Intel announces new FPGA families (opens in new tab)

I wonder if the war has anything to do with this actually. These chips and some larger STM32 like F7 and H7 are dual use and can be used in missiles and drones. I couldn't buy the car manufacturers' excuse. I think MICs around the world have been stockpiling these chips.

I can only imagine how long it will take the fitter to pnr these insanely wide busses. Maybe a 13th gen CPU overclocked to 8GHz would be handy.

I share(d) that concern. Mitigating this is that a lot of (low to mid range) alternatives seems to have gotten steam: Lattice Semi, Efinix, Gowin, and more that I forget at the moment. I'm rooting strongly for the underdogs to become viable alternatives to the current duopoly.

It would of course have been preferable for any of them to directly support the open stack (yosys, nextpnr, and the various bitstream generators); Lattice's ECP5 is nice, but not very fast.

My particular worry was that AMD would cancel Xilinx's CPLD chips, but so far it does not look like it. They are very very useful if you want to replace old 74xxx circuits and greatly simplify development.

For me, one option is to be able to learn processors in depth. You want to understand how one works? You can always emulate it in SW, but on an FPGA you can make them interact with the real world in real time and probe them at the same time, with ready made signal analyzers that the vendors provide, and/or by making your own tools.

Another option, do you want to see by yourself how caching improves code performance? You can make that on an FPGA also.

I have done some time ago a very nice course which is called From Nand to Tetris, that teaches how to go all the way from gates to a processor. My plan for the near future is to implement the NAND2Tetris processor on a Basys 3 board, and make some of the examples I mentioned (like, probing internal registers using HW tools, or seeing by yourself how a cache improves performance). I hope to make some of these things available on my site. I am the OP, but just in case, my website is https://fpgaer.tech

There is vast spread of FPGA sizes, so on affordability side it's definitely possible and some people did it.

The "problem" is really that you can do a LOT with MCUs, and programming FPGAs is very different than microcontrollers serially executing your code so they are pretty niche.

Like, if you for example needed to drive few thousand neopixels techncially FPGA would be perfect use case, just make a blob of memory and a bunch of serializers that feed each chain of LEDs.

... or you can just use some DMA magick on some fast STM32 and hack it out without having to learn new programming language.

When would a hobbyist need processing much faster than a modern beefy CPU provides, but not in a highly parallel fashion which a modern GPU provides?

My only relatively reasonable suggestion is making perfect hardware emulators for old hardware, like vintage computers and their parts.

Alas, FPGA programming model is neither easy nor intuitive, and the tools are mostly proprietary, expensive, and (as I heard) have a steep learning curve.

>Are there any cool hobbyist uses of these things?

Yes, there is, if you're willing to learn.

>Always game to learn a new tech.

Great. A good starting point is NANDLand[0].

Their board is not at all a bad purchase at <$100, particularly to follow their tutorials, but you can do better[1] if you are adventurous. e.g. iCEBreaker or iCESugar.

I'd still recommend to stick to iCE40 and the open source yosys/nextpnr stack, until you become competent enough to actually know what you want/need.

[0] https://nandland.com/

[1] https://www.joelw.id.au/FPGA/CheapFPGADevelopmentBoards

For someone making hobby electronics experiments, an FPGA can be used to make a cheap (but with modest performances) logic analyzer and digital oscilloscope (and test signal generator, if desired).

Many years ago, that would have been the only solution to avoid paying thousands of dollars for a professional instrument.

However, nowadays there exist a number of commercial solutions for such cheap instruments, which are made exactly like this, i.e. with an FPGA surrounded by appropriate I/O circuits and connectors, some USB or Ethernet interface for the PC and some capture software.

These have the advantage that are ready to use out of the box, but making yourself such an instrument can be instructive.

There's always sound or image generation, or programming a game like Pong. That's always looked cool to me. There are some "easy" entry points, e.g. https://maker.pro/arduino/tutorial/getting-started-with-fpga..., but I've never gone as far as actually try them.

With a bit of luck, someone will reply with better options.

Edit: or look at this list: https://hackaday.io/search?term=fpga

FPGA innovation is very incremental. The hyperflex architecture is basically the single biggest innovation in FPGA fabric, and it basically didn’t pay off. Altera also smashed up against the wall of Intel 10nm. So all in all, some reduced precision for AI and some EMIB enabled stuff isn’t too bad.

Competitive in what sense? in which market outside of china? Cyclone V a decade old platform offers similar capability and is a tested platform.

FPGAs would be so much more useful if AMD(Xilinx)/Intel(Altera) just gave up trying to build there own tools and instead supported F4FPGA: https://f4pga.org

Yes, but that's not the problem. The problem is these Agilex devices are insane overkill for that task, and you likely won't be able to buy the SKUs in a reasonable form factor or price for that use case. The Mister worked so well because the actual design of the device has the right blend of peripherals/compute and Intel/terASIC ate the NRE to design and manufacture it, and that's because they intended it use it to teach and educate people and needed a lot of volume for that. Agilex isn't a teaching part. It's their most advanced top-of-the-line part with its own logic design requirements that differentiate it from earlier lines (Arria 10 and prior.)

You likely aren't going to see "consumer-oriented" Agilex-powered devices in similar form factors anytime soon; most of the compute power of the fabric would be completely wasted without fast peripherals/storage anyway (the fabric can be clocked very high if you pipeline enough) and the shipping volume to make the boards attractive/low-price enough would need to be pretty decent. This means the board would actually end up expensive; why pay so much for the FPGA if all the I/O sucks? But good I/O is expensive. The Agilex parts are some of Intel's most complex devices on their latest "Intel 7" process too, it's not cheap.

I guess it's maybe something that would be possible to crowdfund in low volumes, but it's a truly huge amount of work even before you get to price. Designing the board, getting the software and tools working, documentation, that's before you even manufacture. And realistically there are probably other FPGA options that would be cheaper and more readily available. Something like a Kintex-7, or Arria 10 part would probably do just as well for significantly cheaper, I'd think, but it's hard to guess unless you have more concrete requirements in terms of LEs/memory/clock speeds.

Maybe if you pray they will make an Agilex-D PCIe developer kit that "only" costs $2k USD, requires tiny fans that sound like jet engines, and has 1 HDMI out.

I'll try to answer this, though my direct experience is only with Cyclone and Max 10.

The Mister uses the Cyclone V 5CSEBA6U23I7 built on TSMC 28nm. It has 110000 logic elements and also a hard ARM.

These are on Intel 7, so there a significant process improvement.

The smallest Agilex D has the same number of logic elements(1) and the largest has 6x as many (2). Note that Agilex tends to be much more expensive than Cyclone.

Realistic maximum clock rates on the Cyclone V are 50-200MHz, depending on the length of the combinational logic chains. For the Agilex I think that it is more like peak 600MHz. So I would guess say 25% faster and say peak 750MHz?

As to fitting all the logic for n64/PS2, I don't know!

1) Comparing these directly is not quite accurate.

2) https://www.intel.com/content/www/us/en/products/docs/progra...

When you talk about "mid-range" in Intel's current product line up, that usually means the Arria 10 line. Those are typically around $1,000 for just the FPGA package itself, so I don't think we're realistically talking about something that's going to make meaningful impact on the hobbyist emulation market.