It looks like the makers of field programmable gate arrays (FPGAs) are not quite ready to cede the coprocessor market to GPUs just yet. Achronix Semiconductor, an FPGA maker based in San José, just down the highway from chip giant Intel, has announced that it will use Chipzilla's fabs to cook up a future generation of chips that …
FPGAs are lovely.
And to be honest, a future PC with some kind of FPGA board on could be excellent for the gamers (and serious propellorheads) out there.
The game needs some physics coprocessing? No problem. One data transfer later, and the FPGA is a physics coprocessor.
Music program? Well, how about a quick switcheroo and that little chip starts doing DSP work.
Buy a FPGA that's big enough, and you could tell it "be a GPU". Lo and behold, you have a GPU.
Or if you really want something awesome, a 555 timer!
So it's not like GPUs are ever really going to wipe FPGAs out. Different beast, different application. Plus you can't tell a GPU to be an MP3 player with integrated storage and signal processing, like you can with a (large enough) FPGA.
Should really name them ACME chips...
long way off
first you'll need a mobo with a general purpose fpga on it with the ability to download a bitstream onto it. Then you'll need the bitstream or the design. Then the drivers.......
And after all that effort another fpga will come along and render your baby obsolete.
No need to be overly PC
"San José" is the city in Costa Rica, while "San Jose" is the city in Silicon Valley.
Re: No need to be overly PC
Actually, it has upgraded with an accent:
"Intel is well known for being on the cutting edge of wafer-baking tech"
Pity they can't seem to be doing much interesting with all that wafer-baking muscle. Even without the micros~1 straitjacket and with substantial tech transfers from several other companies, itanic sunk like, well, a titanic. To put this in historical perspective: The 8008 was an external commission to an external design that saw its development team plundered for the concurrently developed 4004 because /that/ thing was their own idea and therefore lots cooler to work on.
Still and all, I support this product and/or service, since it might even make something like a fully open source cpu design at reasonably contemporary speeds within reach of the averagely-monied tinkerer.
In the 70's San Jose officially too the accent out. The proper city name is San Jose no accent mark. Just because some choose to use the accent mark on the web site does not change that.
some FPGA are cheap and fun
You can get a couple year old Spartan 3AN starting at $10 in small quantities and it is just a blast to play with. It even comes in a TFQP package you can hand solder (if you've got a steady hand). I never mastered the art of home BGA assembly but I hear you can do it with a toaster oven. The newish Spartan 6 runs about $50. Of course a top of the line Virtex 6 can run $10K. Is it obvious I've only used Xilinx?
@"bring the cost of that 1 million LUT FPGA down to around $400"
That's still ridiculously expensive, which will greatly limit the number of applications it can cost effectively be used in, which in turn will limit the quantity of FPGA's sold, which in turn will keep their price way too high.
To really make an impact we need that 1 million LUT FPGA down to around $4, because at that price point it would start to really open up lots of markets.
Very interesting. One point that seems to have slipped by is that opening their own fab for the FPGA production Intel has a direct supply of high quality, very fast, FPGAs along with the basic recipe to make them. Among other things, Intel can use this knowledge to directly test their HDL before any commitment to silicon. Knowledge is power and the knowledge Intel will gain through this move will be immense.
Good move Intel!