backtop


Print 47 comment(s) - last by rushfan2006.. on Dec 4 at 2:16 PM

45nm on track for 2007

Reuters reports Intel is manufacturing prototypes of its upcoming Penryn processor core. Penryn is the upcoming 45nm die shrink of Intel’s Core 2 architecture. Intel first demonstrated its 45nm fabrication capabilities at the D1D facility in Oregon. Current Penryn prototypes are being produced at Intel’s Oregon factories as well, though a different facility.

In addition to the 45nm die shrink, Penryn is expected to introduce a couple more manufacturing process changes. These changes include shifting away from Silicon Dioxide gate dielectrics to High-k dielectrics. The gate electrodes will also be revamped with metal gates instead of the Polysilicon derivatives currently used.

Intel is currently ontrack to introduce Penryn core processors in the second half of 2007. As Penryn is currently in its prototype stages Intel’s director of process architecture and integration, Mark Bohr stated: "We are processing the first samples of the Penryn design. These samples will go back to the design team to determine if design is working as expected."
 
Nevertheless, Intel is expected to ship 45nm products very soon after AMD makes its final transition to 65nm. Intel is also expected to open up new 45nm production fabs within the next 16 months. Expect Penryn to arrive in 2007 after Intel refreshes its current Conroe based product lineup. Penryn will also be the last product based on Intel’s Core architecture.


Comments     Threshold


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

Correction
By Viditor on 11/28/2006 4:34:09 AM , Rating: 4
quote:
Nevertheless, Intel is expected to ship 45nm products while AMD makes its transition to 65nm


I'm not sure where you got that from, but AMD should be completely transitioned to 65nm before Penryn ships (at least that's what I'm hearing from the Intel people...).
Penryn is to begin shipping (if there are zero problems, and since it's a dumb shrink there shouldn't be any) in Dec 07. AMD is to be completely done with their 65nm conversion by then. AMD will still have ~10-15% of their product at 90nm for legacy items (which they have no plans to convert).




RE: Correction
By KristopherKubicki (blog) on 11/28/2006 4:37:15 AM , Rating: 2
Corrected, thanks.


RE: Correction
By slayerized on 11/28/2006 9:09:18 AM , Rating: 2
quote:
....and since it's a dumb shrink there shouldn't be any) in Dec 07.

I say that this is a rather lose statement. It is true that Penryn is a shrink product; but calling it a dumb shrink is not true. Intel is the first company to transition to a 45nm process and it is always better to transition with a product whose characteristics are well known. Also, since it is a new process there is lack of DfM support in any form. So the only variable (among a zillion other ones) in manufacturing that is eliminated is the design (even that not completely). It is just a sensible engineering decision, and I would call it a smart shrink.


RE: Correction
By Viditor on 11/28/2006 10:41:55 AM , Rating: 3
quote:
It is just a sensible engineering decision, and I would call it a smart shrink


I'm not sure if your speeking tongue in cheek here, but in case you're not...a "dumb shrink" is an industry term for a node change without architectual change. It's not a derogatory term, more a descriptive one...


RE: Correction
By slayerized on 11/28/2006 10:48:11 AM , Rating: 1
The whole point of my last comment was just to emphasize the fact that even though Penryn is a dumb shrink there might be a ton of roadblocks which can push product release. I am well aware that of what a 'dumb shrink' but my point was to emphasize the importance of a 'dumb shrink' for the first 45nm product.


RE: Correction
By Viditor on 11/28/2006 10:56:35 AM , Rating: 2
quote:
my point was to emphasize the importance of a 'dumb shrink' for the first 45nm product


Fair enough...this is a fairly common industry practise which both AMD and Intel adhere to most all of the time.


RE: Correction
By coldpower27 on 11/28/2006 11:32:28 AM , Rating: 3
Well there is a better term then "dumb" simply change the word to optical shrink.

In any case Penryn isn't a straight optical shrink by any means, the added cache changes thing a little bit, as well as the added SSE4 instruction set slated for this processor.


RE: Correction
By Viditor on 11/28/2006 11:37:45 AM , Rating: 2
quote:
In any case Penryn isn't a straight optical shrink by any means, the added cache changes thing a little bit, as well as the added SSE4 instruction set slated for this processor


I believe that the SSE4 will be done in microcode, and the larger cache is actually a non-architectual function...so practically speaking, it is (if you prefer) an optical shrink.


RE: Correction
By coldpower27 on 11/28/2006 11:46:00 AM , Rating: 2
No since both of those features place the Penryn derivatives as something slightly different irregardless they are not pure optical shrinks. While it's close to a straight optical shrink the cache and SSE4 are minor improvements.

Intel's only pure optical shrink to my knowledge has been Prescott-2M to Cedar Mill, the rest have had cache additions or some minor tweaking.



RE: Correction
By Viditor on 11/28/2006 11:48:57 AM , Rating: 2
Potato, potahto...we're picking nits here, mate.
The bottom line is that there isn't much new in the design, which means that there is much less chance of a delay.


RE: Correction
By coldpower27 on 11/28/2006 11:53:06 AM , Rating: 1
Oh were, establishing correct information, so it is important. Correct terminology will likely lead to less issues down the road and reduce subjectivity.

At any rate what Intel is doing with Penryn is not too complex given the minor additions so it should arrive on time.


RE: Correction
By Viditor on 11/28/2006 12:02:04 PM , Rating: 2
quote:
Oh were, establishing correct information, so it is important. Correct terminology will likely lead to less issues down the road and reduce subjectivity


Strangely, I received my terminology from a chip designer at Intel...go figure. I personally can't attest to which is "correct" as I don't work in the field...


RE: Correction
By slayerized on 11/28/2006 12:05:32 PM , Rating: 4
Just to add in! Optical shrink is not possible also for the reasons mentioned in the article with changes to the dielectric and a couple of other process changes. Cache may be a non-architectural change; but it may be a placement issue (so optical shrink? maybe not!)


RE: Correction
By Viditor on 11/28/2006 9:31:35 PM , Rating: 2
quote:
Optical shrink is not possible also for the reasons mentioned in the article with changes to the dielectric and a couple of other process changes

Thanks for pointing that out, mate...
The confusion I have there is that the article DT points to about the changes states:
"Intel said that its first 45nm processor will be Nehalem, which will go into production sometime in 2007 and be introduced in 2008"

Obviously that's incorrect, but what I'm wondering is whether just the name is wrong or if the High-K metal gates will be instituted on Nehelam first...which would mean that Penryn really is more of an optical shrink (or dumb shrink).

On the other hand, if Intel is indeed including the new High-K metal gates in Penryn, then the chances of a delay and low yields go up significantly.
I do know that they are also incorporating new microcode to help with power/performance, but that shouldn't effect production at all...


RE: Correction
By Furen on 11/28/2006 10:40:52 PM , Rating: 2
We have thus decided that using the term "dumb shrink" is demeaning, unfair and prejudicial. As such, we will hereby transition to a more politically correct "optical shrink"...

Seriously, though, Penryn is a dumb shrink. Tweaks here and there don't make it a significant redesign, and neither does expanding expandable cache. SSE4 is very likely in current Conroes, just disabled, since Intel does like to throw stuff into its processors and just flipping a switch at a time of its chosing. Brisbane, too, will be a dumb shrink (a dumber one than Penryn, yes) even though it will add half-multipliers (gasp!) and probably fiddle with the memory controller (every K8 revision has fiddled with the memory controller).


RE: Correction
By DallasTexas on 11/29/06, Rating: 0
RE: Correction
By rushfan2006 on 12/4/2006 2:16:51 PM , Rating: 2
Guys here's an insight....STFU...

We care when the product is out...that's it...end of story.

Dumb shrink, optical shrink, whatever the hell you want to or want NOT to call it...it doesn't MATTER!!


Don't get TOO excited...
By Viditor on 11/28/2006 4:25:27 AM , Rating: 2
Sampling today means that Penryn is on track to begin volume production in Q3 of 2007, and should arrive in stores either at the end of 07 or the beginning of 08.




RE: Don't get TOO excited...
By Master Kenobi (blog) on 11/28/2006 8:13:06 AM , Rating: 1
You forget who your talking about here. Intel does not adhere to the 1 year sampling to availabiliy scheme that AMD tends to adhere to. Given Intel's track record for process shifting, I'd say we could see these as early as June/July.


RE: Don't get TOO excited...
By Viditor on 11/28/2006 8:39:03 AM , Rating: 2
quote:
You forget who your talking about here

I'm sorry to rain on your parade here, but Intel does indeed ship about 1 year after sampling. It's not an adherence to any rule, that's what it takes for everybody to finish getting it checked out (including the 3rd parties).
Besides, Intel just finished publishing (at IDF) that it will be in stores in Q1 08...
Methinks you have a rather idolized view of the Intellista...


RE: Don't get TOO excited...
By Master Kenobi (blog) on 11/28/2006 10:41:13 AM , Rating: 2
Do check your facts nextime. It's set to SHIP in H2'07, which puts it anytime between July and December. More likely to see an August/September availability. Not quite 1 year.

http://www.channelregister.co.uk/2006/11/28/intel_...


RE: Don't get TOO excited...
By Viditor on 11/28/2006 10:51:37 AM , Rating: 2
quote:
Do check your facts nextime

Sigh...I have checked my facts very well.
This is from last month at IDF...Intel's official announcement.
http://techreport.com/etc/2006q4/fall-idf/index.x?...
Please note that Penryn arrives in Q1 08...
Shipping in Dec of 07 means that it arrives in stores for January of 08 (just as Conroe went on sale a little over a month after it started shipping).
It's quite possible that they can push that forward a month, but that really does them no good from a marketing perspective.

I really think you're looking at this through rose coloured glasses, mate...


RE: Don't get TOO excited...
By Master Kenobi (blog) on 11/28/2006 11:14:47 AM , Rating: 2
Nah, your point is valid, I'm leaning towards Intel's track record of getting things out the door as fast as they can. Retail availability might be later, but I'm thinking more of the online retailers which tend to get things fairly quickly. Q1'08 is likely the estimate for systems showing up at your local Circuit City or Best Buy.


RE: Don't get TOO excited...
By Viditor on 11/28/2006 11:33:25 AM , Rating: 2
quote:
I'm thinking more of the online retailers which tend to get things fairly quickly. Q1'08 is likely the estimate for systems showing up at your local Circuit City or Best Buy


Actually, the first to get the parts will be OEMs (and I suspect that Apple will be the first of these). This occurs about 1-2 months (which is the shipping date) prior to launch. The Channel (online retailers and local retailers) get them second, and are usually prohibited from selling until launch date...however they can do pre-orders.


By Master Kenobi (blog) on 11/28/2006 1:01:04 PM , Rating: 2
More or less, I was able to get my paws on an E6600 about 2 days after Dell started taking C2D orders. >.>


In layman's terms...
By jskirwin on 11/28/2006 8:39:42 AM , Rating: 1
What is the value of this change? Are there any performance benefits - or is this more of a price thing since Intel can pump out more chips per wafer?




RE: In layman's terms...
By Viditor on 11/28/2006 8:46:07 AM , Rating: 2
quote:
What is the value of this change? Are there any performance benefits - or is this more of a price thing since Intel can pump out more chips per wafer?

Mainly just a cost saving thing...but it will also allow
1. a slight increase in clockspeed
2. more room on the die for the larger 6MB of cache
3. slightly reduced heat/power


RE: In layman's terms...
By Blauhung on 11/28/2006 10:29:29 AM , Rating: 2
this is because smaller transistors end up with possibilities for faster switching with less leaked current.


RE: In layman's terms...
By Spivonious on 11/28/2006 10:59:09 AM , Rating: 2
If Intel reduces the heat any more, we won't need active cooling solutions for CPUs. Even now I can run the E6600 without the fan running when the CPU is idling. It stays under 55 C.


RE: In layman's terms...
By ajfink on 11/28/2006 12:00:04 PM , Rating: 2
That's something I like hearing. This trend is a beautiful reversal from earlier trends in processors. Now if only the GPU industry will catch on (see: nVidia ~8800).


RE: In layman's terms...
By Master Kenobi (blog) on 11/28/2006 12:59:54 PM , Rating: 2
Yea we won't get into the graphics market problems just yet. But in layman's terms.... they make my stovetop look weak.


What do we know about what's coming after Core?
By Hulk on 11/29/2006 11:22:48 AM , Rating: 2
The end of this article says this 45nm process will be the end of Core. Has anyone read reports on what Intel is looking into for the replacement for the Core architecture?




RE: What do we know about what's coming after Core?
By trivik12 on 11/29/2006 1:13:27 PM , Rating: 2
Mid 2008 should see Nehalem releasing. It is predicted that Nehalem would have CSI and IMC.


http://www.channelregister.co.uk/2006/11/28/intel_...


RE: What do we know about what's coming after Core?
By Viditor on 11/29/2006 5:26:50 PM , Rating: 2
quote:
Mid 2008 should see Nehalem releasing


Actually (to pick nits), Nehalem should begin production in mid 2008, which gives it a release date at the end of 2008/early 2009...

Intel is predicting that from here on out, we will see a major change every year...
A node change every other year, and an architecture change every other year. They will be staggered, so one year it's a shrink and the next it's an architecture change...

AMD is predicting the same thing for themselves, but their change happens mid-year


By Targon on 11/30/2006 8:21:53 AM , Rating: 3
And when you change designs that quickly, there is a lot of potential for problems. Based on the past performance from Intel, they tend to only change designs when their old one runs out of room for improvements in speed.

The Pentium 3 was phased out because Intel couldn't use it to compete with AMD in the MHz race for example. When the Athlon 64 continually was able to stay ahead of the Pentium 4, and Intel ran into problems boosting the speed of the P4, they released Core.

Now, since Core is based on the old Pentium Pro design(which later became the Pentium 2, then 3, then Pentium-M), I personally had my doubts about how far they could ramp that design, even with all the tweaks possible, there WOULD be a limit.

To be fair, the Athlon design is a bit old as well, though not as old as the Pentium Pro design. So I am expecting at some point in the next two years for AMD to release the K9 or K10(I've heard K9 was going to be skipped), and it won't just be a tweaked chip but will be a new design.

The idea that an all new design will be released every other year from Intel is pushing it. Major changes I can see, but a change just to be different if it's not needed seems like a waste of the talents of the designers.

I do find it entertaining that after Intel makes a lot of negative comments about what AMD is doing that they look to be adding a lot of AMD features to their own chips in the future. I bet Intel would have adopted HyperTransport themselves if they could get away with calling it Intel HyperTransport or some other name to make it look like THEY were the one to come up with the concept and get away with it.


I will believe it when I see it.
By mkruer on 11/28/06, Rating: -1
RE: I will believe it when I see it.
By Viditor on 11/28/2006 4:56:22 AM , Rating: 2
Keep in mind that there will be NO architectual changes for Penryn (except added cache). Intel's people are very good at this, and I really consider almost a foregone conclusion that they will be successful...
If there is a Bogeyman for the release, it's going to come from the chipset division...remember that Penryn will be on 1333 MHz. That said, Intel is already using this for their Woodcrest chips...the question will be how well they do on their consumer chipsets.


RE: I will believe it when I see it.
By Marcus Pollice on 11/28/2006 7:38:20 AM , Rating: 2
While there might be no major architectural changes, Penryn will introduce SSE4 instruction set extension with about 50 new commands. Maybe they can achieve that by updating the microcode but I guess it also involves minor changes.

AMD should be converted to 65nm by summer 2007, if they stay on track. Lately AMD had a track record of delivering to their promises.

FSB1333 chipsets shouldn't be the problem as Intel plans to introduce the Bearlake series starting in Q2 2007 (P35, G33) with X38 and G35 following later on in Q3. There will also be 65nm CPUs with FSB1333 in Q3. So that shouldn't be the problem.


RE: I will believe it when I see it.
By Viditor on 11/28/2006 8:43:31 AM , Rating: 2
I tend to agree Marcus...and yes, they should be able to just update the microcode.
I know that Bearlake is due, but we haven't seen it happen yet and it seems to me that it will be more difficult than a dumb shrink (Penryn).
I feel fairly confident that Intel will release on time, but fecal-matter occurs...:)


By Marcus Pollice on 11/28/2006 10:43:07 AM , Rating: 2
Your doubt about chipset availability might be rectified if you look at shortages in the past. While their CPU execution is excellent as of late, they sometimes fail to deliver the same on the chipset side.


RE: I will believe it when I see it.
By coldpower27 on 11/28/2006 11:35:11 AM , Rating: 2
Penryn itself is doubtfully going to be 1.33GHZ FSB as it is the mobile derivative of Intel's 45nm shrink of Core Architecture.

Intel will be refreshing Conroe to be on 1.33GHZ FSB as well.
So the support will be there.

It's good to see the mobile product sampling now however, this bodes well for the desktop and server SKU's slated for this design.


RE: I will believe it when I see it.
By Viditor on 11/28/2006 11:42:02 AM , Rating: 2
quote:
Penryn itself is doubtfully going to be 1.33GHZ FSB as it is the mobile derivative of Intel's 45nm shrink of Core Architecture


Penryn is desktop...though it may be mobile as well.
http://techreport.com/etc/2006q4/fall-idf/index.x?...


RE: I will believe it when I see it.
By coldpower27 on 11/28/2006 11:51:02 AM , Rating: 2
The context used there is that Penryn is refering to all of the 45nm Core 2 being produced by Intel and we know that will not be in the case.

Penryn is indeed slated to be a mobile part, the server and desktop are not called Penryn cores, though they derive the same basic technology.

This is like saying Barcelona architecture and all cores are named that, this is just not the case.

That roadmap is very general right now, so they use one name to describe, the entire 45nm Core Microarchitecture family.



RE: I will believe it when I see it.
By Viditor on 11/28/2006 11:56:28 AM , Rating: 2
quote:
Penryn is indeed slated to be a mobile part, the server and desktop are not called Penryn cores, though they derive the same basic technology


Ummm...not to be a doubting Thomas, but where did you get that from?

quote:
This is like saying Barcelona architecture and all cores are named that, this is just not the case


Well no, it's like calling it K10 architecture and all cores are named that, which is the case. Unless you have special knowledge that hasn't been published yet...?


By intangir on 11/28/2006 11:03:07 PM , Rating: 2
The desktop chips based on Penryn are named Wolfdale and Ridgefield. Penryn, like Merom, refers to both the architecture and the mobile chip.

http://www.tomshardware.com/2005/12/04/top_secret_...


By IntelUser2000 on 11/28/2006 1:05:59 PM , Rating: 4
To be fair, Penryn isn't a complete dumb shrink:
http://sg.vr-zone.com/?i=4159

"Mooly Eden is particularly excited about Penryn and it is more than just a die-shrink. He revealed about lots of micro-tweakings done to make Penryn even more energy efficient and at the same time improving performance."


By trivik12 on 11/28/2006 4:29:20 PM , Rating: 2
I think intel has been amazing when it comes to executing things on schedule this year. 65nm was on time yonah was delivered on time conroe/merom/woodcrest came before time and so did tulsa. Only Montecito was delayed and Itanium is a different beast all together.

Take AMD's case Socket F platform got delayed. Their 65nm schedule changed a lot in last couple of years. 1st they said mid 2008, then 2nd half 2008 then Q4 2008 and December. It just kept pushing things till end.

We have yet to see any AMD 65nm chip and its hard to accept they would move over completely to 65nm in 3 quarters.


"Spreading the rumors, it's very easy because the people who write about Apple want that story, and you can claim its credible because you spoke to someone at Apple." -- Investment guru Jim Cramer

Related Articles
AMD Beyond "Brisbane"
November 14, 2006, 6:31 PM
"Conroe" Refresh Details Unveiled
October 17, 2006, 2:05 PM
Intel Talks 45nm Production
September 26, 2006, 2:19 PM
Intel Focuses on 45 Nanometer Technology
January 25, 2006, 3:05 PM













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