Microsoft Tells ARM Partners to Pick Notebook Vendors

[btarunr]

Windows (PC) will make its first transition to a machine architecture other than x86 in decades with Windows 8 Windows on ARM (WOA), and Microsoft wants to make absolutely sure that it has a well-oiled ecosystem in place to propel its growth. Currently, Microsoft picked three potent players among ARM processor vendors, Qualcomm and Texas Instruments (that have experience and can ship in Zerg volumes), and NVIDIA (which has demonstrated a lot of engineering potential with its latest Tegra products).

Microsoft reportedly asked the three ARM players to pick two notebook vendors each (one major, and one minor) with which they will work to develop some of the first WOA portable computing devices. Qualcomm selected Samsung and Sony, Texas Instruments chose Toshiba and Samsung, while NVIDIA chose Acer and Lenovo. Among these, Samsung, Toshiba, and Lenovo are the major partners. Surprisingly, Taiwan-based companies have an insignificant role in this ecosystem. ASUS, which has thus far been the largest client of NVIDIA for Tegra processors, has been left out. Now that downstream partners are selected, upstream ODMs such as Quanta Computer, Compal Electronics, Wistron and Pegatron Technology, which manufacture for those companies, are getting their R&D teams in shape to compete for the next-generation platform. The finishing line of ARM's marathon run to get into PCs is in sight.

View at TechPowerUp Main Site

 

[Andrei23]

Good thing or bad?

 

[btarunr]

Good thing or bad?

Great thing. It will add choice to the market, and break the x86 duopoly of 80:20 competition between Intel and AMD with PC processors. AMD was always a charity Intel ran (by giving away instruction set, SRAM, and foundry process advancements, albeit delayed) so Intel could stay away from anti-monopolistic regulations.

NVIDIA always wanted to be an x86 processor vendor, but Intel wanted it to stay away, simply because NVIDIA has 10 times the resources as AMD, and can make huge investments to R&D and come up with an x86 processor that's actually competitive with Intel at every level.

The only way NVIDIA can now get back at Intel is Microsoft Windows (PC) supporting another machine architecture than x86. It's done some great work with Tegra so far.

 

[f22a4bandit]

Much faster than expected. Microsoft is making a significant push to dent x86's stranglehold it seems.

I like the choice of Samsung as a major vendor. Surprised that ASUS didn't make the cut with Nvidia.

Good luck, it'll be interesting to watch!

 

[dieterd]

if this is gona be good then there's gona be a price cuts in Intels ultrabooks, no mather what intel wants or says....
At last something is going against intel and my wallet will be happy about it no matter what i will choose

 

[dir_d]

Much faster than expected. Microsoft is making a significant push to dent x86's stranglehold it seems.

I like the choice of Samsung as a major vendor. Surprised that ASUS didn't make the cut with Nvidia.

Good luck, it'll be interesting to watch!

Microsoft is covering its bases and stopping apple at the same time. They could care less if its x86 or ARM they just want to take over another market which they have slipped in. To reiterate its not about x86 or ARM its about being #1 everywhere. They are adapting and playing the game that apple has laid before them.

 

[TheoneandonlyMrK]

wave of doom ahoy , well maybe

i certainly dont long for nvidia forceing their hand in cpu land physx style as their closed door attitude to open standards does my head in its against the concept of pcs

 

[kid41212003]

Great thing. It will add choice to the market, and break the x86 duopoly of 80:20 competition between Intel and AMD with PC processors. AMD was always a charity Intel ran (by giving away instruction set, SRAM, and foundry process advancements, albeit delayed) so Intel could stay away from anti-monopolistic regulations.

NVIDIA always wanted to be an x86 processor vendor, but Intel wanted it to stay away, simply because NVIDIA has 10 times the resources as AMD, and can make huge investments to R&D and come up with an x86 processor that's actually competitive with Intel at every level.

The only way NVIDIA can now get back at Intel is Microsoft Windows (PC) supporting another machine architecture than x86. It's done some great work with Tegra so far.

Good points.

 

[Steevo]

2-5 year adoption rate for any desktops, all these vendors are mobile devices. Once it works and we have the processing power users require from these we will see the move.


2 vendors kissing Samsungs ass? Amazing.

 

[Completely Bonkers]

I'd love to see a REALLY SMART Windows 8 with ARM kernel and x86 libraries. How about that? Intel should stick an ARM into an x86 as not just "another core" but as the primary scheduler. Always on. Low power. etc. Then x86 cores would come live to execute x86 libraries. That would give Intel the best of both worlds, and a time for ARM applications to be designed and compiled as "low power ARM only footprint". Others applications could use a hybrid of ARM + x86 runtime, or legacy, just x86

 

[Steevo]

Why when tests done earlier this year prove that current ARM offerings are 2-3X SLOWER at clock than a AMD mobile chip, and Intel is raping AMD to the tune of 20-30% on IPC, so slowing down and making a efficient chip isn't to hard to do, especially as they develop more energy efficient transistors and the process moves down in size.


We have the convergence of process, technology, materials, and architecture happening.

 

[Completely Bonkers]

Because with an ARM core you could have a wind-up computer ;-)

 

[Marv]

I'd love to see a REALLY SMART Windows 8 with ARM kernel and x86 libraries. How about that? Intel should stick an ARM into an x86 as not just "another core" but as the primary scheduler. Always on. Low power. etc. Then x86 cores would come live to execute x86 libraries. That would give Intel the best of both worlds, and a time for ARM applications to be designed and compiled as "low power ARM only footprint". Others applications could use a hybrid of ARM + x86 runtime, or legacy, just x86

I can't see Intel purchasing a license to use a CPU architecture when they have the largest R&D dept of any CPU designer.

 

[Benetanegia]

Why when tests done earlier this year prove that current ARM offerings are 2-3X SLOWER at clock than a AMD mobile chip, and Intel is raping AMD to the tune of 20-30% on IPC, so slowing down and making a efficient chip isn't to hard to do, especially as they develop more energy efficient transistors and the process moves down in size.


We have the convergence of process, technology, materials, and architecture happening.

Current ARM chips are slower but also consume an order of magnitude less than AMD's "mobile" chip. Ontario is a 9W chip + chipset. ARM chips are compared in mW, not in tens of watts and they are SoCs. A completely different thing. And it's not as if Intel didn't try and fail for years now to make Atom a worthwhile mobile chip.

ARM chips are also much cheaper to produce and the competition ensures low prices and innovation.

ARM Cortex A15 is said to be 2x faster than A9 clock for clock, plus it's designed to go up to 2.5 Ghz, effectively making it 4x faster than current ARM chips. All while mantaining the same TDP bracket. The extent to which that is true remains to be seen, but considering the wattage advantage they have, they can add a lot of performance, by means of making the architecture less efficient, something Intel/AMD can't do. And again whether they get close to AMD/Intel remains to be seen, but IMO they have a greater advantage as they don't need to directly compete with AMD/Intel in raw performance, they only need to reach a certain level so as to be enough for mainstream use. Most people who buy cheap laptops, nettops and even desktops, would gladly get something that is half the price, as long as it is capable enough.

 

[NC37]

The switch is all good and grand. But as a former Mac user, I remember the days when Apple did the switch to RISC. Then again when they went to x86. It was never smooth. Apps were broken, companies didn't update quick enough, and you had people unable to upgrade till this was resolved. Heck the move from OS9 to OSX had the same issues.

Yeah this isn't a complete switch to a new hardware standard, but how long until that is? Windows 9, Windows 10? M$ better be planning for this. They need emulators in dev and ready. Hopefully they'll remember they did acquire VirtualPC. That was originally designed to run on Macs but it was from the time they had RISC hardware. Yeah it was terribly slow and buggy, but it could run CISC on RISC. M$ could refine it much better.

 

[Cruise51]

I'd love to see a REALLY SMART Windows 8 with ARM kernel and x86 libraries. How about that? Intel should stick an ARM into an x86 as not just "another core" but as the primary scheduler. Always on. Low power. etc. Then x86 cores would come live to execute x86 libraries. That would give Intel the best of both worlds, and a time for ARM applications to be designed and compiled as "low power ARM only footprint". Others applications could use a hybrid of ARM + x86 runtime, or legacy, just x86

You just blew my mind.

Doubt it would happen but it'd be awesome.

 

[Completely Bonkers]

Coming back to this thread a day later... and I see how do-able it is. It is happening already in laptops for GPU... a ultra low power weak GPU that drives 2D, and a higher power GPU for 3D. No reason not to do that with the CPU. YEs, the architecture is conceptually simple but a challenge to build and get the software right... but imagine... ALWAYS ON 24/7 PC. With battery backup in case of power failure. Running on 1xAA battery as a 24hr backup. No more need to worry about boot times or "instantish on". It would be permanently on. With a trickle of 1W power to keep the desktop up, and the HDD, SSD and 3DGPU sleeping. Run some heavy x86 runtimes and the x86 cores would kick in and deliver. Otherwise that ARM kernel would just quietly sit there. Your x86 browser would need to get those x86's running... until such time that the browser was x86-free and could run on just ARM code.

Main features of windows "Fusion", kernel, network stack, explorer, 2D desktop, USB, would be ARM. Everything "power" like DirectX, .NET etc would remain on the x86 libraries.

 

[laszlo]

at this rate of interest in ARM maybe we'll see in 2 years also desktop pc's;all depend on soft producers at the end; how fast they can adapt or worst scenario paid by those who have interest in both direction...

 

[Steevo]

Current ARM chips are slower but also consume an order of magnitude less than AMD's "mobile" chip. Ontario is a 9W chip + chipset. ARM chips are compared in mW, not in tens of watts and they are SoCs. A completely different thing. And it's not as if Intel didn't try and fail for years now to make Atom a worthwhile mobile chip.

ARM chips are also much cheaper to produce and the competition ensures low prices and innovation.

ARM Cortex A15 is said to be 2x faster than A9 clock for clock, plus it's designed to go up to 2.5 Ghz, effectively making it 4x faster than current ARM chips. All while mantaining the same TDP bracket. The extent to which that is true remains to be seen, but considering the wattage advantage they have, they can add a lot of performance, by means of making the architecture less efficient, something Intel/AMD can't do. And again whether they get close to AMD/Intel remains to be seen, but IMO they have a greater advantage as they don't need to directly compete with AMD/Intel in raw performance, they only need to reach a certain level so as to be enough for mainstream usee fully featured . Most people who buy cheap laptops, nettops and even desktops, would gladly get something that is half the price, as long as it is capable enough.

No offense but.


bla bla bla

http://en.wikipedia.org/wiki/Instructions_per_second

How much does it cost VS a more fully featured CPU with a larger instruction sets for more fully featured content.


And yes I realize that 80% of computers out there run the hardcore games like farmville. But also know that people tend to buy the best of what they can afford, and people buy cars that go fast only to drive 75 on the interstate just like everyone else.

 

[Benetanegia]

No offense but.


bla bla bla

http://en.wikipedia.org/wiki/Instructions_per_second

How much does it cost VS a more fully featured CPU with a larger instruction sets for more fully featured content.


And yes I realize that 80% of computers out there run the hardcore games like farmville. But also know that people tend to buy the best of what they can afford, and people buy cars that go fast only to drive 75 on the interstate just like everyone else.

IPS per clock per core:

AMD E-350 == 3.125
AMD Phenom 2 X6 1100T == 3.9
Intel Atom == 1.2
ARM Cortex A15 == 3.5
ARM Cortex A9 == 2.5

Seems like it is ok compared to competing products. It's better than both Zacate and Atom and consuming far less and being smaller/cheaper.

Regarding how much cheaper they are to produce:

Nvidia's Tegra 3 with 4x A9 cores (actually 5 counting the low power companion core) is 80mm^2 on 40nm node, Zacate, dual core, is 75 mm^2, same 40nm, but requires an aditional chipset at 28 mm^2. And most competing ARM chips are smaller than Nvidia's Tegra BTW.