News & Analysis
Comment
patrick38
I think that the move of Intel is this direction is very strategic. Security ...
TJones
Of all the conjectures, like DAN's best, but haven't seen anyone mention Intel's ...
Intel to fab FPGAs for startup Achronix
Dylan McGrath
10/31/2010 9:58 PM EDT
Possible game changer
For Achronix, the deal with Intel could be very significant. By beating Xilinx and Altera to the punch at 22-nm, executives say the Speedster22i family could put the company in position to compete for many more design sockets that currently go to standard cell ASICs.
Achronix is one of a number of promising FGPA startup companies that have emerged within the past few years, some of which have since shut their doors. Backed by more than $86 million in venture capital funding, Achronix gained notoriety for the speed of its original 65-nm Speedster family, which operated at a peak performance of 1.5 GHz, far greater than any other FPGAs.
But while Holt said the company has been hitting its internal goals, to date Achronix accounts for a very small percentage of the global programmable logic market. That market is dominated by Xilinx and Altera, which together have more than 85 percent market share.
Even with Intel's process technology, Holt said Achronix would not immediately pose a major threat to the big players. "Xilinx and Altera won't be worried about Achronix, initially," Holt said. "We'll just be the third billion dollar FPGA company soon."
Intel, like many semiconductor companies, once made its own programmable logic devices from the early 80s until the early 90s. There has at times been speculation that Intel would return to the programmable logic market, including a prediction early this year by JP Morgan analyst Christopher Danely that Intel could acquire Xilinx or Altera, which was widely dismissed.
While emphasizing that he could not speak for Intel, Holt said the deal "speaks to how important they [Intel] see FPGAs to the future of the semiconductor industry." But, Holt said, "If Intel wanted to be in the FPGA business they would be already. They certainly have the cash."
The relationship with Achronix could be a precursor to Intel eventually combining programmable logic with its Atom cores on the same die to create a new type of device. Earlier this year both Xilinx and Actel Corp. announced products that combined their programmable logic technology with hard ARM processor cores.
While acknowledging that the Intel deal is a "game changer" that improves Achronix' future prospects, Holt said the company was persevering anyway through a difficult period that has been making it hard on startups, controlling spending and achieving its goal for sales last year, which he described as "single digit millions." Holt said Achronix was originally planning to move from 65-nm directly to 28-nm and would have been taping out a 28-nm device at TSMC in the near future. However, he said, when the strategic agreement with Intel came about the company shifted gears and devoted investment to 22-nm design.
Achronix said the Speedster22i family would establish new standards for FPGAs in terms of size, performance, cost and power consumption. The family will offer devices with up to 2.5 million look up tables (LUTs), twice as many as any other FPGA, Achronix said. Because of Intel's process technology, the devices will also offer a cost advantage of more than 40 percent, according to the company. Like the previous generation of Speedster, the devices will also offer a peak performance of 1.5 GHz, 300 percent faster than other FPGAs, and consume 50 percent lower power, Achronix said.
Achronix plans to target Speedster22i to a wide range of applications in the telecommunication, networking, industrial and consumer markets. The devices will also appeal to emerging applications such as 100G, 400G Ethernet networking and LTE mobile communications, Achronix said.
For Achronix, the deal with Intel could be very significant. By beating Xilinx and Altera to the punch at 22-nm, executives say the Speedster22i family could put the company in position to compete for many more design sockets that currently go to standard cell ASICs.
Achronix is one of a number of promising FGPA startup companies that have emerged within the past few years, some of which have since shut their doors. Backed by more than $86 million in venture capital funding, Achronix gained notoriety for the speed of its original 65-nm Speedster family, which operated at a peak performance of 1.5 GHz, far greater than any other FPGAs.
But while Holt said the company has been hitting its internal goals, to date Achronix accounts for a very small percentage of the global programmable logic market. That market is dominated by Xilinx and Altera, which together have more than 85 percent market share.
Even with Intel's process technology, Holt said Achronix would not immediately pose a major threat to the big players. "Xilinx and Altera won't be worried about Achronix, initially," Holt said. "We'll just be the third billion dollar FPGA company soon."
Intel, like many semiconductor companies, once made its own programmable logic devices from the early 80s until the early 90s. There has at times been speculation that Intel would return to the programmable logic market, including a prediction early this year by JP Morgan analyst Christopher Danely that Intel could acquire Xilinx or Altera, which was widely dismissed.
While emphasizing that he could not speak for Intel, Holt said the deal "speaks to how important they [Intel] see FPGAs to the future of the semiconductor industry." But, Holt said, "If Intel wanted to be in the FPGA business they would be already. They certainly have the cash."
The relationship with Achronix could be a precursor to Intel eventually combining programmable logic with its Atom cores on the same die to create a new type of device. Earlier this year both Xilinx and Actel Corp. announced products that combined their programmable logic technology with hard ARM processor cores.
While acknowledging that the Intel deal is a "game changer" that improves Achronix' future prospects, Holt said the company was persevering anyway through a difficult period that has been making it hard on startups, controlling spending and achieving its goal for sales last year, which he described as "single digit millions." Holt said Achronix was originally planning to move from 65-nm directly to 28-nm and would have been taping out a 28-nm device at TSMC in the near future. However, he said, when the strategic agreement with Intel came about the company shifted gears and devoted investment to 22-nm design.
Achronix said the Speedster22i family would establish new standards for FPGAs in terms of size, performance, cost and power consumption. The family will offer devices with up to 2.5 million look up tables (LUTs), twice as many as any other FPGA, Achronix said. Because of Intel's process technology, the devices will also offer a cost advantage of more than 40 percent, according to the company. Like the previous generation of Speedster, the devices will also offer a peak performance of 1.5 GHz, 300 percent faster than other FPGAs, and consume 50 percent lower power, Achronix said.
Achronix plans to target Speedster22i to a wide range of applications in the telecommunication, networking, industrial and consumer markets. The devices will also appeal to emerging applications such as 100G, 400G Ethernet networking and LTE mobile communications, Achronix said.
|
|
|
|
|
Navigate to related information
TingLu
11/1/2010 1:00 AM EDT
It sounds to me that Intel is getting back to FPGA business. If everything goes well, Intel will likely buy Achronix out.
Sign in to Reply
eewiz
11/1/2010 1:39 AM EDT
Woa! Big Blow to Xilinx and Altera. The only reason I could see for Intel to do this is an Acquisition. As TingLu pointed out, Intel would have already reached an agreement with Achronix that they would buyout the FPGA startup later on. But then the question is why would Intel want to get into FPGA business!! Already they have many things to figure out with McAfee and Infineon WLS acquisition.
Sign in to Reply
resistion
11/1/2010 11:38 AM EDT
Yeah I think going foundry was inevitable path for Intel, as it sees its bread and butter PC business model overwhelmed by apple-driven smartphone business model. Samsung also found it inevitable to relieve itself from the virtually suicidal memory market death spiral.
Sign in to Reply
dylan.mcgrath
11/1/2010 12:00 PM EDT
Great comments guys, thanks. I can't figure this one out myself. You have raised some issues I considered, but I have questions about both.
@eewiz- If Intel's motivation here is to be back in the FPGA business, why do a strategic agreement and not just an acquisition? My first thought was that this was the first step toward an acquisition, but then I wondered, why not just pull the trigger? As Achronix CEO John Lofton Holt pointed out, if Intel wanted to be in the FPGA business, they would be. They would buy Achronix and/or another FPGA startup with (what is for Intel) loose change. So if that's the goal, why wouldn't they just do it?
@resistion- In my opinion, this is more about the strategic importance of programmable logic than it is about Intel going into the foundry business. I think if Intel wanted to go into the foundry business, the company would make a big PR push and announce a division called Intel Foundry Services or something like that. If this was just about Intel generating some extra revenue through foundry work, I just don't think they would start with Achronix, which is, for all of its promising technology, a pretty small player. If Intel wanted to go foundry, there are dozens of other chip companies they could have started with that would generate more revenue. (For that matter, why not start with Xilinx or Altera?)
My bet is this (my colleague Brian Fuller first suggested this to me and I think he is right): This is the first step to an eventual offering that pairs programmable logic with an Atom core in the mold of Actel's SmartFusion and the forthcoming Xilinx chips, both of which offer programmable logic and a hard ARM core.
Sign in to Reply
resistion
11/1/2010 1:10 PM EDT
Indeed, Intel is testing waters in both aspects, FPGA and foundry service provider. It could have worked with Achronix from a distance, letting them use TSMC as foundry, or completely acquired Achronix, as mentioned well.
Sign in to Reply
dylan.mcgrath
11/1/2010 2:15 PM EDT
More info on this: Earlier today I was told that Achronix laid off most of its sales force about a month ago. I spoke to Achronix about this and John Lofton Holt was very clear that Intel is a manufacturing partner only and will not be doing any sales or marketing of Achronix products as a condition of this agreement. Holt also said that within the past 45 days Achronix did lay off several sales people. The reason Holt gave is that the company had been preparing to start selling 28-nm parts in the near future (and was staffing up for it). But when the Intel deal came through, Achronix scrapped 28-nm altogether and instead is preparing to be selling 22-nm parts in about a year. Essentially, they don't need all of those sales people right now because they won't have their new products in the original timeframe. Makes sense to me.
Sign in to Reply
jg_
11/2/2010 6:12 AM EDT
Some brave calls there. Why defer income ?
Unless they have a very large customer who must have 22nm. A customer with enough clout, to 'encourage' intel.
Sign in to Reply
SZA
11/1/2010 2:17 PM EDT
Indeed Industry is finding growing co-importance of Processors and FPGAs for future. And both FPGA vendors and Processor vendors are getting in to it. FPGAs jumped before with hard ARM and now Intel is investigating other way around (perhaps soon we see some stuff from AMD too in that direction). I agree with you Dylan and resistion. Intel is testing different ideas. Few weeks back we saw the Stellarton SIP with Altera, now a more direct sign with Achronix.
Sign in to Reply
shirazkaleel
11/1/2010 3:37 PM EDT
I don't think it is just about FPGAs. Achronix has important asynchronous logic IP that Intel could use in its chips so it could go clockless and thereby improve performance, but they'd have to learn about it and play with it. This would give them an opportunity to do so and overcome the "not invented here" syndrome that may exist amongst its own designers. Designers have shied away from using Asynchronous Logic even though the theory and methodology have been available now for some time; but Achronix has provided a trojan horse by which designers can be tempted to use the technology, the improved throughput being a big incentive.
Sign in to Reply
gutiea
11/1/2010 3:56 PM EDT
I see a clear move towards Foundry Services. The China Fab was the first move in that direction. This announcement is the second. I guess Globalfoundries will encounter even more headwinds and TSMC beginning to see the ascend of a real competitor. Great move!
Sign in to Reply
FPGA_BRAD
11/1/2010 3:57 PM EDT
This is interesting since at IDF they announced a deal with Altera to put an Atom and FPGA in the same package.
"The PC giant also announced Stellarton, a system-in-package that pairs an Altera FPGA with an Atom to enable reconfigurable systems."
Sign in to Reply
fundamentals
11/1/2010 5:16 PM EDT
I think Intel missed a much bigger prize by only a month. They should have bought Actel instead, if their interest in FPGAs is genuine and long term. MicroSemi beat intel to it and scooped up Actel, quite cheaply I might add. No, MicroSemi does not compete with Intel, but this took Actel out of the market as a potential partner. So, Intel had to settle for a start-up, albeit a promising one.
Sign in to Reply
M.O.E.
11/1/2010 6:52 PM EDT
a couple fine points:
Intel did have an ASIC operation built around the 8051 and the 82xx perphieral series back in the 80's.
Intel has also done silicon manufacturing for several companies in the past...just not well advertised:)
Sign in to Reply
hm
11/1/2010 7:00 PM EDT
It will be very nice to have multiple Intel hard wired processor cores inside Achronix FPGA. Achronix and Intel will be able to provide Soc solution to mass consumer and industrial market. Desinger in embedded systems design will have significant advantage with this development. We eagerly await how soon they pur these new parts in market
Will Apple employ these new designs in their novel products?
Sign in to Reply
paul.delamusica
11/1/2010 7:25 PM EDT
Maybe it is just a capacity filler that Intel can turn on and off asynchronously.
Sign in to Reply
goafrit
11/1/2010 7:53 PM EDT
Intel is after Achronix not because of their asynchronous based FPGA, but future development on their product offerings. Intel will buy them to remain competitive because the product driven by asynchronous sequence is delay insensitive and will surely be the right roadmap to keep Moore's law going. This is Intel holding their future.
Sign in to Reply
fatbrain_
11/1/2010 9:41 PM EDT
Intel could be after Achronix IP for an Intel asynchronous processor, the only way to beat ARM?
Sign in to Reply
AliNS
11/2/2010 12:29 AM EDT
Intel couldn't care less about the capacity Achronix might "potentially" bring to the party if it ever does. As a matter of fact Intel probably couldn't care less about the capacity an Altera or Xilinx would be able to drive either. Also, if Intel wanted to get into the FPGA business it could easily buy Altera or Xilinx outright. There has to be something else. The processes Intel develops aren't really optimized for general purpose logic nor FPGAs that require as many metal layers as possible for density optimization. The only angle there may be is for max performance with min power. Yet, it is not clear that Achronix really demonstrated any meaningful performance while minimizing power...to be followed.
Sign in to Reply
jg_
11/2/2010 6:06 AM EDT
This little detail, may be important :
["Because of Intel's considerable U.S. manufacturing presence, Speedster22i will be the only commercial FPGA family manufactured in the U.S., Achronix said."]
Sign in to Reply
Doug the FPGA guy
11/2/2010 6:42 AM EDT
I am cursed with a long memory.
I recall that Intel was the main fab for Altera in the early days (mid-80s) and even had second source rights and competed with Altera for EPLD business. In the mid-90s they SOLD their PLD biz and own R&D (some of it very good too) to Altera.
Maybe others in Intel are also cursed with long memories and miss the old days; it could be just a hobby. Does such a small announcment have to fit into a strategy?
Sign in to Reply
TingLu
11/2/2010 11:11 AM EDT
I will agree what dylan.mcgrath. There is really no point for Intel to get back to FPGA business to compete with Altera and Xilinx. I doubt it will be a winning battle for Intel even it has stack of cash in hand. Intel just had no good tracking record to win business from a mature market not related to CPU business. Their CPU centric view limits their vision and development strategy. Plus, it is not much of silicon anymore in FPGA business. Software becomes so important in this business. It will takes Intel years to catch up. I doubt Intel will go that route.
What my guess is that Intel is eventually going to either license this IP or buy Achronix out. At Achronix size, buyout might be a good choice. They will NOT roll out FPGA. But Intel will add FPGA as co-processor next to its x86 processor to improve its programmability and power. To make support simple, Intel might provide tailor different sets of FPGA program to different markets so end user doesn't have to worry about FPGA at all. This will help Intel target multiple markets with single tape out and not worry about all FPGA tool supports. Xilinx and Altera licensed ARM CPU to put next to their own FPGA. Intel is just starting opposite direction but end products will be similar.
Sign in to Reply
hm
11/7/2010 9:29 PM EST
May be Intel is enticing Apple Inc to come back to them for their future need for manufacturing. Apple designs and Intel foundry may be very good synergy.
Sign in to Reply
dnenni
11/8/2010 12:23 AM EST
Out of the five possibles reasons for Intel opening up their 22nm fab I pick #6:
http://wp.me/psH60-25o
D.A.N.
Sign in to Reply
TJones
11/8/2010 9:25 AM EST
Of all the conjectures, like DAN's best, but haven't seen anyone mention Intel's new direction into....security. Intel is crazy like a fox, here.
Intel stands to make $$$ if it can tout a rock-solid "silicon to software" security solution. It just bought a firm for software, what about the former -- silicon?
Huh? "Secure Silicon"? To a US Govt Security type, "secure" means "Made in USA".
And these days, the biggest customer in the American market is the government's security apparatus.
I see dollars and sense in this Intel move.
Sign in to Reply
patrick38
11/8/2010 11:30 AM EST
I think that the move of Intel is this direction is very strategic. Security might be a reason but I believe it is by far not the main one. WHen you look at the market trend and the evolution of process technology it is clear that 22nm will cost a lot in dev and will need very high volume to make it economically viable. One solution is to create a scale effect in markets where it doesnt't exist, and for this FPGA is a good soluton to make one die fitting several applications. On top of this Xilinx already announced avalability in 2011 of FPGA with 2 cortex A9 inside meaning chips which will be perfectly suitable to be used in mass market application as at hte end the gap in cost between a pure asic approach and a FPGA could not be so high with a much faster time to market for FPGA approach. Taking all of this into acount I wouldn't be surprise that the next step will be availability of FPGA with Atom inside from Achronix in order for intel to get ground on this market. From a pure strategic point of view I think as well that Intel shouldn't buy Achronix at leat for now as it is better for them to be seen as a foundary instead of front competitor to Xilinx and Altera.
Sign in to Reply