Print 57 comment(s) - last by OvErHeAtInG.. on Jun 26 at 11:10 PM

Die shots of Intel's test shuttle 45nm SRAM lithography for "Penryn"

Intel's long-term microarchitecture roadmap

An early tri-gate design from Intel
What's the absolute bleeding edge in Intel CPU design for the next five years? "Penryn," "Nehalem" and "Gesher"

No one would doubt that Intel's Core 2 Duo is perhaps one of the most hyped and anticipated processor launches since the K8.  The processor, expected to start shipping on July 23, 2006, is Intel's first major transition in desktop microarchitecture since the Willamette core was introduced in 2000. 

Conroe and its derivatives are a step away from Intel's former flagship NetBurst, but even these processors are a bit of a dying breed: during Intel's shift to 45nm, the company will no longer focus on derived microprocessor cores in favor of refined unified core architectures.  For example, later this year Intel is expected to release Allendale, a dual core Conroe-based processor with only 2MB of L2 cache.  From what Intel has stated from forward looking statements and roadmaps, this sort of practice will stop. 

One possibility as to how this will work will be in the way Intel packages its processors.  For example Intel's Presler is essentially a Cedar Mill processor with two dice coupled on the same package.  Likewise, Intel's Kentsfield processor is two Conroe dice coupled on a single package to make a quad-core processor.  Even though your server and desktop processor will have the same processor core on Intel's future architectures, the logical way to differentiate the role of these CPUs will be the number of dice per package.

Intel announced the majority of its long-term architecture plans in April of this year. During this same announcement, Intel detailed that the Core microarchitecture would survive into 2008 with the 45nm Penryn core. 

Penryn, which will be based on Intel's lithography process known as P1266, is a 45nm unified core set for launch in 2007 that is also expected to stay into production into 2008.  Intel has already produced SRAM samples for 45nm CPUs, as demonstrated in the lithography shot on the right.  Aside from the process shrink on Penryn, the major divergence in design from Conroe is the new material design.  With P1266, Intel shifts away from Silicon Dioxide gate dielectrics -- a process the company has used since the mid-90s -- to High-k dielectrics.  With new dielectric techniques, the company will also revamp its gate electrodes to metal instead of Polysilicon derivatives.  The last major materials change of such magnitude occurred when Intel moved from Silicon to Strained Silicon in 2002, which is still slated for use in P1266 and beyond.

According to Intel's long term roadmap, Penryn will become the last "Core" (NGMA) microarchitecture, but it will not become the last 45nm generation.  A new architecture, based on the Nehalem core, will make an appearance in 2008.  This Core successor, dubbed by many as NGMA2, will essentially take all of the 45nm manufacturing principles of Penryn and apply them to the totally new Nehalem architecture.  This transition will be very similar to the P6+ transition with Yonah from Dothan to Conroe

If P1266 was radically different from the 65nm Core CPU process, P1268 takes what is normal on 45nm and throws it out the window.  The first P1268 processor will use a CPU codename that Intel has publically announced as Nehalem-CNehalem-C is still based on the Nehalem (NGMA2) microarchitecture, but is a die shrink from 45nm to 32nm. P1268 employs all of the neat manufacturing techniques found on P1266, but may use a totally different lithography process called EUV, or Extreme Ultraviolet. Although it sounds trivial to swap a traditional DUV lithography setup with an EUV one, there is a fundamental problem with EUV that prevents foundries from doing so; the EUV wavelength is so small at 13.4nm that virtually every material absorbs the wave including the lenses and atoms in the air. EUV must be done in a vacuum with reflective surfaces instead of lenses to focus and redirect the lithography. 

However, EUV isn't the only radically new tool in Intel's arsenal for 2009.  According to Intel's forward statements, Nehalem-C is succeeded by Gesher, the third in Intel's next generation microarchitecture designs uniquely dubbed NGMA3.  Intel hinted last week during the 2006 VLSI symposium that it will start using tri-gate transistors on its 32 or 22nm production processors.  Since the 1950s, transistors have been strictly planar designs, with gates that lay flat across the substrate.  A tri-gate design is unique in the fact that a single gate is stacked on top of two vertical gates allowing for essentially three times the surface area for electrons to travel.  Whether or not these tri-gates will appear on Gesher or its 22nm derivative have not been announced.

Even beyond Gesher and its 22nm derivative, Intel has a lot of technologies in its roadmap.  One technology proposed to replace tri-gates is by using silicon nanowires surrounded by a metal gate to increase surface area even more. Intel laboratories have also experimented with carbon nanotubes a mere 1.4nm in diameter as a future transistor material, yet these materials won't show up until 2013 at the earliest. 

If anything, here are the big points to keep in mind for Life After Conroe:

  • Intel will unify its processor architectures -- no more Merom/Conroe/Woodcrest derivatives.  The processor core that you use in your server will essentially be the same chip you use in your notebook.
  • Parallel design -- The Nehalem team is already working on its 45nm processor.  Penryn is a "meet you in the middle" project between Core and Nehalem, which is also undergoing development.  The same leapfrogging will occur with Nehalem and Gesher.
  • P1266 lithography (Penryn family processors) will use radically different manufacturing techniques -- High-k dielectrics, metal gate electrodes
  • P1268 (Gesher and Nehalem-C) switches from 193nm DUV lithography to EUV 13.4nm lithography.

Comments     Threshold

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

EUV in 2009?
By masher2 on 6/20/2006 1:38:46 PM , Rating: 2
I'd hope so...but then, Intel has been saying EUV is "just a few years out" for over a decade now. Some pundits are predicting life for DUV at the 32nm node...with EUV not hitting its stride until 22nm.

RE: EUV in 2009?
By Phynaz on 6/20/2006 2:13:22 PM , Rating: 1
So what? Who care what lithography process is used?

Do you know what process is used to produce the plastic in your keyboard?

RE: EUV in 2009?
By masher2 on 6/20/2006 2:31:53 PM , Rating: 2
> "So what?..."

I would think that answer would be obvious. Process node shifts have been occurring at ever-lengthening intervals for quite some time. A switch to EUV will initially be painful. But the greater litho headroom promises to make subsequent node transitions that much least for the next few nodes. And Intel drives the entire industry.

To simplify-- the faster Intel converts to EUV, the faster the industry as a whole will. And the faster we get to the 22 and 16nm nodes.

RE: EUV in 2009?
By Phynaz on 6/20/06, Rating: -1
RE: EUV in 2009?
By saratoga on 6/20/2006 4:23:01 PM , Rating: 5
Again, I have to ask so what? As users we shouldn't care how a chip is made. We just want the ever faster and cheaper. Who gives a c**p how that happens?

Let me translate his post into moron for you:


Did that help?

RE: EUV in 2009?
By masher2 on 6/20/2006 4:41:52 PM , Rating: 2
> "Let me translate his post into moron for you..."

Lol, best post of the thread there...

RE: EUV in 2009?
By KristopherKubicki on 6/20/2006 11:16:28 PM , Rating: 2
1,000th post Masher. Congrats.

RE: EUV in 2009?
By masher2 on 6/21/2006 7:54:33 AM , Rating: 2
1000? Good lord, how time flies...

RE: EUV in 2009?
By BaronMatrix on 6/21/2006 11:48:56 AM , Rating: 2
I think it is time to look at widening at not shrinking. i thas been shown that more executon enegines provide more power at lower speed.

PCs are plenty fast enough.

RE: EUV in 2009?
By Regs on 6/20/2006 2:55:56 PM , Rating: 2
I care. The manufacturing process is detrimental to all those clock cycles, instructions per cycle, and heat disputation.

RE: EUV in 2009?
By dragonflycms on 6/20/2006 3:03:34 PM , Rating: 2
I think the number one reason people visit this site is that anand provides tons of detailed information. Thanks anand for pouring on the details that us computer freaks love!

RE: EUV in 2009?
By peternelson on 6/20/2006 8:37:38 PM , Rating: 2

"Do you know what process is used to produce the plastic in your keyboard?"

It's called "injection moulding" and very useful it is too. ROFL

RE: EUV in 2009?
By Phynaz on 6/21/2006 10:32:10 AM , Rating: 1
Hey smartass, plastic isn't made by injection molding.

Learn to read before replying.

RE: EUV in 2009?
By PetesEscapade on 6/21/2006 6:57:21 PM , Rating: 2
My guess is that the at least some of the plastic in my keyboard was made by free radical vinyl polymerization of styrene monomer.

RE: EUV in 2009?
By Shadowself on 6/20/2006 4:44:18 PM , Rating: 2
Actually, it's been IBM who has been claiming EUV and soft x-ray have been coming the longest. I did research at Brookhaven at the National Syncrotron Light Source back in the mid 80s and had a chance to interact with a team of IBM researchers doing EUV/soft x-ray litho experiments back then. In the late 80s / early 90s IBM even set up their own syncrotron to do it in house.

From the slow progression of miniaturization, I don't doubt we'll see EUV before 2012. When will we see soft x-ray lithography? Maybe by 2025. By then quantum tunneling will be the designers' biggest concern.

nit picking...
By broly8877 on 6/20/2006 1:00:12 PM , Rating: 2
Allendale (or the whole Core architecure) has a unified L2 cache, not xMB per core. I believe a single core can use all 2MB if need be.

Stopped reading after that.

RE: nit picking...
By KristopherKubicki on 6/20/2006 3:17:08 PM , Rating: 2
Correct - the edit monster butchered that. I updated the text.

RE: nit picking...
By coldpower27 on 6/20/2006 3:46:00 PM , Rating: 1
To my knowledge Allendale is being released now with the Conroe core at launch with the E6400/E6300/E4200.

Are you reporting that is not the case?

RE: nit picking...
By KristopherKubicki on 6/20/2006 5:00:49 PM , Rating: 3
I have not read that Allendale is launching at the same time as Core 2 Duo for desktop, nor am I reporting that.

Intel is launching 2MB and 4MB versions of Conroe, however.

RE: nit picking...
By caboosemoose on 6/20/2006 5:47:25 PM , Rating: 3
Allendale is simply the name for a 2MB Conroe, plain and simple. I have seen a 1.86GHz Core 2 Duo running, and it showed up as Allendale in CPUID.

Another point is that Intel already has its unified architecture. They may have the names Merom, Conroe and Woodcrest, but the chips are identical. This article needs a lot of work, doesn't give me much faith in Dailytech.

RE: nit picking...
By Knish on 6/20/2006 5:56:23 PM , Rating: 3
Another point is that Intel already has its unified architecture.

Ah yes, Dempsey and Woodcrest are totally based off the same core. Not to mention Yonah and Presler. How about Conroe and Montecito?

Real unified there.

RE: nit picking...
By caboosemoose on 6/20/2006 10:51:48 PM , Rating: 3
Ah yes, Dempsey and Woodcrest are totally based off the same core. Not to mention Yonah and Presler. How about Conroe and Montecito?

No you moron, I was referring to the fact that the article says:

Intel will unify its processor architectures -- no more Merom/Conroe/Woodcrest derivatives. The processor core that you use in your server will essentially be the same chip you use in your notebook.

Those three chips are exactly the same, just different packaging and wiring. I'm well aware that Intel still has some other designs on its books other than NGMA.

RE: nit picking...
By shadowzz on 6/20/06, Rating: 0
Interesting stuff...
By ksherman on 6/20/2006 1:07:01 PM , Rating: 2
Though, I would think that by 2013 we woulsnt be using silicon anymore... silicon has its limits, and I bet we are reaching it soon.

I would like to see more work with carbon nano tubes or the like.. arent they supposedly smaller than silicon transistors? I could be wrong...

Neat stuff that intel has coming down the pipe. cant say Im a big fan of a unified architecture. Servers and laptops perform different functions. that either means that notebooks are going to become much more powerful, or servers will become less so... But, kudos to Intel for being so open...

AMD, what you got comming?

RE: Interesting stuff...
By Trisped on 6/20/2006 1:21:33 PM , Rating: 2
No, notebooks will use lower speed cores (though the same design) to save power while servers will use faster cores and more of them to provide more power.

You might think of it as putting 1 core in a laptop, 2 in a desktop, and 8 in a server.

RE: Interesting stuff...
By segagenesis on 6/20/2006 1:40:33 PM , Rating: 2
They said the same about silicon every 5 years since the first nmos ICs and look where we are now. Despite the limitations imposed of going faster it appears that expasion is outward rather than upward in recent years. With the advent of quad core and possibly 8-way CPUs on a single chip in the near future it would be more of an incentive for efficent multi-threaded programming in the future... Your average supercomputer in both the past and present (with the execption of maybe the CDC 6600) did not contain one superfast CPU...

RE: Interesting stuff...
By ksherman on 6/20/2006 2:28:42 PM , Rating: 2
i guess so... and then i just read the article posted about IBM bringing a chip to 500GHz

RE: Interesting stuff...
By Knish on 6/20/2006 3:16:08 PM , Rating: 3
i guess so... and then i just read the article posted about IBM bringing a transistor to 500GHz

Fixed that for you.

By nserra on 6/20/2006 1:17:39 PM , Rating: 2
Who here thinks that Merom/Conroe/Woodcrest are different processors (Isn’t the same chip)?

They have exactly the same specs!!!!!

The only differential factor is the chipset/socket/mobo type they use not the processor core!

Even the K8 (Turion/Athlon/Opteron) is exactly the same.
The Operon is the more different of all, because has the northbridge enabled and independent memory controllers (But that’s part of the K8 design).

RE: What?!?!?
By Trisped on 6/20/2006 1:28:00 PM , Rating: 2
The different CPUs were made from different base cores that allowed optimization for the job they were doing. Laptop CPUs had built in power management and Server CPUs were built for doing fast business calculations while desktops have a little more multi-media focus. Sure, they all have the same capabilities now, but the way they do things is not the same.

RE: What?!?!?
By sotti on 6/20/2006 3:10:25 PM , Rating: 2
Yes laptops sometimes have speed steps, but with opterons and athlons and pentiums and xenons the only difference has been cache size and multiprocessor capabilities.

There where no arcitectural difference that would make one more suited for integer perfomance or "multimediea".

my opteron 175 == athlon X2 4400+
They are exactly the same.

RE: What?!?!?
By Targon on 6/21/2006 9:31:21 AM , Rating: 2
The Opteron has a different memory controller since you want/need ECC memory. That of course is in addition to the additional HT links for multi-processor configurations(which is why you have 200 and 800 series Opteron processors.

You will also see some divergence due to single vs. dual core, and of course, the ability to run multiple operating systems at once will remain a multi-core feature that may not go into budget processors.

There will probably be some cases where two seperate products are the same(the way the Opteron 175 is almost identical to the X2 4400+), but I can see where the feature set will be different between product families, even if they share a very similar design.

RE: What?!?!?
By nserra on 6/21/2006 6:51:15 AM , Rating: 2
Not really...

The difference is at the chipset level like I said,

the chipset (and BIOS) will enable or not, technology there is already inside of all the processors.

And the Server doing more then the laptop or desktop you are joking right?
If the server is faster is thanks to the FSB speed, chipset (again), and maybe better RAM, or even have more CPUs, …

What's the difference
By defter on 6/20/2006 1:14:56 PM , Rating: 2
Intel will unify its processor architectures -- no more Merom/Conroe/Woodcrest derivatives. The processor core that you use in your server will essentially be the same chip you use in your notebook.

Huh? Merom, Conroe and Woodcrest are all using the same core. Only differences are the package and the FSB speed.

RE: What's the difference
By nserra on 6/20/2006 1:21:15 PM , Rating: 3

Maybe what is being said is that they will stop inventing different names for the same thing. :)

RE: What's the difference
By ksherman on 6/20/2006 2:27:31 PM , Rating: 2
All different names, for the same thing...

Its a song from Death Cab for Cutie ;-)

RE: What's the difference
By Mudvillager on 6/21/2006 7:59:40 AM , Rating: 2
Hahaha I almost wanna type "LMAO" (god damn it I just did that)!

Too bad the song sucks. Love their previous work though (especially We Have The Facts And We're Voting Yes).

P1266 != Penryn
By MartinT on 6/21/2006 2:41:49 AM , Rating: 2
Penryn, also known by Intel's lithography team as P1266, is a 45nm unified core set for launch in 2007 that is also expected to stay into production into 2008.

P1266 is the name of Intel's 45nm lith process, Penryn is the codename for a NGMA-based MPU produced on the P1266 process.

You shouldn't use both as if they were the same, cause they're not.

RE: P1266 != Penryn
By KristopherKubicki on 6/21/2006 10:28:38 AM , Rating: 2
Thanks, I cleaned that up.

By Blackraven on 6/23/2006 12:42:26 PM , Rating: 2
Now that Intel is announcing that every year (or maybe even every 6 months), there will be new products that will be released.

So help me out here.

I'm thinking of looking for a successor to my three year old P4 computer (768 DDR RAM, 128 video ram, 845GE Max chipset)

So Intel, are you saying that I should maintain my objective of getting Kentsfield Quad Core (65nm) next year and use a three-year interval for any succeding processor purchases?

Speaking of which, what are the other desktop quad-core processors that would come after "Kentsfield"? Any quad-core units being planned in the near future that would be at least 16nm or below?

Any advice would be appreciated.

Thanks :)

By techhappy on 6/25/2006 7:32:45 PM , Rating: 1
I'm no expert, but I have been pretty lucky with my upgrades. My own advice, upgrade every 4-5 years when pivotal technologies come along. Problem is, recently, things have kicked into cpu overdrive from a somewhat stagnant state of technology for the last 2 years. The dual core adoption by both AMD vs. Intel will probably spurn faster more significatn changes and a slew of benefits for consumers.

Right now, I got a dual core Pentium D system which has been a pleasure. Kentsfield with the quad core design looks to be a monster. If I were you, I would hold off until Kentsfield drops in price a little bit, then upgrade.

You could of course jump onto the dual core bandwagon with the current prices dropping like hot cakes from heaven. But I think you will only regret it when Kentsfield drops a bomb in the cpu world, when it comes out. That is, if it's as good as we all think it will be.

Indy and the Last Crusade: "Choose, but choose wisely..."

3 way gates
By Scabies on 6/20/2006 3:50:08 PM , Rating: 2
Maybe I dont know anything about how chips are made, but wouldnt the 3gate technology lend itself to massive heat issues? A planar etching is simple enough, slap a flat piece of metal on the die and make radiator fins. with two perpendicular surfaces generating heat as well...?

RE: 3 way gates
By masher2 on 6/20/2006 4:49:08 PM , Rating: 2
> "Maybe I dont know anything about how chips are made, but wouldnt the 3gate technology lend itself to massive heat issues?"

No. Heat production is a function of power density; heat dissipation a function of surface area. Tri-gate technology decreases power density and (slightly) increases surface area. So its a win on both counts.

I think you're worried about heat conduction between the a nonplanar die and a heat spreader? If so, remember that such spreaders aren't directly bonded to the conductive die itself; they're separated by a nonconductive epoxy (or other material) that effectively converts the chip to a planar surface.

RE: 3 way gates
By OtakuMax on 6/20/2006 10:57:16 PM , Rating: 1
Tri-gate actually reduces current leakage to the bulk while maintaining the capacitance (Cox) needed to conduct produce the conduction channel. So, power requirement (which translate directly into heat) should be reduced overall.
As for power dissipation, this thing is buried under dielectrics and interconnects anyway. The bottleneck for the heat transfer is from die to air (through the heatsink).
ie. I don't think it matters :P

RE: 3 way gates
By masher2 on 6/21/2006 7:56:15 AM , Rating: 1
Isn't that exactly what I said?

By masher2 on 6/20/2006 1:33:36 PM , Rating: 1
> "Likewise, Intel's Kentsfield processor is two Conroe dice ..."

In the context of lithography (or a stamping die, for that matter) the plural of "die" is "dies", not "dice".

RE: Dies...
By Ryan Smith on 6/20/2006 1:41:40 PM , Rating: 2
When dual-core CPUs were just coming out Anand went to get the definative word on that; dice is perfectly acceptable.

RE: Dies...
By Master Kenobi on 6/20/2006 2:15:30 PM , Rating: 2
For the guy that wants to know more about Tri-Gate. Here is an article from Intel back in '02 that outlines why they need to go with Tri-Gate.

It's fairly technical but they do a good job of explaining it so most people can understand it.

hopefully Intel makes the
By thilanliyan on 6/20/2006 3:20:00 PM , Rating: 2
<in Gandalf's vioce-->"One processor to rule them all!!!"

Sorry, just bored.

By OvErHeAtInG on 6/26/2006 11:10:50 PM , Rating: 2
Actually it sounds much cooler in Saruman's voice.

By Trisped on 6/20/2006 1:19:41 PM , Rating: 2
For example, in 2007 Intel is expected to Allendale,
missing a word or does Allendale mean something? I think it should say, “expected to release Allendale” but it could be anything.

Long post though

And I still don’t get the tri-gate thing. What are they and how do they work and why are they important? Are we fitting 3 gates where we once fit 1 or are we using 3 gates to achieve the results of one?

RE: Tri-Gate?
By Phynaz on 6/20/2006 2:11:10 PM , Rating: 1
Unified Architecture....hmmm
By aguilpa1 on 6/20/2006 1:21:53 PM , Rating: 2
You mean like AMD has had for years in their 64bit chips and unified hypertransport technology for both desktop and server chips, when Intel was still fighting with Netburst and glueing two chips in one package and screaming viola!

By Howard on 6/20/2006 3:22:15 PM , Rating: 2
Oh well.

cronore dice
By ojingoh on 6/20/06, Rating: 0
Not easy to read
By GTVic on 6/20/06, Rating: 0
By UsernameX on 6/20/06, Rating: -1
"There's no chance that the iPhone is going to get any significant market share. No chance." -- Microsoft CEO Steve Ballmer
Related Articles

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki