Intel To Buy Altera For $16.7 Billion

An anonymous reader writes: Today Intel purchased chipmaker Altera for $16.7 billion. This follows another huge purchase in the semiconductor industry last week, when Avago snapped up Broadcom for $37 billion. This has been a record year for consolidation within the industry, as companies struggle to deal with slowing growth and stagnating stock prices. Altera had already rejected an offer from Intel, but shareholders pressured them to reconsider. "Acquiring Altera may help Intel defend and extend its most profitable business: supplying server chips used in data centers. While sales of semiconductors for PCs are declining as more consumers rely on tablets and smartphones to get online, the data centers needed to churn out information and services for those mobile devices are driving orders for higher-end Intel processors and shoring up profitability."

So, what's the plan?

[fuzzyfuzzyfungus]

Given that FPGAs are big, slow, and hot compared to equivalent logic built as a fixed function chip(but with the obvious benefit of not being fixed function), Altera FPGAs manufactured on the fanciest processes available seem like a fairly obvious product of the acquisition.

Any bets on what other purposes they have in mind? FPGAs with one or more QPI links built in, for fast interconnect with Xeons? Xeons with FPGAs on die? Intel NICs with substantially greater packet-mangling capabilities, at full wire speed, thanks to reconfigurable logic?

Merely producing FPGAs on a nice process is logical; but could also be done just by selling them fab services. They presumably have a plan that goes beyond that.

Re:

[Luthair]

I seem to recall that Intel had been selling excess fab space to Altera (along with a few other non-competitors) a few years back.

Re:So, what's the plan?

[Anonymous Coward]

FPGA's can out perform Pentium i7s in certain scenarios. Here is a video showing how a 200 MHz FPGA can perform a discrete wavelet twice as fast as an 2.6 GHz i7.

https://www.youtube.com/watch?v=Er9luiBa32k

The future will be offloading certain tasks to the FPGA, as well as providing downloadable modules that will allow any PC to take on a wide range of roles. One application could be an SDR without any additional hardware, or a data acquisition unit. So, this is all about flexible I/O and optimised processing.

Re:

[Luthair]

When will the Pentium i7 be released? ;)

Re:

[tepples]

Probably at the next process shrink, once tomorrow's Pentium is as fast as today's Core i7.

Re:

[edmudama]

Memory processes in FPGAs aren't very good. You'd have to really be desperate for a few hundred megabytes of extra memory.

Use FPGA to compress a swap file

[tepples]

Would an FPGA be viable as a lossless data compression coprocessor? That way, the computer could swap to a RAM disk and save space on whatever's currently paged out. It'd be like a hardware-accelerated version of zram (or Connectix RAM Doubler before that).

Re:

[tepples]

The point of zram is to take the memory latency hit while continuing to run at far better than HDD speed. I was guessing that an FPGA would reduce the memory latency hit compared to gzipping each swapped-out page in software.

Re:So, what's the plan?

[Lunix Nutcase]

Intel announced Xeons with FPGAs last year.

Re:

[Andy Dodd]

IIRC, wasn't that in partnership with Altera?

Intel and Altera have been partnering on things for quite a while, including Altera being one of the first companies to be allowed access to spare Intel fab production capacity.

Re:

[Lunix Nutcase]

I don't think anyone said specifically but that's what everyone assumed.

Re:

[erice]

Given that FPGAs are big, slow, and hot compared to equivalent logic built as a fixed function chip(but with the obvious benefit of not being fixed function), Altera FPGAs manufactured on the fanciest processes available seem like a fairly obvious product of the acquisition..

That was going to happen anyway. Atlera announced [altera.com] that they were going to manufacture on Intel's newest process back in 2013.

Re:

[Anonymous Coward]

>Xeons with FPGAs on die?

Intel already did it and the speculation was that it was with Altera.

http://www.extremetech.com/extreme/184828-intel-unveils-new-xeon-chip-with-integrated-fpga-touts-20x-performance-boost

Moore's law is basically running out of steam in a variety of ways so everybody is looking for architectural changes that will allow a way to squeeze more performance out of the same number of transistors. There's been a lot of work in working out how to use field programmable logic to accelerat

Re:

[Svartalf]

They're big and slow compared to an ASIC, yes. But the thing is, they're not big and slow overall- they're reconfigurable and you can dynamically change the logic (Witness Altera's OpenCL offering on the higher-end stuff they offer... You don't offer that unless you're competitive with GPUs...) on the fly. They have a place and it's not always custom logic. It's adaptable custom logic- which ASICs **CANT** do. CPUs are slow and plodding in many of the tasks you're talking about

Re:

[asliarun]

Sure, all you need is another $3,000 software package (and
another $2,495 per year to keep it up to date) to let you do
anything with the FPGA... no problem, right?

Everyone has that kind of cash laying around for every box
they own!

Oh, you forgot $10k in annual support cost.

That's absolutely not a problem when you are running computational workloads for your business that has millions riding on it. Doesn't even have to be mission critical stuff. Even for regular analytics (never mind the "big-data" buzzword). $6k for a significant performance boost (even for specific workloads) and reconfigurability is a piddly amount.

Re:

[Anonymous Coward]

Sure, all you need is another $3,000 software package (and
another $2,495 per year to keep it up to date) to let you do
anything with the FPGA... no problem, right?

Everyone has that kind of cash laying around for every box
they own!

A fixed $3,000 that's depreciable and a fixed cost of $2,495 annual support? Relative to the $150,000+ annual cost for an engineer who knows how to use it, that's basically nothing.

L2Business, kid.

Re:

[Megol]

Strange then that the FPGA market is growing and the number of ASIC design starts are shrinking per year. It is almost like FPGAs are increasingly being used as ASIC replacements in both the high end (backbone routers, RADAR equipment, ASIC simulation) and the low end (replacing misc logic and often a microcontroller). Maybe because they are.

Re:

[unixisc]

Strange then that the FPGA market is growing and the number of ASIC design starts are shrinking per year. It is almost like FPGAs are increasingly being used as ASIC replacements in both the high end (backbone routers, RADAR equipment, ASIC simulation) and the low end (replacing misc logic and often a microcontroller). Maybe because they are.

The GP provided the answer - markets where it's not cost effective to build an ASIC to solve a problem (not high volume enough). There are plenty of those markets, and unless those products ramp up to justfiy an ASIC tape-out, they'd remain in FPGAs. Far more cost-effective. Also, FPGAs allow one to iron out any bugs that may be discovered over time in production.

Re:

[TheRaven64]

My guess would be coarse-grained reconfigurable architectures. Altera FPGAs aren't just FPGAs, they also have a load of fixed-function blocks. The kinds of signal processing that the other poster talks about work because there are various floating point blocks on the FPGA and so you're using the programmable part to connect a sequence of these operations together without any instruction fetch/decode or register renaming overhead (you'd be surprised how much of the die area of a modern CPU is register rena

Re:

[Anonymous Coward]

What do they have in mind? They can finally realize technology from 15 years ago! See: http://www.jhauser.us/publications/2000_Callahan_GarpArchAndCompiler.pdf

In all seriousness, yes, we will see CPUs + FPGAs in the same package and eventually on the same die. Altera's next FPGA is slated to work at 700MHz--1GHz with an extreme amount of off chip IO and memory bandwidth, so it should be suitable for doing high speed packet processing + computation. This is what Tabula was supposed to do, and, surprise s

Re:

[gwjgwj]

CPUs + FPGAs in the same package is nothing new. Virtex II-pro in 2004, currently Zynq.

Re:

[Bengie]

CPU+FPGA in $800 gaming rig sounds completely new to me.

Re:

[unixisc]

Making FPGAs in the first place gives Intel a much broader market throughout the semiconductor industry, since anybody making logic circuitry at the semiconductor level can have it manufactured by them. This makes it more stable for Intel than selling Altera their fab services. That way, Intel can deal w/ a wider variety of customers, rather than Altera being their distribution point.

Re:

[GuB-42]

I guess that by including a FPGA in their chips, they could completely get rid of the southbridge/PCH and connect all I/Os directly to the chip. As a result, motherboards will only need to take care about the electrical and physical connections and provide a program to the FPGA for the logic.

Re:

[TheRaven64]

You might want to look at the papers from MS Research on their use of FPGAs in their data centres.

Tricked

[Anonymous Coward]

Altera said "pay us $1000 this month, $2000 next month, and so on for 2 years, doubling each month." Intel thought it was a good deal and accepted before doing the math.

Re:

[Anonymous Coward]

Oh they did the math.... on a Pentium 60.

Takeover Takis

[Impy the Impiuos Imp]

So another computer guy is getting fat on chips?

Layoffs

[Anonymous Coward]

Expect layoffs after the merger. Cost cutting is almost always done after mergers. What kills me is that many of those engineers that will get canned will be considered damaged goods in today's job market: unemployed == doesn't have the skilz.

So, if you are at Altera, start looking for another job ASAP before the mass layoffs happen. It happened to me and when all of us from the layoff hit the job market at once, it became very very difficult. And it was an interesting coincidence that most of us were s

Re:

[TooManyNames]

The $750 million of "annual run rate synergies" and "track record of rapid deleveraging" should give you a sense of how much and how fast Avago will slice from Broadcom:

http://investors.avagotech.com... [avagotech.com]

In my experience, higher-ups for mergers like this aren't afraid to cut until it hurts, then hire back later (if absolutely necessary).

Avago's past practices

[Taco Cowboy]

Disclaimer: I'm one of the early investors of Avago

As far as I know Avago does not carry out the "cut until it hurts" routine

I know the style of Hock Tan, the CEO of Avago --- and from past experiences (from the merger with LSI, et al) the 'cut' were mainly of low level, ie, disposable personels, while key people - those who have been identified to have contributed in key technologies - were often offered plumb hike in salary / stock option to get them to continue to perform

In other words, what Hock Tan l

Re:

[TwoEyedJack]

Actually, I would only expect to see layoffs in areas like accounting and operations. The core engineering between a microprocessor and FPGA gates are quite different. The ecosystems are radically different, as are the customer bases. In fact, there is really not much synergy outside of using an FPGA as an accelerator in an enterprise-class server.

Re:

[TheRaven64]

There's some overlap. Altera FPGAs have lots of fixed-function blocks on them, ranging from simple block RAMs to fast floating point units. There's a good chance that Intel could reuse some of their existing designs (which, after all, are already optimised for their manufacturing process) from things like AVX units and caches on x86 chips. A lot of the FPGAs also include things like PCIe, USB, Ethernet and so on controllers. Again, Intel makes these in their chipset division and, again, they're optimise

Conflict of interest

[erice]

Altera has many customers who compete with Intel. They are not going to want to deal with Altera anymore. Instead of having Altera as a strong #2, Xilinx is going to own the FPGA business. Good for Xilinx, bad for everyone else.

Re:

[Svartalf]

Not a conflict of interest. Just that a competitor just bought your supplier. Big difference. It's a problem that you need to find a new supplier. The drawbacks with FPGAs is that there's nothing other than your sole supplier is just that. You can't readily or easily swap out the FPGAs like you can SoC's in the ARM or MIPS space- or like RAM or eMMC's. There's a bit of "standard" and "open" involved with things there. I consider it necessary evil to be using them be

Re:

[Svartalf]

Biggest problem THERE would be that they'd have to open up the X86 kimono a bit more than they'd really want to do that with NIOS.  I won't be surprised in one way (your meaning of the situation) if they do it and surprised all the same- because they're giving stuff out that can be more readily reverse engineered through the tools, etc. that people would get as a result of that decision.

So 2^24 dollars?

[aybiss]

What was the exact amount?

Intel, Altera, Xilinx and Free Software?

[AndreyFilippov]

We are using Xilinx FPGA for 13+years, when first choosing between Xilinx and Altera the decisive factor was the license terms - while both were proprietary, Xilinx zero-cost software had no expiration, while Altera's one had only 90 days. Our products are based of Free Software/Open Hardware (licensed under GPLv3 and CERN OHL) so it is critical for us to avoid expensive tools as our users would have to use them too just to be able to rebuild the executable image (bitstream in the case of the FPGA) from the