SiliconBlue enters the FPGA fray

Jim,

details their "technology" which is licensed from Kilopass.

Austin

Jim,

It is an anti-fuse array technology.

Austin

Thanks. Any comments on programming Speed, and Yields ? :)

-jg

Jim,

Nope. It also has a shadow SRAM array, so you can program it over and over for debug and test.

Then when you are happy with it, you may program the NVM anti-fuse array, and then set a "can't do anything with SRAM anymore bit" so the device is now like our Coolrunner CPLD's: reads the (eflash) NVM into SRAM on power up.

Not strange to me at all why they chose this set of features.

But, the FPGA business is not just silicon, it is also software development tools, IP, applications, customer service, design service, ....

Without a portfolio of hard IP (MAC, PCIe, USB, SerDes, uP, MMU ...) they are also immediately at a disadvantage.

Austin

Hmmm...

"Please bookmark this page. We will be adding a lot more content here soon."

might have to wait for that...

Austin,

without the "BIG COMPANY" issue and without the need of legacy support, they also have immediate ADVANTAGE and chance to do things right from the beginning. Sure maybe they can not fully use this advantage, but it exist.

Antti

Don't underestimate management ;)

It tend to get worse with size. Some companies have compartmentalised themselfes into several smaller fictitious companies within the "main" company. To mitigate this this issue.

Antti,

Did you that Xilinx is the youngest (newest) successful PLD company? All of the other successful PLD companies are older than we are.

I also like to comment that Xilinx is the "biggest little company I have ever worked for." The meaning of this statement is that although we have more than 3,000 employees now, around the world, everyone acts in a way that benefits the whole (which is more like a small company). If you read in FPGA Journal the interview with Wim and Moshe, you will see what I mean.

By the way, having talked with Wim many times, and having met and talked with Moshe, we are very fortunate that Xilinx has (had) such great management.

I wish SiliconBlue good luck, it is a really tough business (as evidenced by all the failures of PLD companies in the last 24 years).

Austin

Austin - do you refute Antti's post without actually disagreeing?

If you consider SiliconBlue to be a younger (newer) PLD company than Xilinx, the issue is just one of success. Antti's point seems to be that their tool development wouldn't be hampered by the legacy of any successful PLD company that's 20 years or more in age.

I love Xilinx. And I often fight with arcane tools. It doesn't change the lure of the silicon and raw capabilities.

I appreciate small company mentality, but there's only so much that can be done against the history of millions of lines of code.

- John_H

John_H,

Aye, that is the rub: success.

Austin

When we say that (almost 25-year-old) Xilinx is the youngest successful PLD company, we are looking back on decades of exciting announcements by start-ups and also big companies getting into this business. Here is the gist of a slide that I have used for over a year:

"A Maturing Industry...

The two leaders have a combined market share of 86% leaving 14% for Lattice, Actel, Quicklogic, Atmel...

All broad-based suppliers have given up: Intel, T.I., AMD, ATT, Philips, Cypress, NSC, Motorola

"FPGAs demand too much management attention"

Most start-ups vanished: Dynachip, PlusLogic, Triscend, Siliconspice (absorbed) Chameleon, Quicksilver, Morphics, Adaptive Silicon (failed)

Fab-less might make it easy, but there are strong barriers to entry: Tool familiarity, design re-use, risk avoidance by the users, Economy of scale, patents, technical support, availability of sales channels

Where is the next successful FPGA start-up?"

End of quote. I do not necessarily expect this situation to last another 20 years, and it gives us no reason for complacency or arrogance. But it is a sobering thought whenever listening to the newest announcements... Peter Alfke, Xilinx Applications

I wouldn't worry about yield, as Siliconblue/Kilopass should have built in bit error correction when reading / programming the antifuse- NVM.

I would worry more about programming Speed for the NVM, remembering Quicklogic Parts long ago needing 20minutes to program. But again this could be much better because the NVM is seperate and localized - not spread over the chip - so it may also be optimized for speed / parallel Programming.

I hope they have a useable development software environment. Their opportunity is making it easy for engineers to replace other FPGAs with theirs / adapt their FPGA's quickly and being able to deliver stable working parts. They seem a worthy competitor for the mass-FPGA maket. But also the path they have chosen seems easy to pick-up for the big players (on success), leaving a rather small time-to-market advantage for them, compared to the massive market/financial power of the old players.

Raymund Hofmann

Austin,

if their silicon works at all, and is low cost as advertized, it will defenetly find some happy customers just because the PACKAGE options alone:

3500 cells, 10 Kbyte RAM ==> 3x4 mm !!! 15000 cells, 32 Kbyte RAM ==> 5x6 mm

if I compare that with Spartan-3A family ??

Actel is supposed to have now IGLOO15 in 4x4 package but that as tiny as hell, so doesnt really count, so as do not Altera MAX2's in 6x6

Doesnt even matter how bad the tools, for this size:package density options the silicon will find instant customers very quickly in my opinion.

So would Xilinx sell S4 to new customers if only the package options would be better than S3 of course.

Antti

Antti, when a much smaller company (like Actel) wants to compete and survive in this field, it must pick a protected niche. (Ultra- small size, lowest power, non-volatility, different logic-to-pin ratio, you name it!). It is up to Xilinx and Altera to evaluate these "niches" and assess whether they are attractive for us. Sometimes they are, sometimes they are not. What is right for Actel is not necessarily right for Xilinx and Altera. The user, however, benefits from the variety made possible by the wide differences between companies, and their genetic diversification. "Survival of the fittest" combined with a diverse gene-pool. It works for plants and animals... Peter Alfke

Peter,

would you describe "CONSUMER" as a niche market?

sorry, I could not resist.

Xilinx, Altera, Actel, Lattice have nothing to compete with the SiliconBlue, the only one possible having competing solutions is QuickLogic (ArticLink). Sure SiliconBlue is not targetting the general purpose FPGA market, so you can call it niche. But consumer market is a pretty damn big niche!

Antti

and, of course, Xilinx could always Buy them out. They did that to get Coolrunner, and Coolrunner is now looking 'long in the tooth' [in FPGA terms] with no recent advances. MAX IIZ and MachXO are in a space Xilinx is missing.

-jg

Hi Antti,

I have not seen price promises, or package info yet. Got any links for this ? (or send me a PDF ?)

-jg

yes, Xilinx as purchased many things for slow death. in 2004 just before the byout by Xilinx Triscend was about to come to the market with the FPGA devices with ARM+ethernet+ADC+something i dont remember

Spartan-4 (ARM+eth+usb+can) will maybe only announced Q1 2009 with real silicon availability no sooner then H1 2010

so the Xilinx by-out politics made a 6 year vacuum for those FPGA users who wanted to have ARM+ethernet hard-core in FPGA

Antti sorry Xilinx, I was about to obtain those devices from Triscend. Was really disappointed they never realized.

^^^

Surely s/ARM/PPC/ ? If it has any hard CPU core at all, that is. I wouldn't bet on it.

Wim Roelandts originally said that Spartan 4 would be out in 2007, so it seems quite possible that it will be announced in 2008.