Intel Develops 22 nm Atom Silvermont CPU

[Guest]

Intel was not kidding and is apparently already working on a 22 nm Atom CPU that will be taking advantage of the company's recently detailed Trigate transistor technology.

Intel Develops 22 nm Atom Silvermont CPU : Read more

 

[Guest]

wasabiman123: I don't see how using 22nm to achieve the performance and power envelope that ARM can get on 45nm is something to "YAY" about.

 

[PudgyChicken]

I'm too exhausted after giving you corrections on your last article to proofread this one as well. The word of the day is "pathetic."

 

[alikum]

[citation][nom]cynical_douche[/nom]wasabiman123: I don't see how using 22nm to achieve the performance and power envelope that ARM can get on 45nm is something to "YAY" about.[/citation]
Not to mention Brazos will be eating Atoms up in the netbook market. If they eventually decide to move to mobile segment, we may see a 3 corner fight

 

[schmich]

[citation][nom]cynical_douche[/nom]wasabiman123: I don't see how using 22nm to achieve the performance and power envelope that ARM can get on 45nm is something to "YAY" about.[/citation]
Except that an Atom kicks ARM's arse on a per clock basis. It might consume more power but it can do A LOT more work.

 

[johnners2981]

@pudgychicken, don't stop. I'm getting some valuable english lessons from you.

 

[whatisupthere]

[citation][nom]schmich[/nom]Except that an Atom kicks ARM's arse on a per clock basis. It might consume more power but it can do A LOT more work.[/citation]

I agree with you schmich. Its interesting that so many people see the "1ghz" and forget forget about the differences between an arm risc and an intel cisc chip. I for one am really curious how they will compare in performance at similar power usages.

 

[JamesSneed]

I think it would be more interesting if Intel licensed ARM's latest Cortex A15 CPU's and made one on this 22nm process. Its not like Intel hasn't done that before if anyone recalls Xscale.

 

[ta152h]

[citation][nom]PudgyChicken[/nom]I'm too exhausted after giving you corrections on your last article to proofread this one as well. The word of the day is "pathetic."[/citation]

I'd say the word of the day would be "obtuse". It does not flow at all, and is grueling to get through, even forgetting the grammatical issues. On a purely stylistic basis, it offends and disgusts.

 

[darkguset]

[citation][nom]PudgyChicken[/nom]I'm too exhausted after giving you corrections on your last article to proofread this one as well. The word of the day is "pathetic."[/citation]

LOL! What is this about?

 

[vk_87]

Any chance of these coming for the desktop?

 

[silverblue]

Cnet reports that Intel is accelerating the Atom roadmap beyond Moore's Law, which is an unheard of event for Intel all by itself.

This is not exactly difficult considering the first iteration of Atom was supposed to have a 5-year refresh. Silvermont onwards will be refreshed completely every 2 years.

This might be one of those scenarios where Intel may not win on performance, but should do exceptionally well in terms of power consumption. We'll find out in two years, however they simply must go with an out-of-order architecture if they don't want to fall behind the curve too much, and that in itself adds complexity and increased power consumption, though not as bad as originally thought - Bobcat, anyone?

 

[rantoc]

[citation][nom]whatisupthere[/nom]I agree with you schmich. Its interesting that so many people see the "1ghz" and forget forget about the differences between an arm risc and an intel cisc chip. I for one am really curious how they will compare in performance at similar power usages.[/citation]

Agree with you both, would also be interesting to see the processors raw performance & power drain in different scenarios/workloads. For instance h264/mp3 software encoding/decoding, software compression (rar/zip), software ray tracing and software encryption etc between atom, arm and the latest sandy bridges.

Toms Hardware, seems you got work (if your up to it =), and if you decide to go with it - Please measure both with and without dedicated logic for the specific tasks (like the aes/encoding/decoding logic in the sandys)

 

[someguynamedmatt]

Good to see that they're trying to directly attack AMD's Fusion now. Makes me feel a little better knowing that Intel's the one fighting for once, if I'm right about this whole thing...

 

[Guest]

It's not about performance per clock, guys, it's about performance per watt. ARM exponentially kicks the crap out of Atom on performance per watt (or any other architecture). For that matter, Core2, PhenomII, and Core i? all kick the crap out of Atom on performance per watt, Atom has exceptionally poor performance per watt.

When you look at a diagram an Atom die, the CPU core(s) take up most of the die. Now look at an ARM CPU, the actual CPU core is only a tiny part of the SOC. They could easily throw 8 or 16 cores into one of those and it'd still be smaller than an Atom dual core on the same node, which is why ARM is having no trouble scaling their cores and clocks up.

 

[rantoc]

[citation][nom]haha_im_right[/nom]It's not about performance per clock, guys, it's about performance per watt.[/citation]

Depends on what the device the cpu powers are intended to do really. It makes no sense to have a IT staff idle 1-2 hrs a day while the equipment finish the workload to save a few dimes in electricity. On the other hand you don't want that staff to charge their phone battery twenty times/day either...

It will be interesting to see how much Intel will manage to scale down the x86. What they try to do is make a pretty complex cpu architecture with loads of one clock instructions available to the developers consume less fuel, if they manage to get it to very low levels things will get interesting in the hand held market.

I don't care who "wins" as long as my device have good performance, don't weight a ton to support a huge battery and don't have to recharge more than once a day!

 

[loomis86]

[citation][nom]rantoc[/nom]Depends on what the device the cpu powers are intended to do really. It makes no sense to have a IT staff idle 1-2 hrs a day while the equipment finish the workload to save a few dimes in electricity. On the other hand you don't want that staff to charge their phone battery twenty times/day either...It will be interesting to see how much Intel will manage to scale down the x86. What they try to do is make a pretty complex cpu architecture with loads of one clock instructions available to the developers consume less fuel, if they manage to get it to very low levels things will get interesting in the hand held market.I don't care who "wins" as long as my device have good performance, don't weight a ton to support a huge battery and don't have to recharge more than once a day![/citation]

depends on what device???? huh!

wtf do you think an atom is being used for? a freaking server? Get a clue.

 

[loomis86]

microcontrollers and system on chip is the future. Atom is not.

 

[rantoc]

[citation][nom]loomis86[/nom]depends on what device???? huh!wtf do you think an atom is being used for? a freaking server? Get a clue.[/citation]

Actually the normal atom today is used in several nas servers and the like already, many of whom actually have basic webserver / fileserver capability's that are quite enough for regular home users/small offices.

Get a clue what will happen now that Intel manage to get the current atom line more efficient and powerful. Not only by the die shrink but also the new 3d gate tech their rolling out. Sure as hell it won't take the place as the real workhorse/mission critical servers but it could more than well serve loads of other server usages.

I wouldn't be surprised if we see more server clusters with high numbers of weaker low-power cpu's rather than a few monolith ones for several reasons, powersaving by disabling the cpu's completely when in light load, better scalability especially in virtual server environments.

 

[alexanderl]

If Intel were really serious about this segment, they'd have 17nm chips, not 22nm in 2013. For some reason Atom is always 1-2 steps behind the mainstream.

 

[ProDigit10]

the latest ARM processors maybe a bit faster than atom,but the atom processor is definitely the one with lowest TDP when running windows! (Running Windows from ARM requires to run it in a virtual environment, and that sucks even more battery than on an x86 processor)!

The way I see it, ARM is about on par with the Atom processor, if both run a Linux system,and intel actually might be slightly better graphically speaking.

AMD's offer just consumes too much power for a netbook or tablet!

 

[Guest]

rantoc: Real NAS servers (like EMC and Netapp) sure as heck don't run Atom or ARM. NAS "Servers" for home use don't need much in the way of CPU power, a 386 would be sufficient... If anything, ARM draws less power and generates less heat, making it a more logical choice.

The idea that Atom CPUs are robust enough to penetrate the enterprise server market is laughable at best, and anyone who thinks otherwise clearly doesn't manage servers for a living. You'd be better off with a 16 core Bulldozer CPU than a 30 core Atom.

 

[quickmana]

A fundamental change to transistor design is huge, it's the beginning of a new revolution.

 

[Guest]

quickmana: Hi K metal gate transistors were probably more revolutionary than this, and yet the 45 and 32nm CPUs out now aren't really that much faster than the good old-fashioned Q6600 on 65nm, they just use slightly less power per MHZ.

 

[hannibal]

Well that is what competition means... No competition... not so good ATOM CPU's, now when there is some competition Intel has to upgrade producs more agressively... I hope that we newer see monopoly in future, but It may be vain hope...