AMD Announcement: ARMv8 Opterons In 2014

[happysmiles]

I think it's only one side of the story.

Server at first but then it'll trickle down to other products.

 

[cytg111]

..And they solely pursue it for the status of it..

- Yes, because these ppl are all about the pride of it.. and screw the profit.

 

[pelov]

AMD gives up on developing its own server CPU cores. Intel has won. How times have changed with Intel's tick-tock.

They won years ago. AMD just can't compete in perf-per-watt, regardless of how much money they funnel into their R&D because of fab constraints and lagging in node.

Think about it. Most of the big fabs are concentrating on low power ARM-focused node development and advances with the high power stuff lagging behind, typically 1-2 years. This puts AMD 1-2 years behind Intel and likely to be even worse as time goes on. That essentially means the most likely scenario going forward would be AMD having to compete with 28nm against a 14nm Intel node. I don't care how good your microarchitecture, you're going to lose.

They're also roughly on par with ARM as far as server market share goes. The 64-bit hurdle should bump ARM up considerably, though the biggest issue is time to market.

They're also stuck in a tight spot with respect to cash. They don't have the money to compete with Intel on the high end so they've got no other option than to change the rules of the game. Built-ready ARM cores means AMD can accelerate their timeline and spend less cash.

 

[ShintaiDK]

I think it's only one side of the story.

Server at first but then it'll trickle down to other products.

The problem is that they face even greater competition in the ARM market. And by 2014 and later, AMD will face Intel as well in the same segment battling 14nm Atoms with a new uarch.

Jack of trades, master of none. Soon forgotten.

We still have to see if the company even makes it trough 2013. Q3 results needs to be break even. And they expect loses until then.

Then question is, what can AMD offer that others cant.

 

[ViRGE]

Can they still implement their own IP should it benefit the processor? Or are they just directly using the stock CPU design from ARM?

The CPU design would be stock. Of course there are other IP blocks necessary for a complete chip that they're presumably going to source internally.

 

[cbn]

The problem is that they face even greater competition in the ARM market. And by 2014 and later, AMD will face Intel as well in the same segment battling 14nm Atoms with a new uarch.

Jack of trades, master of none. Soon forgotten.

Yes, but they have x86, which I think could give them an advantage against all those other Fabless players if they could somehow integrate the ARM instruction set.

P.S. I noticed the market cap for AMD is really low now ---> http://finance.yahoo.com/q?s=AMD

$1.54 Billion dollars.

I am wonder if they will be bought by someone else? Maybe Texas Instruments (which has x86 license)

 

[ViRGE]

I am wonder if they will be bought by someone else? Maybe Texas Instruments (which has x86 license)

TI just got out of the CPU business (or rather announced their plan of action to do so). So they're just about the last company that would want to buy AMD.

 

[ShintaiDK]

Yes, but they have x86, which I think could give them an advantage against all those other Fabless players if they could somehow integrate the ARM instruction set.

P.S. I noticed the market cap for AMD is really low now ---> http://finance.yahoo.com/q?s=AMD

$1.54 Billion dollars.

I am wonder if they will be bought by someone else? Maybe Texas Instruments (which has x86 license)

You cant mix ARM and x86. You can at best sit back with 2 independent individual systems.

 

[pelov]

Anyone kind of hoping that VIA swoops in and buys AMD for an epic x86 revival?

 

[sontin]

Anything else besides the ARM announcement?

 

[andrescu]

I am wonder if they will be bought by someone else? Maybe Texas Instruments (which has x86 license)

My bet is on one of the Asian corporations. Maybe Samsung?

 

[ShintaiDK]

Buying AMD=no x86. Plus employees you dont need, plus liabilities and so on. So lets just drop those fairy tales.

The only thing worth anything in AMD is the GPU business. And that can be had for scraps.

Also nobody with a working mind would want to compete with Intel in x86 if they could avoid it. Not to mention it would take around 5 years before you even see the first product.

 

[pelov]

http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-eventDetails&EventId=4857382

They've got a PDF if anyone wants to read it.

 

[sontin]

So AMD's savior is Datacenter? Nothing else? That was the big announcement?

nVidia did the same nearly two years ago with Project Denver...

 

[cbn]

You cant mix ARM and x86. You can at best sit back with 2 independent individual systems.

Not sure about the servers, but I am very optimistic about the future of Intel phones.....and those can emulate ARM (to some degree).

 

[MisterMac]

This sounds absolutely crazy.


They seriously cannot expect to compete with a bog standard core + some interconnect secret sauce.


A HSA Arm chip designed by AMD's experience would rock - but this is just.


Yea - this is like--- is this even good for a "potential" dream fairy tale takeover?

 

[ShintaiDK]

Not sure about the servers, but I am very optimistic about the future of Intel phones.....and those can emulate ARM (to some degree).

Smartphone apps? Thats just JAVA.

 

[Centauri]

Anyone kind of hoping that VIA swoops in and buys AMD for an epic x86 revival?

...exactly how large of a company are you under the impression VIA is?

 

[happysmiles]

The problem is that they face even greater competition in the ARM market. And by 2014 and later, AMD will face Intel as well in the same segment battling 14nm Atoms with a new uarch.

Jack of trades, master of none. Soon forgotten.

We still have to see if the company even makes it trough 2013. Q3 results needs to be break even. And they expect loses until then.

Then question is, what can AMD offer that others cant.

GPU

 

[Blitzvogel]

GPU

ARM + on die AMD graphics processors used for GPGPU, could be a formula for win.

 

[ViRGE]

Anything else besides the ARM announcement?

Nope. That was it.

 

[mrmt]

Interesting. AMD isn't designing their own ARM core here; they're just going to be using an ARM design that will be announced tomorrow.

Essentially what I said before. They went for a "me too" ARM core manufactured at a "me too" foundry with "me too" software support, they have only a different interconnect to show. With such a competitive package one can only wonder how low AMD margins are going to sink once this strategy starts to develop.

This is the signature on the certificate of incompetence of AMD management and BoD. And this time, nobody can blame Hector or Dirk. It is on Rory shoulders that this one should stay.This is the most incompetent management team I ever saw.

 

[pelov]

Furthermore, we heard rumors that the company decided to kill their high-performance CPU cores and the Opteron line of products, as it would mean finally giving up on competing with Intel in high performance space, and only utilize its FirePro GPU cores to go against Xeon / Xeon Phi. We spoke with Phil Hughes, who gave us the following statement:

"AMD remains committed to the server market and Opteron. We continue to refine our focus on key workloads in the private and public cloud datacenters and market segments such as dense servers where we can offer unique and compelling solutions."

The unique and compelling solutions is today's announcement of ARM+x86 parts, that's for sure. While the company skipped on giving any details out, the 2014 timeline goes perfectly in line with the arrival of recently announced 14XM process at GlobalFoundries. AMD didn't disclose details about the chips in question, but it is easy to predict that the 2013 will be the year AMD will simply try to remain competitive with Sea Islands and Jaguar/Piledriver-based products, before going all out on its new strategy in 2014.

http://www.brightsideofnews.com/new...2c-x86-chips-for-servers-in-2014-on-14nm.aspx

That's not surprising if true, given they'd be shafted with foundries opting to optimize for low power ARM rather than high power x86.

Essentially what I said before. They went for a "me too" ARM core manufactured at a "me too" foundry with "me too" software support,

What does any of this even mean? Me too foundry? lol

 

[cbn]

In 2016 Intel should be on 10nm lithography (and 3rd generation FinFET + whatever new materials science/xtor design they have in mind)

Meanwhile all the other Fabless companies using TSMC will be stuck with 16nm and TSMC's first generation FinFET.

10nm vs 16nm = Isn't that a 1.5 node difference (plus any xtor design/material science lead by Intel)?

With that kind of difference in process tech, I really have to wonder what choice is better for AMD? Was staying with x86 a better idea or can branching off into ARM still pay off?

P.S. (With respect to phones/Tablets in 2016) I'd imagine Intel's predicted process and design lead has the strong chance of swaying what ISA native apps are written in for Android. If so, then x86 might be the more efficient choice for low cost Android phones/Tablets of the future. I mean, if a Fabless company has to suffer from a 1.5 node lag in lithography compared to Intel (plus any material science/xtor design differences).....wouldn't using x86 help reduce the impact of that lagging process? (At least the apps would be written in the same ISA....even if AMD silicon is 1.5 nodes behind)

 

[pelov]

A technology roadmap published by EE Times indicates that 20-nm chips should start rolling off TSMC's production line next year. Like its current 28-nm process, the 20-nm chips will use traditional planar transistors. 3D transistors, otherwise known as FinFETs, are coming at the 16-nm node. The first silicon based on TSMC's 16-nm process is expected to tape out at the end of next year. That test chip will reportedly be based on ARM's 64-bit V8 processor; the two companies announced a partnership in July to build 64-bit chips based on FinFET technology.

TSMC's 16-nm FinFET process will reportedly be similar to its 20-nm one. The foundry doesn't expect the FinFET process to lower leakage power, but it does say performance will go up by as much as 35%. "Total power consumption" is supposed to fall by the same percentage, according to TSMC R&D Vice President Cliff Hou.

Using a next-gen ARM processor to validate its 16-nm FinFET process should give TSMC's SoC customers confidence in its ability to churn out chips for smartphones and tablets. ARM is also working with GlobalFoundries on FinFETs, which are being used by that foundry's 14-nm "eXtreme Mobility" process. It will be interesting to see how smoothly GloFo and TSMC bring their transistors into the third dimension. Intel, of course, started shipping FinFET transistors in its Ivy Bridge processors earlier this year.

http://techreport.com/news/23768/tsmc-16-nm-finfet-process-coming-next-year

The chasm that exists in x86 in nodes isn't as evident in the low power segment. From tapeout-to-products rolling off the fab, TSMC took less than a year with 28nm. Assuming the same time frame post tapeout, we'd be looking at 2014 16nm FinFET which is when Intel would be on 14nm Broadwell.

The above is the most likely reason AMD is opting with ARM. TSMC, GloFo and the Fishkill alliance won't be rushing any sort of high power node to viability when they can make much more money in the mobile market. As a result, we'd see AMD stuck on 28nm while Intel would be skipping happily along to 14nm. If they wanted to compete, I don't think they'd have much of a choice here other than opting to go with ultra low power.