Intel to Manufacture ARM CPUs for Altera

[krumme]

Jarred take on it:
http://www.anandtech.com/show/7473/intel-opens-fabs-to-competitor-ip

Wow. I am speechless. If that is not news i dont know what news is.

http://www.gsmarena.com/intels_holdout_against_making_arm_chips_comes_to_an_end-news-7097.php

 

[ShintaiDK]

Why is it news? Because someone found out Altera uses ARM in their FPGAs?

Intel seems to have gotten all the high money customers. Altera, Achronix and Xilinx.

 

[Phynaz]

Read: US Military.

 

[seitur]

Why is it news? Because someone found out Altera uses ARM in their FPGAs?

Intel seems to have gotten all the high money customers. Altera, Achronix and Xilinx.

Yeah. I don't get the buzz either.Those companies always chase bleeding edge processes, they can accept low-yields and can pay high price. Intel currently have best process in whole indust, so it grabbedthose high margin client.

 

[Rakehellion]

Yet they still use 5400 rpm drives, sold Core 2 Duo based machines right up until mid 2011 and used Intel GMA 950 :whiste:

7200 RPM drives are awful for laptop and Every MacBook Pro has had a discrete GPU so I'm going to file this under "you don't know what the hell you're talking about."

 

[krumme]

Yeah. I don't get the buzz either.Those companies always chase bleeding edge processes, they can accept low-yields and can pay high price. Intel currently have best process in whole indust, so it grabbedthose high margin client.

So with the best process did you expect Intel to start with low margin customers? Makes no sense. Ofcource they go for that segment.

You dont understand the buzz?. Intel producing arm whats the big deal??
Forbes and the rest of the internet dont understand this is business as usual.

 

[Exophase]

There really is no news, but Intel and ARM have been trumped up as dead-on rivals for so long that something like this is looked at as scandalous.

In other non-news, AMD-owned SeaMicro is still selling Intel-based servers.

 

[Khato]

The single point of interest in this announcement is that it implies a certain level of maturity in the design tool set that Intel's foundry operations provides to their customers. Part of the reason for going after FPGA customers initially was the 'simplicity' of the design. My impression at least is that synthesizing an A53 core into a physical layout is far more involved than your typical FPGA.

 

[Exophase]

The single point of interest in this announcement is that it implies a certain level of maturity in the design tool set that Intel's foundry operations provides to their customers. Part of the reason for going after FPGA customers initially was the 'simplicity' of the design. My impression at least is that synthesizing an A53 core into a physical layout is far more involved than your typical FPGA.

Why would Intel need to make new tools so third parties can design CPUs for processes that have been made with CPUs in mind for decades? Or put another way, if Intel can design processors for their nodes why can't Altera? Intel would have simply made the same tools, libraries, etc they've been using available to other parties. If anything may have increased it'd be Altera's familiarity with these tools.

 

[seitur]

So with the best process did you expect Intel to start with low margin customers? Makes no sense. Ofcource they go for that segment.

You dont understand the buzz?. Intel producing arm whats the big deal??
Forbes and the rest of the internet dont understand this is business as usual.

What? You completly misunderstood. Read again _ i've agreed with Shintai and I do agree with you.
I said the opposite- that Intel starting with FPGA is like you said is preety normal and expected.

 

[krumme]

There really is no news, but Intel and ARM have been trumped up as dead-on rivals for so long that something like this is looked at as scandalous.

In other non-news, AMD-owned SeaMicro is still selling Intel-based servers.

Scandalous? Whats that for a weird strawman?

Its Intel reducing future risk by getting into the arm business in due time using their process knowledge as leverage. Makes perfect sense. And could be good busines for many including Intel.

Buts its a damn huge step strategically.

Would this happen if armv8 was less than stellar?
Would this happen if a57/a53 was average arch?
Would this happen if apple a7 was slow?

Imho no. Its also signs where things is heading. Like seamicro was. And seamicro was big news too. But not anything this caliber.

Who knows perhaps its also signs that eg sgs5 with 20/16nm a57 is comming early spring and the general trend that arm is on a roll now?

Intel is going into arm because they think there is a market.
They do it now.

If anyone had predicted that 3 year ago i would have laughed my head of.

 

[Khato]

Why would Intel need to make new tools so third parties can design CPUs for processes that have been made with CPUs in mind for decades? Or put another way, if Intel can design processors for their nodes why can't Altera? Intel would have simply made the same tools, libraries, etc they've been using available to other parties. If anything may have increased it'd be Altera's familiarity with these tools.

Guess I didn't phrase that too well. The design tool set that would be provided most likely still uses industry standard software, but it's a matter of ensuring that all the collateral, standard cell libraries and the like, that third parties will need are present/creating them when requested. This is, of course, relatively easy when it comes to a straight FPGA, but gets to being more of an issue with complex designs. It's not just a straight matter of plug and play. (I believe such has been mentioned on a few of the quarterly reports during Q&A.)

 

[bullzz]

@krumme - although armv8 would have got intel worried, I dont think this is a reaction to armv8/apple a7. altera announced they will be using intel fabs way back in feb'13

 

[krumme]

@krumme - although armv8 would have got intel worried, I dont think this is a reaction to armv8/apple a7. altera announced they will be using intel fabs way back in feb'13

Agree its not a direct response to that and i know they announced it back. Its a gradual step and we dont know how long Intel will go in the future. If more at all.

I just try to put this in context. Intel is diversifying and thereby reducing risk - no matter if this is the only step or more will come. Sound strategy no matter what happens. Otellini would love the move.

I just try to put this into a larger context - (and i know i look at this from a different perspective than many in this forum). And if it wasnt for that larger context eg armv8 apple a7 perf arm on a roll, there wouldnt be so much noise. If this was only about technicalities eg maturity of tools it would be it. But it obviously isnt. It is tecnicalities and plenty more too. Dont underestimate what is happening very fast now. Look back 3 years! Its crazy how fast its going imho.

 

[krumme]

Quark is another step in the same direction businesswise imho.

All that complaining about quark with 486 talk is off. Its Intel using their competences, process knowledge, tools and superior documentation support and sales as a single solution provider.

= new area of business & diversifying = less risk

 

[Nothingness]

The single point of interest in this announcement is that it implies a certain level of maturity in the design tool set that Intel's foundry operations provides to their customers.

There's another point of interest: Intel probably couldn't propose anything competitive to Altera in place of the Cortex-A53, even though it's their own process.

 

[Exophase]

Scandalous? Whats that for a weird strawman?

Does it matter if I used a bit of hyperbole to describe something how other people are overreacting to and misinterpreting something?

Its Intel reducing future risk by getting into the arm business in due time using their process knowledge as leverage. Makes perfect sense. And could be good busines for many including Intel.

Buts its a damn huge step strategically.

Would this happen if armv8 was less than stellar?
Would this happen if a57/a53 was average arch?
Would this happen if apple a7 was slow?

Imho no. Its also signs where things is heading. Like seamicro was. And seamicro was big news too. But not anything this caliber.

Who knows perhaps its also signs that eg sgs5 with 20/16nm a57 is comming early spring and the general trend that arm is on a roll now?

Intel is going into arm because they think there is a market.
They do it now.

If anyone had predicted that 3 year ago i would have laughed my head of.

Intel isn't getting into the "ARM business." They're allowing an established customer to use ARM in their design for a product that doesn't remotely intersect with any of their own. This is not Intel reacting to anything whatsoever. This has nothing to do with Apple's A7 or ARM at all for that matter, of course it would still be happening if A7 never came out or A7 sucked - why would that have impacted Altera's decision to use Cortex-A53? One that was probably made long before anyone knew anything about A7, even that it'd use ARMv8.

Quark is another step in the same direction businesswise imho.

All that complaining about quark with 486 talk is off. Its Intel using their competences, process knowledge, tools and superior documentation support and sales as a single solution provider.

= new area of business & diversifying = less risk

No, it's Intel using marketing hype to draw attention to themselves over a lousy and nigh pointless product that they probably spent close to nothing developing.

 

[Exophase]

Guess I didn't phrase that too well. The design tool set that would be provided most likely still uses industry standard software, but it's a matter of ensuring that all the collateral, standard cell libraries and the like, that third parties will need are present/creating them when requested. This is, of course, relatively easy when it comes to a straight FPGA, but gets to being more of an issue with complex designs. It's not just a straight matter of plug and play. (I believe such has been mentioned on a few of the quarterly reports during Q&A.)

It's a huge assumption on your part that Intel developed anything at all for this instead of placing the burden on customers in adapting to their existing tools and libraries.

IBM, for example, is pretty well known for making their foundry customers use their tools, which is probably one reason why they don't get that much business. But it's still a non-zero amount.

There's another point of interest: Intel probably couldn't propose anything competitive to Altera in place of the Cortex-A53, even though it's their own process.

That's not news, that's common sense

 

[Khato]

It's a huge assumption on your part that Intel developed anything at all for this instead of placing the burden on customers in adapting to their existing tools and libraries.

Doing a quick search there are indeed a few tidbits pointing in that direction though.

From http://www.engineering.com/Library/...leID/5561/Intel-Custom-Foundry-Interview.aspx

What I took from the discussions with Messrs Rikhi and Hu is that the foundry strategy is well-developed with capabilities planned and executed over the last several years. This is where Altera corroborates the story. Intel’s design enablement for foundry customers has been in the works ever since about 2009.

From http://www.brightsideofnews.com/new...sis-intel-all-but-confirms-foundry-plans.aspx

But the CEO also remarked that the possibility of offering foundry services in the future is clearly there. They use the contemporary fab contracts as learning opportunities, to build up development libraries and tools to enable rapid chip design and modularity which will be required for their SoC business in the future.

And then the best evidence, here's a job listing that's pretty much describing exactly what I said - http://www.jobsinminneapolis.com/j/...-corporation-l-phoenix,-az-jobs-j3440393.html

This is for a position in the ICF (Intel Custom Foundry) division of TMG managing the development and support of Standard Cell Libraries used by Intel's foundry customers.

 

[krumme]

Go for it Exophase ! Its as good as a position in samsung to learn the arm world.

 

[NTMBK]

Well, he has pulled various "Seronx" in the past, like the "no APUs in consoles" one, really fun stuff.

Oh come on, that's totally different. "No APUs" was ShintaiDK's guess and prediction, and he never claimed that it was anything else. Seronx just comes out with complete fabrications and presents them as if they are gospel truth. Everyone makes bad guesses from time to time.

 

[NTMBK]

Read: US Military.

This. Small production runs, extremely high performance FPGAs on the best manufacturing node in the world? Sounds like some fighter jet is getting a new electronics package.

 

[mrmt]

This. Small production runs, extremely high performance FPGAs on the best manufacturing node in the world? Sounds like some fighter jet is getting a new electronics package.

Not fighter I think, but maybe specific hardware in C5I stuff.

 

[Nothingness]

You also need that kind of FPGA for emulators (where "emulation" is the term used in EDA for hardware-based RTL/gate simulation).

 

[roob]

this deal was announced back in february and obviously had been negotiated well before that. So anyone claiming that this Intel reacting to Apple's A7 or anything else recent is simply wrong.

http://newsroom.intel.com/community...nce-fpgas-on-intels-14-nm-tri-gate-technology