Where do intel and AMD get the process size from?

Maximilian

Lifer
Feb 8, 2004
12,604
15
81
Actually where does any semiconductor company get their process size from? How do they get these nanometre numbers? Is it just a sort of boardroom discussion "well guys 45nm went swell, so is 32nm good for everyone?" Why not 30nm or 34nm etc?

Also on a semi related note, where does micron come from? I gather 0.1 microns is 100 nanometres but where does that unit of measurement come from, shouldnt it have been some offshoot of metres like a micrometre or something? Why is it called something totally different.
 

Rifter

Lifer
Oct 9, 1999
11,522
751
126
I've thought about this as well. looking forward to responses from people who know. Maybe its just a big wheel they spin like price is right lol.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Long ago the node label (32nm, 0.13 micron, etc) meant something physical as it related to the minimum physical dimension of a component that existed in the IC.

That naming convention broke down around the time of 0.35um (circa 1997).

Nowadays the node label is mostly arbitrary, like a marketing name.

This is not true for the memory markets, dram and flash memories still adhere to a rigorous definition of the relationship between a node's label and the minimum physical dimensions that exist for a specific structure within the IC.

But to answer your question - its more to do with money and risk appetite. Given XYZ dollars for R&D, what can be reasonably expected from the R&D team? Right now, and for the past few decades, the R&D budgets have scaled such that it was reasonable to expect a new node to reach production around every 2 years and that the new node would deliver roughly a 2x increase in transistor density.

This trend works because R&D spending is scaling such that it is ensured to work. There is nothing magical about the node cadence, it would be faster if companies spent even more, and slower if companies spent less.

Micron (µ)

1879: Micron, symbol µ, was a special name for the micrometer, but it was officially declared obsolete in 1968. [13th CGPM, Resolution 7]

Source http://lamar.colostate.edu/~hillger/unit-definitions.html#micron
 

LiuKangBakinPie

Diamond Member
Jan 31, 2011
3,903
0
0
Moores Law

Gordon_Moore1.jpg


graph1.jpg

http://www.intel.com/technology/mooreslaw/index.htm
 

Maximilian

Lifer
Feb 8, 2004
12,604
15
81
Ah right so its just what is a reasonably acceptable cost, they could shoot lower if they had the cash.

Weird how they kept using 0.18um, 0.13um process etc when microns have been obsolete for 30 years heh. Shouldve been 180nm and 130nm respectively, rather than just wait until it hits 0.09um and starts to get confusing. Ah well good to know :thumbsup:
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
Also on a semi related note, where does micron come from? I gather 0.1 microns is 100 nanometres but where does that unit of measurement come from, shouldnt it have been some offshoot of metres like a micrometre or something? Why is it called something totally different.

The word Micrometer its coming from the Greek words “Μικρό” (small) and “Μέτρο” (Meter)

The word Microns is coming from the Greek word “Μικρά” which is the plural of “Μικρό” and it used to be coming for the word Millimicron (mμ) that was used for the word Nanometer.

Nanometer is coming for the Greek words “Νάνος” meaning Dwarf (Very small people) and the word Meter.
 

Mopetar

Diamond Member
Jan 31, 2011
8,497
7,753
136
There's also a tendency to target what other fabs are going to use so that it's easier to get contracts for chips that have been designed with a certain process in mind. The targets are also released far in advance to allow the chip designers to plan out new architectures or tweak old ones for a die shrink.

There's no reason why we couldn't have companies using 41nm, 38nm, 35nm, 30nm processes right now, but it would make more of a mess for companies interested in using multiple fabs to get similar parts.
 

Soleron

Senior member
May 10, 2009
337
0
71
There's also a tendency to target what other fabs are going to use so that it's easier to get contracts for chips that have been designed with a certain process in mind. The targets are also released far in advance to allow the chip designers to plan out new architectures or tweak old ones for a die shrink.

There's no reason why we couldn't have companies using 41nm, 38nm, 35nm, 30nm processes right now, but it would make more of a mess for companies interested in using multiple fabs to get similar parts.

If Intel made a 31nm process, it would be easier to port to from 32nm than to port from Intel 32nm to Globalfoundries 32nm. Similarly named nodes are dissimilar in actual characteristics.

The reason the names are the same is to make it look like certain companies are at parity with each other, when actually they are not.
 

nismotigerwvu

Golden Member
May 13, 2004
1,568
33
91
I can personally attest to micron still being used extensively in science. Most filter media are denoted as x micron for the size of their pores.
 

Ben90

Platinum Member
Jun 14, 2009
2,866
3
0
wikipedia said:
A micrometre is one-millionth of a metre (1/1000 of a millimetre, or 0.001mm). Its unit symbol in the International System of Units (SI) is μm.

The term micron and the symbol μ, representing the micrometre were officially accepted between 1879 and 1967, but officially revoked in 1967.[1] In practice it is still used, especially in American English, astronomy, and the semiconductor industry. The micrometre is a common unit of measurement for wavelengths of infrared radiation. The use of the term micron also helps differentiate the unit from the measuring device of the same name.
''
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
If you thought the micron was quirky in its relation to the SI then you should know about the unit Angstrom (Å).

The angstrom is useful in chemistry and physics as the conventional definitions of bond lengths and atomic diameters are usually referenced in angstrom's.

An angstrom is 1/10 of a nanometer (the 1 nm = 10 Å) of 100x more than a picometer (1 Å = 100 pm).

There are far more entertaining units to be aware of...such as the milliHelen which is the quantity of beauty required to launch 1 ship as one Helen (of Troy) was reputed to be so beautiful that her face launched 1000 ships.

(so 1 Helen = 1000 ships, thus 1 ship = 0.001 Helen or 1 milliHelen)

Which leads one to the ageless question of what, exactly, a milliButtaface would launch...