Microsoft Licenses ARM Architecture

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
ARM and Microsoft Corp. today announced that they have signed a new licensing agreement for the ARM® architecture. The agreement extends the collaborative relationship between the two companies. Since 1997 Microsoft and ARM have worked together on software and devices across the embedded, consumer and mobile spaces, enabling many companies to deliver user experiences on a broad portfolio of ARM-based products.

“Microsoft is an important member of the ARM ecosystem, and has been for many years,” said Mike Muller, CTO ARM. “With this architecture license, Microsoft will be at the forefront of applying and working with ARM technology in concert with a broad range of businesses addressing multiple application areas.”
 
Meanwhile a few more cracks appeared in the x86 ivory tower.
Not really, at least in this case. ARM was and still is used in many consumer devices and embedded products. What this means is ARM is holding its ground in the segments it rules. The smaller part of this IMO is to counter Android smartbooks. Windows mobile/CE is desperately in need of a major update and it doesn't seem to be coming until next year.

At this point Intel is probably reassessing Atom's embedded future. It ain't going into masses of cell phones, MP3 players, embedded devices or Apple products/clones.
 
The atom isn't a *bad* processor, but intel had to burn a ton of die space on the x86 decoder, space that burns power. The cortex a9 should easily embarrass this processor, no matter what intel does with it.

I'm guessing MS is finally going to get into the handheld gaming market. I think anyone with a clue should realize that people prefer portable gaming versus sitting anchored in front of a TV.
 
The atom isn't a *bad* processor, but intel had to burn a ton of die space on the x86 decoder, space that burns power. The cortex a9 should easily embarrass this processor, no matter what intel does with it.
Only in power consumption. In performance, the lowly Atom 1.6GHz still runs rings around even a dual core 2GHz A9, according to the company making high density, low power servers that can run on both Atom and ARM processors.

BTW, the Atom core, excluding L2 cache, is about 9mm^2 in area on a 45nm process. The decoder in Atom really isn't the problem in area or power consumption. The whole FEC minus L1 I-Cache occupies about 1mm^2, and the instruction decoder itself looks like it occupies about 1/2 that area. Silverthorne has a lot of die space used for interfacing with the north bridge, about 1/4 of the total area. For comparison, a single core Cortex A9 is about 1.5mm^2 in area in a 65nm process excluding caches. The difference is huge, but so are the capabilities. ;) Unfortunately for Atom, most of those things aren't really necessary for handhelds/handsets.

Even with 32nm SoC Atoms coming next year, I really can't think of any new customers who will be asking for it. Netbooks? Yes. Handhelds, handsets, embedded devices? Probably a few. Smartbooks? Probably fewer. An 8-10 hour or more portable ARM device = great. A 3-4 hour portable Atom device = lame.
 
One would think that intel would put the R&D in to build something that would work in this market. They must have determined that it wasn't profitable enough.
 
Even with 32nm SoC Atoms coming next year, I really can't think of any new customers who will be asking for it. Netbooks? Yes. Handhelds, handsets, embedded devices? Probably a few. Smartbooks? Probably fewer. An 8-10 hour or more portable ARM device = great. A 3-4 hour portable Atom device = lame.

3-4 hours on a portable atom device? Where did you get these numbers from?

I used to own a Viliv S5 and it gets 6 hours of life with an atom 1.33 and a 4.8 inch screen. The newer atoms will surely get better life than this. In a phone, this life would be even greater because of the smaller screen and slower cpu speed.
 
Another thing. The new atoms coming out in 2011 are supposed to use about 1/3 the power.

Here are some estimates for the new cpu with a 1500 mAh battery.

Intel estimates that Moorestown based devices will last about 5 hours when browsing the web on WiFi. Talk times are expected in the 4 - 5 hour range over 3G, and 8 - 10 hours on 2G.
 
No people, this is NOT a crack in x86's desktop hold. ARM is neat but they have nothing that comes anywhere near touching desktop chips performance wise. Their very highest end doesn't even keep up with Atom chips.

Also as a practical matter, the fixation with x86 is compatibility. If someone came and made a new CPU that was just as fast as an i7 but cheaper and used less energy that'd be great. However if said CPU used a different instruction set, it wouldn't be so useful. All your stuff, OS, apps, etc, would have to be recompiled for it and tested. As you might guess that's a lot of work and anyone who plays games should know how lazy developers are about updates (tell me again how many 64-bit native games there are).

ARM has a firm hold in the mobile and embedded markets and will continue to be huge there. However at this point they seem to have no interest in designing desktop CPUs. This is not a real surprise, it is a highly competitive market and would be hard to break in to, even if they had an x86 license. Without x86, well nobody would care.

I know it is trendy to hate on Intel, but they really are one of the very best chip design firms in the world. What that means is the performance you get out of them isn't something someone else could easily beat. It isn't like ARM could come in and magically make a desktop chip 50% faster than a i7 that uses half the power and costs $50. There are real limits to performance on a given lithography.
 
3-4 hours on a portable atom device? Where did you get these numbers from?
Um, Anand Chandrasekher, head of Intel’s Ultra Mobility Group. But maybe you have a better source than the guy in charge of Atom. :p

Two months ago in a talk he pointed out this for optimistic Moorestown platform:
As well as a 50-times idle power-use reduction, enabling more than 10 days of standby, audio and video improvements should provide up to two days of audio playback and 4-5 hours of browsing and video battery life, he said.
It's puffed up, so I put more realistic figures in my post. Sorry, there is not going to be a small factor PMP that gets 48 hours of music playback, even if the screen is always off. It might be possible with an old, bulky Archos-style brick, but nobody would buy one.

If you managed to get 6 hours on your device doing something, great. For the same form factor and battery size, ARM does a lot better.
 
Back
Top