FPGA boards,which one you use?Xilinx,lattice?Microsemi,or Altera?

[david-IT]

FPGA boards,which one you use?Xilinx,lattice?Microsemi,or Altera?And could you give me some suggestions?

[danadak]

May not seem like it but open up Excel and make a feature vs board list,
then rank the feature set according to your needs.

Decision matrix is very effective t answering questions like this.

Regards, Dana.

[suicidaleggroll]

Xilinx

[rstofer]

May not seem like it but open up Excel and make a feature vs board list,
then rank the feature set according to your needs.

Decision matrix is very effective t answering questions like this.

Regards, Dana.

Otherwise, it is Chevy vs Ford.

There are two aspects (maybe 3):  Which chip is being used and does it have sufficient resources for the project and, second, which board features are important to the project.
Years back, Digilent would make Spartan 3 boards with multiple 50 pin headers.  Then, on the Nexys2 they went to a single 100 pin FX header plus (4) 12 pin PMOD headers.  The Basys3 board I bought the other day has (4) 12 pin PMOD headers and nothing else.  Headers are getting scarce and headers are the only feasible way to attach something like a Compact Flash device.  Sure, I could use an SD device but that's a lot harder to do (yes, cores are available) and I'm not sure about the difference in transfer rate.

Then there are the ztex.de boards which have almost nothing on the board (some have SDRAM) but 100 pins of GPIO.  The headers are stackable so there is some hope of building a motherboard with all the desired expansions (like my Compact Flash).

I tend to buy my boards from Digilent but over the years, I see the boards becoming more all-inclusive and a LOT more expensive.  There are only a very few boards I can actually afford.  Yes, they give substantial student discounts but that's of no value to a retiree.  Maybe I should go back to school!

So, when you make out the spreadsheet, include things like cell count, BlockRAM size, DSPs (if desired), multipliers (if needed) and other features of the chip and then create a sheet for boards and list their characteristics while referring back to the chip.  There are a lot of boards out there that use truly minimal chips.  Somewhere around 19,000 equivalent logic cells seems like a nice place to play.  My Nexys2 board has this.  The new Basys3 has 33,000 equivalent logic cells - a viirtual dumpster for logic.

A third issue is the toolchain and, more to the point, the licensing.  At one time Altera issued a 'free' annual license with the understanding that it might not be renewed.  Well, why would I invest the time to learn the package if I would lose access down the line.  USCD did this with their UCSD Pascal licenses and it was fairly disruptive when entire businesses were based on writing code using their package.  So, I skipped Altera in favor of Xilinx WebPack ISE and have used it for several years.  It is now at the end-of-life.  Version 14.7 is the last there will ever be but it is still available and the only version that works with the older boards.  Vivado has a WebPack (free) version but I think the range of devices is fairly limited.  At some point you need a license just to implement a design.  That's fine, you can get a node and board locked license for 'free' (maybe $10) with some of the Digilent boards.  This is important!  This is the new toolchain, you need to learn it because the old ISE doesn't support the new chips and Vivado doesn't support the old chips.  But the license is a big deal.  It is both node and BOARD locked.  At this point, even Altera's license seems like a better deal.  And that license you get for Vivado because you bought the board?  It's apparently only valid for a year.  Then what?  Throw away the board?  This has gotten ugly!  Or maybe it's only potentially ugly!

Before you settle on a chip, check into the toolchain and the licensing!

I don't expect to get access to all of the manufacturer's IP with the 'free' version of the toolchain.  But I expect it to compile my code and program my boards.  With Vivado, this is no longer true.  On the old WebPack ISE, it was pretty clear that I didn't get that access but it would program a wide range of chips in an even wider variety of boards.

I like Xilinx, I like Digilent, I hate the new licensing and I'm none too fond of Vivado.  It takes a very powerful PC just to get up to the region I call SLOW.

[uChip]

Boards are okay for fiddling around but soon you need to move to your own circuit.  A board that has every possible display, communications and peripheral option will be expensive.  The tools and documentation that help you move from a reference board to your own design should be a large part of the choice.  Also, chip price is a factor. 

I went to Digi-Key and searched CPLDs and FPGAs.  I chose the cheapest parts that would do what I needed for my project.  Then I evaluated the synthesis, simulation and programming tools.  These are largely a matter of taste since they all do well enough to get designs out the door.  Once I decided on tools THEN I bought a reference board.  Seriously the reference board lasted about a day before I needed to either build my project around the reference board (not feasible) or move to my own design.  I made miss-steps along this path but ultimately have progressed to my own functional PCB.  For me the steps (including miss-steps) lead as follows.

Digi-Key search -> Xilinx CPLD, Xilinx & Lattice FPGAs
Tool evaluation -> I preferred Xilinx ISE Design Suite
Chip selection -> Xilinx CPLD
Programming tool -> Digilent HS2
Reference board -> Digilent CoolRunner-II
Ran into clock gen issues with CPLD.  Realized that FPGA was a cheaper system solution.
Re-evaluated Xilinx & Lattice specs in light of clock gen.
Chip selection -> Xilinx FPGA
Reference board -> none.  By now I was feeling the limits of the reference board and eager to move to my own design.  I went straight to my own circuit and fab'd PCB (OSHPark).

YMMV,
  - Chip

[poorchava]

I've mainly had experience with Xilinx and Lattice.  Altera always turned out to be more expensive for the required functionality.

Sent from my HTC One M8s using Tapatalk.

[izx]

My experience only extends to high-performance, relatively cost-effective FPGAs for DSP so I'd have to say Xilinx.

[TiN]

or Altera

You mean Intel? 

I like Quartus flow, found it easier to use.

[uncle_bob]

Hi

If you are looking at eval boards, you are doing stuff with the tools as much as you are doing anything with the chips. Do you want to learn the latest tools? (Vivado is new, ISE is on it's death bed).

If you want the latest parts, you will be buying "manufacturer boards" and not eBay items. That's fine, the price will be higher.

If you buy the "cheap boards" take a hard look at what the modern tools will support. Older chips generally mean older versions of the tools if you want the "free tool". You could easily be learning all the in's and out's of a tool that is already dead (pre-Eclipse Quartus or the already mentioned Vivado / ISE).

It all depends on what you are trying to do and what your ultimate goals are.

Bob

[obiwanjacobi]

Look at the IDEs first - that is where you spent your time.

[2c]