Tuesday, August 19th 2008

IBM and AMD First to Reach the 22 nm Silicon Fabrication Mark

IBM and its chip development partners announced today that they've developed the first functional 22nm silicon fabricated SRAM cell. This puts them ahead of Intel, which had announced its technological entry into the 32 nm domain in September, 2007. SRAM is usually the first semiconductor device a chip-maker tests a new fabrication-process on, before working on microprocessors. These devices were developed and manufactured by AMD, Freescale, IBM STMicroelectronics, Toshiba and the College of Nanoscale Science and Engineering (CNSE). They were built in the conventional 6-transistor design and on a 300 mm wafer. This level of miniaturization made the SRAM cell shrink to a mere 0.1 sq. μm, compare this to the SRAM cells that go into making caches on the 45 nm Intel processors, 0.346 sq. μm.

22 nm chips are two generations into the future, and AMD is yet to catch up with Intel in commercially available 45 nm silicon fabricated CPUs. Intel on the other hand has demonstrated 32 nm fabrication last year and would probably show off its 32 nm CPU prototypes at the IDF this week.

Intel has responded to this development saying that while it indeed is an entry into the 22 nm domain by IBM and its partners, merely producing a SRAM cell doesn't validate that, it is merely an exercise in lithography scaling. The technology can be validated only when they come up with SRAM arrays of any given density like Intel did back in September when it announced its 32nm conquest. Demonstrating a single cell doesn't represent reality, where SRAM occurs as arrays and not single cells. The 32 nm array Intel demonstrated had 290 million cells.
Source: TG Daily
Add your own comment

18 Comments on IBM and AMD First to Reach the 22 nm Silicon Fabrication Mark

#1
candle_86
if AMD can build a 22nm SRAM chip why not get the phenom move over ASAP to 22nm and sign a deal with IBM
Posted on Reply
#2
btarunr
Editor & Senior Moderator
candle_86if AMD can build a 22nm SRAM chip why not get the phenom move over ASAP to 22nm and sign a deal with IBM
Because 22nm is still a zygote. According to Intel, they didn't even fully bake the cake before putting it up on the bakery's shelves.
Posted on Reply
#3
candle_86
but why not get it started ya know. If they can make working SRAM at 22nm id say in 3-4 months IBM will throw the Power PC CPU's on it.
Posted on Reply
#5
btarunr
Editor & Senior Moderator
[I.R.A]_FBiall talk and no evidence then?
Ok, a very casual explanation:

An SRAM array (of a said size (that goes into making the CPU cache)) is made of SRAM cells. Each cell holds a small bit of information.

Intel says: "you just shrunk a SRAM cell into 22 nm, it's just like using a crayon to write really fine (by sharpening it). You still have to prove you opened a can of 22nm by building and demonstrating a stable array using those cells like we did back in September. We shot our mouths only after making sure the thing actually worked at 32nm."
Posted on Reply
#7
WarEagleAU
Bird of Prey
Its still a remarkable feat, lets see if we hear any more news from them.
Posted on Reply
#8
Darkrealms
That responce from Intel kinda suprises me. I'd think they wouldn't want to piss off IBM. Isn't IBM already using a decent number of AMD server chips?
Posted on Reply
#9
tkpenalty
OOF. This is unexpected; I think a majority of people expected intel to come up with this first. Even if its not polished and feasible, IBM DOES work with AMD therefore, R&D should not take that long
Posted on Reply
#11
kwchang007
Sweett, although to be completely fair, I'm guessing between all those companies they have more money to put into all this
Posted on Reply
#12
brian.ca
tkpenaltyOOF. This is unexpected; I think a majority of people expected intel to come up with this first. Even if its not polished and feasible, IBM DOES work with AMD therefore, R&D should not take that long
Did't IBM & AMD announce another milestone along these lines a few months back? Also from what I gathered back then was that Intel was still dealing with logistics problems that were keeping them from getting the tools they needed to perform some R&D on sub 32nm processes. I'm not sure if they're still dealing with that but I wasn't too surprised by this.

What will remain to be seen is how this edge will translate down the road for AMD's operations (Will they catch up to Intel, pass them, keep the edge but still get beat? etc.)
Posted on Reply
#13
PCpraiser100
Intel's response:

"CALL MICROSOFT!! See if they will bug it!"
Posted on Reply
#14
a111087
Bulldozer on 22nm would be awesome :))
Posted on Reply
#15
[I.R.A]_FBi
bull dozer? we dont even know if it can shovel sh*t yet.
Posted on Reply
#16
X1REME
[I.R.A]_FBibull dozer? we dont even know if it can shovel sh*t yet.
best one i heard by far this week although am a amd fan, moron
Posted on Reply
#17
laszlo
is not easy to shrink the SOI technology has big power leakage;intel's tri-gate seems to support better the shrink but even they have problems... soon they reach the end but new approaches are already in work remember :shyamkol.blogspot.com/2007/08/500-ghz-processor-worlds-fastest.html

i think they'll stop shrinking once and concentrate on new design with new materials
Posted on Reply
#18
Steevo
Intel is just pissed as AMD got in like Sven with IBM and immersion lithography and is going to be taking the whole process two jumps in a short time.
Posted on Reply
Add your own comment
Dec 26th, 2024 11:44 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts