AMD Announcement: ARMv8 Opterons In 2014

[piesquared]

http://www.xbitlabs.com/news/cpu/di...mportant_Strategy_Announcement_Next_Week.html

Interesting times ahead..

AMD likely to announce ARM-based server on Monday

AMD CEO Rory Read is expected at the event, and an ARM spokesman confirmed Thursday that ARM CEO Warren East will be there as a "special guest." AMD is also arranging press briefings at the event with Andrew Feldman, who was CEO of SeaMicro before AMD bought the company this year.

 

[ViRGE]

This has long been expected, no? We've known that Seamicro was going to use ARM since almost the day AMD acquired them.

 

[mrmt]

http://www.xbitlabs.com/news/cpu/di...mportant_Strategy_Announcement_Next_Week.html

Interesting times ahead.

Whatever the announcement is it should not impact revenues in the short term or costs, or it would have been announced before the EC. It is something ARM or Seamicro related, something to give us a clearer picture of where the company wants to be in the medium/long term.

 

[Olikan]

my bet:
ARM will change it's mali gpu to a radeon gpu
:biggrin:

 

[sontin]

Why should they? ARM gets revenue from their Mali-IP. Maybe AMD will kill Radeon and using Mali for their future Low-Power-SoCs.

 

[Olikan]

Why should they? ARM gets revenue from their Mali-IP. Maybe AMD will kill Radeon and using Mali for their feature Low-Power-SoCs.

well, it's hard to find a company that actually uses Mali...
anyway, it was a joke

 

[sontin]

well, it's hard to find a company that actually uses Mali...
anyway, it was a joke

Huh? Samsung is using Mali in nearly every new SoC.
And then there are a lot more companies in asia.

 

[BrightCandle]

I really hate announcements announcing announcements, they are far from news worthy.

 

[Idontcare]

I really hate announcements announcing announcements, they are far from news worthy.

Thank you for announcing this to the rest of us

This has long been expected, no? We've known that Seamicro was going to use ARM since almost the day AMD acquired them.

At the time of the seamicro announcement, I wondered if it made sense for AMD to make a move to acquire Calxeda or if it made more sense to invest those monies into fleshing out a jaguar-based server lineup.

Does it make sense for AMD to attempt to pursue both angles (ARM & x86 low-power server products) simultaneously, given their precarious economic situation?

 

[podspi]

Thank you for announcing this to the rest of us



At the time of the seamicro announcement, I wondered if it made sense for AMD to make a move to acquire Calxeda or if it made more sense to invest those monies into fleshing out a jaguar-based server lineup.

Does it make sense for AMD to attempt to pursue both angles (ARM & x86 low-power server products) simultaneously, given their precarious economic situation?

Well, if the rumors are to be believed, AMD's strategy is re-usable blocks of IP. So AMD doesn't have to dedicate much to release an ARM product if it has already developed a Jaguar-based SoC, because they just rip out the Jaguar cores and slap in bog-standard A15 ones.


How is that is in practice vs. theory and how true that actually is I do not know.

 

[Idontcare]

Well, if the rumors are to be believed, AMD's strategy is re-usable blocks of IP. So AMD doesn't have to dedicate much to release an ARM product if it has already developed a Jaguar-based SoC, because they just rip out the Jaguar cores and slap in bog-standard A15 ones.


How is that is in practice vs. theory and how true that actually is I do not know.

This is the part I don't get. I understand why everyone else on the planet has to pursue using bog-standard A15 cores embedded in their SoC, including Calxeda, but I don't understand why AMD would.

AMD has an x86 license, and they have already created the IP blocks involving x86 to a degree that is unrivaled in that power footprint - neither Intel nor Via can touch them at this time in that segment (that may change with future Atom iterations).

Nvidia has no choice, they must make ARM viable if they want to get into the existing x86 low-power sea-of-cores space. Same with Calxeda. But AMD not so much, they have the low-power core and all the software infrastructure exists.

They just need to start producing sea-of-core products using Jaguar derivatives. At this point in time, ARM would definitely be a step backwards as well as a distraction of limited resources and focus.

But AMD might decide to go after the handset market, ala qualcomm and broadcom style, and for that they would do fine just dropping in a bog-standard ARM core surrounded by their own IP (specifically the GPU IP).

 

[KompuKare]

Yes all very puzzling. If AMD's R&D budget is too small to compete with Intel (although I'm sure Bulldozer used up a fair budget), why do they think they can compete in a market they have no experience in? Looks like someone on the board has got either a tablet or ARM obsession.

Still, if their modular Jaguar and /or Bulldozer designs are truly able to be 'slotted' in with minimal redesign, maybe the automated design (while still as inefficient as reported) makes some sense. Those designs may not be as good as handcrafted ones but AMD are used to selling chips with 20%+ larger dies than the competition - not good for their margins though.

But AMD might decide to go after the handset market, ala qualcomm and broadcom style, and for that they would do fine just dropping in a bog-standard ARM core surrounded by their own IP (specifically the GPU IP).

Wonder if Qualcomm didn't get a time limited agreement for them not to come after that market when they bought Imageon though. So AMD may not even be able to do that.

 

[Olikan]

This is the part I don't get. I understand why everyone else on the planet has to pursue using bog-standard A15 cores embedded in their SoC, including Calxeda, but I don't understand why AMD would.

well, if charlie is right... AMD will move to a "custom chip designer"...hence the IP blocks, if someone wants a 3 cpu "block" + 10 gpu "block"...amd will ask " 3 cpu blocks of ARM or x86?"
now... AMD selling ARM on theyr on, is indeed strange

this is actually a smart way to make someone else pay your R&D :biggrin:
(aka sony and 3d stacking)

 

[Idontcare]

well, if charlie is right... AMD will move to a "custom chip designer"...hence the IP blocks, if someone wants a 3 cpu "block" + 10 gpu "block"...amd will ask " 3 cpu blocks of ARM or x86?"
now... AMD selling ARM on theyr on, is indeed strange

this is actually a smart way to make someone else pay your R&D :biggrin:
(aka sony and 3d stacking)

Interesting...so AMD's plan could be to become a foundry of IC SoC designs.

Instead of fabless companies going to TSMC with a design in hand for production we will see a rise of not just fabless, but also design-less, companies that want someone to design the chip and someone else to build the chip, and at the end of the day the design-less company is just paying royalties to resell the final chip in whatever product the chip is getting bundled into.

Isn't that how Intel got into x86 in the first place? Some company came to Intel asking them to develop a chip for electronic typewriters (and produce the chip too) and they didn't care what Intel did in terms of recycling the IP in future chips?

I could see that working for AMD. Companies that want to compete with qualcomm but don't have the revenue to justify keeping an entire design team on the books year round. Use a design foundry, then use a fab foundry, get your chips black-box style.

 

[Lonyo]

This is the part I don't get. I understand why everyone else on the planet has to pursue using bog-standard A15 cores embedded in their SoC, including Calxeda, but I don't understand why AMD would.

AMD has an x86 license, and they have already created the IP blocks involving x86 to a degree that is unrivaled in that power footprint - neither Intel nor Via can touch them at this time in that segment (that may change with future Atom iterations).

Nvidia has no choice, they must make ARM viable if they want to get into the existing x86 low-power sea-of-cores space. Same with Calxeda. But AMD not so much, they have the low-power core and all the software infrastructure exists.

They just need to start producing sea-of-core products using Jaguar derivatives. At this point in time, ARM would definitely be a step backwards as well as a distraction of limited resources and focus.

But AMD might decide to go after the handset market, ala qualcomm and broadcom style, and for that they would do fine just dropping in a bog-standard ARM core surrounded by their own IP (specifically the GPU IP).

Really? There's STB software for x86, and TV software for x86?
There are many markets where the software infrastructure does NOT exist for x86, and where ARM is currently found, and where GPU focused ARM products may be very beneficial (leveraging their ATI assets).

 

[Idontcare]

Really? There's STB software for x86, and TV software for x86?
There are many markets where the software infrastructure does NOT exist for x86, and where ARM is currently found, and where GPU focused ARM products may be very beneficial (leveraging their ATI assets).

I was speaking loosely, not "all" as in "ALL, 100%, every single one", and I was intending to only discuss the seamicro marketspace (dense servers). But your point stands, and is more technically correct than mine, I concede.

 

[podspi]

They just need to start producing sea-of-core products using Jaguar derivatives. At this point in time, ARM would definitely be a step backwards as well as a distraction of limited resources and focus.

But AMD might decide to go after the handset market, ala qualcomm and broadcom style, and for that they would do fine just dropping in a bog-standard ARM core surrounded by their own IP (specifically the GPU IP).

I think (hope) we will know more after the 'strategy announcement' next week. I think AMD has woefully underutilized Bobcat. It is a great product, but they essentially stuck it in Netbooks and let it languish in overpriced 15.6" Acer laptops. Being build on Bobcat, I am sure Jaguar will be solid, but the real question is how will it compete with the 22nm Atom?

Interesting...so AMD's plan could be to become a foundry of IC SoC designs.

Instead of fabless companies going to TSMC with a design in hand for production we will see a rise of not just fabless, but also design-less, companies that want someone to design the chip and someone else to build the chip, and at the end of the day the design-less company is just paying royalties to resell the final chip in whatever product the chip is getting bundled into.

Isn't that how Intel got into x86 in the first place? Some company came to Intel asking them to develop a chip for electronic typewriters (and produce the chip too) and they didn't care what Intel did in terms of recycling the IP in future chips?

I could see that working for AMD. Companies that want to compete with qualcomm but don't have the revenue to justify keeping an entire design team on the books year round. Use a design foundry, then use a fab foundry, get your chips black-box style.

This is the impression I am getting from the rumors. I think AMD's biggest issue is that it isn't entirely clear how much that other IP is actually worth. Looking at the mobile GPU market, it almost seems to me that one of their best assets with regards to IP is their experience in software and drivers. They are (relatively) good at getting their GPUs to work with Windows and Windows programs. Take all that away, and are their Radeon graphic cores that good compared to the competition? Same thing with Tegra and how Geforce mobile graphics did not blow everybody away...

 

[Olikan]

Interesting...so AMD's plan could be to become a foundry of IC SoC designs.

Instead of fabless companies going to TSMC with a design in hand for production we will see a rise of not just fabless, but also design-less, companies that want someone to design the chip and someone else to build the chip, and at the end of the day the design-less company is just paying royalties to resell the final chip in whatever product the chip is getting bundled into.

Isn't that how Intel got into x86 in the first place? Some company came to Intel asking them to develop a chip for electronic typewriters (and produce the chip too) and they didn't care what Intel did in terms of recycling the IP in future chips?

I could see that working for AMD. Companies that want to compete with qualcomm but don't have the revenue to justify keeping an entire design team on the books year round. Use a design foundry, then use a fab foundry, get your chips black-box style.

there is a thread on semiaccurate forums, for "more" information...
http://semiaccurate.com/forums/showthread.php?p=170352#post170352

 

[OBLAMA2009]

for their sake i hope its about servers and not mobile stuff. i predict a mobile overload in the near future

 

[piesquared]

I really hate announcements announcing announcements, they are far from news worthy.

SeaMicro has been working with ARM long before AMD aquired them.

And you think possibly having the CEO of ARM, Warren East, attend the conference call is far from news worthy? I sense panic from the short club.

 

[Idontcare]

there is a thread on semiaccurate forums, for "more" information...
http://semiaccurate.com/forums/showthread.php?p=170352#post170352

Cool :thumbsup: thanks for the link, just read it, sounds like a very smart move for AMD.

 

[sontin]

Take all that away, and are their Radeon graphic cores that good compared to the competition? Same thing with Tegra and how Geforce mobile graphics did not blow everybody away...

The competition is using much bigger Dies:
Tegra 3 is 85mm^2, A5 120mm^2, A5X 160mm^2. Samsung 45nm DualCore with Mali is 120mm^2 like A5.

 

[Blitzvogel]

Becoming an "integrator" of sorts makes for an interesting paradigm, especially when you can leverage your own experience and IPs to make another licensed product better for a specific task or market.

 

[Lonyo]

I was speaking loosely, not "all" as in "ALL, 100%, every single one", and I was intending to only discuss the seamicro marketspace (dense servers). But your point stands, and is more technically correct than mine, I concede.

Well, my point mainly came due to reading the ARM financial statements yesterday which showed significant increases in marketshare in some areas, specifically TVs and STBs, as well as much less relevant areas like harddrives.

If you want to look at the future potentials of ARM, the financial statements have a lot of general industry discussion in them talking about segment penetration, growth, market size, chips per device and future prospects.

 

[MisterMac]

Becoming an "integrator" of sorts makes for an interesting paradigm, especially when you can leverage your own experience and IPs to make another licensed product better for a specific task or market.

This. x 10.
I think this will be a major winning strategy in the future.

Unfourtunately that pretty much also means withdrawal from the dedication to it's current major markets.
Which won't exactly be great for competition and technology.