AMD details first ARM-based server chip, codenamed "Seattle"

[Jos]

Late last year, following months of speculation, AMD finally revealed plans to build low-power server chips based on the ARM architecture. The move marked an interesting shift for the company in its attempt to challenge Intel’s dominance of the steadily growing data center market. AMD’s first 64-bit ARM-based chip isn’t expected to arrive until sometime in the second half of 2014 but this week the company offered up a few early details.

Codenamed Seattle, the upcoming server part will be built around ARM's 64-bit Cortex-A57 and a 28nm manufacturing process. It will be made available in an 8-core configuration initially, with a 16-core variant to follow later, and is expected to reach clock speeds of at least 2GHz. Performance-wise, AMD says Seattle should be between 2 and 4 times faster than its Opteron X-Series with improvements in compute per watt.

Other features mentioned in AMD’s presentation include 128GB DRAM support, extensive offload engines for better power efficiency and reduced CPU loading, server calibre encryption and compression, and legacy networking including integrated 10GbE. This will also be the first AMD processor to utilize its Freedom Fabric technology for high-density low-power servers, acquired from SeaMicro last year.

AMD is confident ARM designs will win out in the long run and expects the architecture to have double digit market share by 2016. That doesn’t mean they are abandoning their x86 roots anytime soon, however. The company also plans release the “Berlin” and “Warsaw” x86 processor lines in the first half of 2014.

It’ll be interesting to see how this move plays out for AMD in the server market and if / when any of it will translate to other product lines. After all, only a couple of months ago the company hinted at an eventual ARM-Radeon combo for the embedded market, it’s not unreasonable to think the same is possible for tablets and phones. That, however, is merely conjecture at this point and if anything such a move might still be a bit far off.

Permalink to story.

https://www.techspot.com/news/52972-amd-details-first-arm-based-server-chip-codenamed-seattle.html

 

[howzz1854]

I would love to see microsoft develop a ARM based OS variant for windows. I love the efficiency of ARM processors and would love to see it utilized on a desktop platform one day.

 

[VitalyT]

I would love to see microsoft develop a ARM based OS variant for windows.

For anyone living under a rock in the last couple years, Windows RT is there since last year, though it found little love out there - where have you been?

 

[Jad Chaar]

64bit ARM. Awesome.

 

[howzz1854]

I would love to see microsoft develop a ARM based OS variant for windows.

For anyone living under a rock in the last couple years, Windows RT is there since last year, though it found little love out there - where have you been?

Microsoft really needs to do a better job at advertising that's where I've been. instead of showing a bunch of crazy nuns dancing on the table, they should be advertising the efficiency and capability of ARM based on RT that's where I've been.

 

[dividebyzero]

AMD's PowerPoint slide decks never fail to raise a smile...

"Industry's only 64-bit ARM Server SoC from a proven server processor supplier......Sampling in Q1 2014"

Kind of implies that Calxeda who are shipping 64-bit ARM Server SoC's now aren't proven ?
Nice slap in the face from AMD marketing considering Calxeda likely have more experience with ARM architectures than AMD.

 

[Mister_K]

Woo go Team AMD!

 

[Puiu]

I would love to see microsoft develop a ARM based OS variant for windows.

For anyone living under a rock in the last couple years, Windows RT is there since last year, though it found little love out there - where have you been?

Yeah they have RT, but it's not a desktop OS. When ARM will get good enough for mid to high range laptops we'll need something similar to Win 8, not Win RT. (but software for it will be a problem, so I don't think it will sell too well)

 

[lawfer]

I would love to see microsoft develop a ARM based OS variant for windows.

For anyone living under a rock in the last couple years, Windows RT is there since last year, though it found little love out there - where have you been?

Yeah they have RT, but it's not a desktop OS. When ARM will get good enough for mid to high range laptops we'll need something similar to Win 8, not Win RT. (but software for it will be a problem, so I don't think it will sell too well)

It's not only Windows RT. Windows 8 in itself supports ARM. RT was simply an attempt to provide superior battery life on a wide range of non-desktop form factors by stripping down certain core things.

So yes, the Windows 8 desktop OS does support ARM.

 

[Puiu]

It's not only Windows RT. Windows 8 in itself supports ARM. RT was simply an attempt to provide superior battery life on a wide range of non-desktop form factors by stripping down certain core things.

So yes, the Windows 8 desktop OS does support ARM.

This is the most important difference between Windows RT and Windows 8: Windows RT only runs on ARM-powered devices, while Windows 8 only runs on x86 devices.
They can't port the legacy software that makes up 90% of windows 8 (think paint, wordpad, etc) to ARM.

If they ported the entire windows 8 desktop to ARM then it's news to me and I would like a link so that I won't remain ignorant.

 

[lawfer]

This is the most important difference between Windows RT and Windows 8: Windows RT only runs on ARM-powered devices, while Windows 8 only runs on x86 devices.
They can't port the legacy software that makes up 90% of windows 8 (think paint, wordpad, etc) to ARM.

If they ported the entire windows 8 desktop to ARM then it's news to me and I would like a link so that I won't remain ignorant.

No, you're right in that RT only runs in ARM devices and Windows 8 in x86 machines. The point I was trying to clarify is that this software distinction was not in place because Windows 8 doesn't support ARM, but because if it did battery life would be affected.

So my point is that Windows 8 does support ARM, and that's in the form of RT. RT is the same kernel just without x86 support for, like I said, battery life reasons. At least that's what MS says.

 

[ET3D]

It would be easy to port legacy software to ARM, and if ARM catches on as a general purpose CPU architecture I'm sure we'll see that. For now there's a distinction which I think has more to do with marketing (where Microsoft wants to see the OS) rather than technical issues. Granted some of it was that the C++ compiler wasn't optimised enough for ARM, but it's still mainly an issue of image (Microsoft not wanting its ARM offering to feel too slow or power hungry) rather than technical limitations.

I do hope that Microsoft will open up the ARM side of the OS sooner rather than later.

 

[jraymp]

AMD - First to hit 5Ghz on stock speed.
AMD - First ARM-based server chip.
AMD - First to combine CPU & GPU in one chip. [APU]
AMD - First 8-core CPU.

What else AMD.... ?

 

[Maximum Payne]

AMD - First to hit 5Ghz on stock speed.
AMD - First ARM-based server chip.
AMD - First to combine CPU & GPU in one chip. [APU]
AMD - First 8-core CPU.

What else AMD.... ?

First dual core , First true native quad core, first 64 bit, first to reach 1ghz....

 

[dividebyzero]

AMD - First to hit 5Ghz on stock speed.

No. IBM's Power6 achieved that goal five years ago

AMD - First ARM-based server chip.

No. AppliedMicro and Calxeda already been there, done that.

AMD - First to combine CPU & GPU in one chip. [APU]

No. Intel's Sandy Bridge was demonstrated eight months before AMD's Zacate (Brazos)

AMD - First 8-core CPU.

No. AMD's FX-8150 launched October 2011. Intel's Beckton based Xeons (Nehalem-EX) launched in March 2010.

First dual core , First true native quad core, first 64 bit, first to reach 1ghz....

Yes. Now that's closer to the mark.

 

[Vrmithrax]

This is the most important difference between Windows RT and Windows 8: Windows RT only runs on ARM-powered devices, while Windows 8 only runs on x86 devices.
They can't port the legacy software that makes up 90% of windows 8 (think paint, wordpad, etc) to ARM.

If they ported the entire windows 8 desktop to ARM then it's news to me and I would like a link so that I won't remain ignorant.

No, you're right in that RT only runs in ARM devices and Windows 8 in x86 machines. The point I was trying to clarify is that this software distinction was not in place because Windows 8 doesn't support ARM, but because if it did battery life would be affected.

So my point is that Windows 8 does support ARM, and that's in the form of RT. RT is the same kernel just without x86 support for, like I said, battery life reasons. At least that's what MS says.

Windows 8 by itself does not actually support ARM hardware. It is a pure x86 native OS that has been completely rebuilt, but does actually still rely on many legacy-based software and driver relics from previous x86-based versions of Windows. I think where the confusion lies is in the fact that there were some major considerations and adjustments made to parallel coding with 8 and RT, and not just to keep the interfaces similar. They kept much of the basic APIs and libraries for new apps in a "platform independent" form, which was intended to create a situation where any app you made to run on the new "Metro" app you built would run on either RT or 8. In practice, it's turning out that pretty much any RT app you build can run on 8, but not every 8 app can run on RT. Some of that is by design from the developers, some of it is limitations of the ARM kernel. (I didn't know most of this until a buddy of mine involved in the Win8/RT development process filled me in).

I am assuming that Win8 & RT was just the starting salvo. If anyone remembers, original announcements of Win8 had it being all-encompassing and cross-platform. Reality crashed in, and it was obvious that was too ambitious to take on up front. I wouldn't be surprised if Win9 or 10 makes the full transition and is truly completely x86 & ARM cross-compatible, though. If nothing else, it would take the consumer confusion out of having 2 versions of what looks like the exact same OS...

 

[Jad Chaar]

No. IBM's Power6 achieved that goal five years ago

No. AppliedMicro and Calxeda already been there, done that.

No. Intel's Sandy Bridge was demonstrated eight months before AMD's Zacate (Brazos)

No. AMD's FX-8150 launched October 2011. Intel's Beckton based Xeons (Nehalem-EX) launched in March 2010.


Yes. Now that's closer to the mark.

I have a old Celeron D from like 2003 with integrated graphics. How was SB the first? Also, isnt the Power 6 not for consumers?

 

[3DCGMODELER]

They spelled Seattle wrong, Its spelled "Spokane"....

 

[spydercanopus]

Is there Windows RT server software like Access, Active Directory, or even IIS?

 

[dividebyzero]

I have a old Celeron D from like 2003 with integrated graphics. How was SB the first?

Whaaaaat? Celeron D is a CPU. The integrated graphics (2003 era would have been Brookdale/Springdale 845 or 865 chipset) was incorporated into the chipset of the motherboard.

Also, isnt the Power 6 not for consumers?

Immaterial. jraymp's claim was:

AMD - First to hit 5Ghz on stock speed

It makes no distinction between architectures or market sector. If consumer desktop CPU were the claim, then they would be correct...when/if it actually becomes available (as would claiming the first desktop CPU to have a TDP exceeding 200 watts...IBM's Power5+ would likely claim that particular record for all comers)

 

[Jad Chaar]

Whaaaaat? Celeron D is a CPU. The integrated graphics (2003 era would have been Brookdale/Springdale 845 or 865 chipset) was incorporated into the chipset of the motherboard.

Immaterial. jraymp's claim was:

It makes no distinction between architectures or market sector. If consumer desktop CPU were the claim, then they would be correct...when/if it actually becomes available (as would claiming the first desktop CPU to have a TDP exceeding 200 watts...IBM's Power5+ would likely claim that particular record for all comers)

Right.

 

[DarkMantle]

No. AMD's FX-8150 launched October 2011. Intel's Beckton based Xeons (Nehalem-EX) launched in March 2010.

No. Intel's first 8 cores Xeons were released on March 30 2010, AMD's first 8 cores Opterons were released March 29 2010. So... Yes, AMD first 8 cores by a day

 

[dividebyzero]

No. Intel's first 8 cores Xeons were released on March 30 2010, AMD's first 8 cores Opterons were released March 29 2010. So... Yes, AMD first 8 cores by a day

Sorry, but that is incorrect. AMD's Opteron 6100's (Magny-Cours) aren't 8 (or 12) core CPUs- they are two 4 or 6 core CPU's on the same piece of substrate (in other words a multi-chip module or MCM). If you take MCM's into account then IBM was doing "8-core" CPUs 10 years ago.

 

[lawfer]

Windows 8 by itself does not actually support ARM hardware. It is a pure x86 native OS that has been completely rebuilt, but does actually still rely on many legacy-based software and driver relics from previous x86-based versions of Windows. I think where the confusion lies is in the fact that there were some major considerations and adjustments made to parallel coding with 8 and RT, and not just to keep the interfaces similar. They kept much of the basic APIs and libraries for new apps in a "platform independent" form, which was intended to create a situation where any app you made to run on the new "Metro" app you built would run on either RT or 8. In practice, it's turning out that pretty much any RT app you build can run on 8, but not every 8 app can run on RT. Some of that is by design from the developers, some of it is limitations of the ARM kernel. (I didn't know most of this until a buddy of mine involved in the Win8/RT development process filled me in).

I am assuming that Win8 & RT was just the starting salvo. If anyone remembers, original announcements of Win8 had it being all-encompassing and cross-platform. Reality crashed in, and it was obvious that was too ambitious to take on up front. I wouldn't be surprised if Win9 or 10 makes the full transition and is truly completely x86 & ARM cross-compatible, though. If nothing else, it would take the consumer confusion out of having 2 versions of what looks like the exact same OS...

Windows 8 and RT are, architecturally-speaking, the same OS. The underlying kernel is the same; in fact, code between 8 and RT is so similar, only really the exclusion of x86 support makes its difference apparent to both the consumer and the developer. Such is inevitable as RT is tailored to run under an ARMv7 instruction set architecture.

However, RT tablets have even been hacked to run x86 software, showing that the lack of this feature was implemented for both performance/platform reasons, and not for a limitation, or difference of the kernel.

Windows 8 was never announced as all-encompassing and cross-platform. If you recall correctly, RT was at the beginning called Windows On ARM or WOA. During an analyst Q&A MS was asked if the then-WOA supported x86 applications, to which I believe Sinofsky outright said no.

So the point isn't that RT is an OS built under a difference kernel that happens to operate very much like Windows 8's kernel and yet doesn't support x86 applications; it also isn't the case that Windows 8 was somehow supposed to be both ARM and x86 complaint all in one build. The case is that both Windows 8 and RT are essentially the same OS, but one is "optimized" for the SoC platform. x86 binaries are in fact still found in RT tablets, used by the OS to function, but not to be ran by the user. It seems that other than ARM-specific optimizations, MS restricted binaries--other than Office and IE, which are included--for compatibility, performance or platform (e.g. touch-input) reasons. Who knows.