Your opinions on the FPGA market in the next few years

[filssavi]

Anyway, this question half me trying to figure out what FPGA maker/toolchain to go with and half me being curious about the industry and where it's going. If you have any insight, it would be cool to hear!

I don't really have any insight  , but I'll dump my thoughts here.

First of all, moore's law - most likely- is coming to an end and we're processing more and more stuff, e.g. Augmented reality stuff, Pokemon Go, Speech, AI, more video processing etc. So going for FPGAs is the right thing and I see all those ARM cores in mobiles, tablets, etc. getting 'FPGA-extensions' very soon (especially ZYNQ-like SoCs).

For high volume consumer devices an FPGA is about the worst choice. An ASIC is much more cost effective and consumes less power. An FPGA is a niche solution for projects where you can't use a SoC alone.

It all depends on the complexity/performance of the design if a  modern process is required (32nm or lower) the volume must be very High (i mean the First Nvidia gsync module shipped to consumers using an altera arria FPGA, and Nvidia can definitely do it's silcon just fine

[frogmaster]

For high volume consumer devices an FPGA is about the worst choice. An ASIC is much more cost effective and consumes less power. An FPGA is a niche solution for projects where you can't use a SoC alone.

This is hard to argue with, but I hope FPGAs will become more efficient in the future.

'Niche solutions' I find hard to define these days where we have many different compression and crypto algorithms. Also some phones connect to dummy 'display+keyboard-laptops', there may be some protocols that could possibly be updated after shipping. I could think of some applications for networking? idk. It'd be cool to have those devices just to see what software people would come up with.

[uncle_bob]

Hi

It has been a "Ford vs Chevy" race for the last two decades at least. Some love Altera, some love Xilinx. There has always been a couple of "other guys" in the race that might pull ahead someday. Each of them go through generations of their tool suite. When they do there is much mashing of teeth and "how could they make this crap" going on. A few years later, it all calms down and the world wonders how they ever got along without the new tools. Every so often one or the other moves ahead in a significant way. A few years later things are same /same between them. Right now Altera has a Max 10 with built in flash, one would guess that Xilinx will have one "real soon now".

What hasn't changed much over the years:

1) Altera really loves to sell you various licenses for the IP, Xilinx is a bit more willing to give you stuff like DSP for free.

2) Altera will let you do things in a schematic entry format, Xilinx has never been real happy about this sort of thing. Both really like the idea of some sort of high level language coupled with their "wizards".

3) Altera seems to be more tolerant of "clone" programming hardware. They still nuke the stuff from time to time. It's a close race in terms of who is more hostile ...

Bottom line, you can do just fine with one or the other. If you flip back and forth, there is a lot of learning in terms of their latest toolchain.

Will this change in the future?

Altera *might* get cheaper based on "last generation" Intel fab utilization ... it also could get more expensive. Cheap isn't Intel's biggest feature.

Xilinx could get bought by Taco Bell. Who knows what happens as a result.

CPU <-> FPGA links could become a big deal. So far nobody really has that stuff worked out down all the way through the tool chain. In fairness, getting that all worked out is a really complicated task.

"Somebody Else" could pull ahead of the big 2. Since everybody is bound by the same fab issues, that's going to be tough.

Best guess, past performance does predict the future in this case.

Bob

[Scrts]

We've got 5 floating Quartus licenses plus all the Altera IPs for free when committed to buy at least 10k parts/year for 5 years.

[uncle_bob]

We've got 5 floating Quartus licenses plus all the Altera IPs for free when committed to buy at least 10k parts/year for 5 years.

Hi

.... yup... and in 5 years those licenses go to $2K to $20K each per year (depending on the IP) . Don't ask how I know this....

Bob