10 Things You Need To Know About Intel’s New Atom

35

Comments

+ Add a Comment
avatar

d3v

Please fix this:

"Up against four competing ARM chips, Intel says at the *same power use*, Silvermont will be from 1.6 times to 2.3 times faster and consume from *3 to 5.8 times less power*."

avatar

roberto.tomas

The comparisons in the charts here are likely made against A9's since they are mostly dual cores. So they are also like 40nm. Power consumption is down in half since the switch to A15 and 28nm. This is a no-show for Intel .. optimistically the Silvermont will compare to the Tegra 4 @ 2Ghz.

Intel has to keep ahead in the die size in order to compete.. that is ending this year. For just about 6 months TSMC will be on 20nm. After that, late in Q4 2013, they are starting production on 16nm. Intel won't get to 14nm in the Atom line until Q4/2014 (it will see 14nm in the mainstream line Q1/2014 though) .. so unless Intel will *skip* 10nm and go straight to 7nm, it won't enjoy a power savings over ARM anytime in the next few years.

Almost all of that newer-processes by TSMC will be on ARMv8 (A57/A53) designs .. which do basically the sae thing Intel is doing here, relative to their power and compute projections. So basically, the way the market has been the past 5 years, is how it is going to continue for the next few. Assuming that the new peripheral compatibilities of the next gen ARM chips don't elevate them to compare well with mainstream laptop chips, that is.

avatar

Slowman

Intel sucks and atom sucks more! x86 is so inefficient that when a dud I know turns on his Wintel machine it throws the bearing out of the wheel in his electric meter. Intel might as well give up making chips and start selling yarn and knitting supply because they are so out of date. Just use the ultrapowerful A15 cortex chip in your workstation and desktops. Any company with any fore sight uses ARM and......What's that? The Ipad 6 will use Airmont? WTF is that!?! Oh it's an Intel atom chip. Weeellllll, of course they will. Intel has massive R&D and are years ahead of everyone, years I mean decades, when it comes to their fabrication process. I have Moore's law tattooed on my chest and further more...

It doesn't matter how good Intel makes the atom. And it doesn't matter how many phones they sell in Kenya. They need some Flagship products using their chips to be taken seriously.

avatar

Hey.That_Dude

And when 64 bit ARM comes out who will perform better?

avatar

fteoOpty64

Until such time as Intel published some audited benchmarks of the performance and battery life from an exactual prototype tablet or phone, then the real comments can be seen.
It has taken Intel almost 7 years to evolve to the state they claimed it will do. I think it is possible but I think a lot of hype has been infused in those powerpoints. A video of the prototype board with some preliminary numbers of known benches would be broad indicators of such claims.
We will have to wait for some official benchmarks before anyone can be convinced. Judgement reserved until evidence comes ....

avatar

fygquas

If actual performance is not known, how can you know that "it won't suck"?

Please...

avatar

thematejka

Sounds interesting. Given Intel's history, it is only a matter of time before this kind of chip outperforms AMD. I like my AMD chips, but Intel's are often way tastier ;).

But seriously, I'm just happy to see Intel pursuing this Atom thing. As long as AMD has something to battle against (and vice versa), we will see more development. Intel and AMD are both great chip makers, they need each other to keep developing better hardware.

I'm not mentioning ARM because if x86 can prove worthy of whatever efficiency, then rock on all three of you (Intel, AMD, ARM). Rock on.

avatar

Chronologist

Again... The point of this Atom is to target the smartphone and mobile platforms. Gordon has gone from subtle innuendo to outright stating that this is targeting ARM. Intel's aiming straight at the likes of the Samsung Exynos, the heterogeneous octo-core mentioned in the 2nd point. This situation is x86 vs ARM. Intel offers the new Atom, AMD is putting out the Z60.

avatar

Refuge88

It wasn't just Gordon, Intel in no so blunt words said the exact same thing. They are purposing this one directly for Smartphones, then tablets, and then they plan on using its modularity to create low power server chips.

Although they have tried before, I feel this is their first true step into the mobile market. (Competitively atleast)

avatar

Jared G

They can't possibly think of competing with trinity. They are out of their league in this portion is of CPUs.

avatar

Chronologist

Ah yes, because an Atom is supposed to compete with 100 watt TDP CPU's as a smartphone/tablet SoC.

If they pulled that off, ARM can just quit and off themselves.

avatar

Refuge88

This isn't aimed at the desktop market at all, thats for the core series. Which draw WAY more power than these are *Expected* to. As well have much different pipelines, and throttling parameters that sacrifice uneeded perforamce for large efficiency gains.

Or atleast thats how I understand it. Correct me if I'm wrong please. :)

avatar

Morete

Well, Trinity is currently at 32 nm. If there is a way that this APU could scale down to 14 nm, I don't see why it wouldn't fit into a handheld device.

avatar

Chronologist

It's harder than you think. Global Foundries won't be at that level for at least a good year, and likely longer. Fabs are expensive, you know.

avatar

Peanut Fox

I think we'll here news about what AMD is up to very soon. As it is there's a road map showing Kaveri replacing Trinity APUs, with Kabini and Samara going into tablets and smartphones.

I think late this year and early next things are going to get very interesting in the mobile space.

avatar

astrocramp

Microsoft must have seen the roadmap when developing Win8. If all this Silvermont dreaminess really pans out, we'll get super thin ARM-like hardware with almost i5 performance. There will be almost zero reason to use a limited ARM device when you can take advantage of similar form factors with a full featured OS.

avatar

Refuge88

Price is a big reason.

avatar

Peanut Fox

If that's the case why did they even bother trotting out RT? I'm sure they had predictions, but if the knew Intel and AMD were going to have decent low power chips that could compete with ARM I don't think they would have bothered with that experiment.

avatar

QuadraQ

I guess the real question will be if Intel can overcome the terrible CISC instruction set that is x86, compared to ARM's clean RISC instruction set. Ultimately, I think those fundamental technological differences will matter. The relatively open nature of ARM is another factor, but not a technological one.

avatar

wumpus

Intel overcame the kludge that is x86 compared to the clean RISC competition (for non-power constrained devices) with the P6 (Pentium Pro & Pentium 3). This was Intel's first Out-of-Order x86. Of course, Intel also was riding the mass volume and software base so it could take on all RISC comers with somewhat less performance than an Alpha, and still customers would jump ship to run high performance applications on a wintel base. The shoes a bit on the other foot now and Intel has to convince phone makers of clear superiority as selling Office on a phone isn't going to work (and with Surface RT, it's cheaper on ARM anyway).

So if the switch from in-order to out-of-order in non power-constrained devices has any indication on power-constrained devices, I would expect Intel to be mostly clear of the x86 kludge. One thing to remember is that both ARM and AMD64 have 16 registers: that was a large part of the x86 kludge post-P5. A lot of the x86 decoding mess can be quickly handled by burning a few bits in the L2 to split the code between decoders, and I don't expect that decoding for A15 is quite as straightforward as when ARM was new.

I've heard that ARMs have some software differences, but judging from lack of screaming about Android I'm guessing that it is about the same level as Intel on the user level (Intel just loves to mask out features to make different SKUs). I don't think that any need to recompile system software is going to push phone makers to Intel.

avatar

xXhallelujahXx

So confused on how many Alias' this chip has : Silvermont, Silvermount, Silverton and Silverston.

Whoa!

avatar

harz3000gt

BLAH BLAH BLAH! Will it play Crysis?

avatar

vrmlbasic

#10 is impossible. The Atom will always suck, the atom has always sucked. If Intel wants to create a chip that doesn't suck then they must change the name.

The performance-draining prowess of the Atom brand name is so strong that if they branded the top-of-the-line pending Haswell chip an Atom it would plummet to Pentium 4 levels of performance.

avatar

gordonung

I agree. I have suggested to Intel several times that the Atom-brand is pretty burnt to people. "Atom, that's the crappy CPU I got in the Netbook from Verizion! That thing sucks!" We will conduct tests with Atom to see if the negative performance of it will actually slow down other CPUs in the general vicinity.

avatar

Refuge88

Only people who can't read archetectural breakdowns on Anand or even read a decent benchmark would turn away from a high performance atom.

To the people who pay attention to their hardware, its always the facts that matter not a name.

avatar

gordonung

Yeah, but we're not talking MPC readers. The mass public remembers brands and what burned them before. They won't see Atom and think new improved Atom. For what we know, Intel may rebrand it anyway. It could be Atom II or Atom i7

avatar

crwlngkngsnk

Yup. Like I remember the original Celeron.

avatar

Peanut Fox

I like Atom II, or fold it into the current naming convention and call it an Core i2. Atom i7 goes against the performance moniker it stands for.

avatar

foxmuldr

"Up against four competing ARM chips, Intel says at the same power use, Silvermont will be from 1.6 times to 2.3 times faster and consume from 3 to 5.8 the power."

"At the same power use"
--and--
"from 3 to 5.8 the power"

??

avatar

jimmthang

 

Sentence should read, "Silvermont will be from 1.6 times to 2.3 times faster and consume from 3 to 5.8 times less power."

 

:)

 

avatar

foxmuldr

There is no comparison between anything x86-related and ARM in terms of power savings versus performance. x86 has always been faster on the high-end. It's also been more power hungry. That may be changing, but at the low end, when it's mostly in idle, ARM absolutely buries x86 in terms of power consumption.

How much did Intel pay you to write this piece?

avatar

gordonung

The truth is out there Fox.

avatar

Peanut Fox

All the tech sites are posting about Silvermont.

avatar

n2ic8k

do you guys ever proof read your articles?
" but the new [Silvermont] version of Atom should offer far more performance ever seen."

avatar

jimmthang

d'oh! Thanks for pointing that out. Should be fixed now! :)

Log in to MaximumPC directly or log in using Facebook

Forgot your username or password?
Click here for help.

Login with Facebook
Log in using Facebook to share comments and articles easily with your Facebook feed.