Bookmark and Share


AMD today announced a roadmap of near- and mid-term computing solutions that harness the best characteristics of both the x86 and ARM ecosystems, called "ambidextrous computing." The cornerstone of this roadmap is the announcement of AMD's 64-bit ARM architecture license for the development of custom high-performance cores for high-growth markets. Today's announcement also provides a forward-looking glimpse into AMD's development plans to deliver truly unmatched ambidextrous computing and graphics performance using a shared, flexible infrastructure to enable its customers to blaze new paths of innovation for the embedded, server and client markets as well as semi-custom solutions.

"Before today, AMD was the only company in the world to deliver high performance and low-power x86 with leadership graphics. AMD now takes a bold step forward and has become the only company that can provide high-performance 64-bit ARM and x86 CPU cores paired with world-class graphics," said Rory Read, AMD president and CEO. "Our innovative ambidextrous design capability, combined with our portfolio of IP and expertise with high-performance SoCs, means that AMD is set to deliver ambidextrous solutions that enable our customers to change the world in more efficient and powerful ways."

The market for ARM- and x86-based processors is expected to grow to more than $85 billion by 2017.

AMD is uniquely positioned as the only company delivering differentiated solutions capable of addressing the breadth of this market. This is the first time a major processor provider has created the IP path to allow others to leverage innovation across both ARM and x86 ecosystems.

AMD's ambidextrous computing roadmap includes:

Project SkyBridge - This design framework, available starting in 2015, will feature a new family of 20 nanometer APUs and SoCs that are expected to be the world's first pin-compatible ARM and x86 processors. The 64-bit ARM variant of "Project SkyBridge" will be based on the ARM Cortex®-A57 core and is AMD's first Heterogeneous System Architecture ("HSA") platform for Android; the x86 variant will feature next-generation "Puma+" CPU cores. The "Project SkyBridge" family will feature full SoC integration, AMD Graphics Core Next technology, HSA, and AMD Secure Technology via a dedicated Platform Security Processor (PSP).

K12 - A new high-performance, low-power ARM-based core that takes deep advantage of AMD's ARM architectural license, extensive 64-bit design expertise, and a core development team led by Chief CPU Architect Jim Keller. The first products based on "K12" are planned for introduction in 2016.

"At ARM we are dedicated to working with partners who revolutionize and transform experiences everywhere from sensors to servers," said Simon Segars, CEO at ARM. "AMD's market reach and proven experience in leading industry transitions to 64-bit computing in client and server environments, combined with ARM's low power expertise and server base system architecture (SBSA) standard, will deliver new capabilities and drive innovation across multiple high growth markets."

AMD today also publicly demonstrated for the first time its 64-bit ARM-based AMD Opteron™ A-Series processor, codenamed "Seattle," running a Linux environment derived from the Fedora Project.

The Fedora Project is a Red Hat-sponsored, community-driven Linux distribution, providing a familiar, enterprise class operating environment to developers and IT administrators worldwide. This Fedora Project-based Linux environment enables companies to transition to ARM-based servers without the need to integrate entirely new tools and software platforms to their IT environments. This demonstration represents a significant step forward in expanding the footprint of ultra-efficient 64-bit ARM processers within the data center.

Tags: AMD, Opteron, x86, ARM, Cortex A57, K12


Comments currently: 24
Discussion started: 05/06/14 01:24:58 AM
Latest comment: 06/02/14 01:45:42 PM
Expand all threads | Collapse all threads


Well done the cat is well and truly amongst the pigeons now.
1 2 [Posted by: tedstoy  | Date: 05/06/14 01:24:58 AM]

Once AMD delivers then we can determine the results. It's way too early to get excited.
4 1 [Posted by: beenthere  | Date: 05/06/14 08:36:56 AM]

This looks like a winning strategy. It was always going to take a few years before Jim Keller's touch would come to fruition.
3 1 [Posted by: linuxlowdown  | Date: 05/06/14 08:46:15 AM]
- collapse thread

AMD hasn't completely abandon the higher end market yet.

Here's an interesting read.

AMD Confirms Development of High-Performance x86 Core With Completely New Architecture

AMD’s top chip architect Jim Keller has confirmed during the AMD’s Core Innovation press conference that they are developing a new high-performance x86 core which will replace the current high-performance x86 cores that include Steamroller, Piledriver and Bulldozer.

With Jim Keller onboard, AMD is finally developing a new core architecture that’s built from scratch and means that AMD won’t make the same mistakes they made with Bulldozer. The architecture is still far from being ready and we are looking at a launch scenario of atleast 2016-2017 so that’s still a long time to go from where we are currently. It is likely that if AMD succeeds in developing the new cores in time, then we won’t see AMD’s fourth generation modular x86 Excavator cores in 2015 since AMD will skip it for the new core introduction otherwise AMD’s 2015 line of APUs will stick with Excavator which will be the last modular AMD core before it hops onboard the SMT design.

Performance numbers are a mystery and so is the codename but atleast we got a conformation from AMD that they haven’t abandoned the high-performance desktop market at all. AMD has several APUs planned for Opteron and desktop lines with new cores but one can now imagine much faster next generation desktop processors and APUs which will fuse AMD’s next high-performance x86 core architecture.
2 1 [Posted by: xenocea  | Date: 05/06/14 09:38:14 AM]
"AMD is finally developing a new core architecture that’s built from scratch and means that AMD won’t make the same mistakes they made with Bulldozer"
With Bulldozer, they corrected mistakes made with Phenoms. They are innovative company, isn't ?
2 0 [Posted by: Tristan  | Date: 05/06/14 10:55:11 AM]
no they went two steps forward and one step back with bulldozer. the phenom based chips were actually faster then the first gen bulldozer chips were in some benchmarks. So in retro spec the bulldozer had problems out of the gate.
0 0 [Posted by: SteelCity1981  | Date: 05/06/14 11:46:11 AM]
Its not all 'd gate' thing. what become of bulldozer is ill fated from all those scratch that scratch this feature and weird agreements with Intel. If they did parallel development along with K0 then i guess server market wont adoptig it as we see today but at least they could test and deliver some software optimization for it
Bulldozer is failure from software perspective along with all that troubles with GloFo transitioning, poor process, transistor optimizations done manually etc.
They actually claim transfer for automated optimization on 65nm dumb shrink of 90nm K8. And after they delayed K10, bugfixing it for a year when it was introduced. Recalling early K10 server products and rebranding them as X2 7750BE X2 7850BE and similar weird marketing moves.
AMD problem for a small guy is that they always try to suck up into server market's arse. And that paid well with two failed launchups after they hadnt right for "another mistake" as they did in 2008
There will be no soft to improve either Piledriver or Steamroller performance and all that fiddling into cores parallel workload is plain bugfixing to get back onto Intels path. And not bringing anything spectacular as they promised before Bulldozer launchap. If they implement same strategy with some 4x128b FPU + 4x128b FMAC in "Future After Devastator" and finally improve those LSUs and L1 associativity in Devastator we could maybe get a glimpse of hope for some actual x86 improvement. While this future server roadmap fully oriented towards "cheap" ARMs ARMv8 IP doesnt smell too god for any feasible future for AMDs x86-64.
After all we saw great job done, thou reluctantly executed, with K8's essential incarnations inside Bobcats and K10's inside Jaguars/(Pumas). Thus we have real proof that parallel projects on K10/K10.5 along on "this new Bulldozer HPC architecture" could be feasible in past if AMD really had vision at the right moment. As they certainly had much better resources back then.

Lack of vision
And traditional lack of software
(is what bring demise to every hardware company)
0 0 [Posted by: OmegaHuman  | Date: 06/02/14 12:57:26 PM]
show the post
0 4 [Posted by: amdzorz  | Date: 05/08/14 10:04:09 AM]
This is good news to hear from AMD, so when is AMD going to join IBM's Open Power Foundation, and get access to the Power8 ISA/IP for AMD's SeaMicro server division. AMD needs to use any SOC/CPU IP it can license for its servers and its PCs and mobile devices products. IBM is opening up the Power8 ISA/IP to the market(Arm Holdings style) and it is seeding Samsung and Global foundries with some of its advanced fab IP, in advance of IBM getting out of the fab business(IBM's research Fabs/Labs not included). The IBM power based CPUs(Not to be confused with PowerPC) are monsters at high end computation, and IBM is loosing money on CPU hardware production. IBM is trying to create a larger economy of scale with its Power/Power8 being marketed more ARM Holdings like, to give IBM a steady supply of low cost Powre8 parts for its servers, and IBM's real money maker, its OS/services/software business.

IBM through its hardware IP shareing has Helped Samsung develop its 14nm process, which Samsung is now providing GlobalFoundries to potentally provide IBM, and indirectly the entire market CPUs/GPUs/SOC with a more competative with Intel chip fab process. AMD needs to leverage its chip design, and SOC integration knowhow to more than just ARM and x86 ISAs. Nvidia is partnering with IBM, who helped Nvidia develop NvLink, to utilize Nvidia's GPU as accelerators for the POWER8 server CPUs. AMD needs to get SeaMicro to offer Power8, as well as the other SeaMicro options of Xeon, Opteron(x86 and Arm based) SKUs. AMD should use any Licenseable CPU IP, and add its graphics and HSA/hUMA ability to these CPUs ISAs, and produce systems that can use GPGPU from mobile SOCs to Servers, and HPC/supercomputing. Google already has a testing motherboard built for power8, and x86 is going to have some serious competition in the high end server(heavy data analytics) market that requires major CPU horsepower, and with ARM based servers taking over general web page serving workloads. x86 will still be used for PC/Laptop computing, but even there the Power ISA could make inroads should someone take the power ISA and modify it for PCs and Laptops.

IBM is trying to achieve with its Power8 ISA IP shareing the same economy of scale/competative pricing that it needed for its IBM personal computers, when it forced Intel, as a pre-condition of getting x86 onto the IBM PC, to cross-license the x86 16/32 bit ISA with AMD and Others. What goes around comes around, and the CPU hardware market is completely becoming a commodity market, where CPU ISA will be licensed to any and all with the money and resources to roll their own SOCs/CPUs.
0 1 [Posted by: BigChiefRunAmok  | Date: 05/09/14 09:30:38 AM]
"Google already has a testing motherboard built for power8." Just testing does mean a damn. IBM cough up more than $3B for its desperate attempt to stay relevant. They made a few test racks, send to Google with a blank check, sure they will get tested. It does not mean anything. IBM saw their Power architect market share dropped from 90% to 5%, and this is just a last ditch effort.
0 0 [Posted by: Tukee44  | Date: 05/10/14 01:03:18 AM]
IBM won't be selling the Power8 that Google Uses, it will be maybe Samsung, or someone else that licenses the IP from IBM and uses GlobalFoundries as a fab partner, or Samsung's fabs. The Power CPUs(That you are confusing with the PowerPC) are not the same, and PowerPC never had that much marketshare. But just look at how much marketshare The ARM industry has in mobile, and all that is Licensed from Arm Holdings, including the custom ARMv8 ISA based CPUs that only run the ARMv8 instruction set(they License the ISA), but are otherwise completly different from Arm Holdings' refrence designs under the hood. Now take that Licensing Model and bring it to the Power8 IP/ISA, well Google could hire a custom System Integrator and take the Power8 refrence design and customize it for Googles Servers, Google could Add Licensed GPUs from a variety of Licensed IP suppliers, or just go with the refrence design Power8, and have Globalfoundries, or Samsung fab the parts at their chip foundries, all at a fraction of the cost of a Xeon part, and Power8 CPUs rival Intel's server CPU offerings. Apple with its P.A. simiconductor engineers could take the Power(Like thay had in the past with PowerPC) and use the Power server CPU in their Mac Pro, and not use the Xeon that they currently do, and those Apple In House P.A. Simi folks, the ones that created that Apple A7(Custom Wide order superscalar ARMv8 ISA based SOC) can work wonders with a CPU design like the Power8. You could add to that list AMD, and Nvidia, Samsung, or any other company that has in House chip/SOC design engineers, or any company that has the funds to hire any of the chip design houses that cater to needs of the Licensed CPU/GPU/SOC industry.
0 0 [Posted by: BigChiefRunAmok  | Date: 05/10/14 08:24:01 AM]

AMD is preparing to fight
2 1 [Posted by: Tristan  | Date: 05/06/14 10:46:49 AM]
- collapse thread

I endorse your humour
0 1 [Posted by: tedstoy  | Date: 05/06/14 09:25:47 PM]

All the best to AMD as we SERIOUSLY need competition against the Intel i5 and i7s although they haven't remotely come close to Intel quite yet so I hope this pushes them ahead as its been so many years (mid-2000s?) when they were about 1:1 with price/performance. Consumers need another option to keep prices down.
2 0 [Posted by: thudo  | Date: 05/06/14 10:49:08 AM]
- collapse thread

competition leads to aggressive price cuts, in which intel doesn't have in it's mid to highend chips. something that hopefully will change in the next couple of years. that and intel has gotten lazy on the performance crown in the past few years considering there is no competition for them.
2 0 [Posted by: SteelCity1981  | Date: 05/06/14 11:48:18 AM]
I suppose but I don't consider a pre-unlocked i7-4770K going for $359 CDN to be expensive considering the benchmarks. Next steps up are indeed pricey so no sense there. Offer me a comparable CPU for much less then bang I am there.
1 1 [Posted by: thudo  | Date: 05/06/14 02:23:45 PM]
but imagine if amd matched that performance and cut that price down to under 300 dollars that would force intel to do the same with the 4770k.
2 0 [Posted by: SteelCity1981  | Date: 05/06/14 02:35:13 PM]
Absolutely! I just wanted that to happen sooner so waiting until 2016 to see if AMD come really come back in this department is a killer. Consumers need a strong competitor in this field like we once had what seems like a decade ago.
1 0 [Posted by: thudo  | Date: 05/06/14 04:18:18 PM]
While AMd sticks to their pragma that small users could only buy crappy leftovers from server table AMD could never match Intel power and performance wise on HPC market. Historically they were pretty tied up at same nodes. Even AMD performed better with K10.5 against Core2Quads but that was when AMD had IMC inside and Intel relied on pretty damn good, old NetBurst IP chipsets and their QP FSB. And even then K10.5 sucked more juice than C2Q+NB together. Not to mention AMD had also NB+SB setups back then, and NB was only PCIe links (+optional GPU) + SB-link

So until AMD really invests into good HPC arch improvement along that already great K10.5 reincarnations in Jaguar/Puma we should expect anyy price cuts. Not to mention also that AMD burned their money, sell real estates and fabs and pay for every chip they produce. So any price cuts arent really an option to AMD as you suggests
0 0 [Posted by: OmegaHuman  | Date: 06/02/14 01:19:15 PM]
Well, considering that the only reason Intel even offers K sku's is likely because of AMD.

Well, anything unlocked under $1000.
0 0 [Posted by: xrror  | Date: 05/06/14 11:42:29 PM]

Could this be the pre-cursor to an Ambidextrous O/S.'s with even more hassles than the present O/S which is trying to be ambidextrous, or is it one of the reasons that the newer O/S.'s are the beginnings of a radical shift in computing as we have known it since the early 1990's when the industry settled on the the architecture and type of systems as we now use. If one cares to look back to the precursors of the present x86 environment there was another environment which dominated and it never died and that environment is now coming to dominate once again and we all will have to change again.
1 0 [Posted by: tedstoy  | Date: 05/07/14 08:34:44 PM]
- collapse thread

In M$'s case that's not Ambidextrous but bimodal! M$ having an App store runtime (stapled) on top of its OS, and borking the user experience/UI, can not be compared to having CPU/SOC hardware that is Pin compatable between AMD ARM ISA based SKUs, and Its x86 ISA based SKUs. What AMD is doing is giving the servers based on AMD server/other "Ambidextrous" motherboards the ability to host both ARM and x86 based CPUs/SOCs without having to buy a new motherboard, This will save the owners of server farms millions, and give them the flexability to tailor these AMD SOCs/CPUs to the workloads that ARM may be more cost efficient at, or the workloads that x86 may be better at, buy just swapping the x86 CPU/SOC for a pin compatable ARM based SOC/CPU. Just swap the chip/s plug the motherboard back in and the system software/OS/VM automatically re-classifies the SOC/CPU abilties and balences the workload, just Swap, Plug, and Serve/Play/number cuunch. This makes the hardware/motherboard HSA aware, and maybe in the future GPUs(this is a Moot point once server chips become SOCs with integrated GPUs/vector processors) could be made to be swappable into motherboard sockets also without the need for PCIe based GPUs, so the PCIe slots are freed up for SSD, or phase change based flash. In the future whole computing clusters may come on mezzanine/socketed modules including GPUs(vector processors), for server/HPC workloads.

M$'s Bimodal 8 OS, a jack of two trades, a productivity disaster, that no one can master, Let's take it commode-all and flush it on down, I'm still using 7, no need to update, got rid of that TIFKAM before it's too late.
1 0 [Posted by: BigChiefRunAmok  | Date: 05/09/14 02:59:58 PM]

Dudes pin compatible arm , YESSSS.
I want that for a tong time now.
They go up 8 cores , up 2.0 ghz , they are 64 bit , their TDP is low as hell, ill take 2 of them
0 0 [Posted by: kingpin  | Date: 05/13/14 03:29:53 PM]
- collapse thread

And they're for server market. Low availability in retail channels and pricey. YESSSSSS
Not too much to be happy for.

Well at least they're TRUE EIGHT MP CORES. Something AMD could done long ago if they just copied what they leased to SPS4/XBOX chip designs and put EIGHT Jaguars cores in current AMD offerings.

Well that would probably raise already pricey production costs and Athlon AM1s are already tablets/netbooks rejects and in that market there are no true Microsoft support inside their weaslishly bad product remakes like Win8/Win8.1 so thats another reason why there are no E1/A6 eight core offerings. Along that it would be too similar with hardware inside newish and poorly adopted XBOX ONE. And would need additional 64b IMC. And would need for more layered board which budget notebooks/Pc boards cannot afford if they want them to be sold with considerable revenue and still be somewhat attractive.
0 0 [Posted by: OmegaHuman  | Date: 06/02/14 01:45:42 PM]


Add your Comment

Related news

Latest News

Monday, July 28, 2014

6:02 pm | Microsoft’s Mobile Strategy Seem to Fail: Sales of Lumia and Surface Remain Low. Microsoft Still Cannot Make Windows a Popular Mobile Platform

12:11 pm | Intel Core i7-5960X “Haswell-E” De-Lidded: Twelve Cores and Alloy-Based Thermal Interface. Intel Core i7-5960X Uses “Haswell-EP” Die, Promises Good Overclocking Potential

Tuesday, July 22, 2014

10:40 pm | ARM Preps Second-Generation “Artemis” and “Maya” 64-Bit ARMv8-A Offerings. ARM Readies 64-Bit Cores for Non-Traditional Applications

7:38 pm | AMD Vows to Introduce 20nm Products Next Year. AMD’s 20nm APUs, GPUs and Embedded Chips to Arrive in 2015

4:08 am | Microsoft to Unify All Windows Operating Systems for Client PCs. One Windows OS will Power PCs, Tablets and Smartphones