TSMC and ARM Tape-Out First ARM Cortex-A57 Processor on 16 nm FinFET Technology

[StrangerGuy]

Yes .. but ... ARM is a much bigger fish than MIPS. Yes ARM will have soon 64bit CPUs for server purposes, that is similar to MIPS, but there are still lots of differences.

Just think about who has licenced the Cortex A57 design and will provide chips:
AMD, Broadcom, Calxeda, HiSilicon, Samsung and STMicroelectronics.

Additionally there are nvidia, AppliedMicro and 2 anonymous companies (probably Qualcomm and Apple) who will design their own ARM64 cores.

Together these companies have a "critical mass" and there will be several, different kinds of SoCs aiming at different market niches. The customer can choose among several solutions and get the best one.

Even Intel will learn to fear that momentum. Over a longer time period, Intel could even be pushed out to the big-tin server / HPC market and compete with IBM/Oracle only. But lets see how their Server-Atom SoC will perform. Intel's always good for a surprise, too, and they have a manufacturing advantage.

The danger that ARM poses stems from the fact ARM has next to nothing to lose and everything to gain by eating Intel's server share. If ARM provides a better overall benefits/price over Intel, customers switch to ARM and Intel loses, but if Intel makes their low-end chips too good to compete with ARM then there is a very real risk they will eat into their own higher-end chips. Lose-lose situation.

 

[Khato]

The danger that ARM poses stems from the fact ARM has next to nothing to lose and everything to gain by eating Intel's server share. If ARM provides a better overall benefits/price over Intel, customers switch to ARM and Intel loses, but if Intel makes their low-end chips too good to compete with ARM then there is a very real risk they will eat into their own higher-end chips. Lose-lose situation.

Not necessarily. If they make their low-end chips too good they simply up the prices to maintain the exact same or possibly better margins than what they get on their core based Xeons.

It's definitely going to be interesting to see how it plays out though. Currently available data appears to show that 32nm A15 improved performance by a marked amount over 40nm A9's while keeping comparable efficiency... and in an ideal scenario Calxeda's 40nm A9 based server offers slightly better efficiency than 32nm Sandybridge Xeons. So how will 22nm Ivybridge Xeons and 22nm Silvermont-based Atom Xeons compare to 28nm A15's? Fun times ahead!

 

[R0H1T]

Not necessarily. If they make their low-end chips too good they simply up the prices to maintain the exact same or possibly better margins than what they get on their core based Xeons.

It's definitely going to be interesting to see how it plays out though. Currently available data appears to show that 32nm A15 improved performance by a marked amount over 40nm A9's while keeping comparable efficiency... and in an ideal scenario Calxeda's 40nm A9 based server offers slightly better efficiency than 32nm Sandybridge Xeons. So how will 22nm Ivybridge Xeons and 22nm Silvermont-based Atom Xeons compare to 28nm A15's? Fun times ahead!

The future indeed looks promising for ARM atleast if these numbers are any indication of things to come ~

 

[Khato]

The future indeed looks promising for ARM atleast if these numbers are any indication of things to come ~

Well, there's no need to compare using those numbers when we now have benchmark data for the Snapdragon 600 thanks to the HTC One review. Which at least implies a marked improvement over the Snapdragon S4 at similar power usage... Though one must remember that the Snapdragon S4 performance wasn't all too impressive to start with - this is merely bringing its performance up to Medfield levels.

 

[Arachnotronic]

The danger that ARM poses stems from the fact ARM has next to nothing to lose and everything to gain by eating Intel's server share. If ARM provides a better overall benefits/price over Intel, customers switch to ARM and Intel loses, but if Intel makes their low-end chips too good to compete with ARM then there is a very real risk they will eat into their own higher-end chips. Lose-lose situation.

Bwahahaha, by that logic, AMD's real well positioned - it only has 4% of the server market!

In this business, you have to have the best chip on a performance/watt basis, but you need to be at the right performance levels. Good luck to any of the other chip vendors coming close to Intel in high performance land.

 

[R0H1T]

Well, there's no need to compare using those numbers when we now have benchmark data for the Snapdragon 600 thanks to the HTC One review. Which at least implies a marked improvement over the Snapdragon S4 at similar power usage... Though one must remember that the Snapdragon S4 performance wasn't all too impressive to start with - this is merely bringing its performance up to Medfield levels.

The S4 was king of the hill when it was launched, a marked improvement over Tegra3 which again was alot better than anything before it ! Whatever spin you put on it the ARM architecture has been cruising for a good five years now & looks good for another five, now whether "it'll be enough for servers" can only be answered when we have working samples of Cortex-A57 but ARM smokes anything & everything in the mobile/tablet arena at this point in time!

 

[Khato]

The S4 was king of the hill when it was launched, a marked improvement over Tegra3 which again was alot better than anything before it ! Whatever spin you put on it the ARM architecture has been cruising for a good five years now & looks good for another five, now whether "it'll be enough for servers" can only be answered when we have working samples of Cortex-A57 but ARM smokes anything & everything in the mobile/tablet arena at this point in time!

Correct, the Snapdragon S4 beat the ~2 year old A9 core by a decent margin. However compared to the first A15 chip that came out at the end of the same year it's nothing special. The Krait 300 core appears to bring Qualcomm up to the level of A15 more or less.

As for Cortex A57, I believe the projection was for a 20-30% performance improvement over A15 no? No question it's a good bump, especially if it maintains current efficiency. But it's nothing awesome. It's definitely indicative of the fact that they're quickly running out of the low hanging fruit when it comes to improving performance without lowering efficiency.

 

[SocketF]

The danger that ARM poses stems from the fact ARM has next to nothing to lose and everything to gain by eating Intel's server share. If ARM provides a better overall benefits/price over Intel, customers switch to ARM and Intel loses, but if Intel makes their low-end chips too good to compete with ARM then there is a very real risk they will eat into their own higher-end chips. Lose-lose situation.

Yes that is true, too. Just imagine a cheap but speedy Atom .. who needs the ULV chips then? Probably a few specialists, but the mass market ...

Here are some more points from AMD's Seamicro / Server manager:
http://www.theinquirer.net/inquirer...-designed-arm-chips-will-give-intel-headaches

Seems that ARM SoCs are so easy, fast and cheap to design & manufacture, that high-volume custumers ( Amazon, google, etc), may what their "own" designs with features they need.

 

[NTMBK]

Yes that is true, too. Just imagine a cheap but speedy Atom .. who needs the ULV chips then? Probably a few specialists, but the mass market ...

Here are some more points from AMD's Seamicro / Server manager:
http://www.theinquirer.net/inquirer...-designed-arm-chips-will-give-intel-headaches

Seems that ARM SoCs are so easy, fast and cheap to design & manufacture, that high-volume custumers ( Amazon, google, etc), may what their "own" designs with features they need.

This is entirely the reasoning behind AMD's Kabini chip. Small, cheap, fast enough, and can get shoved into laptops. If it gives performance roughly equivalent to a Sandy Bridge ULV i3, but at a much lower price, people will be very happy. Of course, we have to wait for the benches to find out how it does.

 

[krumme]

Bwahahaha, by that logic, AMD's real well positioned - it only has 4% of the server market!

In this business, you have to have the best chip on a performance/watt basis, but you need to be at the right performance levels. Good luck to any of the other chip vendors coming close to Intel in high performance land.

You just ignored how Intel themselves won the server market. Cost.

Muhahahahhahahahahahahahahahaha

 

[mrmt]

And how does the ARM architecture offset the huge validation costs that all server parts have to go through? You all know that server validation is a bit more expensive than phone validation, right?

 

[Phynaz]

You just ignored how Intel themselves won the server market. Cost.

Muhahahahhahahahahahahahahahaha

Intel servers used to cost $25K less than IBM. Do ARM servers cost $25K less than Intel servers?

 

[SocketF]

Intel servers used to cost $25K less than IBM. Do ARM servers cost $25K less than Intel servers?

If you think about power-consumption .. maybe?
But I would agree that performance/power will be interesting ...

 

[StrangerGuy]

This is entirely the reasoning behind AMD's Kabini chip. Small, cheap, fast enough, and can get shoved into laptops. If it gives performance roughly equivalent to a Sandy Bridge ULV i3, but at a much lower price, people will be very happy. Of course, we have to wait for the benches to find out how it does.

I can't emphasize enough on that. Too many people here are waaaaay too hung up on performance or even performance/watt that anything that didn't top the charts will be an instant failure. Brazos was plenty successful enough simply by the fact it is adequate enough for the needs of most consumers at a low enough pricetag, which the original Atoms clearly isn't especially that it chokes in Flash.

So why would ARM be any different on the server space? Let's say if absolute power draw and price is my overriding concern for a web server would I really care if the 95W $1000 Intel chip has better much performance and better performance/watt when a sub 5W $50 ARM SoC would suffice?

 

[Arachnotronic]

I can't emphasize enough on that. Too many people here are waaaaay too hung up on performance or even performance/watt that anything that didn't top the charts will be an instant failure. Brazos was plenty successful enough simply by the fact it is adequate enough for the needs of most consumers at a low enough pricetag, which the original Atoms clearly isn't especially that it chokes in Flash.

So why would ARM be any different on the server space? Let's say if absolute power draw and price is my overriding concern for a web server would I really care if the 95W $1000 Intel chip has better much performance and better performance/watt when a sub 5W $50 ARM SoC would suffice?

You really, really don't understand this industry, do you?

Educate yourself by looking at Intel's processor price list before making silly claims.

http://intc.com/priceList.cfm

 

[ShintaiDK]

If you think about power-consumption .. maybe?
But I would agree that performance/power will be interesting ...

They dont even fit into performance/watt either, simply due to the rest of the platform. Atom/ARM servers only fit a tiny niche, hence their essentially nonexistant marketshare.

It seems we are back to the classic "work your way back". Here is a product, how do we make it fit into the world to our hopes.

 

[krumme]

And how does the ARM architecture offset the huge validation costs that all server parts have to go through? You all know that server validation is a bit more expensive than phone validation, right?

I dont think its relevant to discuss arm servers before we see what a57 is capable of compared to the new Atom. And even at that time ofcource its of minor importance. I agree with you. My post was just a reaction to the typical answer completely ignoring the history even for Intel. Even the server market reach maturty at some time, i just cant see it comming the next 5 or even 10 years.

But rest assured Intel themselves dont rest and assume everything will be fine just they do the same. They learned themselves by bitter experience thats not the case. And that paranoia is what makes Intel strong today. I watched Intel capture the server market. And there is more to it than TCO.

 

[krumme]

You really, really don't understand this industry, do you?

Educate yourself by looking at Intel's processor price list before making silly claims.

http://intc.com/priceList.cfm

What is your point?

 

[beginner99]

I think the situation now just shows how bad, bad it was for Intel - and AMD - to come to late to the mobile market. TSMC, Qualcomm and Samsung is no small players. They dwarf Intel the same way Intel dwarf AMD.

Is it that bad? The RAZR I phones with a 5 year old Atom uarch at 32 nm usually gets the best battery life ratings in reviews and at the same time reviewers praise fast web page load times. It also fares better than the RAZR M, same model but with ARM. The only thing it sucks at is GPU performance and efficiency, eg. battery life sucks on video playback but great on idle and talk time. Also note it's a midrange phone and much cheaper than iPhone 5 or such.

So with an outdated uArch and an node disadvantage the Atom is IMHO actually competitive. Maybe I'm a dreamer but I do believe silvermont will "core-2" the ARM camp, at least on CPU side.

 

[Tuna-Fish]

And how does the ARM architecture offset the huge validation costs that all server parts have to go through? You all know that server validation is a bit more expensive than phone validation, right?

Most of the low-end server market doesn't actually care all that much about validation. These things run PHP scripts, not bank transaction databases. The low-end doesn't even use ECC ram ffs.

 

[MisterMac]

Most of the low-end server market doesn't actually care all that much about validation. These things run PHP scripts, not bank transaction databases. The low-end doesn't even use ECC ram ffs.

The LOW end that uses desktop chips?
And sells desktop fitted machines as dedicated boxes?

Consider how big of shipments that low end actually is.


Consider it because - those types are normally included in normal desktop sales.


Then you'll see what a fraction ARM is currently prepping itself to battle.

 

[krumme]

Is it that bad? The RAZR I phones with a 5 year old Atom uarch at 32 nm usually gets the best battery life ratings in reviews and at the same time reviewers praise fast web page load times. It also fares better than the RAZR M, same model but with ARM. The only thing it sucks at is GPU performance and efficiency, eg. battery life sucks on video playback but great on idle and talk time. Also note it's a midrange phone and much cheaper than iPhone 5 or such.

So with an outdated uArch and an node disadvantage the Atom is IMHO actually competitive. Maybe I'm a dreamer but I do believe silvermont will "core-2" the ARM camp, at least on CPU side.

Yes its bad. Because if they have moved big time 8-10 years ago, they would owned the market, now they are left outside, because no one actually wants them to be there and pay their shareholders. They prefer to pay their own

Intel does not sell to the market. And its not a catatrophy for them, as their cpu is outright huge, and not suited for the purpose. I guess, as you, the new Atom will be a blast for the cpu part, but who is to pay for it? - and actually care?

Apple is moving on the GPU part because thats what the consumers want. They dont give damn about Anands or Brians talk about single threadded performance. They just want to play. And what excatly is it there a future a57 does not deliver on the cpu side to support that?

And then if we start talking of gpu part, that takes up the far majority of the mm2, Intels solutions is just bad right now. They have a lot to improve before this is a viable solution.

What excatly is Intels strategy here on the mobile side, and how do they want to execute it?

I think its a bold move they take, but the competition can only help us consumers. Surely if the new nodes will not sell on the mobile side, they will just move it to desktop or server. We win there too

 

[Lonbjerg]

What is your point?

Stop being stupid on purpose....

You really, really don't understand this industry, do you?

Educate yourself by looking at Intel's processor price list before making silly claims.

http://intc.com/priceList.cfm

Intel servers used to cost $25K less than IBM. Do ARM servers cost $25K less than Intel servers?

I am tried of anti-Intel peeps making mountains of molehills.

ARM is not going to kill Intel and save AMD...sorry to burst your bubble.

"Stop being stupid on purpose"? Lon, we've talked about this...
-ViRGE

 

[beginner99]

Yes its bad. Because if they have moved big time 8-10 years ago, they would owned the market, now they are left outside, because no one actually wants them to be there and pay their shareholders. They prefer to pay their own

Intel does not sell to the market. And its not a catatrophy for them, as their cpu is outright huge, and not suited for the purpose. I guess, as you, the new Atom will be a blast for the cpu part, but who is to pay for it? - and actually care?

Apple is moving on the GPU part because thats what the consumers want. They dont give damn about Anands or Brians talk about single threadded performance. They just want to play. And what excatly is it there a future a57 does not deliver on the cpu side to support that?

And then if we start talking of gpu part, that takes up the far majority of the mm2, Intels solutions is just bad right now. They have a lot to improve before this is a viable solution.

What excatly is Intels strategy here on the mobile side, and how do they want to execute it?

I think its a bold move they take, but the competition can only help us consumers. Surely if the new nodes will not sell on the mobile side, they will just move it to desktop or server. We win there too

Well if the new atom is fast and the phones are not pricier than comparable ARM based phones I would sure prefer it because I would mainly be browsing on such a phone and better single threaded performance is obviously very important for that. RAZR I is priced well IMHO, not more expensive than similar phones.

I agree about the GPU part. However AFAIK all ported games also run on crappy gpu of RAZR I so no big issue there except lower quality settings.

 

[Phynaz]

ARM is not going to kill Intel and save AMD...sorry to burst your bubble.

Oh, I fully believe ARM is a major threat to Intel's dominance. Just not in server.