backtop

Print 47 comment(s) - last by typo101.. on Jan 18 at 7:17 PM
More good news for those anticipating 45nm processors

Six days after DailyTech broke the news of Intel's 45nm Penryn tape-out, the company responded today with an official announcement regarding the story.

A statement sent to DailyTech reads "Paul [Otellini] said we have working samples of the Penryn processors -- and in a short period of time have already booted 4 operating systems (Windows Vista, XP, Linux and Apple Mac OS-X)."

The original DailyTech report only specified that Windows had booted off the new CPU. 

The company would not comment on whether or not Penryn's launch window was moved up, but added "These are good indicators of how healthy our 45 nm manufacturing and future product designs are looking so far as we plan to begin manufacturing production in the second half of the year ramping to three 300mm factories in 2008."

The two of the three 300mm facility mentioned are Fab 32 Chandler, Arizona and Fab 28 in Qiryat Gat, Israel.  Development for 45nm processors is currently done at Intel's D1D facility in Oregon.  All three facilities are expected to be fully production-ready for 300mm wafers on 45nm nodes by 2008.

In early 2006 the company announced it will build a massive 300mm wafer facility in Ho Chi Minh City, Vietnam.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Question regarding tape-outs
By Blackraven on 1/17/2007 8:15:34 AM , Rating: 2
How does it work?

Is it by chance that a tape-out can occur?
Is it due to technological advancements that lead to higher yields in early stages or production?
Are tape-outs doable (esp. for Intel's future processors)?

Questions like those.




RE: Question regarding tape-outs
By Viditor on 1/17/2007 8:33:29 AM , Rating: 2
A first tape out is when all of the designs that the various teams have been working on is first made into a mask and silicon is produced.
Having it be bootable is certainly a cause for celebration because you have to remember that none of those designs have actually been built together before...

That said, this is nowhere near what will be released at the end of the year yet. There are still at least 2 turns of tweaks and modifications that need to be performed before volume production can occur (a "turn" is the period it takes to turn a silicon wafer into chips...about 90 days).


RE: Question regarding tape-outs
By IntelUser2000 on 1/17/2007 9:33:59 AM , Rating: 1
quote:
That said, this is nowhere near what will be released at the end of the year yet. There are still at least 2 turns of tweaks and modifications that need to be performed before volume production can occur (a "turn" is the period it takes to turn a silicon wafer into chips...about 90 days).


For those that don't know or forgot, Intel's 45nm Penryn TAPED OUT NOVEMBER 2006.


RE: Question regarding tape-outs
By Viditor on 1/17/2007 9:56:25 AM , Rating: 2
quote:
For those that don't know or forgot, Intel's 45nm Penryn TAPED OUT NOVEMBER 2006


Exactly...the rule of thumb is that products are released about 1 year after first tape out. Remember that they have to begin volume production at least 4-6 months prior to release (3 months to turn out a wafer, and then you have to build inventory for release).

We should expect to see Penryn on the shelves at the end of 2007, right on schedule.


RE: Question regarding tape-outs
By therealnickdanger on 1/17/2007 10:12:25 AM , Rating: 2
We'll probably see 45nm engineering samples being abused on Xtremesystems.org by this August.

Core3Duo? Yummy.


RE: Question regarding tape-outs
By Zandros on 1/17/2007 5:10:51 PM , Rating: 2
No, Core 3 Duo/Quad would be Nehalem if anything.


RE: Question regarding tape-outs
By masher2 (blog) on 1/17/2007 12:12:43 PM , Rating: 2
> "the rule of thumb is that products are released about 1 year after first tape out..."

The industry average is 9-12 months....but it all depends on the level of debugging required. Given that the first Penryn tapeout seems extraordinarily stable, I'm betting we'll see it a good bit sooner than the end of 2007.


RE: Question regarding tape-outs
By Viditor on 1/18/2007 9:44:12 AM , Rating: 2
quote:
The industry average is 9-12 months


Ummm...no. Certainly not for CPUs (though maybe so if you're talking about all semiconductors, as things like memory tape-outs require very little tweaking afterwards). 1 year is considered the minimum standard for anything but a paper launch in CPUs...


RE: Question regarding tape-outs
By MrBowmore on 1/18/2007 7:43:17 AM , Rating: 2
It isnt THAT fantastic. It´s build on roughly the same architechture, whith "only" few enhancements such as approx. 50 new SSE intructions, I give em creds for that, but the architechture itself is not new, so they know most of the parts work allready.
They got the litography to work right, but to have that to work and also a totally new arcitechture... thats something else, I dont think "Core 3 Quad" Nehalem, will have the same destiny with the 32 nm process node. maybe, causs it will first implement itself on the 45 nm process node. Their roadmaps look really interesting anyhow.


ok
By sprockkets on 1/16/2007 11:01:34 PM , Rating: 3
Just keep in mind, frequencies are getting that much better, and the only way processors are getting better is by adding more cores.

Again, how many people again need dual core processors in their computers anyhow? I like how there are s939 2.2ghz AMD 64 OEM processors for $55.




RE: ok
By Regs on 1/17/2007 8:12:55 AM , Rating: 1
Exactly. I can see some limited uses of dual core in multi-tasking for a everyday user like myself who maybe runs photoshop and some other cpu intensive media program in the background. Even then you are still getting the same performance except you don't get the transition penalty half the time.

We still live in the age of single thread x86 applications and that's not going to change anytime soon. So simply slapping on another core is more or less a lateral move more so than a upgrade and I don't care how the marketeers or marketarrets like to spin it. Just like how you have the AMD X2 vs the Intel Core Duo, which one would you pick? The "real" dual core or the "fake" dual core that does everything at least 30% faster?


RE: ok
By masher2 (blog) on 1/17/2007 9:39:55 AM , Rating: 4
> "We still live in the age of single thread x86 applications ..."

You're guilty of assuming that your own personal usage is reflective of the entire world. There are literally thousands of applications that not only make good use of dual cores, but four or more. Pretty much any modeling, simulation, scientific, transcoding, or high-end financial analysis package fits this profile....along with many others. An application I regularly use works well with 326+ cores...and I've had a quad-cpu workstation for 6 or 7 years now.





RE: ok
By masher2 (blog) on 1/17/2007 10:24:43 AM , Rating: 2
edit: 326+ = 32+


RE: ok
By IntelUser2000 on 1/17/2007 11:13:30 AM , Rating: 2
quote:
We still live in the age of single thread x86 applications and that's not going to change anytime soon. So simply slapping on another core is more or less a lateral move more so than a upgrade and I don't care how the marketeers or marketarrets like to spin it. Just like how you have the AMD X2 vs the Intel Core Duo, which one would you pick? The "real" dual core or the "fake" dual core that does everything at least 30% faster?


"Fake" dual core?? X2 is less of a dual core than Core Duo. And its the Core 2 Duo that's 30% faster.


RE: ok
By FITCamaro on 1/17/2007 4:28:44 PM , Rating: 1
Uh...X2 and Conroe are both true dual cores. If you wanna get down to it though, AMDs design is better since both processors can directly talk to each other via their HT links. Even with Conroe, the cores have to talk to each other over the FSB. It's not enough of a bottleneck to slow them down a lot obviously, but think about what'd they do if they could talk directly to one another. And as Intel adds more cores, they'll eventually have to ditch the FSB. AMDs got a leg up in that arena. They just have to add more HT links.

So how is X2 less of a dual core?


RE: ok
By MrBowmore on 1/18/2007 3:26:16 AM , Rating: 3
Hold ya horses!
The Core 2 Quad "talks" through the fsb, the core 2 duo is a solid dual core which have a unified cache, it´s a SINGLE die. The cores DON´T have to go through the fsb to get necessary information. So, no, you are wrong, get it right next time by reading the facts next time.
OnO


god help amd
By Pirks on 1/16/07, Rating: -1
RE: god help amd
By archcommus on 1/16/2007 10:19:11 PM , Rating: 5
We really don't need the "poor AMD" and "OMG what is AMD going to do" comments on EVERY single article about Intel's progress. Yes, Intel is progressing, but AMD will be out with 65 nm mid this year, and so will be about a half generation behind Intel just like they've always been. Fact is, they've got good CPUs, great chipsets, and great GPUs, and are doing just fine.


RE: god help amd
By xFlankerx on 1/16/2007 10:29:23 PM , Rating: 2
Its strange that you call it "half a generation" when Intel's scheduled to release 45nm processors only one quarter after AMD's 65nm. Considering AMD's product will be a C2D competitor, and Intel's will be a successor (as demonstrated by 3.5-4.0Ghz clock speeds), I can't be too optimistic about AMD.

Disclaimer; Typing this from an AMD system. It may have been the product of choice in '05, but times change.


RE: god help amd
By nerdye on 1/16/2007 10:29:44 PM , Rating: 2
I agree that we hear, omg amd is finished all the time due to announcements of intel's current developments. I can't refute that 65nm runs pretty darn cool compared to 90nm cpus, the same will be true with 65nm compared to 45nm. Amd is not finished, am2 was a rough transition to ddr2 that perhaps only awarded the am2 adopters the ability to migrate their am2's ddr2 memory to their next amd rig (who wants to run an am2 mobo on a yet to be released hypertransport 3.0 cpu, seems silly), but amd still lives. Performance is the key to profits in the tech industry, yet we have some exceptions based on reputation, p4 netburst anyone? Amd64 has a great reputation and that is what amd will live and breath on until they have a true competitor to core 2 duo. Amd is not out for the count, only time will tell.


RE: god help amd
By melgross on 1/17/2007 12:01:36 AM , Rating: 2
AMD is about a generation behind. They're trying to move to half a generation. They hope to accomplish that by end of 2008.

Right now, they have announced 65 nm, but supply is almost non-existent. Don't expect to see ample supplies of their first unoptimised chips for another two months.

Intel will be moving off it shortly aftwewards.


RE: god help amd
By ChipDude on 1/17/2007 1:05:52 AM , Rating: 2
God help you fools who can't get a clue.

F32 = 1 factory
F28 = 2 factories
D1D = 3 factories.

Do any of you read. The place in Vietnam is a low tech assembly site where american companies pay a bunch of asian women to sweat for 10 hours a day for a less then a buck an hour.


RE: god help amd
By paydirt on 1/17/2007 10:00:53 AM , Rating: 2
A buck an hour is better than a buck a day working as a farmer. Ten years down the road, the growth in income will still be strong for open-market Asian countries.


RE: god help amd
By ChipDude on 1/17/2007 7:04:56 PM , Rating: 2
A buck an hour is good.

I see the logic. I have visited expats working in China. They all brag about their fancy living quarters and maid that they pay a few bucks a day to clean their house, fold their cloths, pick up their used underwear, shop for their food, cook. Damm I can't wait to hear if they wipe their butts and shake the dick after they pee.

They feel its all good as they are offering them a good wage, 20x what they make in the country.

The saddiest part some of the most blatant abusers of it are Chinese returning to China.



Intel is desperate to fool Wall Street types
By cornfedone on 1/16/07, Rating: -1
RE: Intel is desperate to fool Wall Street types
By Viditor on 1/17/2007 9:04:58 AM , Rating: 3
quote:
Intel actually did much better this quarter than analysts expected


The key point that investors are looking for is Intel's Gross Margin. When C2D was released, Intel promised that the company's GN was going to return to the mid 50s by year's end...it ended up as 49.6% (down from 61.8% a year ago).
The reason that analysts look at GM as the most important is because it is less sensitive to products in the channel, which may not be accounted for properly. Intel's accounting states that products shipped into the channel is counted as sold, even if they don't get paid for it...and quite often a lot of that product is returned.
There's nothing wrong with this, but it does skew the EPS and profit numbers significantly.

quote:
Plenty of margin to continue applying price-screws on AMD

Sure there is...Hell, they could drop the GM down to 5% if they wanted to. The question is, what are they getting for all of that money? Intel also guided in Q2 that they expected to get back to marketshare numbers of a couple of years ago by years end...unfortunately for them, the opposite has occured. AMD has continued to gain marketshare, even with inferior products at the high end of desktop and mobile.

When K10 comes out next quarter, if it's good enough to meet or beat Intel's C2D (and at least the specs suggest that this is the case), then I expect that Intel will have to keep their GM low for quite some time to maintain marketshare levels. Of course this will hurt AMD as well, but I think there are some considerations you are forgetting here...
1. AMD's costs will drop significantly this quarter as they have turned over the halfway mark on 300mm, and 65nm should be halfway done this quarter as well. They are to be fully 65nm by mid-year...
2. The purchase of ATI (while expensive) garnered many things for AMD...it gets them a huge amount of R&D, they cut the cost of developing a chipset for their K10s (typically, AMD has spent more money on developing chipsets for their CPUs than on the CPUs themselves), they open up a new market for themselves (CE and Graphics), and they fill the last advantage that Intel had...being able to design a whole platform in-house.


By Khato on 1/17/2007 12:09:37 PM , Rating: 2
quote:
AMD has continued to gain marketshare, even with inferior products at the high end of desktop and mobile.


Q4 marketshare numbers have been released? I was only aware of the Q3 numbers, in which that statement was true. My impression was that Q4 numbers weren't to be available for another week or two?


RE: Intel is desperate to fool Wall Street types
By melgross on 1/17/2007 12:04:47 AM , Rating: 1
Reality is that AMD is in worse shape. A huge profit slide. Stock dropping daily. Their newest chip lines still are behind.

Competition is always good for us. No doubt about that. And it hurts Intel, but it hammers AMD.


RE: Intel is desperate to fool Wall Street types
By Targon on 1/17/2007 7:55:43 AM , Rating: 1
It has to be noted that while Intel may be a good short-term buy due to the news about their process technology, that doesn't make up for having a total SYSTEM architecture that is and has been a bottleneck when it comes to overall performance.

Don't get me wrong, Intel has the performance lead right now, but Intel is still focused on improvements to each individual area of the computer, and trying to make each one "better" while AMD looks at how they can make the overall system run faster. This is why AMD was able to beat the P4 for most of the A64 vs P4 competition, and why so many people expect that the K8L may allow AMD to catch back up or beat Intel in terms of performance.

In my opinion, AMD has made a mistake in not releasing K8L on a 90nm process before mastering 65nm.


RE: Intel is desperate to fool Wall Street types
By stromgald on 1/17/2007 11:41:20 AM , Rating: 2
Wrong. IIRC, Intel was the first to adopt a platform/system level strategy when they introduced the Centrino platform for notebooks in 2003. Intel continued that with the Viiv platform early last year. AMD matched that with their AMD Live! platform at about the same time. I think AMD's is edging out the Viiv by a little bit, but I'm not sure of the numbers. I don't think either side has really adopted a 'system' level perspective for their desktop systems.

The A64 beat the P4 due to the onboard memory controller (which doesn't really constitue a platform/system perspective IMO) and power efficiency. The P4 Prescotts were inefficient and produced alot of heat, which limited the speed of the processor and allowed the more efficient AMD64s to beat the Prescott with similar cooling.


RE: Intel is desperate to fool Wall Street types
By Motley on 1/17/2007 12:20:19 PM , Rating: 1
The first paragraph is correct, but your analysis of the AMD/P4 matchup is a little off. The memory controller has very little to do with it, although AMD (and their fanboys) love to use it as the best thing since sliced bread. If the memory controller really made that much of a difference, why is AMD getting their a$$ handed to them while intel still hasn't got an integrated memory controller?

The problem was that intel was never able to solve their heat problems (Which I give you is related to the power efficiency). The P4 would have been a fantastic processor if they would have been able to crank up the speed of it, especially when dealing with tight loops of code processing lots and lots of data (Hence the name netburst). You can see the changes, from reducing the IA32 instructions to micro-ops, and caching the micro-ops themselves, the out of order execution units, the very fat/fast memory choice (Remember they originally were pushing RAMBUS which has many times the throughput of SDRAM at the time).


By FITCamaro on 1/17/2007 4:17:02 PM , Rating: 2
quote:
If the memory controller really made that much of a difference, why is AMD getting their a$$ handed to them while intel still hasn't got an integrated memory controller?


Because C2D is a 4 instruction per clock processor vs. AMDs 3. Processor's are the same as graphics cards. You don't compare one generation to the previous. AMDs chips are by no means slow. Just Intel got out its next generation first. Props to Intel for it and creating a great processor (have an E6600 myself). In the 4P server and above space though AMD is still highly competitive with Intel due to their much lower latency from the onboard memory controller and the processor's being able to talk to each other via HT vs. Intel having to talk through the FSB. Having buttloads of cache as the Xeon's now do doesn't help that.

Right now the main advantage of Intel in that space is quad cores but AMD will have that patched up relatively soon, and with a native quad core vs. Intel's thrown together one. The quad core Core 2s are by no means slow, but they'd probably be faster if it wasn't two dual cores slapped on a die. Intel rushed it to market because they knew they needed to beat AMD to it.

Yes AMD is in a hard place right now but that's where they've always been. I think AMD is more used to it than Intel. Intel's long considered AMD a nuisance much like Microsoft has considered Apple one. But both of the underdogs are coming back fighting hard (hate saying it since I hate Apple).


RE: Intel is desperate to fool Wall Street types
By Russell on 1/17/2007 1:00:40 PM , Rating: 2
I would hardly call Centrino a platform level strategy in a technical sense. Centrino is a marketing platform, nothing more.


By typo101 on 1/18/2007 7:17:51 PM , Rating: 2
took the words out of my mouth


RE: Intel is desperate to fool Wall Street types
By slickr on 1/17/07, Rating: -1
RE: Intel is desperate to fool Wall Street types
By Pirks on 1/17/07, Rating: -1
By Oregonian2 on 1/17/2007 3:15:42 PM , Rating: 1
Well, you do need to change AMD mobo's every few months when AMD changes the socket they use.

After being on the AMD bandwagon since I upgraded my 486-50 a long time ago, I've finally jumped back with the Conroe processor. AMD's top of the X2 line still doesn't quite perform as fast as my middle-line Conroe. AMD has dropped prices quite a bit to compete since I made my upgrade (they weren't even close last year when the Conroe's were relatively new), but they're still changing sockets. AM2/AM3 will stay for a while fer sure! Right.



By FITCamaro on 1/17/2007 4:19:06 PM , Rating: 2
Do you enjoy looking like a 12 year old kid on a caffeine high?


By ChipDude on 1/17/2007 1:10:19 AM , Rating: 2
Desperate.. heah?
How much did they spend buying back shares
How much did they pay out in divends
How much did they invest in R&D
How much did they spend building 65nm and 45nm factories

And they still made how much.

Their stock sucks... but any investor would know not to invest in INTEL. They were the sorriest stock in the DOW last years. But as a business they continue to be a pretty solid one. WHo knows maybe they'll make a run like GM did going from dog to leader. ALl it takes is for 45nm Penrym to be a hit in 2007, AMD to release Barcelona in 2nd half 2007 and only be 20% faster then current offerings and you could easily see INTEL be at 30 bucks by EOY if the economy recovers.

I'd agree last year going into 2006 INTEL looked worst then desperate.. this year it couldn't look better!



By ScythedBlade on 1/17/2007 3:39:03 AM , Rating: 3
... When they're announcing that they're firing employees ... sometimes that may be a good thing I heard... That's usually when more people will buy stocks of that company.

And, no, they're not BSing. And Netburst did not contribute to SO much global warming. (Dude, a hairdryer in friken 2 minutes consumes more electricity than a Netburst for 2 hours!)


RE: Intel is desperate to fool Wall Street types
By PrinceGaz on 1/17/2007 6:07:04 AM , Rating: 2
quote:
Dude, a hairdryer in friken 2 minutes consumes more electricity than a Netburst for 2 hours!

I hope that wasn't meant to be taken literally because it is untrue.

A typical hairdryer generally uses a maximum of 2KW of power. For the hairdryer to use more electricity in 2 minutes than a Netburst CPU in 2 hours, it would require that the Netburst use less than 2KW/60, or under 33 watts.

Most Netburst proccesors I'm aware of use rather more than 33 watts.

What you said would be true if you said a hairdryer uses more elecricity in 5 minutes than a Netburst CPU in 2 hours if you are talking power consumed by the CPU alone and exclude the hottest running Netburst models. If you factor in PSU efficiency and other losses and consider the hottest running Netburst processors, then you would need to have the hairdryer running over 10 minutes before it used more electricity than what the computer used in 2 hours to feed the CPU alone.


By theapparition on 1/17/2007 7:26:48 AM , Rating: 2
In all fairness, the OP did state a 2 min op time for the hairdryer. Using your own numbers, that would be 66.6W, which is more in line for most netburst processors, excluding presscott.

My challange to you is to find any woman who doesn't use the hairdryer for more than 5 minutes


By theapparition on 1/17/2007 8:09:15 AM , Rating: 2
correction,
LESS than 5 minutes


By ScythedBlade on 1/17/2007 10:40:07 AM , Rating: 3
*poke* ... dude my hair dryer has a ... (err is TDP the correct term?) label with 2.5 kW ... and yes its from that conair company i think ... (don't ask me about my hair dryer, i just know i have one and i was bored one day and decided to look at it)


By masher2 (blog) on 1/17/2007 12:06:26 PM , Rating: 1
> "*poke* ... dude my hair dryer has a ... (err is TDP the correct term?) label with 2.5 kW ... and yes its from that conair company i think ..."

Then its probably a dual-voltage model. Plug it into a 240VAC euro socket and it'll draw 2500w. But in the US, its limited to 1650w (1875 if you want to assume the max 125 VAC).


"So, I think the same thing of the music industry. They can't say that they're losing money, you know what I'm saying. They just probably don't have the same surplus that they had." -- Wu-Tang Clan founder RZA











botimage
Copyright 2012 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki