Jump to content

Talk:Multi-core (computing)

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

Compatibility/Support Lists[edit]

I've been looking for a while, and I don't see a list of any kind anywhere showing what OSs or apps or what handle multi-core and multi-socket configurations. For instance, I was reading on a message board a few days ago where noone knew whether Win98 could handle an dual-core Athlon64 or not, or how many cores a Win2K machine could handle, or whether XP would know how divide cpu time among apps in a reasonably efficient manner. Stuff like this is almost impossible to find straight dope on without first trolling through an incredible amount of useless data. Sweetfreek 01:41, 15 November 2007 (UTC)[reply]

Embedded[edit]

There is major separation between generalist processors (Intel, AMD - and perhaps- Cell) and Embeddded processors (especially DSP & NPU).

I added a sectrion on this and a little discussion. It probably needs more, particularly about the differences in software environment & the implications.

--Rupert baines 14:58, 20 May 2007 (UTC)[reply]

QUESTION[edit]

I wanna know that how these multicore(tiled or shared) processor communicate with each other and whether core simply refers toa CPU ?


—The preceding unsigned comment was added by 125.23.52.99 (talk) 00:02, 20 February 2007 (UTC).[reply]

If you have questions that aren't directly related to improving an article, please consult the reference desk. The simple answer is: electrical connections, and generally yes. -- mattb @ 2007-02-20T01:30Z

I asked the above question because someone told me that core genrally refers to the L1 cahe and CPU is one and acts virtually as two if it were two?

-He was probably talking about Hyper-Threading. Either you didn't get what he said (core refers to L1 cache?) or what he told you was non-sense. —Preceding unsigned comment added by 84.161.185.217 (talkcontribs) 06:55, February 23, 2007

Suggestion[edit]

I think it would be beneficial in the short term (until it is standard) to create a list of software or applications that are actually designed for multi-core computing.

And another suggestion[edit]

My Suggestion is Each CPU has a front side and a backside. Intel or AMD should place a CPU on each side. This would make the todays dye into 4 CPU. (Most of todays CPU only place 2 CPU on one side) —Preceding unsigned comment added by Troy Frei (talkcontribs) 14:58, June 27, 2006

First this is a discussion page for the Multi-core article and no place for "suggestions what should be done". If you want to make suggestions then go to a Intel / AMD Forum and post them there. Plus 3D die stacking is already in the works so your suggestion is not only in the wrong place but also obsolete. Perhaps a link to 3D die Stacking wouldn't be so bad (if such an article exists)? 84.161.185.217 13:50, 23 February 2007 (UTC)[reply]

Suggestion[edit]

I recently spent a lot of time writing a paper on multicore and reconfigurable computing -- including a diagram of the various options that I think is unique. You can find it on my site at http://www.diycalculator.com/sp-compuniverse.shtml. I think it provides a very useful background piece that's worth an external link. But I'll leave that up to whoever is in charge. Cheers -- Max (max@diycalculator.com).

I have a question, many people on the ineternet says that a dual core processor has the capability of running twice as fast as one core, is that true? Wouldn't there be limitations of amount of data allowed to be transfered or something? This was not in the article so if someone finds an answer could you please add it? 65.92.53.225 21:34, 17 March 2006 (UTC)[reply]

The basic answer is "definitely not". Anyway, this isn't the appropriate place to ask questions like this. You ought to read up on topics like parallel computing, instruction level parallelism, and thread-level parallelism. -- uberpenguin 02:22, 18 March 2006 (UTC)[reply]

TODO[edit]

I would like some feedback on the following suggestions:

• A lot can be said in the "software impact" section that isn't specific to multicores, but is rather generally true for a multiprocessor. The software impact of multicores specifically is less significant/interesting than the impact of multiprocessors in general.

The significant thing, I think, about multicore is getting multiprocessor technology into normal users machines for client apps --PdDemeter 15:33, 27 September 2006 (UTC)[reply]

• A new section that relates(compares/contrasts) multicores with general multiprocessors is needed.

• I think the article should go further to make it absolutely clear that a multicore chip is just a more tightly integrated multiprocessor.

• Wherever possible, refer to the multiprocessor article in order to avoid confusion. For example, TLP is mentioned in the introduction - multicores in particular don't do much for TLP that multiprocessors in general don't.

• Introduce the notion of marketing as related to "hyping" the word multicore and the notion that the idea is not new.

Merger[edit]

I went ahead and merged the dual core and multi core articles into this one, since there is absolutely nothing architecture-wise about the former that cannot be included in a more general article about the latter. I reworked the intro a bit, but the article still needs serious improvement and rewriting. -- uberpenguin 22:44, 22 December 2005 (UTC)[reply]

good work! --68.92.212.251 13:43, 28 December 2005 (UTC)[reply]

There is also a mergefrom tag on Quad core, proposing a merge to this page. Perhaps someone with more knowledge of the subject than me can handle it? Kcordina 15:46, 14 March 2006 (UTC)[reply]

I just redirected quad core to here. There was pretty much zero useful information in that article. -- uberpenguin 17:24, 14 March 2006 (UTC)[reply]

Spelling[edit]

Is it "Multi-core" or "Multicore"? Afair the old article was "Multicore". Because I'm neither a native speaker nor full of knowledge about English grammar and spelling rules, somebody who meets both requirements might write some explanation to enlighten me: which variant is correct, "multi-core", "multicore" or "multi core"? From what I know, multi is a regularly used prefix, so it should be "multicore" - though I'm not, for reasons already described, sure about this. --Christianhujer 23:33, 18 March 2006 (UTC)[reply]

Multi-core is a technical slang for Multiple core, so you can use any word related with it. But multicore can be take like a single world so it can create some confusion, so it's better to use multi - core o multi core.

I agree that "multicore" would be a single word, but I see nothing at all wrong with using "multicore", since "multi-" is an oft-used prefix (in its own right, not as an abbreviation of the word "multiple"). However, given that the choice settled on for this article is "multi-core" (with a hyphen), it is odd that "manycore" (without a hyphen) is used, since "many" seems to be seldom employed in that manner. Jim 04:22, 11 September 2007 (UTC)[reply]

Spelling[edit]

It should be "dye", not "die".
http://dictionary.reference.com/search?q=die
http://dictionary.reference.com/search?q=dye

Quoting your first cited reference: 2. <electronics> Plural: dies. An unpackaged integrated circuit.' "Die" is correct. -- uberpenguin @ 2006-05-30 02:09Z

Copyright infringement[edit]

This edit is plagiarized directly from this copyrighted website.

Edit: oops, forgot to sign. J. Finkelstein 15:52, 28 July 2006 (UTC)[reply]

question[edit]

does a dual-core processor running at 1.5 ghz = a single-core processor running at 3 ghz?

No. If you read the article, it says that the theoretical maximum performance increase is 70%. This is uncited, but I assume it's fairly close to the mark. As with Symmetric multiprocessing (multiple separate processors), to get the most out of multicore you have to have software that executes instructions in parallel. If you're running multiple programs/tasks at once (multitasking), SMP-aware operating systems can make use of the additional processors (or cores) automatically, but if only one program is actually doing anything, and it isn't capable of executing instructions in parallel, there is very little (if any) performance increase. --Smithfarm 09:08, 20 August 2006 (UTC)[reply]

Software impact section[edit]

Are any of the issues discussed in this section different than for SMP? If there are different issues, wouldn't it be good to discuss how they differ from SMP? If there aren't, then wouldn't it be better to just link to the SMP article to avoid duplication? --Smithfarm 09:53, 20 August 2006 (UTC)[reply]

I also don't understand the difference between this and SMP. It should be explained somewhere. Sloverlord 21:46, 12 September 2006 (UTC)[reply]

from the SMP article: "SMP is one of many styles of multiprocessor machine achitecture"... i.e., SMP is a subset of multiprocessor architectures... if anything, the SMP article would link to this one, but it seems to me there are sufficient differences and concerns in each to warrant separate articles - SMP is just one (simplistic) way of using multiple processors, and unfortunately has very limited application in the real world - not everyone needs to use their computer to predict global weather or simulate nuclear explosions 69.235.255.45 17:45, 22 July 2007 (UTC)[reply]

Disadvantages???[edit]

The "disadvantages" listed aren't much disadvantages at all. "Raw processing power is not the only constraint on system performance"... well yeah, DUH. But this isn't a disadvantage, it's just a limitation on real-world impact and expectations. It's like saying that "money can't buy happiness" is a disadvantage of being rich! The others are questionable too. Either this section should be renamed, or some of these points should be moved to a different section.Thelastemperor 04:45, 19 October 2006 (UTC)[reply]

You're reading too much content and not enough context. that argument is most certainly a disadvantage, especially if you're looking into the history of Video Game Consoles, more specifically the Sega Saturn, which, among other things, featured two processors, and shared RAM between the two processors. the particular argument you mentioned can be a rather large disadvantage, if the system kernel is not programmed to properly make use of both cores --Pandora Xero 19:53, 30 October 2006 (UTC)[reply]
I have to agree though. These "disadvantages" look bogus... Games not being multithreaded (i.e. optimized for the architecture they're running on) isn't the fault of the architecture. Why should production yields matter to the one using the multicore? More difficult to manage thermally? It doesn't provide a link to back up that claim. Risk of obsolescence? Someone please explain that. I say rename the Disadvantages section or at least tack some "quotation needed" on the arguments. --89.61.76.181 02:26, 23 March 2007 (UTC)[reply]

multi cores?[edit]

I dont understand the '9core' stuff regarding the xbox360/ps3, when pcs are still on dual core? Do those consoles actually have '9 paths' to run things? And then wouldn't there be massive overheating problems and much more power needed to run the unit too then? —Preceding unsigned comment added by 144.131.144.102 (talkcontribs)

Not "nine paths" so much as the ability to run nine separate programs simulaneously (where are you getting the number nine from? Xbox 360 is tri-core, PlayStation 3 is a single PPC + 7 usable mini SIMD processors). The power/heat issue is a big 'maybe' and depends on the particular design. A huge superscalar microprocessor can easily use as much power as ten core microprocessor. -- mattb @ 2006-11-08T14:55Z

Multi-core problem[edit]

Why does my dual-core CPU computer (1496MHz) shows on the System Information only 598MHz?

May I suggest consulting Google for the answer to your problem? Wikipedia talk pages are generally intended for discussing items relevant to their associated articles. -- mattb @ 2006-11-10T00:44Z
Just to answer your question: Your 'problem' has nothing to do with the presence of two cores on your processor. If your computer is a laptop, then I would expect that it is because of Intel Speedstep/AMD Cool 'n Quiet technology, which reduces (throttles) the processor clock when it is idle. If you check System Information again when you are doing something processor intensive, it will show the 'correct' value. I'm not entirely sure about this, but Intel's Core 2 Duo processors for desktops also reduce their clock multiplier when they are idle - for example, the C2D E6600 reduces it from 9 to 6. -- Jiahui1992 08:37, 21 January 2007 (UTC)[reply]

Tiled Multi Core and Shared Cache Multi Core[edit]

This article fails to mention that actually there are two types of Multi Core architecture being used in Industry as of now. Tiled Multicore processor are simply when a simple uniprocessor core is replicated on the die. The two cores don't share cache. This is the architecture of the following processors Intel Pentium D, Dual Core Opetron, Intel Montecito, SUN UltraSparc IV etc.

What this article mention is the sahred cache multicore processors which is used by the follwoing processors : Intel woodcrest, Intel conroe , SUN Niagra, IBM power 4, IBM power 5 etc.


—The preceding unsigned comment was added by 125.23.52.99 (talk) 00:00, 20 February 2007 (UTC).[reply]

Missing Multi-core example[edit]

In looking for technical information for a specific chip (the Vega 1 & Vega 2), I found no mention of them in this article. Apparently they boast 24 and 48 core systems (I found out about them through a ZDNet posting) however at first glance I do not see any information about them (these systems are not cheap either). If anyone has any information about these (and time to write/read about it), it would be fantastic! --Stux 23:05, 4 December 2006 (UTC)[reply]

Okay, let me be more specific: The reason I brought these two chips up is that the bulk of this article deals with 2-core and 4-core systems, and makes absolutely no mention of 4+ core systems. For a while I thought this might be possible, but had never seen any examples of said technology. After running into the ZNET posting I wanted to know more about these chips on wikipedia, but unfortunately there is no information about them here or in their site about them. I am really interested to know more details about these chips! (Besides how expensive they are!) --Stux 17:48, 3 January 2007 (UTC)[reply]
I added a mention of Chuck Moore's prolific multi-core work. The x18 series had a 5x5 grid of cores and his latest seaForth-24 has 24 cores. --IanOsgood 00:31, 31 January 2007 (UTC)[reply]
I agree. Recently I've seen too much breathless hype over 2-core and 4-core systems. I assume it's because these reporters don't know that 5-core chips have existed for over a decade, and 8-core, 24-core, and 256-core chips now exist. (I'm referring to the 5-core TMS320C80 MVP, which was introduced sometime before 1996. Why doesn't it have an article already? Also the Kilocore). --75.19.73.101 19:22, 9 November 2007 (UTC)[reply]

Missing a mention of current art in this area[edit]

I read this section with great interest. I fully admit I represent a commercial entity, but the fact is there is a Java framework that was built to address exactly this mulicore arms race we're in. So I added it to the list of applications supporting parallelism.

Actually, it's not an app that supports it, it's a framework that enables any data-intensive Java app to enjoy parallelism when you code with the framework. So maybe I put it in the wrong list?

Emilio 23:33, 14 December 2006 (UTC)EmilioB[reply]

Link to SWARM[edit]

I removed a link to a multi-core library/toolkit (diff) because it seems to be self-promotion and not particularly useful to the audience of this article. Inasmuch as I can tell, the software isn't a particularly well known or popular, so its notability is also in question. However, the author of the software package has added the link back and emailed me asking for me not to remove it again. I believe this may be a conflict of interest situation, and I don't believe the link passes WP:EL because the software in question isn't any more notable than the myriad of other parallel programming toolkits that we could link. Please discuss. -- mattb @ 2006-12-20T00:47Z

I agree with you, but it seems he's done it again. I imagine it's someone posing as the true author, but it certainly appears to be a conflict of interest issue. I'm going to remove it again and ask that Dbader discuss the issue here before trying to add the link in again. Upon reviewing his Dbader's edits he first added the section and an external link on 25 November 2006, they were removed and he added them again 19 December 2006, and just recently added them back in yesterday (27 February 2007). Littleman TAMU 23:09, 28 February 2007 (UTC)[reply]

Intel status[edit]

The article says, "Quad-core processors are announced by Intel[2] to be produced in 2006...." Can this be updated to reflect what, if anything, Intel is actually doing now? JamesMLane t c 17:40, 22 December 2006 (UTC)[reply]

Commercial Examples[edit]

  • In this section it states that Microsoft's Xbox 360 has a triple core processor. Surely this should be referred to as a 'triple chip module' ,since it is actually 3 chips each with a form of Hyper Threading present not just one chip with three cores, as it references these designs as this in the actual article.

Also Sony's Playstation 3 CELL processor design as a 9 core design when in actually fact it is effectively one core (PPE) with 8 (7 enabled) in a sense 'Mini-cores' (SPEs) Gamer2325 21:28, 19 February 2007 (UTC)[reply]

    • I have decided to edit the bit on the console's if anyone objects just change it back. Gamer2325 21:42, 19 February 2007 (UTC)[reply]
Your information on the Xbox 360 microprocessor (aka Xenon, Waternoose) is not accurate. It is, in fact, a triple core IBM PPC design, each core being SMT-enabled. Hyperthreading is an Intel trademark for their particular implementation of SMT, and is therefore not the appropriate term to use in relation to this IBM design. I've edited the article accordingly. -- mattb @ 2007-02-19T23:05Z

First step in the multi core revolution edited[edit]

i modified this section to include AMD, after all this article is about Multi-core processors, not Intel multi-core processors, i also removed the hyperthreading as intel no longer uses it or no longer plans to use it in future cpus, as it is less engery effecient than dual cores, or adding more cores, here is a quote from the wikipedia hyper-threading article "recently Hyper-Threading has been branded as energy inefficient. For example, specialist low power CPU design company ARM has stated SMT can use up to 46% more power than dual CPU designs. Furthermore, they claim SMT increases cache thrashing by 42%, whereas dual core results in a 37% decrease[1]. These considerations are claimed to be the reason Intel has dropped SMT from new cores."--Superchad 00:21, 21 February 2007 (UTC)superchad[reply]

dual core[edit]

would a dual core processor that runs at 1.86Ghz be able to run a game that requires a sinle core processor that runs at 2.6Ghz —The preceding unsigned comment was added by 60.234.107.90 (talk) 06:14, 7 March 2007 (UTC).[reply]

Without further information, probably not. However, this is not the place to receive help with technical questions; I'd suggest asking at WP:RD/C. -- mattb @ 2007-03-07T06:28Z
Yes. If you are talking about a Pentium 4 2.6GHz to a E6300 1.83GHz, then the answer is definetely less bacause that number you are quoting does not reflect actual performance. If you are not talking about those chips, the answer is still most likely less. I personally plan to upgrade from a 3.0GHz to a 2.1 GHz.86.150.14.125 08:25, 10 April 2007 (UTC)[reply]
Chances are high it will run. The requirement for a single 2.6 GHz CPU is very likely not taking current processors into account that deliver much higher performance per GHz then older types. Like the previous contributor i changed from a 3.0 GHz system to one with a lower clockrate (2.66 GHz in my case) and had a very significant boost in performance (dual cores aside).--B marl 17:31, 16 April 2007 (UTC)[reply]

"merge" of the old "chip-level multiprocesing" article[edit]

I am adding information from the old "chip-level multiprocessing: article. The subject matter of that article is isomorphic with our article, toi the extent that I feel that no discussin is warrented. I have redirected that article to here. If there is in fact a difference of opinion, I humbly apologize and we can revert to the usual merge process. -Arch dude 01:52, 9 May 2007 (UTC)[reply]

Multi-core Vs Hyper-threading[edit]

Can someone add a section describing the difference between the functions of a multi core processor and hyper-threading? How are they different? Can a multi core processor have hyper threading?

What is a core?[edit]

This article begins with "A multi-core CPU (or chip-level multiprocessor, CMP) combines two or more independent cores", but doesn't mention what is a core. The page core doesn't help either. We cannot assume that the reader is a computer science student. 203.158.89.10 (talk) 07:38, 27 November 2007 (UTC)[reply]

Wikiproject Computing[edit]

I have removed the WikiProject tag, as this article is either a redirect or deleted. If you oppose, please restore the tag. Thank you, fahadsadah (talk,contribs) 15:54, 30 March 2009 (UTC)[reply]