Jump to content

Talk:Dual-core

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

As far as I can tell, this article is completely accurate and doesn't leave out any important considerations. On the other hand, in six months, instead of being a statement in fact, it will be history. I agree that the wording tends to sound like a proponent of a technology, but strictly speaking, it's all true. Today, that is. If the wording is changed to mark the state of things at a specific time, then the tone of proponency will also be improved. Dennis 18:48, 18 Dec 2004 (UTC)

The claims of thermal advantages were questionable so I removed them. They should have some support if you put them back. Dennis 21:30, 18 Dec 2004 (UTC)


The thermal claims lie in what you are comparing to. The motivation for dual-core, at least in the case of Intel, is to offset the failure of continued frequency-scaling. If a chip can't be shrunk and frequency-scaled, whether for circuit reasons or thermal reasons, then dual-core is the response. These chips are not displacing dual- or quad-processor SMP designs, but rather displacing single-core processors in workstations and (now) high-end desktops. So the advantage is that a dual-core processor will run cooler (because of heat density, rather than overall output) than an equivalent technology processor running at (e.g.) 170% of the clock speed. A dual core CPU will dissipate about 2x the single-core power, whereas a single-core CPU of similar technology running at 170% of the clock speed will theoretically (if it's possible at all) would dissipate much more than 2x the heat. I will leave this comment here for a few weeks, and in the absence of other commentary, will produce a version of it for inclusion in the article. gnetwerker -- Dec 22, 2004

The thermal claims are always spun in terms of comparable separate chip equivalent designs. At 90nm, leakage currents become a major source of power dissipation as opposed to the frequency-dependent dissipation that is typical of CMOS. This somewhat offsets the gains by reducing core voltages and gate capacitance. Additionally, the higher density chips mean that the heat is generated in a smaller area, making the mechanical solution to dissipate the energy much more difficult. This is why you see bare dies directly on heatsinks in high-end chips. Finally, the smaller geometries make the IC more sensitive to high-end temperatures. Once, I found 125C to be a common max junction temp, but in the newest designs (and in fact the one I'm working on now using a Freescale 7447A) the max junction temp is much lower (in this case, 105C). This lowered headroom makes the thermal solution much more difficult, especially in embedded designs (as opposed to a desktop PC).

I agree that thermal claims can be made correct if in context. Dennis 00:24, 23 Dec 2004 (UTC)

Dual core

[edit]

Hai guys

I am a technology enthusiast and an AMD fan

I am using AMD processors in my PC since AMD K6 and I frequently upgraded my PC to later releases from AMD like Athlon,Athlon XP,Sempron,Athlon 64,Athlon 64 FX 55 and now the big brother Athlon 64 X2.

I am extremely satisfied with the latest piece of silicon from my ever favourite AMD

It worths its price .

The first IBM Clone CPUs of mine were Cyrix then AMD (before that I always used Motorola). My latest machine is my first Intel, being a Dual-Xeon 604 HTT 2.66 GHz with 533 MHz FSB. Comparing them to the Athlon XP 1800+ that I've had before, I'm extremely disappointed by the Xeon's performance. I will definitely go back to AMD in my next machine.


This is a lovely anecdote but you'll need to quote some independent, quantitative data for it to have a meaningful impact on the article. Not arguing with you (I don't have the data), but I think the article is more about the concept of dual-core than the performance of any particular implementation. -- Gnetwerker 00:39, 8 October 2005 (UTC)[reply]

Correct Spelling: "Dual-core" or "Dual core" or "Dualcore"? Please unify with Multicore and similar.

[edit]

This article is named "Dual-core" while it's sister article is titled "Multicore" (Multicore). This is very inconsistent. So what is the correct spelling, is it "Dual-core", "Dual core" or "Dualcore"? And please unify this with Multicore and similar articles.

I prefer "Dual core". I think the inconsistent spellings are acceptable because multi is a prefix, and dual is not. SGDentarthurdent 20:26, 11 November 2005 (UTC)[reply]
Just my two cents: A quick "google test" reveals that Apple, AMD, and Intel all use "dual-core" in their web docs, and that the hyphenated form is much more common than the non-hyphen variant (the latter observation done by eyeball search, since no search engine I know of is able to discriminate between results based on punctuation). --Wernher 17:18, 13 November 2005 (UTC)[reply]

Difference between chip and die

[edit]

I hope someone would explain the difference between a die and a chip and a silicon. Some people claim multicore is many cores on a die, a chip, a silicon, or a package. Unless there is clear distinction among these, there will confusion as to what multicore really is.

A die is a single piece of silicon. Chip officially means die, but sometimes people use it to mean package. A package is something that contains one or more die. Most packages have one die, so it usually OK to call a package a chip. If a package has more than one die it should be called a "multi-chip-module".

In theory, multicore refers to multiple cores in a single chip. In real life, there's always put only one single die in a chip, so practically multicore is usually multiple cores on a single die as well. The single term silicon sounds cool but is pretty ambiguous and could refer to silicon die as well as to wafer, so it shouldn't be used in that context anyway. Also, not all chips are made of silicon, some are made of other elements, with Germanium, Gallium and Arsenid being amongst the most popular alternatives to silicon. So the term silicon should be scratched completely from articles like this.
Ehh... No. I can think of several examples off the top of my head of packages ("chips") with multiple dies in them, even multiple CPU dies. IBM's MCM packaging for POWER4 and POWER5 are easily the most recognizable... -- uberpenguin 21:40, 11 December 2005 (UTC)[reply]

Just FYI, the original packages at Intel with more than one die in them were called "dual-well". Don't remember exactly but this was probably in the early 1990s. Also, I concur: a single-unit piece of silicon is a die, a packaged integrated circuit was always referred to at Intel as a chip. There should be no disagreement on "die" -- everyone in the business knows what that is. Chip may be more ill-defined. No one at Intel every referred to anything with the noun silicon. Except when referring to the element on the periodic table, silicon was an adjective modifying "wafer" or something else. Intel does not equal universal truth, however, just so no one feels the need to point that out. -- Gnetwerker 07:43, 22 December 2005 (UTC)[reply]

Merge with Multicore

[edit]

... and redirect. There is nothing in a discussion of dual-core ICs that isn't encompassed by the term 'multicore.' There's some unneccessary duality in these articles going on. -- uberpenguin 21:43, 11 December 2005 (UTC)[reply]