News Intel launches 45nm processor

[WilHarris]

http://www.bit-tech.net/news/2006/01..._45nm_process/

[Aphex_]

amd catch up? they are still in the lead last time i looked and from the s939 opterons they have proved they still have a lot of ability in 90nm nevermind 65nm. intel is trying to catch up with amd by shrinking as oppossed to actually being clever.

everyone talks about conroe this conroe that but at the end of the day do u think that amd are going to stand still gawping at it?? are they hell, in all likelihood they'll bring out something just before conroe or just after and nobody will be talking about conroe.

[DeX]

I was looking around yesterday but couldn't find an answer to this: What distance does the oft. cited 90nm, 65nm and 45nm refer to? What is the minimum theoretical distance?

[ch424]

Aphex_, shrinking processes and therefore die sizes is a major part of developing chips. As Intel moves away from the current netburst architecture, and towards the shorter-pipeline, lower power consumption Conroe and Merom CPUs, the only place they'll loose out to AMD is their archaic FSB technology.Once they get a decent interconnect like HyperTransport working, they'll be just as good as AMD for much lower power consumption. The transition to 65 and 45nm is just along the way. As AMD haven't started showing off 65nm yet, it would imply that they're going to be behind in the next generation of smaller, faster, and most importantly silent media centres and laptops, which is where the market is headed.

That's why people are excited about Conroe. And if it inspires even cooler stuff from AMD, that's even better. However, quite a few people have noticed that AMD don't seem to have a long term vision like Intel do: Intel has marketing campaigns and platforms like VIIV and Centrino, while AMD is just "We make awesome server and gaming CPUs. We just need to keep making them faster for the future."

Dex, the absolute minimum is about 17nm, but 34 is practically more probable.

ch424

[Hamish]

Originally Posted by Aphex_

amd catch up? they are still in the lead last time i looked and from the s939 opterons they have proved they still have a lot of ability in 90nm nevermind 65nm. intel is trying to catch up with amd by shrinking as oppossed to actually being clever.

everyone talks about conroe this conroe that but at the end of the day do u think that amd are going to stand still gawping at it?? are they hell, in all likelihood they'll bring out something just before conroe or just after and nobody will be talking about conroe.

intel are miles ahead in terms of manufacturing technology and capacity, this is a fact theres no disputing it

amd chips are still better tho

[eek]

Originally Posted by DeX

I was looking around yesterday but couldn't find an answer to this: What distance does the oft. cited 90nm, 65nm and 45nm refer to? What is the minimum theoretical distance?

I'm guessing that it relates to the lambda value associated with VLSI.

If it is then it is essentially a measurement of how close you can get things together and multiples of this value are used to determine widths and distances apart of things.

E.g. the minimum metal width is 3xlambda, theminimum poly contact to poly edge spacing is 1xlambda

The lower the value, the smaller things can be thus the smaller chips can be (or the more you can fit on a chip depending on how you look at it).

The problem with smaller sizes is increased leakage (the walls between the wires become thin enough to allow electrons to go between wires). This can be solved by using an increased htreshold voltage, substrate bias and high-k gate dielectrics.

I beleive ATI currently use this high-k gate dielectrics process and now it seems Intel is going there with their 45nm process.

Hope this helps.

[DXR_13KE]

amd is ahead because they have a shorter pipeline, have a direct comunication channel between the memory and cpu, no northbridge here, and has no botleneck exiting and entering the cpu via the bus.

if intel uses this tricks on their new cpu, well, i dont know if amd will be ahead for long.
intel will eventualy have the fastest one because they can not be that dumb.

[jakenbake]

I was looking around yesterday but couldn't find an answer to this: What distance does the oft. cited 90nm, 65nm and 45nm refer to? What is the minimum theoretical distance?

i don't know what your talking about with this lamda stuff, but 90nm, 65nm, and 45nm refers to how big the actual transistor is. it can't get much smaller because the atoms that make up the elements of the transistors (silicon, germanium, ect) are like 20-30 nm in size, for just one atom. the smallest particle is an atom, so you can't go any smaller than that without finding a new process. if you wanna know more google something like "smallest transistor" or something like that.

[eek]

Just did a search for it and it appears that the size actually corresponds to the 'average feature size'. Lambda refers to the smallest possible feature size.

[Firehed]

We're talking about two different "ahead"s. Intel has an insane lead as far as technological processes and sheer market domination, and fabs to back it up. AMD processors are ahead in performance.

Now overall it's only a good thing. It either means more cores in a package, or it could translate to faster clock speeds or more cache. Whatever. In theory they could ramp up the clocks again with the smaller process, as they tend to move beyond the whole limit of how far an electron can travel in one three-billionth of a second (about a millimeter) because they wouldn't have to travel as far. OTOH, adding more cores is going to be much more effective than increasing clock speeds, at least in terms of general usability, and performance in the ever-increasing number of threaded apps.

jakenbake - half true. While the smallest chunk of silicon we could lay down would be one atom of it, there are things far smaller. In fact you can go something like 26 more levels of magnitude smaller before you start reaching the feasable limit of measurable existance. And yes, subatomic particles could theoretically be used in processors.

[Neo40k]

AMD has no chance

Intel + MAC + Flash memory right on the MOBO = Unstoppable!!

[Da Dego]

Originally Posted by Neo40k

AMD has no chance

Intel + MAC + Flash memory right on the MOBO = Unstoppable!!

Neo40k + Intel thread = Fanboy!!

Now, can we get back to the discussion at hand?!

[Moriquendi]

Originally Posted by jakenbake

it can't get much smaller because the atoms that make up the elements of the transistors (silicon, germanium, ect) are like 20-30 nm in size, for just one atom. the smallest particle is an atom, so you can't go any smaller than that without finding a new process.

sorry, i dont think thats right, the atomic radius of silicon is 117^-12 m or 0.117^-9 m which is 0.117 nm and not 20-30nm, i doubt you could form a transistor with four or nine atoms

Did anyone else read about the magnetic logic gate someone made,could be shrunk to about 10nm (i'm not sure on this ill look for the article)

Moriquendi

Edit: just checked, it was in last weeks New Scientist, currently the magnetic gate is 330nm across but they think they could get them to 15nm across :Edit

[automagsrock]

Originally Posted by Moriquendi

sorry, i dont think thats right, the atomic radius of silicon is 117^-12 m or 0.117^-9 m which is 0.117 nm and not 20-30nm, i doubt you could form a transistor with four or nine atoms

Did anyone else read about the magnetic logic gate someone made,could be shrunk to about 10nm (i'm not sure on this ill look for the article)

Moriquendi

Edit: just checked, it was in last weeks New Scientist, currently the magnetic gate is 330nm across but they think they could get them to 15nm across :Edit

I have no idea what any of that means....


I am pulling for Intel. Once they get to 45nm and Conroe comes out I have a feeling AMD will be saying "WTF?" as Intel gains the crown again.

[hitman012]

Originally Posted by DeX

I was looking around yesterday but couldn't find an answer to this: What distance does the oft. cited 90nm, 65nm and 45nm refer to? What is the minimum theoretical distance?

The process designation refers to the the distance between the source and drain in the FETs (transistors) on a processor. Keep in mind that this distance is by no means the smallest thing in the processor - the actual gate oxide layer is tiny by comparison, with Intel's 65nm process having only 1.2nm of the stuff. That's less than 11 atoms thick

With regard to the next-gen architectures, I think that Intel are still seriously lagging behind without an on-die interconnect. Using the FSB was a serious mistake on their part, and was probably due to them rushing the dual-core offerings to beat AMD. Even Conroe still uses this system, although the speed has been increased to offset any serious bottlenecking.

I don't think we'll see Intel truly start to outpace AMD until they get a full core-speed on-die communication system with a more efficient memory controller. Their new interlinking system (CPU->NB) is apparently set to whip HyperTransport, although that remains to be seen

[Kaze22]

Originally Posted by automagsrock

I have no idea what any of that means....


I am pulling for Intel. Once they get to 45nm and Conroe comes out I have a feeling AMD will be saying "WTF?" as Intel gains the crown again.

Yes thats the benefit of having unlimited cash flow and resources, Intel is a hard nut to crack I doubt AMD will ever catch up. The best they can hope for is to play second fiddle, which isn't bad considering somebody's gotta fill that position

[Moriquendi]

just saying that the value given for the size of a silicon atom was off and that if a silicon atom was 20 or 30nm across a current 90nm transistor could only be three atoms across and a 65nm transistor could only be two atoms across.

[hitman012]

Originally Posted by Kaze22

Yes thats the benefit of having unlimited cash flow and resources, Intel is a hard nut to crack I doubt AMD will ever catch up. The best they can hope for is to play second fiddle, which isn't bad considering somebody's gotta fill that position

I have doubts about that - everyone who says these things about AMD seem to assume that AMD are going to sit still and do nothing. They won't - they are there to make money, and to do so involves being extremely competitive. They might have the edge at the moment, but they can't ride on that wave forever. AMD tend to be quieter about product annoucements, and I've no doubt that they have a longer term roadmap that some people think.

Intel will always (for the forseeable future) have the advantage of market presence and brand recognition - who hasn't heard of their processors? They force every advertiser who flogs a PC with a P4-gen chip to shout the name out on TV for the average Joe to hear. Their Viiv and Centrino monikers are certainly very valuable assets to their market expansion.

Even though Conroe looks to be an excellent performer, it still has a limitation in the form of an FSB bottleneck - a problem that AMD lacks. Despite this, I think that they'll be able to top AMD for a fair period before AM2 matures into a better platform. Both companies have some of the smartest people in the world working for them, and it is indeed going to be a very tough battle.

[Veles]

Originally Posted by Moriquendi

sorry, i dont think thats right, the atomic radius of silicon is 117^-12 m or 0.117^-9 m which is 0.117 nm and not 20-30nm, i doubt you could form a transistor with four or nine atoms

Heh was just about to say that

Originally Posted by hitman012

Intel will always (for the forseeable future) have the advantage of market presence and brand recognition - who hasn't heard of their processors? They force every advertiser who flogs a PC with a P4-gen chip to shout the name out on TV for the average Joe to hear. Their Viiv and Centrino monikers are certainly very valuable assets to their market expansion.

Always the way to make lots of money, doesn't matter if there is a better product out there. A good advertising campaign will grab you the attention of the ignorant masses. Since intel are they only ones who do some decent advertising, every average joe who walks into a PC world will be wanting a pentium. Because pentiums are so popular, the shops then make a big song and dance in thier adverts about intel processors, making other processors unknown to most people. Making p4s sell more, so more sales mean more sing ing and dancing, more sales, more dancing, it goes on and on.

Its the "I've never heard of that, it must be crap" syndrome. Worked for iPods and it's worked for intel.

[Cthippo]

Originally Posted by hitman012

Intel will always (for the forseeable future) have the advantage of market presence and brand recognition - who hasn't heard of their processors? They force every advertiser who flogs a PC with a P4-gen chip to shout the name out on TV for the average Joe to hear. Their Viiv and Centrino monikers are certainly very valuable assets to their market expansion.

This has been true essentially forever, and yet AMD continues to chip away at their market share. Granted, AMD got a boost when Intel bunged the Dual core release, but still, they were able to take advantage of it because they had a better product out already.

Call me a fanboy if you like, but Intel is going to have to come out with somthing at lease an order of magnitude better, as stable, before I'll buy another of their products.