Talk:Central processing unit

From Wikipedia, the free encyclopedia

Former featured article Central processing unit is a former featured article. Please see the links under Article milestones below for its original nomination page (for older articles, check the nomination archive) and why it was removed.
Main Page trophy This article appeared on Wikipedia's Main Page as Today's featured article on March 4, 2006.

This is the talk page for discussing improvements to the Central processing unit article.
Article policies
Archives: 1, 2
Peer review This Engtech article has been selected for Version 0.5 and subsequent release versions of Wikipedia. It has been rated B-Class on the assessment scale (comments).
WikiProject Spoken Wikipedia The spoken word version of this article is part of WikiProject Spoken Wikipedia, an attempt to produce recordings of Wikipedia articles. To participate, visit the project page.

Archive 1 Archive 2

Processor was merged into Central processing unit (this article). That history now exists at Talk:Central processing unit/Processor article history.--Commander Keane 07:36, 14 January 2006 (UTC)

renn


Contents

[edit] List of CPU flags?

Is there a list somewhere (in Wikipedia or not) of common CPU flags (like SSE, APIC, MMX, 3dnow...) and what they mean? (Other than the one I started building at User:Dcljr/Sandbox#List of CPU flags, of course.) - dcljr (talk) 17:31, 23 May 2006 (UTC)

Uhh... For what architecture? Try reading the programmer's manual for whatever ISA you're using (looks like late x86 to me). -- uberpenguin @ 2006-05-23 18:36Z
That list is a mixed bag of terms, some of which refer to features (such as MMX, SSE, 3DNow!), some of which refer to components (such as APIC and MTRR), some of which refer to I/O buses (such as MCA and VME), some of which refer to instructions (such as SYSCALL) - there's really nothing they all have in common other than being computer hardware terms (and some of them might not even be computer hardware terms - "DE" goes to a disambiguation page, and the only computer term there is "Desktop Environment"). Guy Harris 21:26, 23 May 2006 (UTC)
Yeah, it's a mixed bag exactly because I don't know what they all mean! <g> Anyway, most of these are from the "flags" entry of "cpuinfo" in Linux (see, for example, [1]). - dcljr (talk) 19:31, 24 May 2006 (UTC)
The ones from "cpuinfo" are probably the flags you get from the x86 CPUID instruction, listing the capabilities of the processor. Lists of them can be found in the Intel and AMD documentation of the x86/IA-32 (including EM64T) and AMD64 instruction sets; see, for example, the description of the CPUID instruction in IA-32 Intel® Architecture Software Developer’s Manual Volume 2A: Instruction Set Reference, A-M. "MCA" and "VME", unfortunately, are ambiguous; they can refer to the MCA and VME buses, which are system I/O buses and are characteristics of the system as a whole rather than of the processor, or they could refer to "Machine Check Architecture" and "Virtual 8086 Mode Enhancements", which are x86 CPU features and have nothing to do with the MCA or VME buses - the x86 CPU features are the ones reported by "cpuinfo". Guy Harris 19:35, 24 May 2006 (UTC)
Aha! I just found List of computing and IT abbreviations, which would seem to be the place for this info, but the few terms I've looked for so far haven't been in the list. - dcljr (talk) 06:18, 25 May 2006 (UTC)
Actually, I think CPUID would be an even better place for this information (please add any details it leaves out). --70.189.77.59 17:54, 27 October 2006 (UTC)

[edit] Good article...

Sorry for cluttering the talk page, I just want to say this is a really well-written and informitive article. 128.208.41.109 05:25, 25 May 2006 (UTC)

[edit] Peformance comparison of common processors

It would be useful to read a comparison of the performances of common processors (either in a table in the article, or an external link).

I don't care for the idea. Performance comparisons are very involved and detailed discussions, are often quite subjective and subject to testing bias, and would bulk up the article significantly. Furthermore, a fair sample of performance comparisons from the entire history of CPUs could be very difficult to contain in a table. That being said, I think it might be an interesting and relevant factoid to include a short sample comparison of, say, the integer performance of a very early von Neumann computer to a modern microprocessor. If you want to collect some figures on modern microprocessors, I can look up the data for old computers. -- mattb @ 2006-09-14T04:41Z
I think there is room for such a table *somewhere* in Wikipedia, but I agree central processing unit is not the place. Please stick that information in benchmark (computing) until we find a better place? --70.189.77.59 17:54, 27 October 2006 (UTC)
Check out iCOMP and PR rating. BebopBob 01:56, 21 December 2006 (UTC)

[edit] manufacturers of central processing units?

Who are the main manufacturers of CPUs and how is performance measured

1. Too many to briefly enumerate. 2. In too many ways to briefly enumerate. Reading CPU design may provide a little enlightenment. -- mattb @ 2006-10-22T02:16Z
I like the CPU design article. It mentions As of 2004, only four companies are actively designing and fabricating state of the art general purpose computing CPU chips. So -- are 4 too many to enumerate? Or do we need to update that article to include those other companies you are thinking about?
Those are excellent questions. Please help us improve Wikipedia to make the answers better and to make the answers easier to find. Many companies that design CPUs are Fabless semiconductor company, who pass their design over to some other Semiconductor companies, which actally fabricate the CPU. CPU performance is measured using benchmarks -- but be aware that there is no one "performance" number -- one CPU may perform one benchmark faster, but some other CPU may perform your actual application faster. Do we need to mention/link to these things in the article? --70.189.77.59 17:54, 27 October 2006 (UTC)

I think it would be nifty if this article mentioned the "top" 8 or so CPU design companies and fabrication companies, in terms of the number of CPUs shipped.

I see that

  • ARM Limited says "In 2007, 2.9 billion chips based on an ARM design were manufactured."
  • PIC microcontroller says "Microchip recently announced the shipment of its five billionth PIC processor."
  • 68k says "CPU32 and ColdFire microcontrollers have been manufactured in the millions ..."
  • Data General Nova says "eventually 50,000 units were sold."
  • MIPS_architecture says "... By ... 1997 the 48-millionth MIPS-based CPU shipped, making it the first RISC CPU to outship the famous 68k family." —Preceding unsigned comment added by 68.0.124.33 (talk) 16:27, 15 March 2008 (UTC)
  • Advanced Micro Devices ... how many?
  • Intel Corporation ... how many?
  • Microprocessor says "About 55% of all CPUs sold in the world are 8-bit microcontrollers. Over 2 billion 8-bit microcontrollers were sold in 1997."

Would such a table (with a few more numbers filled in) be appropriate for this CPU article? --68.0.124.33 (talk) 02:29, 14 February 2008 (UTC)

[edit] Different CPU models?

What is physically different between similar processors, eg. one that is 2 ghz, and one that is 3 ghz?

There may be absolutely nothing physically different about them, or they may be totally different. You'd need more information than just clock speed to say. -- mattb @ 2006-10-19T19:03Z
Say they were the exact same model, and only the clock speed is different? Would only the multipilier be changed in the hardware?
Well, in some form or another the global clock signal will have its period decreased. Whether that means changing a clock multiplier or the target frequency of some oscillator is application-specific. -- mattb @ 2006-10-22T02:13Z
Certainly one *can* run a 3 GHz processor at 2 GHz, so there is not necessarily a physical difference.
However, most 2 GHz processors cannot run at 3 GHz -- there *is* a physical difference.
So, from least amount of difference to most difference, we have:
  • only the external clock speed is different -- no internal difference.
  • only the internal multipler is different.
  • They were manufactured the same way, with the same photomask, but chip-to-chip and wafer-to-wafer variations in blurriness and defects made some chips fail to run at the designed speed ("weak transistors"), but the chips passed the test at a slower speed.
  • They were manufactured according to the same layout, but a photomask shrink produced smaller, faster transistors and shorter, faster wires.
  • The layout was tweaked slightly ("transistor sizing", "strengthening transistors") to improve the critical path.
  • the chip was completely re-designed (deeper pipelines, improved clocking tree, smaller cache, etc.) to shorten the critical path.
Does that answer your question?
(Should I move this to the CPU design article?) --70.189.77.59 17:54, 27 October 2006 (UTC)

[edit] Link inappropriate?

The external link to cpu-collection.de, a documentary website about the history of microprocessors, has been removed for being "inappropriate". In an article about Central Processing Unit - isn't this link helpful and therefor appropriate? What do you think? —The preceding unsigned comment was added by Morkork (talk • contribs) .

It was removed partially because you were adding it to several articles without first discussing its inclusion. That is generally viewed as WP:SPAM, especially if you have something to do with the website you are adding. I'd say that link may be appropriate for microprocessor (ask on the talk page), but not this article. -- mattb @ 2006-11-08T01:25Z

[edit] Why???

Why does this get vandalized so much, anyway??? What makes this a popular target??? 170.215.83.212 07:44, 17 November 2006 (UTC)

You got me... I guess it's just people trying to mess up a good thing. -- mattb @ 2006-11-17T16:07Z

All people who have vandalized this page should be IP blocked. Especially the monumental asshole with IP 82.42.146.28 who wanted to destroy the whole article! —Preceding unsigned comment added by 132.203.109.196 (talk) 23:58, 21 September 2007 (UTC)

[edit] Moving some content from CPU Design?

I created a new section called Markets in the CPU design article. I think that section probably belongs here. What do people think?

Even more important, the CPU Design article has a list of micro-architectural concepts, similar but not exactly the same as the Design and Implementation section of this article. This article doesn't mention RISC nor cache, for example. All of these ought to collected in one place. I don't have an opinion on where this micro-architectural stuff goes - either here or CPU design. Opinions?

Previously, it was discussed that the CPU design article should concentrate on the actual implementation task of designing a CPU. I agree with that. Things like micro-architecture and markets straddle the line between architecture and implementation so its difficult to figure out where this content belongs. Dyl 17:03, 18 February 2007 (UTC)

I think the content you added should rightly stay in the CPU design article. This is an overview article and it is far more appropriate for it to describe the operation of CPUs at a high level rather than muck around with a lot (and there is a LOT to be said) of detail. This article actually does mention caches, but it defers to the article on cache rather than taking a lot of space to explain it here. -- mattb @ 2007-02-19T14:44Z
Ah, I see the overview aspect now. Previously, I did not see it as an overview as it is poorly titled. The titles of the sections mention the names ILP and TLP but don't explicitly state what they are, instead they name specific sub-topics. I'm renaming the sections to show they are different types of parallelism. I believe that's the aspect the overview is discussing. --Dyl 08:40, 20 February 2007 (UTC)
That's fine, but the new names you've given to the headings are a bit longer than preferable. Could we shorten them a bit? How about just "Instruction level parallelism", "Thread level parallelism", and "Data parallelism". -- mattb @ 2007-02-20T15:11Z

[edit] CPU Operation

I dono what happened to the CPU Operation section, but it's filled with example pictures and crap text now. 24.7.201.100 19:56, 18 February 2007 (UTC)

That was probably vandalism you saw that has since been removed. -- mattb @ 2007-02-19T14:45Z

[edit] grammar

Someone please change 'generally expense" to "general expense" in the fourth paragraph of the "integer range" section of the Design and Implementation section in this article. Thank you. v

[edit] Something of wich wikipedians go crazy about,

pointless question. I'm thinking about buying a computer(in pieces) and assembling it, now if something(ArmA) requires a 3.0 GHz prosessor can I use a 2.4?

  • RAM 2GB 800GHZ
  • HD 500
  • GeForce 8800GTX (752) —The preceding unsigned comment was added by 84.250.110.93 (talk) 09:02, August 22, 2007 (UTC)
This is not the place to ask such questions, but Wikipedia:Reference desk/Computing is! This is a lesser known part of Wikipedia where you can ask questions like this. At the Wikipedia:Reference desk you can also ask questions about many other non computer related subjects such as medicine, philosophy and many other subjects. Mahjongg 00:31, 25 October 2007 (UTC)

[edit] Missing Pictures all over the place!

The two pictures with Intel chips do not appear. Is it broken? Has it been vandalized? —Preceding unsigned comment added by 132.203.109.196 (talk) 20:09, 15 September 2007 (UTC)

Hello! There are a lot of missing pictures in this article as well as in many other ones related to computer hardware. That would be cool if the people in charge of Wikipedia could fix these things! —Preceding unsigned comment added by 132.203.109.205 (talk) 16:58, 16 September 2007 (UTC)

[edit] homebrew CPU article

A homebrew CPU is a central processing unit constructed using a number of simple integrated circuits, usually from the 7400 (TTL) Series. There used to be an article about it at Homebrew CPU, as can be seen in the history of the page [2]. However somebody removed all the articles content, (which was quite interesting) and made a simple redirect to Central processing unit out of the page. I discovered this when I read about the "Magic-1", and decided to create the page "homebrew CPU" by myself, only to find out it already existed as a redirect to "Central processing unit". However, the Central processing unit article itself talks about CPU's built from TTL only in a historical context, and fails to even mention that even today some people are building their own Central processing units from scratch. I really think that "Homebrew CPU" warrants an article all of itself, or failing that at least a cursory mention in this article. Mahjongg 00:44, 25 October 2007 (UTC)

I don't think the subject belongs in this article (which has plenty to cover as it is), but I might believe it warrants coverage, and that would probably be most appropriate standing alone. I suggest you just restore the pre-redirect version, and then expand and improve it. (Don't homebrewers use FPGAs or something like that now?) -R. S. Shaw 06:08, 25 October 2007 (UTC)
I personally find homebrew CPUs fascinating.
If it turns out that homebrew CPUs aren't "notable enough" for Wikipedia,
perhaps Wikibooks:Microprocessor_Design/Wire_Wrap
would be a better place to discuss them.
It currently contains all the text that used to be in the Homebrew CPU article on Wikipedia.
--68.0.124.33 (talk) 08:39, 21 January 2008 (UTC)

[edit] Request for partial protection

This article was vandalized on Monday, November 12th, 2007 at approx. 9:00 a.m. Someone changed various sections of the article from "Central Processing Unit" to "Central Cock Sucking Unit" and committed various other vandalisms. I have tried to fix the article as best I could, but perhaps the article should be reverted to a previous state.

Additionally, I move that this page be partially protected for a while in order to deter the vandal from changing the page again. In addition, I move that the IP of the vandal (or his/her username) be given a temporary ban.

The Ice Inside 14:10, 12 November 2007 (UTC)

in the world of computing raw facts has to be accepted as an input, processed, and used as information for decision making ( out put). fully explain how this process takes place and explian how central processing unit plays a vital role in this process —Preceding unsigned comment added by 41.242.25.134 (talk) 17:19, 22 April 2008 (UTC)


[edit] CPU architecture

(move from Chipset dicussion) discussion started by ramu50 Most people today don't know that CPU itself doesn't run in GigaHertz, the Gigahertz is calculated by (ALU + NPU + FPU) *FSB* max IOPS = CPU speed (Note: my calculations may not be correct, but I know it is through some sort of forumla). Hence it is assist by chipset to achieve that speed. ALU, NPU and FPU are the internal core of CPU and initally round around 300MHz thus proceed onto 800MHz. Strictly speaking CPU and all the other families of RISC, MIPS are not processors at all, they are microprocessor. The only true CPU is MAJC, because it can handle ANY type of data. 19:59, 27 May 2008 (UTC)

[Error statement fixed]~~~Sorry for the inconvience--Ramu50 (talk) 19:54, 29 May 2008 (UTC)

Uh... what??? Do you have any reference for that? Are you talking about the use of clock multipliers or something? Your statement about MAJC being the only "true processor" makes no sense; traditionally, the distinction between a "processor" and a "microprocessor" was that the former was implemented with discrete logic (SSI or MSI chips), while the latter was implemented in a single VLSI chip. ǝɹʎℲxoɯ (contrib) 22:12, 27 May 2008 (UTC)
Nope, CPUs certainly do run in the GHz range (IBM Power6) and GHz is certainly not calculated by "(ALU + NPU + FPU) *FSB* max IOPS = CPU speed". Chipsets have nothing to do with "assisting" the CPU to achieve high rates. There are processors such as the DECchip 21064, that take an external clock signal from a crystal oscillator before divinding it and outputing it as the system clock to the chipset. As for your claims that on MAJC is the only "true processor" becuase it processes all types of instruction in one unit simply does not make sense. The MAJC has circuitry that is not shared between different instructions (fixed point, floating point) just like any other processor, except for that the other processors separate the circuitry used for each type of instruction into units. Rilak (talk) 08:58, 28 May 2008 (UTC)

Sorry I meant by the current design architecture isn't a CPU. MAJC is a CPU, while the current CPU should be considered as a microprocessor, sorry I said before "The only true processor is MAJC" I forgot to add the word "Central."

CPU = Central Processing Unit and Central = Management, well when you are processing intensive logic processing (you get something like this 80% APU, 18% NPU, 2% FPU), is that management? Unable to balance it efficiently thus producing 45~50 degree Celsius of heat is not management. The design architecture of MAJC is they can handle any type of data, in one single unit, they don't require subunit and in processing sometimes you get interlocks, C = A + B, E = C + D where the second instruction must be read before the entire thing can be process. In a regular CPU today, you have NPU doing all the work while the ALU + FPU is doing nothing, but in MAJC the entire unit is processing thus, require less power because the amount of processing require is distrbuted evenly. Says if the previous interlocks require 2 clock cycle, both MAJC and CPU are capable with scheduling something that takes only less than 2 clock cycle.

But the problem arise is what happen when you are doing abstraction (in graphics) when most of the data are arthimetic, then you can't schedule because the NPU is full, but in MAJC each transistor is only partial work so they are able process more work thus able to manage things more efficiently.

MAJC I think isn't recognize today, because I think they just started in 1990s and they are still researching on processor technologies (like Intel Extended Memory 64 technology, SpeedStep...etc.). But I think the industry should recognize the architecture of MAJC, because if the current Intel, AMD, Via, IBM, PowerPC is consider a CPU, then explain to me what difference is it between a GPU, NPU (Network Processing Unit in Killer NIC). GPU has a subunit NPU and FPU, just because it has less subunit doesn't mean its central, each processor has its own different function so which subunit they need is obviously different from each other.

  • Forget about the CPU in Gigahertz, I didn't save my the source, plus that was about 3 years ago when I found it. --Ramu50 (talk) 18:10, 29 May 2008 (UTC)
I am afraid I don't understand what NPUs (I assume that NPU in this context refers to Network Processing Unit) has to do with CPUs or why APU (ALU?), FPUs, GPUs and NPUs are the same thing. This discussion is also getting a little off topic (this talk page is for the discussion of how to improve the article about Chipsets) but if you wish, you can take it to my user talk page. Rilak (talk) 07:04, 30 May 2008 (UTC)

FPU, ALU, NPU in CPU subunits is known as
FPU = Floating Point Processing Unit
ALU = Alogrithmic Processing Unit (meaning logic)
NPU = Numeric Processing Unit

Common ones GPU Graphic Processing Unit (currently being considered as a partial APU, because of the discovery of Geometric mapping of sound DSP (Digital Signaling Processor)--use in Optical Mouse


APU = Audio Processing Units
PPU = Physics Processing Unit (very dependant)
NPU = Networking Processing Unit (see BigFoot Network, Killer NIC cards)

I think this is more related to CPU, NOT chipset. I will move this entire discussion to CPU discussion. --Ramu50 (talk) 16:04, 30 May 2008 (UTC)