Told You So!

[olsenn]

I've been saying for years that FPGA's are the future of technology and that it would be wise to invest in companies that produce them. Unfortunately, I chose Xilinx and it appears that Altera is getting the contracts, but it still shows promise.

http://www.tomshardware.com/news/Intel-Altera-FPGA-tri-gate-14nm,21283.html

[Rufus]

I've been saying for years that FPGA's are the future of technology

Yes they have been the (claimed) future of technology for many years and will continue to be.

The reality is compared to ASICs FPGAs are inefficient in power and silicon usage and expensive in volume which limits their application. Improvements in process technology apply to both and don't change that reality.

[AndyC_772]

It limits their application in highly cost-sensitive, high volume consumer products. For other markets, in which volumes are lower and the price is less of an issue, FPGAs are absolutely commonplace, essential pieces of technology, and I use them all the time. You have to sell a *lot* of products in order to justify the development cost of a £5 ASIC when a £15 FPGA will do the job, and the NRE to develop the part is in the thousands as opposed to the millions.

[Rufus]

It limits their application in highly cost-sensitive, high volume consumer products.

If you look at semiconductor market share by revenue Altera and Xlinix don't make the top 25 (Xilinx was 25th last year). Given the cost of FPGAs market share by parts shipped would look much worse.

I don't deny FPGAs have a future, I deny they are the future. Their future will be much the same as their recent past because nothing (including process shrinks) is going to make them significantly more competitive against alternate solutions.

[olsenn]

If you look at semiconductor market share by revenue Altera and Xlinix don't make the top 25 (Xilinx was 25th last year). Given the cost of FPGAs market share by parts shipped would look much worse.

I don't deny FPGAs have a future, I deny they are the future. Their future will be much the same as their recent past because nothing (including process shrinks) is going to make them significantly more competitive against alternate solutions.

I can certainly see your reasoning behind that; but I'm not so sure FPGA's wont improve more quickly than ASICs and eventually catch up and surpass standard microprocessors in terms of sheer ability.

Since hardware is fairly general purpose (verilog may be ported from one FPGA to another as long as it doesn't use specific hardened logic [built in PLLs/SerDes...etc]) I think it will be more capable for developers to pick the modules/cores they need and use whatever FPGA (cheap and small to big and expensive) best suits their needs. Currently, if you develop software for an x86 platform, you're stuck with that platform. C and other higher level languages are reasonably portable, but you need compilers, libraries, and everything else written in low level code.

[EEVblog]

I've been saying for years that FPGA's are the future of technology and that it would be wise to invest in companies that produce them. Unfortunately, I chose Xilinx and it appears that Altera is getting the contracts, but it still shows promise.

They have been saying FPGA's are the future for 20 years, and doesn't even look like happening yet.
Altium for example bet the entire company's fortune on FPGA's over more than a decade, and they lost every cent of that fortune. 

Dave.

[EEVblog]

I can certainly see your reasoning behind that; but I'm not so sure FPGA's wont improve more quickly than ASICs and eventually catch up and surpass standard microprocessors in terms of sheer ability.

You don't get it, the ability and performance doesn't matter. FPGA's beat the pants of processors for many tasks, and have done so for deacdes.
It's the inherent flexible reconfigurability of FPGA's that is their downfall in terms of replacing "hard" logic processors. If you think FPGA's will replace processors (as Altium did), then you are wrong, and will always be wrong. They will always be too expensive and wasteful for general applications that hard processors are optimised for.
You'll never get a $1 FPGA to replace a $1 processor, ever.

Dave.

[Bored@Work]

There is also something else. You can teach ever bum on the street to somehow program a conventional processor / computer. Operating systems like Windows show this. Movements like Raspberry PI are even centered around the idea.

That hasn't happened with FPGAs. If you need to get an FPGA program you won't find someone at every corner of the street.

[bfritz]

By definition an FPGA is inefficient at use of silicon area.  There is lots of overhead for the ability to route between the logic cells.  There is overhead in programming that interconnect logic.  Add to that the difficulty in using all the logic, by definition you will not achieve 100% utilization.  (Unless we are discussing some academic exercise rather than a real project.)

In addition, with the overhead in an FPGA, this drives the total power consumption up, as the overhead to make it programmable takes power to drive.

In high volumes, an ASIC is cheaper.  There is a whole business segment that takes FPGA designs, and turns them into ASIC's, as over some volume level, it pays to do so.

Sorry, but with the overhead in an FPGA, I don't see how you came to the conclusion they will overthrow ASIC's or processors.

[AndyC_772]

They're not going to supersede microprocessors or ASICs, but they have absolutely superseded the piles of 74 series logic which we used to see scattered all over boards.

The consumer market is dominated by microprocessors which are designed for a specific application, and which include all the peripherals (eg. USB, SD, Ethernet, LCD controllers etc) which that application is likely to use. They're designed to be a good fit for a certain type of product, and volumes are high so they're cheap. ASICs are, of course, even more specialised.

An FPGA is great for any application where there's no off-the-shelf processor that includes just the right set of peripherals. Need an unusual interface? Put it in an FPGA. Got two devices that need to communicate but which don't share pins with the same names that you can just join up? Put the glue logic in an FPGA. Massively parallel scientific number crunching application? Duh.

The idea of using FPGAs instead of general purpose CPUs is one which has gained some attention in the popular press, but it's really not that relevant an application for them. They make it possible to design a huge variety of boards which it just wouldn't be practical to design any other way, and that's their real strength.