Spartan 7

General discussion about the Mojo
otzen42
Posts: 46
Joined: March 1st, 2014, 2:37 pm

Spartan 7

Post by otzen42 » September 30th, 2017, 7:38 pm

Any thoughts of upgrading the Mojo to a Spartan 7 or Artix 7 FPGA? It would require a BGA package it looks like, but it would be nice to upgrade to something that can use the Vivido development environment. ISE isn't supported on anything newer than a 2.6 Linux kernel, so support for it is running out. Vivado runs on the latest 4.x kernels which is much nicer for us Linux users (or in my case Mac users who don't want to pay for a copy of Windows just to run Xilinx tools). VIvado is also a MUCH nicer tool, and does a way better job when it comes to routing designs, particularly if the FPGA is full or has tight timing (it actually mathematically pre-calculates and optimizes where to place the logic before starting iterative placement, rather than just randomly dropping all the logic down and trying to fix the resulting timing failures like ISE does). It also uses a Tcl command interface similar to PlanAhead (with way more functionality), so I expect it should plug in to the Mojo IDE fairly easily (can't say for sure as I have not yet used the Mojo IDE).

I have also heard that ISE does not run well on Windows 10, so I think its days are limited. Strange that Xilinx still makes Spartan 6 devices without a development tool that can run on a modern OS...

embmicro
Site Admin
Posts: 834
Joined: March 24th, 2013, 12:45 pm

Re: Spartan 7

Post by embmicro » October 30th, 2017, 12:59 pm

We have one in the works. It'll be based on the Artix 7 XC7A35T.

ISE is a huge pain especially with all the work-arounds needed to get it working on modern OS's. We should be able to say goodby for good fairly soon.

We will likely have an announcement with more info on it by the end of the year.

Flux
Posts: 9
Joined: June 29th, 2017, 1:00 am

Re: Spartan 7

Post by Flux » November 3rd, 2017, 8:41 am

That's good news. It will be interesting to see how it compares to the Digilent Arty.

cerkit
Posts: 27
Joined: March 30th, 2017, 10:27 am

Re: Spartan 7

Post by cerkit » November 9th, 2017, 10:19 pm

embmicro wrote:We have one in the works. It'll be based on the Artix 7 XC7A35T.

ISE is a huge pain especially with all the work-arounds needed to get it working on modern OS's. We should be able to say goodby for good fairly soon.

We will likely have an announcement with more info on it by the end of the year.
This is welcome news. I was looking at the Digilent offerings because of the issues I've been having with ISE. I'd prefer to stick with Embedded Micro.

Will the form-factor remain the same and retain compatibility with the shields or are they specific to the Spartans?
Michael Earls
https://cerkit.com

nalexopo
Posts: 5
Joined: November 27th, 2016, 11:23 am

Re: Spartan 7

Post by nalexopo » November 18th, 2017, 2:27 pm

Are you gonna make Vivado tutorials as well? I want to start with vivado but I dont have the courage to do it out of the blue. I did make some projects with Mojo and tutorials here helped me a lot to get started. I will definitely would like to buy a new mojo if it comes with vivado tutorials.

embmicro
Site Admin
Posts: 834
Joined: March 24th, 2013, 12:45 pm

Re: Spartan 7

Post by embmicro » November 29th, 2017, 12:41 pm

I need to figure out Vivado as well but I'm sure we'll have some basic tutorials on getting setup and working though a project. While it seems to be an improvement over ISE in many ways it has its own weird interface quirks.

User avatar
GNF
Posts: 17
Joined: May 7th, 2016, 3:27 pm

Re: Spartan 7

Post by GNF » January 1st, 2018, 5:11 am

embmicro wrote:We have one in the works. It'll be based on the Artix 7 XC7A35T.

ISE is a huge pain especially with all the work-arounds needed to get it working on modern OS's. We should be able to say goodby for good fairly soon.

We will likely have an announcement with more info on it by the end of the year.
Mmmm.... not sure if this is good or bad news. Are you telling us that you will not be supporting the Mojo V3 anymore "fairly soon"? That's sad news.

Although I like the advantages of a more modern FPGA like the Artix 7 XC7A35T I think they're quite expensive compared to a Spartan 6 XC6SLX9. I bought a Mojo for learning how to code Verilog and VHDL on a reasonable priced starter board. And I just ordered the SDRAM-, IO- and Proto-shields. And I'm currently waiting for the new HDMI shield to be released (any updates on that?) . It did seem like a good investment at the time. If and when I finally grasp all aspects of this FPGA there are all kinds of XC6SLX9 based boards to play with starting from USD 25,-. But when reading this I'm wondering if I did the right thing. When I accidentally fry an Arduino it will cost me USD 3,- for a new ATMEGA328. Frying a Mojo with Spartan 6 XC6SLX9 is a more costly event. Frying an Artix 7 XC7A35T will be even more costly.

I'm not against progress though. I think planning the next step in Mojo evolution is wise. But a quick scan on eBay reveals that even the cheapest Artix 7 XC7A35T boards cost USD 90,- and up and there aren't that many to choose from. If you already know what you're doing and need performance then an Artix 7 XC7A35T would be a wise choice I guess. But If you're in the lets-blink a-led-stage it would be overkill.

I always saw the Mojo as an affordable learning platform for starters. So I hope you will keep things affordable. I think more people will try a Mojo if prices went down, not up. And abandoning a Mojo v3 in favor of a more expensive Artix 7 XC7A35T FPGA will probably cost you a lot of user base. But that's just my two cents.

About the Vivado software: Is there a free licence for Vivado just like for ISE Webpack?

If your target audience are mostly developers starting with FPGA's (is it? looks like that to me), why not go for e.g. a Lattice iCE40 together with IceStorm and your IDE on top? Drawback would be that IceStorm, for now, is only available for Linux-based platforms. Note I that I never actually used IceStorm but considered it when starting my FPGA quest (and then I stumbled upon the Mojo via Kickstarter).

To sum up:
Artix 7 XC7A35T, looks expensive (but maybe I'm wrong here).
Abandoning ISE Webpack, not a good idea for those who already own a Mojo or recently bought one.
I would like to see a more future proof (devkit/IDE supported) but mostly cheap FPGA board for the next evolution of Mojo. More affordable probably results in more people buying it, more people using it, larger community, larger user base, more FPGA adoption, etc. etc.
And this is all just my personal opinion. Based on gut feeling and not any fact or marketing research whatsoever :)

Best regards,

GNF
-- The source doesn't lie, Use the source! --

Flux
Posts: 9
Joined: June 29th, 2017, 1:00 am

Re: Spartan 7

Post by Flux » January 5th, 2018, 10:51 am

Artix-7 FPGAs are significantly more powerful than their Spartan-6 equivalents. They also work with Vivado on Windows 10 and modern Linux distros. Given the fact that Xilinx have abandoned development of ISE I think Embedded Micro have to move on too. That's not to say existing boards and ISE installs can't continue working, but for those buying an FPGA board today or starting a new project I don't think it makes sense to use Spartan-6.

However, I don't see why the Artix should affect board prices significantly. The XC7A35T-1FTG256C only costs around $35/€30/£27 in single quantities from the big electronic supplier (e.g. Farnell). Larger quantities will obviously be cheaper. Plus Digilent's Artix-7 board using the XC7A35T-1FTG256C costs $99.
  • The Spartan-6 XC6SLX9 has 5,720 LUTs, 11,440 flip-flops, and 576Kb block ram.
  • The Artix-7 XC7A35T has 20,800 LUTs, 41,600 flip-flops, and 1,800Kb block ram.

tcmichals
Posts: 1
Joined: January 5th, 2018, 4:09 pm

Re: Spartan 7

Post by tcmichals » January 6th, 2018, 2:59 pm

What is the ETA of this board?

User avatar
GNF
Posts: 17
Joined: May 7th, 2016, 3:27 pm

Re: Spartan 7

Post by GNF » January 6th, 2018, 3:06 pm

Flux wrote:Artix-7 FPGAs are significantly more powerful than their Spartan-6 equivalents.
I agree. No discussion here.
Flux wrote:They also work with Vivado on Windows 10 and modern Linux distros.
Also no discussion here.
Flux wrote:Given the fact that Xilinx have abandoned development of ISE I think Embedded Micro have to move on too.
Again, I agree.
That's not to say existing boards and ISE installs can't continue working, but for those buying an FPGA board today or starting a new project I don't think it makes sense to use Spartan-6.
This is were our opinions start to differ. For me it all depends on who Embedded Micro's target audience is. I suspect it is electronics hobbyists on a low budget just starting with FPGA's coming from Arduino or Raspberry PI. Also,

I also suspect that if it's that hard to get ISE working on modern OS's it will be also hard to maintain Mojo IDE for it. So I don't think maintaining two sets of integration (with Vivado and ISE) will be viable for long. But maybe I'm wrong. I hope I am. Question is will the next step be 1. something cheaper with maybe less functionality but targeting a larger user base, of will it be 2. more expensive, have more functionality, intended for a the more advanced users. I'm all for option 1. That's why I opt for the cheapest FPGA available that still allows you to do the tutorials. And maybe an open source IDE is the way to go. That get's a community going (I think it's dying now because you and I are just about are the only ones responding to messages the last few months).
However, I don't see why the Artix should affect board prices significantly. The XC7A35T-1FTG256C only costs around $35/€30/£27 in single quantities from the big electronic supplier (e.g. Farnell). Larger quantities will obviously be cheaper. Plus Digilent's Artix-7 board using the XC7A35T-1FTG256C costs $99.
I think an FPGA starter board with on board SDRAM should cost somewhere around EUR 35,-. You can find them for that price on eBay form Chinese sellers. Don't know about the quality though. The reason why I think starter boards should be this cheap is to introduce as many users as possible to the wonderful world of FPGA's. If the bare FPGA already costs EUR 30,- taking into account manufacturing costs, design time/labor, certification, VAT and margins for Embedded Micro to make it worth while, that will make the board at least cost EUR 100,- to 140,-. I think that's to steep a price for beginners and students. Those of us with jobs and electronics hobbies can afford them, sure. But there aren't that many FPGA hobbyists to begin with because, I.M.H.O. the boards are too expensive and most tutorials are not aimed at absolute beginners. That needs to change. And I think Embedded Micro is doing a hell of a good job and is the most accessible place to start of all places I encountered thus far.
  • The Spartan-6 XC6SLX9 has 5,720 LUTs, 11,440 flip-flops, and 576Kb block ram.
  • The Artix-7 XC7A35T has 20,800 LUTs, 41,600 flip-flops, and 1,800Kb block ram.
I don't think this is relevant for beginners. It is relevant if you've already learned Verilog, Lucid or VHDL and are ready to scale up. Most newbies will focus on the Embedded Micro tutorials and those will keep them occupied for the first couple of months of hobbying. After that, when you think you know what you are doing and know what you actually need for your own projects, you'll probably find the FPGA-board that is best suited for the job. That might not be a Mojo.

GNF
-- The source doesn't lie, Use the source! --

Post Reply