massive intel lay off

[dannyf]

Not a small design house with 300 people, I wonder if these even exist

Plenty of them actually. What you get with your licensing fee to ARM is not just the right to use the core but also help in customizing the design to your own process, not to mention the marketing value and brand recognition of ARM.

The intel vs. arm fight on the mobile front isn't a fight of product capabilities but a fight of a business model. Intel's business model is one of vertical integration and sole supplier. ARM's business model is one of intermediation and multi-sourcing. ARM basically recognized that there is ***NO*** value in core and the best way to make money is to sell the IP to others so they can integrate their value-add into the final products. In essence, ARM has turned the mcu vendors into their marketing departments.

we will have to see which strategy will prevail - I think on the high-end, Intel's approach has legs and on the low-end, ARM seems to be winning. Either way, fat margins for PC CPUs are likely in the past.

[RGB255_0_0]

Either way, fat margins for PC CPUs are likely in the past.

If the rumoured $1500 (probably between $1000-1200 IMO) for the decacore Broadwell-EP is true then nope. There will be plenty who buy that just for the e-points; and the Xeons still sell like hotcakes to those who need to shave another 30 seconds from their renders.

[AntiProtonBoy]

I used to think gaming could keep PCs going, but not any more. That market has shifted to consoles and game developers seem to be less and less inclined to spend money making the PC versions of their games any better than they have to.

Been on Steam lately? PC gaming has been as strong as ever, if not stronger.

[lukier]

Sorry but to my knowledge the ARM core is just an important part of a uC but the peripherals are very important and difficult also. And they are often proprietary designs.
Therefore you will only see huge firms implementing and building ARM cores to a commercial "cheap" microcontroller product, like NXP, ST, Broadcom to name just a few.
And even these huge firms screw up sometimes (I2C peripheral with ST for example).
Not a small design house with 300 people, I wonder if these even exist and what kind of peripherals they would use?

Not these days. I believe you can do a SoC today with a team of 300 people and some Cadence software + hardware (simulators). You buy particular core license from ARM, ARM also can supply quite a bit of peripherals (check PrimeCell series like PL011, PL022, PL061 etc). If not ARM then Synopsys DesignWare would happily sell you peripheral IPs (e.g. USB controller cores). If you need GPU or video acceleration then ARM has Mali, Imagination has PowerVR and also video decoder core and so on. At this point your job is integrating the whole shebang, simulating it inside out in System C, on Cadence Palladium and FPGA boards (see DiniGroup) and sending the design to TSMC or other fab. Check Allwinner or HiSilicon SoCs, these things are pretty much a collection of off the shelf IP cores (and rightfully so - makes it much easier to port drivers sometimes, sometimes not as even if they wanted to open source their GPU driver their hands are tied, as the GPU is an external IP under NDA).

Of course all of this isn't cheap and that's why NXP, ST, Broadcom or TI etc make their money, for small product runs (< millions), without any special requirements, off the shelf SoC is a sensible solution. Situation changes if you are Apple or Qualcomm. Then you buy the most expensive ARM licence - the right to the architecture - and you design your stuff from scratch, using your preferred silicon process and geometry etc.

[RGB255_0_0]

I used to think gaming could keep PCs going, but not any more. That market has shifted to consoles and game developers seem to be less and less inclined to spend money making the PC versions of their games any better than they have to.

Been on Steam lately? PC gaming has been as strong as ever, if not stronger.

PC gaming might be doing fine, and if you look at AAA titles, sales are quite strong, with indies doing extremely well.

But, if you look at http://store.steampowered.com/hwsurvey?platform=pc there are more people running on a dual core and using HD 4000 graphics than the high end gear. Only 0.9% running a 980 Ti. Not saying Steam Hardware is the best way to gauge sales; when I get asked to participate I decline, but statistically it's the best we've got. So what I would call gaming hardware (AMD 270+; nVidia 960 Ti+; quadcore Ivy Bridge+) doesn't sell as much as Steam suggests.

[Mechanical Menace]

But, if you look at http://store.steampowered.com/hwsurvey?platform=pc there are more people running on a dual core and using HD 4000 graphics than the high end gear. Only 0.9% running a 980 Ti. Not saying Steam Hardware is the best way to gauge sales; when I get asked to participate I decline, but statistically it's the best we've got. So what I would call gaming hardware (AMD 270+; nVidia 960 Ti+; quadcore Ivy Bridge+) doesn't sell as much as Steam suggests.

70% are running DX12 compatible cards, only 25% are using a HD4000 machine. And how many of that 25% are peoples laptops they have Steam installed on as well as on their main machine?

Also how the hardware survey info is presented is hard to properly read. I have Steam on at least 5 machines, the two that get serious gaming use have a 690 and 980tis, the others all have Intel graphics. All my machines will show up at least twice as they all have at least two OSs installed. Just that would give you the misleading impression that over half of users run Linux and only 40% aren't running Intel graphics.

Luckily the "active user" stats go by active accounts and not number of client installs.

[RGB255_0_0]

But, if you look at http://store.steampowered.com/hwsurvey?platform=pc there are more people running on a dual core and using HD 4000 graphics than the high end gear. Only 0.9% running a 980 Ti. Not saying Steam Hardware is the best way to gauge sales; when I get asked to participate I decline, but statistically it's the best we've got. So what I would call gaming hardware (AMD 270+; nVidia 960 Ti+; quadcore Ivy Bridge+) doesn't sell as much as Steam suggests.

70% are running DX12 compatible cards, only 25% are using a HD4000 machine. And how many of that 25% are peoples laptops they have Steam installed on as well as on their main machine?

Also how they present hardware survey info is presented is hard to properly read. I have Steam on at least 5 machines, the two that get serious gaming use have a 690 and 980tis, the others all have Intel graphics. All my machines will show up at least twice as they all have at least two OSs installed. Just that would give you the misleading impression that over half of users run Linux and only 40% aren't running Intel graphics.

Luckily the "active user" stats go by active accounts and not number of client installs.

DX12 doesn't say anything about performance though.

And of those DX12 cards, how many are actually capable? There are a bunch of Kepler cards there that are barely classified as better than the HD4000: GT 620M, 630, GTX 650 is better but that card is only really suitable for games like League of Legends and CS:GO.

The 970, 980, 980 Ti and Titans; GTX 750 Ti and above; and GTX 680, possibly the GTX 670 4GB and their respective slimmed down mobile parts are the only real GPUs that can be considered gaming capable from nVIDIA at least.

[dannyf]

Intel  basically made s strategic mistake. They banked heavily on the PC mkt without realizing or for seeing the shift to small mobile processors.

It is ironic in that they practically invented the mcu mkt and 8051 is still today the most widely used mcu.

Fortunately for arm, Intel became too focused on protecting their desktop mkt to see that there is a brand new world out there.

[Mechanical Menace]

And of those DX12 cards, how many are actually capable? There are a bunch of Kepler cards there that are barely classified as better than the HD4000: GT 620M, 630, GTX 650 is better but that card is only really suitable for games like League of Legends and CS:GO.

The 620M is still more powerful than the PS4 and Xbox One GPU. The PS4.5/PS4 Neo will just about match it according to Sony's own figures. Maybe not capable of running new games maxed out but still capable of giving current consoles a run for their money.


EDIT: Corrected by coppice

[rdl]

If you read carefully you will see that I didn't actually say that PC gaming was dead or dying.

[coppice]

And of those DX12 cards, how many are actually capable? There are a bunch of Kepler cards there that are barely classified as better than the HD4000: GT 620M, 630, GTX 650 is better but that card is only really suitable for games like League of Legends and CS:GO.

The 620M is still more powerful than the PS4 and Xbox One GPU. The PS4.5/PS4 Neo will just about match it according to Sony's own figures. Maybe not capable of running new games maxed out but still capable of giving current consoles a run for their money.

The PS4 is more like a GTX660 in its graphics capabilities. Its really quite big step up from the PS3. I understand the XBox One is somewhat weaker.

[Mechanical Menace]

The PS4 is more like a GTX660 in its graphics capabilities. Its really quite big step up from the PS3. I understand the XBox One is somewhat weaker.

Yeah you're right, I got the wrong figures. Sorry.

[RGB255_0_0]

And of those DX12 cards, how many are actually capable? There are a bunch of Kepler cards there that are barely classified as better than the HD4000: GT 620M, 630, GTX 650 is better but that card is only really suitable for games like League of Legends and CS:GO.

The 620M is still more powerful than the PS4 and Xbox One GPU. The PS4.5/PS4 Neo will just about match it according to Sony's own figures. Maybe not capable of running new games maxed out but still capable of giving current consoles a run for their money.

http://gpuboss.com/gpus/Radeon-HD-7870-vs-GeForce-GT-620M http://www.futuremark.com/hardware/gpu/AMD+Radeon+HD+7870/review

http://www.extremetech.com/extreme/171375-reverse-engineered-ps4-apu-reveals-the-consoles-real-cpu-and-gpu-specs

In no way is the PS4's GPU in any way comparable to the 620M let alone when we take into account the PS4 devs have access to bare metal coding and optimisation to maximise the GPU's performance vs the multiple layers PC game devs have to live with. If we factor in clock speed, it's closer to a 780M.

[miguelvp]

Console vs PC gaming belongs somewhere else other than this thread since Intel was never or had been on the running in the console market and the PC gaming market even if healthy-ish is just a drop as far as Intel is concerned and has been that way for the last decade and a half.

Footprint on data centers is where their potential is at. The consumer part is dwindling I will dare to say that at least 90% of home computing can be fulfilled with cheaper ARM based systems at lower cost and power requirements.

[djacobow]

Intel  basically made s strategic mistake. They banked heavily on the PC mkt without realizing or for seeing the shift to small mobile processors.
...
Fortunately for arm, Intel became too focused on protecting their desktop mkt to see that there is a brand new world out there.

I agree that they made a strategic mistake, but I'm not sure it was for lack of realizing the shift to small mobile processors. They knew it was happening, they just could not figure out a way in that market to make the fat margins that their investors expect. Investors would have pounded them for saying "we're going to this much less lucrative business because it's the future." And as far as I can tell, it _is_ much less lucrative, for all the reasons you have described in other posts. The vendors of ARM parts do not have the differentiation that would command high margins.

Andy Grove had written about a somewhat similar situation in the past, when they were getting crushed in the memory business (previous bread and butter) and made a strategic pivot to processors being their main thing. But even that was an easier transition than this because they were going up market, not down.

The current situation is maybe more like what happened to SGI with graphics cards or the minicomputer companies with PCs. They ran upmarket, but the slice that could command margins got smaller and smaller. This is the "disruptive technology" that Clayton Christensen has made a career talking about (https://en.wikipedia.org/wiki/The_Innovator%27s_Dilemma). I'm not sure there's really a way out for companies caught in this situation, except to make strategic investments in the upstarts and to cash cow their business as long as they can and hope to figure something out before it dries up.

Now they're talking all about IoT, which is kind of interesting, but I don't see how SoCs for IoT won't be a race straight to the commodity bottom.

Note: not everyone believes in the disruptive innovation concept. Here's an interesting critique: http://www.newyorker.com/magazine/2014/06/23/the-disruption-machine

It is ironic in that they practically invented the mcu mkt and 8051 is still today the most widely used mcu.

They also had a lot of opportunities to make something of the MCU market. 196/296, i960, etc. Some of those were fine parts for their day. Intel just had this funny habit of canning embedded processor lines, pissing off their customers immeasurably.

When I was there I worked for awhile on embedded controllers that mostly went into hard drives. The strategy of the entire division was 1) find some productive use for old fabs, 2) make things nobody else is doing well that are necessary to ship computers (computer needs an HD, HD needs a processor, so let's make that). USB came out of that division, too. That's a fine strategy, but it's a little weird, too, in they did not see a meaningful business for embedded stuff in its own right. It was all through the x86 lens. And in fairness, it made them a sh!t-ton money. Definitely catching up to them now, though.

[miguelvp]

Just adding some PC market study links:

http://www.gartner.com/newsroom/id/3280626
http://www.idc.com/getdoc.jsp?containerId=prUS41176916

Seems like these studies reflect both commercial and consumer products, but the market seems to be shrinking.
Not to a horrible halt, but the writing is in the wall so I guess that's why Intel decided to change their strategy.

Not too sure on their IoT vision, but they do need to dominate the commercial and cloud services if they are going to succeed.
FPGAs on chip was part of that effort I guess (the cloud) and the reason they did acquire Altera.

http://www.theregister.co.uk/2016/03/14/intel_xeon_fpga/

[uncle_bob]

Hi
The gotcha with ARM is this:
Say I'm a large user of CPU's. (Fill in the name of any major tablet or phone outfit here). Rather than buying CPU chips from (say) Intel (with an ARM in them), I go out and buy a small design house. Those 30 to 300 people drop an ARM plus a few odd parts into a chip design. They cut out all of the stuff that my tablet or phone has no use for.

Sorry but to my knowledge the ARM core is just an important part of a uC but the peripherals are very important and difficult also. And they are often proprietary designs.
Therefore you will only see huge firms implementing and building ARM cores to a commercial "cheap" microcontroller product, like NXP, ST, Broadcom to name just a few.
And even these huge firms screw up sometimes (I2C peripheral with ST for example).
Not a small design house with 300 people, I wonder if these even exist and what kind of peripherals they would use?
Perhaps if you take an FPGA or something like that, but those are incredible expensive building blocks, a totally different market segment as the standard ARM core uC's.

Hi

ARM will indeed sell you a reasonable set of peripherals. They aren't going to include a graphics sub system that will run the next generation of gaming consoles. There are also a *lot* of people who will quite happily sell you various peripheral designs and include support as well as performance guarantees. If you look at a full blown chip from (say) Freescale and it's complement of peripherals (and multiplexing and clock options and this and that) it *would* be a daunting task for a small group to duplicate. The secret is that they don't duplicate it. There is no need for pin multiplexing, they only have one pinout. Multiple clock subsystems likewise are not needed. They have one clock scheme and it gets throttled in a specific way. They don't need a dozen UART's, a dozen I2C's, a dozen SPI's, and five different types of external RAM. Their CPU still is quite powerful. The chip does just what is needed. Is this "cheating"? In a sense yes it is.

As a reference, this is the outfit I was thinking of as I wrote the post:

https://en.wikipedia.org/wiki/P.A._Semi

At the time Apple bought them, they had 150 engineers on the design team. They were doing well enough in the "fabless CPU" business to have multiple designs under their belt.  They are hardly the only people doing this.

ARM holdings *total* head count is a bit under 4,000 people. Effectively, their entire workforce is < 1/4 of Intel's layoff count.

Does ARM manufacture and test chips? Of course not. Does the 150 guy outfit manufacture and test chips? No they don't do that either. They contract that out to a fab and a test house. There's likely a contract packaging house (or three) in the mix as well.

Bob

[coppice]

Intel  basically made s strategic mistake. They banked heavily on the PC mkt without realizing or for seeing the shift to small mobile processors.

Intel DID realise mobile was the future in the early 90s, and developed a smart phone platform very early on. Recognising needs has not been their problem. Its all about the execution.

Kodak realised digital cameras were the future in the early 70s, as they saw the video camera business develop. They made the first working digital stills camera in the mid 70s. They invested heavily in sensor development, CDRs and other components of the digital camera environment. They did very well in some of these for a time, and they were definitely identifying good places to put their investments. They were not agile enough to stay ahead of things, though. Others passed them by, and they have gone.

Intel looks a lot like Kodak right now. They invested billions to develop a phone platform in the 90s that failed so badly in the marketplace that they wrote off their billions. Xscale was one part of that which hung on for a time, and was the main apps processor in most early 2000s smartphones. They knew what they needed to do, but again they just haven't been agile enough. I assume they are trying to hang on until the market matures sufficiently that innovation slows down and they can keep up. That might be happening right now. This year's phones are not massive improvements on last year's.

[Kjelt]

ARM will indeed sell you a reasonable set of peripherals. .............
The chip does just what is needed. Is this "cheating"? In a sense yes it is.
As a reference, this is the outfit I was thinking of as I wrote the post:
https://en.wikipedia.org/wiki/P.A._Semi 

Thanks for clarifying. 

Intel looks a lot like Kodak right now.

Intel is betting on multiple technologies. Processors is only one part, they also produce flash chips and sell them succesfully in their own brand of ssd's, and they also now seem to have the next generation of memory together with Micron in their grasps: xpoint. If they don't screw it up like they did with Rambus a decade ago this could become a huge new technology.
And with a promise to be 1000 times faster than a ssd and cheaper this could well become a next gen pc upgrade enabler.
http://www.pcworld.com/article/2973549/storage/intels-crazy-fast-3d-xpoint-optane-memory-heads-for-ddr-slots-but-with-a-catch.html
Besides they still make billions of profit, I know dozens of companies that would "kill" for that.

[coppice]

Intel looks a lot like Kodak right now.

Intel is betting on multiple technologies. Processors is only one part, they also produce flash chips and sell them succesfully in their own brand of ssd's, and they also now seem to have the next generation of memory together with Micron in their grasps: xpoint. If they don't screw it up like they did with Rambus a decade ago this could become a huge new technology.
And with a promise to be 1000 times faster than a ssd and cheaper this could well become a next gen pc upgrade enabler.
http://www.pcworld.com/article/2973549/storage/intels-crazy-fast-3d-xpoint-optane-memory-heads-for-ddr-slots-but-with-a-catch.html
Besides they still make billions of profit, I know dozens of companies that would "kill" for that.

Mobile has been Intel's biggest bet, but they put a lot of money into a variety of things, such as electronic toys and home routers. Very little has gone well for them. Although the XPoint work with Micron looks interesting, if you look at recent quarters they have been loosing money in their flash business.

[uncle_bob]

Intel  basically made s strategic mistake. They banked heavily on the PC mkt without realizing or for seeing the shift to small mobile processors.

Intel DID realise mobile was the future in the early 90s, and developed a smart phone platform very early on. Recognising needs has not been their problem. Its all about the execution.

Kodak realised digital cameras were the future in the early 70s, as they saw the video camera business develop. They made the first working digital stills camera in the mid 70s. They invested heavily in sensor development, CDRs and other components of the digital camera environment. They did very well in some of these for a time, and they were definitely identifying good places to put their investments. They were not agile enough to stay ahead of things, though. Others passed them by, and they have gone.

Intel looks a lot like Kodak right now. They invested billions to develop a phone platform in the 90s that failed so badly in the marketplace that they wrote off their billions. Xscale was one part of that which hung on for a time, and was the main apps processor in most early 2000s smartphones. They knew what they needed to do, but again they just haven't been agile enough. I assume they are trying to hang on until the market matures sufficiently that innovation slows down and they can keep up. That might be happening right now. This year's phones are not massive improvements on last year's.

Hi

I worked for Kodak back then ... they most certainly did *not* understand the digital side of things. Their experience was all in building giant building sized machines that ran for decades making product. Once they built the machine, they were set for 40 year in this or that business. The only thing they ever made money on was film and paper. The other stuff was given away at a loss to get people to use the film. With digital, there is a new this or that every two or four years. A fab costs *more* than the machines they were used to and is obsolete in a few years. They also had a couple of leaders who just plain didn't do their jobs (leading the company to success).

====

Intel is in a bit different boat. Their industry (CPU's) has not vanished. It hasn't been replaced by something entirely different. They are certainly seeing the PC market shrink. The whole market ... growing. They need to adjust, not close the doors.

Bob

[miguelvp]

Hardware alone doesn't mean much without the proper marketing.

So how about AMD? A coworker just pointed me to this article:

http://www.pcworld.com/article/3060273/components-processors/a-new-amd-licensing-deal-could-create-more-x86-rivals-for-intel.html

Maybe this was part of the Intel decision, if they knew this was being dealt behind their backs.

[dannyf]

The Chinese deal basically means and wants to take the arm model rather than the Intel model. We will see how it works out. If it is successful, it will hurt Intel greatly given how much of their business is in China.

I wonder about if there are IP related issues.

[TerraHertz]

On quick research this appears to be a 10% employee cut back or so. How does that rank with other 'massive' lay offs of major corporations?

http://www.zerohedge.com/news/2016-04-22/halliburton-fires-one-third-global-staff-what-we-are-experiencing-today-far-beyond-h
Halliburton Fires One Third Of Global Staff: "What We Are Experiencing Today Is Unsustainable"

http://www.zerohedge.com/news/2016-04-21/retailer-bankruptcies-are-hailing-down-us-economy
Retailer Bankruptcies Are Hailing Down on the US Economy

It's probably a mistake to try analyzing Intel's problems entirely from the perspective of machine architecture choices and consumer preferences. Wider economic factors are likely to predominate. If you don't know what I mean, I think by the end of 2016 it should be much clearer to you.

[uncle_bob]

On quick research this appears to be a 10% employee cut back or so. How does that rank with other 'massive' lay offs of major corporations?

http://www.zerohedge.com/news/2016-04-22/halliburton-fires-one-third-global-staff-what-we-are-experiencing-today-far-beyond-h
Halliburton Fires One Third Of Global Staff: "What We Are Experiencing Today Is Unsustainable"

http://www.zerohedge.com/news/2016-04-21/retailer-bankruptcies-are-hailing-down-us-economy
Retailer Bankruptcies Are Hailing Down on the US Economy

It's probably a mistake to try analyzing Intel's problems entirely from the perspective of machine architecture choices and consumer preferences. Wider economic factors are likely to predominate. If you don't know what I mean, I think by the end of 2016 it should be much clearer to you.

Hi

Some companies observe that the headline "firing 10% of workforce" causes the stock to go up. They then seem to then get into a regular process of layoff announcements. For some reason, the process of "headline -> stock price" keeps working. After a while, you sort of wonder - is this real? Back to the numbers and look at their total employment. Oddly enough, despite laying off "10% of workforce" every 3, 6, 9 or 12 months, their total number of people employed keeps climbing > 10% year over year.

Yes, I realize the very concept that headlines *might* be less than the whole story comes as a great shock to all of us

In the context of Intel, the key question is (obviously) - are they hiring a boatload of people to do "the next big thing" at the same time they are moving a bunch of poor sods out?  Indeed, only time will tell. Yes, if they were in the oil business, there would be no need to wonder about the answer.

Bob