backtop


Print 60 comment(s) - last by SlyNine.. on Sep 15 at 8:20 AM


Surprise! Microsoft's Internet Explorer 10 preview was gracefully running an ARM CPU, unbeknownst to the audience. Microsoft employees let this little secret out later at the conference.  (Source: Engadget)
Watch out Intel and AMD, power efficient ARM processors will soon be able to run Windows

At CES 2011, Microsoft Corp. (MSFT) CEO Steve Ballmer showed off an early build of a next generation Windows operating system running on an ARM architecture CPU.  This week at Microsoft's MIX Developer Conference in Las Vegas, the company gave developers a surprise Easter egg -- a preview build of Internet Explorer 10 and its underlying version of Windows were running on a 1 GHz ARM processor.

Samsung Electronics (005930), Texas Instruments Inc. (TXN), Qualcomm Inc. (QCOM), NVIDIA Corp. (NVDA), and other ARM chipmakers have all been hard at work cooking up power savvy multicore offerings, which would be perfect for a netbook or notebook.  

Versus similarly clocked x86 processors from Intel or AMD, ARM processors would likely squeeze out an hour or two of extra battery life.  While die shrinks and the ever-rising leakage current may eventually largely negate this advantage, in the short term ARM presents the first compelling consumer alternative to x86 in decades.

Windows 8 is expected to insert Microsoft's Ribbon UI element into more locations, including Windows Explorer.  It is also expected to have deeper touch integration and tie together the PC version of Windows with the Metro UI that Microsoft developed for the defunct Zune and Windows Phone 7.

But the addition of ARM support is perhaps the most anticipated feature.

While ARM currently offers power advantages, how compelling a buy Windows ARM portables will be still remains to be seen.  By offering base Windows support, including access to its Office suite and other enterprise tools, Microsoft makes ARM accessible to the everyday consumer.

But exactly how far Microsoft is able to go with its compatibility efforts remains to be seen.  If Microsoft can add ARM support for the Direct X and sound libraries, for example, it would be a relatively trivial exercise for developers to recompile their executables for ARM-architecture Windows 8 computers.

Microsoft makes the world's most used development environment, Microsoft Visual Studio.  By adding tools to make it quick and easy to switch from x86 to ARM builds, Microsoft could make applications compatibility complaints largely a moot point.  

Likewise, if Microsoft can embed an ARM-specific virtual machine in the OS with an x86 emulation layer, it might be possible to run native x86 apps, as is, without recompilation.  This would be helpful in cases where a company didn't have the source and the application developer was unresponsive or unwilling to make the change.  Implementing the same sort of system to provide ARM emulation in x86 Windows would be even more helpful to ARM, because it would allow developers to effectively target the more efficient ARM architecture, while ignoring x86.

Ultimately the question also still remains how low Intel can price its options and how big the true gap in power efficiency will be.  Unlike in the past, Intel may now find its pricing ability hindered by new international scrutiny that prevents it from resorting to anti-competitive arrangements to try to stomp out pesky rivals like ARM. But the exact picture is unclear.

Even more unclear is the fate of Microsoft tablets.  Even if ARM takes off in the notebook space, it may do little to help Microsoft sell Windows tablets, with Apple and Android so deeply entrenched.  In that regard, Microsoft may find that it's just given ARM a free ride to major expansion.  If that's the case Microsoft's customers should still reap minor gains -- a positive for the company -- but Microsoft itself may not make significant in-roads in its market expansion hopes.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Power vs clock
By Varun on 4/13/2011 2:41:08 PM , Rating: 2
Yes exactly. If Intel can get their idle power usage really low, then ARM doesn't really stand a chance. Even if Intel uses more power while active, the performance advantage would mean it would be able to get the task completed sooner, and back to an idle state much quicker. Just look at how much better IE9 is on battery life compared to the other browsers for the same reason.

I would guess that x86 will likely beat out ARM in the near future, just as it has done with all other RISC ISAs. Intel has only recently even cared about power usage, and look at the amazing differences in only the last couple of years.

The key with ARM is that software will have to be recompiled or even re-written to work with ARM. I just don't see this as a real threat, when x86 will likely just move down to the smaller platforms instead.


RE: Power vs clock
By DanNeely on 4/13/2011 2:45:36 PM , Rating: 2
Yup. This only makes sense in terms of MS's "tablet strategy", in that it lets them use the same kernel/os libraries/etc for both PC and tablet software. It doesn't change the need for a completely new GUI to be written; but it lets them share the rest of the OS stack. Depending on how well Intel's 32nm atoms do on power consumption it may become a moot point for windows tablets though.


RE: Power vs clock
By raddude9 on 4/13/2011 3:57:09 PM , Rating: 1
quote:
If Intel can get their idle power usage really low, then ARM doesn't really stand a chance


Yea, it's not like ARM could compete on price or integration of components.

quote:
Just look at how much better IE9 is on battery life compared to the other browsers


I don't get your point, IE9 was very marginally ahead of (the faster) Firefox4 in terms of battery usage, and that's from looking at miscosofts own research.

quote:
I would guess that x86 will likely beat out ARM in the near future, just as it has done with all other RISC ISAs.


Not at all, IBM's power7 is a RISC chip and it manages to get 264.96 GFLOPS per chip, but the best core-i7 chip can only manage 69 GFLOPS.

quote:
Intel has only recently even cared about power usage, and look at the amazing differences in only the last couple of years


I'm not sure what you mean again?
The Pentium 4 used 115W back in 2005 (7 years ago) and their current flagsip core i7 uses 130W.

Or maybe you mean low-power:
The ULV Pentium M used just 7W back in 2003 (9 years ago) which I admit is a bit more than the newest oak-trail Atom. But it's hardly what I would call amazing performance in the past decade! never mind couple of years.

Sounds like you are drinking too much Intel kool-aid.


RE: Power vs clock
By Da W on 4/13/11, Rating: 0
RE: Power vs clock
By vol7ron on 4/13/2011 5:02:58 PM , Rating: 2
Nope, I agree with him. x86 is inferior to RISC for the sheer fact that RISC is a smaller instruction set and x86 includes support for legacy systems (not always a bad thing, but really, really old systems).

With current silicon, the procs will hit a die-shrink wall and ARM will catch up. This is a fact, but we do realize that there are other feasible alternatives to silicon, so that might be for some time.

I'm curious how the Linux community will handle this. I'm not sure if there's a RedHat or Ubuntu support for ARM.


RE: Power vs clock
By omnicronx on 4/13/2011 5:32:54 PM , Rating: 5
Thats clearly a bias one sided way to look at things. Both RISC and CISC have their advantages/disadvantages, but to say one is inferior to another with little to no analysis other than instruction set is flat out incorrect.

The Power series in particular are not consumer desktop parts, they are clearly enterprise parts meant for high-end enterprise servers (these things are huge, I'm not even sure if you could even fit this thing in most desktops). So why the op would be comparing an i7 to an 8 core Power 7 I'm not sure.

Clearly not Apples to Apples.. Want to make a fair comparison, it would be the Nehalem-EX or Itanium line vs the Power 7. (and once again, both have their advantages and disadvantages, the Power 7 may have the best raw performance, but the both Intel lines are more flexible in terms of scalability and also cost a LOT less..)


RE: Power vs clock
By EclipsedAurora on 4/14/2011 1:10:40 AM , Rating: 2
quote:
The Power series in particular are not consumer desktop parts, they are clearly enterprise parts meant for high-end enterprise servers (these things are huge, I'm not even sure if you could even fit this thing in most desktops). So why the op would be comparing an i7 to an 8 core Power 7 I'm not sure.


Not really. All 3 consoles, PS3, 360 and Wii are powered by IBM POWER family CPUs. Also, POWER and Xeon competite in server field as well. IBM mainframe Z11 accept both Xeon or POWER7 as accelerator module. However it is obviously that Xeon competitely lose out to POWER no matter in performance or RAS! Xeon only have the price advantage!

quote:
Thats clearly a bias one sided way to look at things. Both RISC and CISC have their advantages/disadvantages, but to say one is inferior to another with little to no analysis other than instruction set is flat out incorrect.

I'm afraid to tell you that since original Pentium, Inel x86 CPU is an RISC core inside as well. All they did was just transcoding CISC input into RISC processing and than output back to CISC. That's part of the reason why so many transistor count and power consumption is wasted!


RE: Power vs clock
By SlyNine on 4/14/2011 10:12:28 AM , Rating: 3
Again you are not talking about the same CPU, All the CPU's in the consoles would get their asses handed to them by a Core I7,I5, or even I3. So try again.

You are over emphasizing the impact and amount of transistors that backwards compatibility has.


RE: Power vs clock
By Zingam on 4/23/2011 4:23:56 AM , Rating: 2
i3 is at least 5 years newer than any of the console chips!


RE: Power vs clock
By SlyNine on 9/15/2011 8:20:07 AM , Rating: 2
The I7 came out in 2008, the PS3 came out in 06. The I3 is based pretty much on the same tech and is a BUDGET cpu, the Cell was supposed to be a super highend CPU.

My old Core I7 1366 is still high end enough to beat the crap out of a brand new 2011 Core I3.


RE: Power vs clock
By vignyan on 4/13/2011 6:51:00 PM , Rating: 1
Whoa... I can tell that you don't really understand the complications in computer architecture... You think you do, but you don't... Trust me, I am a geek! :)


RE: Power vs clock
By Azethoth on 4/14/2011 12:59:40 AM , Rating: 3
No. You read an article in the early 90's about how RISC (Reduced Instruction Set) was going to dominate over CISC (Complex Instruction Set). Then when that did not actually happen you did not do your homework as to why.

The answer is that the instruction set you send to a CPU only makes up a tiny part of the die which is used to decode it into the actual instruction set used in the CPU. After that step there really is not much separating the two architectures / no intrinsic reason for separation. They can both make use of all the same tricks / latest innovations / design choices etc.

So instead of Intel dying, they beat RISC at its own game and relegated the "legacy" aspects of their design to a tiny tax that pales in comparison to the relentless tick tock of Moore's law at Intel.

Why they used their monopoly power to kick AMD in the face is a mystery to me. They compete quite well without such shenanigans.


RE: Power vs clock
By Strunf on 4/14/2011 7:58:39 AM , Rating: 2
Exactly, the latests CISC CPU are only CISC on the looks, once the decode is done it's just like any other RISC.

The difference I see is that on a CISC CPU the instructions are broken into smaller ones at the CPU level whereas in the case of RISC they are broken into smaller ones at the software level.


RE: Power vs clock
By Kary on 4/14/2011 5:36:12 PM , Rating: 2
Actually I am one of the ones who read the books in the 90s (I was studying electrical engineering specializing in computer systems).
As I recall they said that RISC chips were easier to design and smaller, but produced higher IO on the memory bus and required more RAM for the same program. The RISC instructions are typically function something like this:
Load Register A with RAM location x89030084
CISC chips were more complex to build and tended to be larger since they had to support more instructions, but a single instruction could do something like this:
Load from Memory location x03895783, Add Register A, Multiply by Register B, then save in Register C. Note, those don't have to be separate instructions...that could be done by ONE instruction.....1 read for the instruction...1 read of the data location is all the RAM access needed....versus one (or more) RAM reads per step for RISC.

And yes, Intel switched to RISC internally (the term "micro ops" is preferred here I believe...it's a CPU within the CPU so to speak) so they basically chose the best of both worlds.

...any chance of a complete article along these lines? Seems like a subject that comes up often and is gaining more attention.


RE: Power vs clock
By DanNeely on 4/13/2011 4:54:07 PM , Rating: 2
Average/idle power consumption is much more important than peak TDP in mobile devices. Intel's improved by large amounts at that in its most recent generation of mobile systems. Oaktrail is another major step forward for intel in that regard, being made on a low power (~10x lower idle power) process instead of a high performance (~10% faster) process like all their previous x86 CPUs.


RE: Power vs clock
By omnicronx on 4/13/2011 5:38:48 PM , Rating: 2
quote:
Not at all, IBM's power7 is a RISC chip and it manages to get 264.96 GFLOPS per chip, but the best core-i7 chip can only manage 69 GFLOPS.
And my 1992 Honda Civic only goes 120MPH while my Ferrari goes 400MPH!

You just compared a $10000+ 4GHZ 8 core Power 7 enterprise system vs a high end desktop chip that costs a fraction of the price.

Great comparison buddy, you clearly pulled that right from Wikipedia (yes others can use it too) and have no idea what you are talking about.


RE: Power vs clock
By raddude9 on 4/14/2011 7:35:03 AM , Rating: 2
The original comment said that x86 could beat every RISC architecture. Regardless of the performance statistics and price points, that's a very inaccurate thing to say


RE: Power vs clock
By SlyNine on 4/14/2011 10:16:01 AM , Rating: 2
So is basing your opinion on the theoretical flops number. But that didn't anyone.


RE: Power vs clock
By omnicronx on 4/14/2011 3:43:25 PM , Rating: 2
He said no such thing, you just interpreted it as such.

I'm sure he was implying on an Apples to Apples basis. (Notice how he talks about beating ARM in the near future, that clearly shows he is comparing mobile space architectures as ARM is nowhere close to Intel in the desktop space)

I was not necessarily implying you are wrong, in terms of RAW performance high end RISC chips such as the Power 7 series can be more powerful, but as stated in other threads, there are advantages/disadvantages to both.

So to merely post stats from Wikipedia and to make a Apples to Oranges comparison does not help anyone or prove anything.


RE: Power vs clock
By SPOOFE on 4/13/2011 5:42:16 PM , Rating: 2
quote:
Yea, it's not like ARM could compete on price or integration of components.

Sure, but on the other side of the coin they lack mindshare or consumer awareness. While their products are hugely ubiquitous, relative to Intel they're complete unknowns. Consumers don't get excited about the latest ARM product, they get excited by phones; meanwhile, you say "Intel" and most of them immediately know what you're talking about.

This is a balance to the playing field; ARM needs to be significantly better/cheaper in order to beat out Intel's marketing penetration.

quote:
Not at all, IBM's power7 is a RISC chip and it manages to get 264.96 GFLOPS per chip, but the best core-i7 chip can only manage 69 GFLOPS.

Power7 is a completely different market, though; it's not aimed at consumer-oriented devices, where Intel and ARM are going to do battle. And in that market, Intel has the advantage.
quote:
The Pentium 4 used 115W back in 2005 (7 years ago) and their current flagsip core i7 uses 130W.

Awful comparison; in 2005, most of Intel's P4's used north of 100 watts. Nowadays only their highest-end chips have TDP's that high.

quote:
Sounds like you are drinking too much Intel kool-aid.

Ah, I see you establish your bias at the END of your post instead of the beginning. Clever; I almost took your post seriously. :)


RE: Power vs clock
By raddude9 on 4/14/2011 7:40:29 AM , Rating: 2
quote:
Power7 is a completely different market, though; it's not aimed at consumer-oriented devices, where Intel and ARM are going to do battle. And in that market, Intel has the advantage.


Doesn't matter, the original poster said that x86 could beat very RISC architecture. My example is extreme, but it shows that it clearly does no such thing.

quote:
Awful comparison; in 2005, most of Intel's P4's used north of 100 watts. Nowadays only their highest-end chips have TDP's that high.


Is it, never mind high-end, the total amount of power used by mid-range chips has not moved that much in the last 5 years

quote:
Ah, I see you establish your bias at the END of your post instead of the beginning. Clever; I almost took your post seriously. :)


Thanks :-) I was only trying to rain on the parade of happiest intel cheerleader that I'd ever seen.


RE: Power vs clock
By SlyNine on 4/14/2011 10:20:32 AM , Rating: 2
Purely theoretical numbers mean nothing, prove nothing and since that is what you based your argument around (a clearly fallacious one at that) those numbers, Your argument means nothing/ proves nothing.


RE: Power vs clock
By SlyNine on 4/14/2011 10:22:22 AM , Rating: 2
I mean hell, by that account a 5770 for 100$ is so much better then either at 1 Tflop. Maybe we should just stick with GPUs and not have CPUs.


RE: Power vs clock
By raddude9 on 4/14/2011 10:31:20 AM , Rating: 2
If you look at the reason for my comment I wasn't trying to prove anything. I was merely disproving the original "x86 beats RISC" statement by the original poster. Why didn't you get annoyed by that statement instead? That was clearly more inaccurate and misleading than mine.


RE: Power vs clock
By vignyan on 4/13/2011 7:25:19 PM , Rating: 2
well let me also take a crack at you...
quote:
Not at all, IBM's power7 is a RISC chip and it manages to get 264.96 GFLOPS per chip, but the best core-i7 chip can only manage 69 GFLOPS.

This is the theoretical max limit as someone correctly pointed out. For that matter Power6 was also better than Nehalem-EX in the theoretical max, but performed poorly in actual benchmarks. Anyways, they have customized software that makes it good only for servers and nothing else runs on it... so not a fair comparison. Also, You don't understand the effect of RISC on performance - so stop attributing the theoretical max performance to RISC.
quote:
The Pentium 4 used 115W back in 2005 (7 years ago) and their current flagsip core i7 uses 130W.

Hmm... If you remember, it was for a single core. while the flagship Intel core i7 990X at 130W is a hexacore processor. So you see, there was a reduction from 115W per core to ~22W per core and it comes with performance gains. Appreciate that!

Ah... that was awesome... Let me get another kool-aid! :P


RE: Power vs clock
By raddude9 on 4/14/2011 11:02:58 AM , Rating: 2
quote:
Anyways, they have customized software that makes it good only for servers and nothing else runs on it... so not a fair comparison


So you think the original posters comment where he compared x86 to ARM was fair? Actually all I was trying to show was that the original comment of "x86 beats RISC" is not justified and it's easy to find numbers that show otherwise. With that in mind, there is no reason why the absolute performance of an ARM chip cannot match the performance of an x86 chip. There is nothing about ARM's RISC'ness that precludes this.

quote:
Hmm... If you remember, it was for a single core.


Again, the original poster did not mention performace, or performance per watt, just "power usage", and in terms of total power usage, Intel has gone from a "5W to 115W range" 10 years ago to about "2W to 130W" today, i.e. power usage has gone almost nowhere (ok, idle power has improved a bit, but that's a different story...).


RE: Power vs clock
By Varun on 4/13/2011 10:25:42 PM , Rating: 2
No, ARM likely can not compete on price or integration. Once again, Intel is the bigger player here, with the better fabs, and the best technology. They are always at least one fab cycle ahead of AMD, and that means more chips per wafer, and therefore lower costs per chip.

My point about IE9 was that if you have great performance, you get your work done faster and get your chip back to idle faster. You may use more power to get the work done, but over time, it is lower. My comment was not about IE vs FF, but thanks for that.

Others have already commented on your other amazing finds on Wikipedia, but still I find it funny that you counter my argument on Intel doing much better on performance per watt, with a P4 vs i7 TDP number. Great job there. You proved my point.


RE: Power vs clock
By raddude9 on 4/14/2011 7:54:04 AM , Rating: 2
quote:
No, ARM likely can not compete on price or integration.....They are always at least one fab cycle ahead of AMD, and that means more chips per wafer, and therefore lower costs per chip.


ARM does not need to.. ARM do not make the chips. And your point about AMD is irrelevant. In the market for phone and tablet chips, the way it is shaping up, Intel are not competing against a single chip maker any more, they have to compete against EVERY chipmaker that makes ARM chips, an entirely different proposition. It's also a market where price is more important, all that money spent on the latest fabs has to be recouped with higher prices.

quote:
I find it funny that you counter my argument on Intel doing much better on performance per watt, with a P4 vs i7 TDP numbe


Nope, you are remembering it wrong, your original post said nothing about performance-per-watt, all you mentioned was "power usage". If your point was about performance-per-watt you should have said that.


RE: Power vs clock
By EclipsedAurora on 4/14/2011 1:10:42 AM , Rating: 2
quote:
The ULV Pentium M used just 7W back in 2003 (9 years ago) which I admit is a bit more than the newest oak-trail Atom. But it's hardly what I would call amazing performance in the past decade! never mind couple of years.


Well, your information is still wrong. ULM Pentium 3 back in 2000 and 1st gen ULV Penitum M in 2003 use just 5W. Core2Duo SU series adds the power requirement to 10W, while the latest SandyBridge is using 18W.

However, the Atom speed is roughly equal to a 2004 5W 1.2GHz pentium M ULV only.

quote:
Not at all, IBM's power7 is a RISC chip and it manages to get 264.96 GFLOPS per chip, but the best core-i7 chip can only manage 69 GFLOPS.

U should metion IBM's Cell BE also used in PS3 as well!


RE: Power vs clock
By Da W on 4/13/2011 4:02:52 PM , Rating: 2
If Windows tablets catch on, then ARM netbook/notebook/desktop have a chance;
Else if Windows tablets fails, then ARM netbook/notebook/desktop fail.

By the way, Android 3.0 on tablets IS NOT DEEPLY ENTRENCHED IN THE MARKET AT ALL!!!! So far it's still Apple's playground.


RE: Power vs clock
By SPOOFE on 4/13/2011 5:49:34 PM , Rating: 2
quote:
Else if Windows tablets fails, then ARM netbook/notebook/desktop fail.

I don't think that's a given at all.

quote:
So far it's still Apple's playground.

Apple uses ARM.


RE: Power vs clock
By Hoser McMoose on 4/14/2011 11:24:44 AM , Rating: 2
quote:
Else if Windows tablets fails, then ARM netbook/notebook/desktop fail.

If tablets of any OS succeed, ARM chips will succeed in that market. Apple uses ARM. RIM uses ARM. Android uses ARM. Damn near everyone uses ARM.

The things are pretty darn ubiquitous these days. Even most PCs probably have at least one ARM core in them somewhere.


RE: Power vs clock
By Aloonatic on 4/14/2011 3:27:47 AM , Rating: 2
Your comment seems to have a lot of IFs in it. If Intel does this and that then...

Good points I'm sure, but you seem to be assuming that ARM have made their chips and are now down the pub, celebrating how great they are and waiting for the cash to roll in, assuming that Intel re not going to change much, so the don't have to.


"Folks that want porn can buy an Android phone." -- Steve Jobs














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki