspect that they will try to move into the holes being left by A and X.

The picture is complicated by the supply problems, and this makes it even harder to guess which way things will go.

I think there is huge scope for low power sub 50k LUT FPGAs in the future. A and X are betting the farm on the really big ones - I expect both markets to survive but that eventually A and X will disappear as discrete entitites and be completey absorbed into their owners, or possibly even spun out as independent again.

MK

It totally makes sense michaelkellett. True that FPGA company entities have been changing around probably because of niche market segments like space and high frequency. So called small players are really growing bringing solution in their own way for specific applications and going really affordable which will be useful for entrants.

Usually people still think FPGAs are very specific, complex and overkill for general solutions. Nowadays, applications of them in power electronics and algorithm implementation is coming up and surely something to keep an eye at.

Xilinx is most interested in the high-end high-profit market.  But that leaves room at the bottom for other companies.  My favorite Xilinx family was the Spartan-3A, a relatively simple architecture that was pretty easy to understand.  The Spartan-6 clocking alone is a steep learning curve.  I haven't delved into Spartan 7.

These days I like the Lattice iCE40 FPGAs which Lattice acquired from Silicon Blue.  They have 4-input LUTs like the Spartan-3A.  They also have open-source design tools.  So while 4-input LUTs may be obsolete as far Xilinx is concerned, you can get small cheap FPGAs from others.  Similarly, Intel specializes in gigantic hot multi-core high-performance processors even though they invented the microcontroller long ago with the 8048.  They leave the low-profit microcontroller business to STM and others.

Interesting outlook. I had never thought about it. 

iCE40s are really nice from what I have seen from some people using. I have thought of getting hands-on with one, but just felt that there's already an established xilinx toolchain and documentation which is widely used and also saw contests coming up using them. And I once thought ok, now let me understand Zynq 7000 in detail, but as you mentioned, its too deep and what's in documentation is just a very high level abstraction and proprietaryness. They have kind of allowed us not to worry about internals much bringing flashy HLS, Microblaze, Block design and abstractions like that.

True that open source toolchains and simplistic architecture will be of great use here allowing user customization and feature addition to some extent atleast. I will now take a chance to explore opensource tools and low-end FPGA. Let's see, its just a matter of penetration to market for these FPGAs I feel.