Sign in with
Sign up | Sign in
Your question

Intel 32nm by 2009

Last response: in CPUs
Share
September 19, 2007 1:49:31 PM

Some good info straight from the horses mouth...

During a keynote at the Intel Developer Forum in San Francisco, Intel president and CEO Paul Otellini showed a 300mm wafer built using the 32-nm manufacturing technology. The chip will house more than 1.9 billion transistors and its increased performance will enable "true to life entertainment and real-life graphics capabilities," Otellini said in his keynote.

The chips will be an upgrade over processors built using the 45-nm process Intel is incorporating in its Penryn processor, due in November, and Silverthorne and Nehalem processors, slated to appear early and in the second half of next year, respectively.

Intel currently uses a 65-nm process to manufacture chips, and Penryn is the code name given to the 45-nm "shrink" of Intel's current chip designs. The measurements refer to the size of the features on the silicon chip.

In the first public demonstration of the Nehalem processor, Otellini said it will deliver better performance per watt and better system performance through its QuickPath Interconnect system architecture, which will include an integrated memory controller and improved communication links between system components.

Otellini also announced a Penryn dual-core processor operating at 25 watts that will be available on the upcoming Montevina platform, which will also include WiMax technology. To meet multiple computing needs, Otellini said Intel also plans to introduce 15 new 45-nm processors by the end of the year and 20 in the first quarter of 2008.

Intel isn't the first to announce 32-nm chip technology. In May, a group of chipmakers led by IBM agreed to further collaborate to jointly develop 32-nm semiconductor production technology. Other companies in the collaboration include Freescale Semiconductor, Chartered Semiconductor Manufacturing, Infineon Technologies, and Samsung Electronics.

There was nothing surprising in Otellini's keynote, said Doug Freedman, an analyst at American Technology Research.

Intel's road map tends to be conservative, and the company is well on track to meeting its time line, he said. With the new chips, users will continue to get more processing performance at a similar price point, he said.


http://www.infoworld.com/article/07/09/18/Intel-chips-s...

Regards

Chad

More about : intel 32nm 2009

September 19, 2007 3:06:31 PM

wasn't Nelhelm supposed to be a 45nm chip with a 32nm shrink following a year later?
September 19, 2007 3:20:43 PM

jwolf24601 said:
wasn't Nelhelm supposed to be a 45nm chip with a 32nm shrink following a year later?

Presler = 65nm
Conroe = 65nm - new core
Penryn = 45nm
Nehalem = 45nm - new core
Nehalem-C = 32nm
Gesher = 32nm - new core
Related resources
September 19, 2007 3:27:03 PM

This is not that new although it is still interesting.

I doubt AMD will have 32nm by then because they are still just getting going with their 65 nm while Intel is about to release their 45 nm. Intel will only have I believe two major cores (Penryn and Nehalem). After those two, then they drop down to 32 nm.

Right now AMD simple dose not have the money to change their line again unless they merge or get bought by either IBM or Samsung.
September 19, 2007 6:29:07 PM

So MrsFUD, Intel already has 32nm demonstrated (that's kind of what the article mentions) - IBM still hasn't shown functional 45nm logic (they only released a "ME TOO!" statement when Intel released theirs, but without proof). AMD's not exactly the same as IBM, even though IBM does most of AMD's R&D for them, so AMD's 45nm schedule that's been released is probably too aggressive and 32nm certainly won't be there in 2009 for AMD.

I'll add onto what enewmen stated:

2005 - 65nm Core based Arch (Pentium D for desktop)
2006 - 65nm Core 2 based Arch (Conroe for desktop, Woodcrest/Clovertown for Server)
2007 - 45nm Core 2 shrink with bonus goodies like High-k Metal Gates(Penryn)
2008 - 45nm Nehalem based Arch (note that at least server processors will eliminate the FSB)
2009 - 32nm Nehalem shrink (Westmere)
2010 - 32nm Sandy Bridge based Arch (Gesher was the old name, but had some political undertones)
2011 - 22nm shrink of Sandy Bridge (assumed)

Intel's schedule is aggressive for Intel and may not be doable. To think that AMD's schedule is on track (where they supposidly are closing the gap in 65nm -> 45nm -> 32nm -> etc. to Intel, since they're pretty far behind) is pretty much just unbelievable.
September 19, 2007 7:04:11 PM

There was a post at the other forum detailing this too.

Apparently, since IBM is the supplier for AMD's 65nm process, its a lot harder for AMD to optimize the process once sold to them. According to that same poster, there was some heated debate between IBM and AMD engineers, and a lot of finger pointing ensued.

On the Intel's side, since they develop the process as well as using them to manufacture, its a lot easier, and cheaper for them to optimize and correct any deficiency in yields.

While I do not know the validity of such statement, the logic is correct. If IBM sold the process to AMD, and refused to optimize, and/or correct any deficiency due to costs concern, AMD is pretty much alone. They don't have a team of engineers specializing in fab processes, and they simply don't have the financial mean to actually develop one of their own.
September 20, 2007 2:43:52 PM

I agree that the entire in house approach by Intel gives them the benefits you mention (as well as others). However, one thing that AMD's position buys them is they can often make the factories they partner with eat lost wafers/die through the manufacturing process. In other words, if an Intel fab screws up some product halfway through the line, Intel always pays the price. If Chartered screws up AMD's products halfway through the line, it's not always AMD that has to pay - in fact Chartered may need to eat a good portion of the cost just to encourage AMD not to shop around for a better partner fab.

And I do think AMD has a team of people trying to optimize the process (not necessarily the order of steps, but how to balance factory constraints, resources, etc - which is actually a big contributer to how long it takes to produce a chip). They advertise their automated system on their website pretty heavily - I would imagine they had at least a hand in that. I could be wrong, though.
September 20, 2007 5:27:55 PM

wolverinero79 said:
I agree that the entire in house approach by Intel gives them the benefits you mention (as well as others). However, one thing that AMD's position buys them is they can often make the factories they partner with eat lost wafers/die through the manufacturing process. In other words, if an Intel fab screws up some product halfway through the line, Intel always pays the price. If Chartered screws up AMD's products halfway through the line, it's not always AMD that has to pay - in fact Chartered may need to eat a good portion of the cost just to encourage AMD not to shop around for a better partner fab.


True, but I believe most dies being produced in Chartered are Brisbanes. The newer Barcelona dies are being produced in Dresden, which is where the problem originated. If you look at IBM's track record on their manufacturing technology, you'll notice that they offer excellent technology, but not so in manufacturing them. Besides AMD, Sony was also a victim of IBM's low yielding manufacturing process. When PS3 came out, the cell processors could only be yielding at 10~20%, and IBM considered that "lucky".

http://www.xbitlabs.com/news/cpu/display/20060713074825...

Besides the fact that Cell processors are extremely sophisticated, the manufacturing process must also played a significant role.

Since Brisbanes, as well as Barcelona, are built based on the same (not exactly the same) process. the rumors of AMD yielding Barcelona at 30% might be true.


And I do think AMD has a team of people trying to optimize the process (not necessarily the order of steps, but how to balance factory constraints, resources, etc - which is actually a big contributer to how long it takes to produce a chip). They advertise their automated system on their website pretty heavily - I would imagine they had at least a hand in that. I could be wrong, though. said:

And I do think AMD has a team of people trying to optimize the process (not necessarily the order of steps, but how to balance factory constraints, resources, etc - which is actually a big contributer to how long it takes to produce a chip). They advertise their automated system on their website pretty heavily - I would imagine they had at least a hand in that. I could be wrong, though.


Also true, but as one of the posters from other forum points out, once IBM developed the process in East Fishkill, AMD will have to transfer that to Dresden, which have completely different climate. This means AMD will have to optimize that on their own, without IBM's help. Since AMD haven't developed its own manufacturing process before, it would be a lot more time-consuming and difficult for them to optimize the process.

You might be right that AMD has a team working on optimization, but I wonder if that's enough to tweak the process enough to get a better yield.
!