Intel shows off 32nm chip, offers update on Penryn, Nehalem

[AppleInsider]

Presenting at its Intel Developer Forum in San Francisco on Tuesday, Intel Corp. demonstrated the industry's first 32-nanometer (nm) chip and also offered a status update on its upcoming "Penryn" family of processors and next-generation "Nehalem" architecture.

Chief executive Paul Otellini said chips based on its upcoming 32nm technology are due out in 2009 and will include transistors so small that more than 4 million of them could fit on the period at the end of this sentence. He also provided an overview of the advantages of the upcoming 45nm family of Penryn processors, which are based on the chipmaker's high-k metal gate transistor technology and slated to ship this November.

"Our tick-tock strategy of alternating next generation silicon technology and a new microprocessor architecture -- year after year -- is accelerating the pace of innovation in the industry," said Otellini. "Tick-tock is the engine creating today's most advanced technologies and keeps them coming out at a rapid cadence. Our customers and computer users around the world can count on Intel's innovation engine and manufacturing capability to deliver state-of-the-art performance that rapidly becomes mainstream."

Penryn

When Intel introduces Penryn in a few weeks, it will be the world's first high-volume 45nm processor, the company said. Along with the Silverthorne family of 45nm processors due next year, Penryn will have the small feature size, low-power requirements and high-performance capabilities to meet a wide variety of computing needs from handheld Internet computers to high-end servers.

Intel, which already claims to have achieved more than 750 design wins for the Penryn processor, plans to quickly ramp the technology, introducing 15 new 45nm processors by the end of the year and another 20 in the first quarter of 2008.

"We expect our Penryn processors to provide up to a 20 percent performance increase while improving energy efficiency," he said. "Intel's breakthrough 45nm silicon process technology allows us to provide low-cost, extremely low-power processors for innovative small form factor devices while delivering high-performance, multi-core, multi-featured processors used in the most advanced systems."


Otellini also announced that Intel's 45nm processors and 65nm chipsets would use halogen-free packaging technology beginning next year, yielding chips that will not only be more energy efficient but also better for the environment.

Nehalem

Looking to 2008, the Intel chief also conducted the first public demonstration of Intel's Nehalem processor and said the company is on track to deliver the new processor design in the second half of the year.

The Nehalem architecture was conceived with the company's leadership in performance-per-watt in mind, and will give way to the first chip to utilize Intel's QuickPath Interconnect system architecture, which includes integrated memory controller technology and improved communication links between system components to significantly improve overall system performance.

"Nehalem is an entirely new architecture that leverages Intel's Core Microarchitecture, bringing leading-edge performance advantages, power efficiency and important new server features to market just a year after Intel leads the industry to 45nm technology," said Otellini.

32nm technology

Describing other advanced Intel technologies destined to quickly come to market, Otellini showed the world's first 300mm wafer built using next-generation 32nm process technology. The development of advanced test chips serves as a critical milestone in the company's march toward high-volume manufacturing of 32nm process technology, which it plans for 2009.

Intel's 32nm test chips incorporate logic and memory (static random access memory --SRAM) to house more than 1.9 billion transistors, Otellini said. The 32nm process also uses the company's second-generation high-k and metal gate transistor technology.

This additional performance made possible by Intel's push to drive chip design and manufacturing technology forward will not only be seen in computing, but will enable more true-to-life entertainment and realistic graphics capabilities. As a result, the company said it will be placing increased emphasis on using the power of its processors to enhance key technologies such as visual computing and graphics.

"Satisfying demand for ever-greater computer performance increases means we need to move rapidly to the next manufacturing technology." said Otellini. "Intel engineers and researchers deserve a great deal of credit for setting the pace for the industry. As our advanced technology reaches consumers and businesses in the next couple of years the amount of computing power they'll be able to harness will help them become even more productive, creative and innovative."

25 watt Penryn

During his presentation, Otellini also announced that a version of a Penryn dual-core processor operating at 25 watts will be available on the upcoming Montevina platform, which will include Intel's mobile WiMAX silicon. He said several equipment manufacturers are already planning to introduce Montevina-based notebook PCs starting next year when the platform is introduced.

[ View this article at AppleInsider.com ]

[melgross]

One thing I don't understand from the article published on Anand, which is more detailed, is why Nehalen is coming out late 2008, rather than somewhat earlier. According to Intel, it taped out three weeks ago, and they have it up and running already. They demo'ed it with Windows, and said it was up on OS X as well.

I know this is a new process, but I would have expected to see this by June, latest. Consider that Intel is several months AHEAD of schedule with 45nm, as they have been for the past 18 months with everything else.

[Zandros]

Originally Posted by melgross

One thing I don't understand from the article published on Anand, which is more detailed, is why Nehalen is coming out late 2008, rather than somewhat earlier. According to Intel, it taped out three weeks ago, and they have it up and running already. They demo'ed it with Windows, and said it was up on OS X as well.

I know this is a new process, but I would have expected to see this by June, latest. Consider that Intel is several months AHEAD of schedule with 45nm, as they have been for the past 18 months with everything else.

I'm not sure why, but chips always seem to be released a year after tape out. I guess there's a lot of fine tuning to do, as well as getting the fabs up to date, etc.

Nehalem's where my next Mac is at.

/Adrian

[Haiduc210]

When they say Penryn will be dropping in a few weeks, does that mean that's when it'll first become available to manufacturers, or when we'll start seeing the first consumer products employing the new chips?

I ask only because I'm trying to gauge the amount of time it'll take for Apple to begin introducing these chips in their lineup. Would they be able to have designs finalized and in production in time for MWSF?

[palple]

I believe that if the Montevina chipset will be available early next year Apple will ship their first Penryn MacBook Pro with Montevina (say in March) then will switch to Nehalem in the fall (when I'll be on the market to update my iBook G4, and I'm not going to buy a MacBook as long as it will ship with integrated graphics).

[macinthe408]

Can someone reconcile the following statement:

"Looking to 2008, the Intel chief also conducted the first public demonstration of Intel's Nehalem processor and said the company is on track to deliver the new processor design in the second half of the year."

How does one "demonstrate" a new processor? Do they mean 'first public showing'? If it's a demonstration, do they use a desktop outfitted with the Nehalem and show the 'Tubes' screensaver running really, really fast, as you hear attendees whispering to each other, "Wow, notice how much faster the tubes are redrawn compared to the 45nm processors? Amazing!"

[Lafe]

Originally Posted by macinthe408

Can someone reconcile the following statement:

"Looking to 2008, the Intel chief also conducted the first public demonstration of Intel's Nehalem processor and said the company is on track to deliver the new processor design in the second half of the year."

How does one "demonstrate" a new processor? Do they mean 'first public showing'? If it's a demonstration, do they use a desktop outfitted with the Nehalem and show the 'Tubes' screensaver running really, really fast, as you hear attendees whispering to each other, "Wow, notice how much faster the tubes are redrawn compared to the 45nm processors? Amazing!"

They put a screen up showing a benchmark of floating point calculations, integer calculations, etc., compared to the previous (or different) chip.

For all you know, it could be a Keynote presentation, but it looks nice.

[melgross]

Originally Posted by Zandros

I'm not sure why, but chips always seem to be released a year after tape out. I guess there's a lot of fine tuning to do, as well as getting the fabs up to date, etc.

Nehalem's where my next Mac is at.

/Adrian

Not necessarily. Often chips go into production 3 months after tape-out.

this can easily be illustrated by the showing of actual computers running Nehalen with Windows, and being told that OS X was up and running on it as well.

We see this in the graphics industry all the time, where chips are in production before three months after tape-out.

I was expecting to see Nehalen in production by May, and possible new machines from Apple using them by the Dev conf.

Now, we may not see them until late fall, possibly later—Macworld, though I hope not.

[melgross]

Originally Posted by Haiduc210

When they say Penryn will be dropping in a few weeks, does that mean that's when it'll first become available to manufacturers, or when we'll start seeing the first consumer products employing the new chips?

I ask only because I'm trying to gauge the amount of time it'll take for Apple to begin introducing these chips in their lineup. Would they be able to have designs finalized and in production in time for MWSF?

Manufacturers.

[Splinemodel]

Originally Posted by AppleInsider

During his presentation, Otellini also announced that a version of a Penryn dual-core processor operating at 25 watts will be available on the upcoming Montevina platform, which will include Intel's mobile WiMAX silicon. He said several equipment manufacturers are already planning to introduce Montevina-based notebook PCs starting next year when the platform is introduced.

[ View this article at AppleInsider.com ]

About a week ago I made a post about what I thought it would take to make a really great subnotebook (or ultra-compact, or whatever). It looks like this chip answers two of the points.

On a separate note, this should coincide well with my hardware upgrade periodicity. So. . . iMac, Mac Pro, or MacBook Pro. . . Hmm.

[melgross]

Originally Posted by Splinemodel

About a week ago I made a post about what I thought it would take to make a really great subnotebook (or ultra-compact, or whatever). It looks like this chip answers two of the points.

Looks to be a good year, next year.

[TBaggins]

Think we might see Penryn Macs in time to coincide with the Leopard launch?

.

[melgross]

Originally Posted by TBaggins

Think we might see Penryn Macs in time to coincide with the Leopard launch?

.

Only if Leopard is delayed again by another two months or so.

[shanmugam]

what i wish to see, price drop in Mobile CPUs ...

if desktop and mobile are selling 50-50 then why mobile cpus should be priced higher than desktop parts...

Performance per watt
performance per price - whether desktop or mobile cpus ..

iMac/MacBooks will be cheaper ... then AIO iMacs makes more sense

[TBaggins]

Originally Posted by melgross

Only if Leopard is delayed again by another two months or so.

Check. Am gonna delay any Leopard Mac purchases in that case.

.

[David Stevenson]

Originally Posted by melgross

One thing I don't understand from the article published on Anand, which is more detailed, is why Nehalen is coming out late 2008, rather than somewhat earlier. According to Intel, it taped out three weeks ago, and they have it up and running already. They demo'ed it with Windows, and said it was up on OS X as well.

Short answer: bugs in the design. Even taking an existing design and "adding" a few features can introduce hardware bugs that can take a year to track down and iron out. Speaking from experience working for a microprocessor manufacturer (Zilog) many years ago, where the process was much more labor intensive, but with the quadzillion circuit increase, probably much the same time frame.

[melgross]

Originally Posted by David Stevenson

Short answer: bugs in the design. Even taking an existing design and "adding" a few features can introduce hardware bugs that can take a year to track down and iron out. Speaking from experience working for a microprocessor manufacturer (Zilog) many years ago, where the process was much more labor intensive, but with the quadzillion circuit increase, probably much the same time frame.

I'm not so sure. They didn't say that they did the first tapeout. They had silicon running. Usually, you don't do that until you have confidence the design is pretty much secured.

That doesn't mean that they might not want to do some efficiency re-working. But, it shouldn't take another 11 months. It should be through by February.

I think they just want to give the first 45nm gen a chance to pay out.

[murk]

OK, I will ask the question everyone is dying to: Should I update my G3 iMac to Penryn or wait for Nehalem? On second thought maybe I should just wait until the 32nm shrink.

[onlooker]

Originally Posted by murk

OK, I will ask the question everyone is dying to: Should I update my G3 iMac to Penryn or wait for Nehalem? On second thought maybe I should just wait until the 32nm shrink.

At your update rate you may as well wait for 1nm processors.

[King Chung Huang]

Originally Posted by onlooker

At your update rate you may as well wait for 1nm processors.

Yes, after waiting this long, what's another decade or two?

[Splinemodel]

Originally Posted by onlooker

At your update rate you may as well wait for 1nm processors.

I'm surprised that UV photolith has made it this far, but 1nm. . . no fucking way. (?)

[crentist]

Originally Posted by Splinemodel

I'm surprised that UV photolith has made it this far, but 1nm. . . no fucking way. (?)

Please watch your language. Thank you.

[murk]

Originally Posted by onlooker

At your update rate you may as well wait for 1nm processors.

When can I expect to get one? By the way, I was making a joke. I'm quite happy with my G3 and OS 9 still works.

[Splinemodel]

Originally Posted by crentist

Please watch your language. Thank you.

what the fuck are you talking about?


I imagine that one's lost on you. Oh well.

[melgross]

Originally Posted by murk

When can I expect to get one? By the way, I was making a joke. I'm quite happy with my G3 and OS 9 still works.

You are an IRON man. I couldn't wait that long.

[otayranchdweeb]

I usually buy a new Mac when performance doubles.

IIci --> Quadra 650
Quadra 650 -- > PM 7600/120
PM 7600/120 -- > G3/266
G3/266 -- > G4/Dual 1.42
G4/Dual 1.42 -- > G5/Dual 2.7

PowerBook 145B --> PowerBook G3/500
PowerBook G3/500 --> PowerBook G4/1GHz
PowerBook G4/1GHz --> MacBook Pro 2.16GHz

I bought the dual G5 in May '05 and now I feel stuck... XBench scores on the top-end dual core Intels don't seem anywhere near double the G5 speed in CPU and memory scores. Any idea when we can expect systems that really do double the performance of a G5/Dual 2.7? (An octo-processor Mac doesn't interest me as 99% of the apps I run won't use the extra procs anyway.)

[Splinemodel]

Originally Posted by otayranchdweeb

I usually buy a new Mac when performance doubles.

I bought the dual G5 in May '05 and now I feel stuck... XBench scores on the top-end dual core Intels don't seem anywhere near double the G5 speed in CPU and memory scores. Any idea when we can expect systems that really do double the performance of a G5/Dual 2.7? (An octo-processor Mac doesn't interest me as 99% of the apps I run won't use the extra procs anyway.)

What apps would those be?

Moreover, if the G5 works, then keep it.

[SpinDrift]

Originally Posted by Splinemodel

what the fuck are you talking about?


I imagine that one's lost on you. Oh well.

He's not the only one who doesn't appreciate bad language. I'm sure you're able to articulate yourself far better than that. There's just no need for swear words.

[Royboy]

Originally Posted by SpinDrift

He's not the only one who doesn't appreciate bad language. I'm sure you're able to articulate yourself far better than that. There's just no need for swear words.

Here's some words that are less inflammatory, but will get across your message with similar fervor:

be intimate, breed, copulate, fool around, fornicate, go all the way, go to bed with, have sexual intercourse, have sexual relations, lay, make out, mate, procreate, sleep together, make love, have sex

[Fairly]

Ah. Elephants can remember when fanboys used to deride people with computers with Intel processors. Fanboys can't. Kool-Aid has been medically proven to induce "benign amnesia".

[rtdunham]

Originally Posted by Splinemodel

what the fuck are you talking about?


I imagine that one's lost on you. Oh well.

Instead of this degenerating into a "don't say that"/"ooh, I'll say it again!" exchange, let's see if we can get something educational out of it: How old are you? It'd be ****ing great if you'd answer honestly, let the chips fall where they may, whether it reaffirms or shatters expectations. Can you man up to that honest question?

[nvidia2008]

Originally Posted by Fairly

Ah. Elephants can remember when fanboys used to deride people with computers with Intel processors. Fanboys can't. Kool-Aid has been medically proven to induce "benign amnesia".

Eh? Us "fanboy" "kool-aid drinkers" know when CPUs that are good are good. G4 and G5s whipped the Pentium 4's ass 5 days from Sunday (or whatever the phrase is). Notice your non-Apple-fanboy-Kool-Aid-free AMD fans who, rightly, lorded over the craptastic Pentium 4 especially with the impressively overclockable Athlon64s.

It was well known that the G5 though had numerous problems and that IBM-led endeavour had to end soon. So Apple got out in time, unlike oh, MS' XBOX360 which is similar to the G5 and surprise, surprise, is facing a lot of heat issues.

[nvidia2008]

I'm not saying Apple is now the greatest and smartest companie on Earth because they switched to Intel, but, for the most part, it has been good for everyone.

[nvidia2008]

Originally Posted by Splinemodel

I'm surprised that UV photolith has made it this far, but 1nm. . . no fucking way. (?)

At what nm do we have to switch to optical? Like when the quantum thingys cause to much unpredictability because the transistor thingys are too close together?

25nm? 10nm? 1nm? 0.1nm? 0.001nm? Melgross?

[palegolas]

Is this what the new PowerBook 12" replacement has been waiting for all these years?

[onlooker]

Originally Posted by rtdunham

Instead of this degenerating into a "don't say that"/"ooh, I'll say it again!" exchange, let's see if we can get something educational out of it: How old are you? It'd be ****ing great if you'd answer honestly, let the chips fall where they may, whether it reaffirms or shatters expectations. Can you man up to that honest question?

PLease use PM for this stuff, and take it out of a hardware thread. Your totally off topic.

[melgross]

Originally Posted by Fairly

Ah. Elephants can remember when fanboys used to deride people with computers with Intel processors. Fanboys can't. Kool-Aid has been medically proven to induce "benign amnesia".

Umm. At one time, before the end of 2001, there was no question that the PPC was a much better chip than the x86, from any manufacturer.

During the middle 90's the PPC was expected to take the PC market away from the x86. It might have done so if MS hadn't decided to withdraw the PPC version of NT right after it was finished.

When the G4 came out, it had a good 30% advantage in integer, and a good 40% advantage in float over any x86 designs out there. Altivec was much more advanced then MS and Intels' weak competitor, offering up to 200% faster manipulation of vectorized functions.

Only when Moto couldn't raise the G4's speeds at the same time Intel went into MHz overtime, did that change. But, it still took a 1GHz x86 to finally overcome the much slower G4.

With the G5, IBM had a much superior chip than the Netburst line from Intel. The chips from AMD performed a bit better still, but IBM seemed committed to raising the bar at a good rate, and did begin to do so.

The 90nm speed barrier got in the way of all manufacturers.

Even so, IBM speeded the G5 up at a greater pace than either AMD or Intel.

The problem there was that Apple's sales were simply not enough to pay for all of the required R&D needed to keep up. IBM lost interest at that point, and Apple left.

But, the PPC platform was architecturally superior to the x86 designs. They carried none of the 8 and 16 bit baggage both AMD and Intel felt they had to maintain.

But now with 64 bit, that baggage is of less importance, because 64 bit code totally ignores any older code present. Newer 64 bit chips won't include most of that backwards compatible code at all.

Apple has an advantage here because its software, and that of Apple third party developers, haven't used anything other than 32 bit "clean" code for many years. That fits well with 64 bit chips, and Apple's OS, which allows either 32 bit OR 64 bit programming, and doesn't need a separate version, as does Vista (or XP).

Intel had reversed its mistake two years ago, and understands what must be done. IBM could never afford to put nearly as much R&D into its cpu designs as Intel does.

Apple made the right choice at the right time.

It has to be understood that when Jobs first made the announcement about Intel that fateful day, no one outside of those "in the know" was aware of Intel's subsequent announcements that were to rock the world in the area of cpu design. That's why we were skeptical about the whole thing. I, for one, was curious about why Jobs would do this when the G5 was creeping up in speed faster than either AMD or Intel, and seemed to soon take the performance crown.

Only after Intel revealed its new plans, and presented the world with the first of its new chips, did we see the reason.

Now we know that the decision was required if Apple was not to be totally left out in the cold.

[melgross]

Originally Posted by nvidia2008

At what nm do we have to switch to optical? Like when the quantum thingys cause to much unpredictability because the transistor thingys are too close together?

25nm? 10nm? 1nm? 0.1nm? 0.001nm? Melgross?

Your deep technical understanding of these issues always amazes me, as does your ability to use some of the more esoteric phraseology.

Actually, a number of experts in process technology think that there could be problems between 32 and 22nm, the step after.

Even if that is managed without too much turmoil, the step down to 15 isn't assured. There are still major problems with getting to that size lithography. The standards aren't agreed upon, and the technologies for getting there are still being investigated.

I'm not even talking about the electronic implications of going to that small size, but just the possibility of drawing the masks! If they can't do that, the rest doesn't matter.

If they do make it to 15, the step to 10 is thought by many to possibly be out of reach, but that isn't universally agreed upon yet.

As far as anything else goes, there a a bunch of technologies that are being investigated that will bypass all of this.

But none of it is anywhere near usability as yet, much less being at the point where it is known whether they can be mass manufactured. We won't get a hint of it for a good five more years.

Optical is not a panacea, as it is pretty much subject to the same physical laws. There has been a big advance by Intel in making silicon laze effectively. This is considered to be seriously important for any optical computing breakthrough, because other elements are simply too difficult to work with, too expensive for mass replications of chip technologies, or both.

But, this is also in a very early stage of development.

[jbh0001]

Originally Posted by Haiduc210

When they say Penryn will be dropping in a few weeks, does that mean that's when it'll first become available to manufacturers, or when we'll start seeing the first consumer products employing the new chips?

I ask only because I'm trying to gauge the amount of time it'll take for Apple to begin introducing these chips in their lineup. Would they be able to have designs finalized and in production in time for MWSF?

Right now, Penryn will officially launch on November 12, 2007. Don't hold your breath for Apple to incorporate Penryn or Nehalem. We have been waiting since May 9, 2007 for the Macs that rely on integrated graphics (MacBook, MacMini) to be upgraded to the Santa Rosa platform, with GMA X3100 integrated graphics to replace the current GMA 950 integrated graphics. A refresh of the Santa Rosa chips is now also expected to accompany the Penryn launch on November 12, 2007.

Apple waited about 18 months to update the MacMini from Core Duo to Core 2 Duo, when all that was required was a simple swap of a pin-compatible CPU (socket M).

Maybe Apple will update the MacBook and MacMini for MacWorld 2008. But I'll believe it when I see it. After waiting so long for Santa Rosa to materialize in a MacBook or MacMini, I might as well limp along until MacWorld 2009 and hope for a Gilo+Calpella MacBook and/or MacMini. But I won't hold my breath for it.

[melgross]

Originally Posted by jbh0001

Right now, Penryn will officially launch on November 12, 2007. Don't hold your breath for Apple to incorporate Penryn or Nehalem. We have been waiting since May 9, 2007 for the Macs that rely on integrated graphics (MacBook, MacMini) to be upgraded to the Santa Rosa platform, with GMA X3100 integrated graphics to replace the current GMA 950 integrated graphics. A refresh of the Santa Rosa chips is now also expected to accompany the Penryn launch on November 12, 2007.

Apple waited about 18 months to update the MacMini from Core Duo to Core 2 Duo, when all that was required was a simple swap of a pin-compatible CPU (socket M).

Maybe Apple will update the MacBook and MacMini for MacWorld 2008. But I'll believe it when I see it. After waiting so long for Santa Rosa to materialize in a MacBook or MacMini, I might as well limp along until MacWorld 2009 and hope for a Gilo+Calpella MacBook and/or MacMini. But I won't hold my breath for it.

I agree with what Has happened, but I'm not so sure that I can agree completely with what you say WILL happen.

While no one knows what Apple intends as the long term goal for the Mini, sink or swim, we do know that Apple intends to stay with its sales leading product, the MB.

I think that Apple didn't see the need to revise the MB in the second half with the Santa Rosa.

In looking at that decision, we must look at what was expected from that chip, and what was found to be the reality.

Was what expected was a fair amount of improved performance, and significant battery life enhancement.

Neither occurred.

Apple's decision to not use it made sense, therefore, because the improved graphics has also proven to be better, but not so much that games would be actually significantly more playable. to go from atrocious to less atrocious is not much of an improvement.

I do see Apple updating the MB with the newer chips, because Apple must, if they expect to be at parity. These are major changes, unlike that of Santa rosa, and so Apple will make them.