Intel-Altera, again

You probably have a few in your house.

Windows, and Windows programs, have to run on all of the Intel and AMD machines. That must be a nightmare.

The point is not that FPGAs show up in some phones, the volumes don't make a blip on the FPGA dollar radar screen. Their market is still largely comms and that is not expected to change.

I think Intel is more concerned with selling more CPUs by leveraging the FPGA advantage than they are selling more FPGAs with a CPU connection. I'm hoping someone in the middle of this will find a way to justify a more rational CM4 type CPU in the FPGA device. Nothing wrong with a CA type CPU, but the horsepower is often not needed and there are more and more low power apps looking for such a device.

Microsemi has such a chip, but they are not mainstream enough to market it I believe. It has languished for quite awhile. I can't even remember the name of the FPGA company they bought.

So this is about a lecture one guy Aswath Damodaran gives. Why should I believe him over anyone else's opinion?

It's back. And I think I know now how to fix it if it happens again.

They don't sell so many as CPU chips like ARMs do, but I understand they are used in a *lot* of comm ASICs. I seriously doubt they are near ARM or x86 volumes though. I'm a bit surprised Xilinx picked an ARM rather than the MIPS given their customer base. I'm guessing they found their customers would not reject the part because it had an ARM and the non-comms apps have a preference for the ARM. Who knows for sure. Xilinx won't be telling.

Yeah, almost as much fun as trying to run on all the different ARM CPUs.

No worst than running Android (ARM) emulator on x86.

FM is kind of interesting. Band limiting results in AM. But that's easily filtered, so no one cares :-)

Tim

The fact that MIPS architecture is in some CPUs wasn't the issue. The issue was that I don't believe they are anywhere near the top two in CPUs shipped, nor do I believe that Intel only ships 1% of all CPUs.

In that case, you're welcome to do the research and post the numbers.

According to my very rough googling research, Intel sells about 400-500 million, but there are about 15,000 million ARM cores sold a year. MIPS sells around 500 million cores a year. Then there are all the smaller cpus in microcontrollers - I don't think ARM is dominant there. There is are a great many 8-bit (and even 4-bit) cpus made.

I would have guessed that MIPS sold a fair bit more than Intel, rather than just a similar quantity. I suppose more accurate research is needed.

The 1% figure may be right if all CPUs are considered. There are an awful lot of 4 and 8 bit processors shipped. Intel has the big dollar share because the ASP is so much larger, at least 100x I would think, possibly near 100x when all CPUs are considered.

her

ly

anet,

y work

the

the

lure.

ig

ed.

Yes, Intel has more than 90% share of the $1k big iron CPUs, and is increas ing, not decreasing. ARM might have a 20% power advantage, but Intel has a 30% fab advantage. Things might change if and when ARM buyouts Samsung, o r perhaps the other ways around.

We buy an LPC1764 ARM, pre-programmed by the distributor with our code, for about $4 in modest quantities.

It runs a pretty complex ADC/PID/filtering/SPI loop at 100 KHz, and we turned down the CPU clock rate to half max to save power. There are 13 on this board, 12 of them isolated from ground:

Happened across this, but nothing really new:

I've also heard the 1% figure, just looking around the house it makes sense you might have one or two computers with an Intel CPU, but then there's the monitors, TVs, dvd players, microwaveoven, thermostat, alarmclocks, watches, battery powered tools, toys, cellphones, 10's of cpus in a car, etc. etc.

-Lasse

Why would it matter who owns what? Each company is doing what they do best and if ARM were majorly involved in manufacturing ARM CPUs, that would create a problem for many of their customers.

I don't see Intel's fab advantage as being so great other than the fact that it lets them lower their ASP or more importantly make more money. Does their process advantage really give the parts much more speed or lower power? Not that I can tell. I'm not sure CPUs have gotten much faster over the last 5 and possibly 10 years.

Nice analogy, but as you say, nothing we didn't already know.

I don't agree with this, "We can, however, expect to see a discontinuation of ARM-based SoC FPGAs for a line of hybrids based on Atom and Core architectures."

I think that would be foolish of Intel. They only way I can see that happening is if they introduce an x86 integrated FPGA product line and the ARM line withers because the x86 line becomes so popular. I doubt that would happen as there is a lot of accumulated knowledge of using ARM devices in the customer base while there is relatively little for the x86. Although maybe I am wrong about that. My customer uses an x86 mother board running Linux in their products, including the one that uses my boards.

Intel can design chips tuned to their fab, and build fab most suitable for their chip. When you have fabless, you have less integration by definition.

They do have enough of an edge to keep locking customers in. For the same speed or power, chips are getting denser: more cores and most importantly more cache to match them. We just don't really see real competition in production yet.

I think the advantage of designing a CPU architecture tuned to a fab process is pretty minimal. Yes, Intel optimizes their process for their own fabs, but not very many ARMs are made fabless. So ARM vendors can still tune their processes for their fabs as well.

I'm not sure which customers you are referring to. My understanding is that Intel doesn't have many fab customers. On the CPUs the architecture provides speed and being a process behind is not the determining factor in speed. AMD lost the x86 battle because of the inability to keep up with Intel's fab expenditures as the cost of a new fab rose. Not because they couldn't get the speed, but because their costs were higher than Intel's. So now they have switched architectures to one that won't be at such a disadvantage. Smaller die and lower power means AMD can possibly rise from their impending ashes.

I'm not sure the Bear v. Wolf analogy is as accurate as the Bear v. the Bee. Nimble and quick the bee can sting at will, but will never kill a bear. The bear only has to want the honey enough to take the stings.