Talk:Advanced Micro Devices/Archive 03

From Wikipedia, the free encyclopedia

Archive This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

This is Archive 3, which contains articles started in 2006.

Contents

Grammar

This article has various grammatical issues, ranging from minor to severe. There are instances where commas are inserted unnecessarily, adverbs are carelessly strung onto sentences, and there are phrases that just don't make a lot of sense. The article needs to be audited for its grammar.—Kbolino 05:43, 26 September 2005 (UTC)

I wrote most of it, and all the language tests I've done typically put me in the top 1-2% of the population. Writing is as much about style, as applying strict rules. Even the Oxford University grammar guidelines now state that split infinitives can be accepted in certain contexts for stylistic reasons i.e. even world experts on English say there is more to good writing than typing like a grammar robot. As for not making sense, if you don't know anything about technology, thats not really my problem. The page is written in relatively simple laymen terms. Its hardly technical. Timharwoodx 23:06, 30 August 2006 (UTC)

Add 2900 bit-slice history

AMD's 2900 bit-slice chip family was legendary for its flexibility, as you could make a processor as wide as you wanted, with your own instruction set. The price for this was the slow speed imposed by inter-chip time (RC) delay.

Paperfoam packaging

The paper wiki article mentions that AMD have begun using 'paperfoam' packaging. Is this worth mentioning in the AMD article? --Aidan 01:54, 1 March 2006 (UTC)

that green crap that gets all over the place? man, i hate that stuff! -Lordraydens 18:25, 16 March 2006 (UTC)


Missing history, missing products, missing concept

this article is good within the limited scope the author has staked out but IMHO falls a bit short as a general history of the firm. the article concentrates mostly on AMD's history as a second source of intel processors, and treats the recent intel-beaters as mearly an outgroath of that relationship. i have a different take on the story. i see amd as an independant processor designer that became a second source for the intel x86 line and over time learned to gain market acceptance for it's own internal architecture by accepting the x86 instruction set as input. a lingua franca for processors, if you will.

as noted in another comment the 2900 series is missing. what falls by the way side with it is the more general notion that AMD was a respected semiconductor firm with a history of quality in house design before they started second sourcing x86 chips. (as noted in another wikipedia article, am2900 products were used in the PDP-11.)

there is no mention of am29K processor in the article per se, even though there is a wikipedia article on it which is linked to at the end of this article. the am29K was an important product for AMD. during it's production lifetime it was one of the most successful risc processors.

the acquisition of nextgen and nextgen's contribution to the AMD product line is given some attention, but a major concept is missing from the treatment.

to the best of my recollection (and i'll admit it's been a while) nextgen's processor understood x86 instructions but used a different internal architecture. it translated on the fly. they're internal architecture wasn't fast enough to do that translation plus it's own work and still yeild the kind of performance needed to win broad market acceptance. they're processor did not take the world by storm, but they had a powerfull idea and experience implementing it.

AMD had experience with the x86 instruction set. they also had a very fast, cheap processor with it's own instruction set that was limited to niche markets because of the installed base of x86 software. what nextgen brought to the party was a way of putting those things together.

as described in the wikipedia K5 article, internally the AMD K series processors in large part constited of sections of the 29K. nextgen style technology was used to translate the x86 instruction set. this time the internal architecture was fast enough to handle the translation and provide a performance gain on top of it.

the rest, as they say, is history. AMD gained flexibility in design was not restricted to simply cloning another vendors product.

Nextgen and AMD transformed the published x86 instruction set from simply the instruction set of a particular processor family to a generalized interface to processors of various designs. since then other vendors have used it this way including national semiconducton, transmeta and via.

wasn't there a compatability issue with early K series CPUs?

i may be wrong about this, but i seem to recall that at the time of the release of the early K chips some popular commercial had exploited undocumented intel x86 features. the K chips which only supported the published spec didn't run this stuff (and i've forgotten what it was) reliably.

This is highly unlikely. When AMD reverse engineered that Intel instruction set, they did not just go with what was documented. They did complete pin signal level reverse engineering, as well as a lot of software reverse engineering to figure out the entire instruction tree of the Pentium family of CPUs. So AMD was fully aware of even the undocumented instructions.
Starting with the P6 family however, Intel introduced the concept of "MSRs" or "model specific registers" which are not feasible to fully reverse engineer (there are 4 billion of them, and each may be associated with an internal state machine of arbitrary complexity.) Fortunately, the architectural specification of these registers state that their meaning may change with processor version and that general software should *NOT* rely on their consistent definition -- i.e., only Operating Systems, drivers and special tools should ever use them. Nevertheless in the Athlon, AMD implemented the special "performance monitoring" MSRs as well as write combining and certain cache control MSRs which were either documented, or exposed by software which AMD was able to reverse engineer. In real software, however, there should not be any noticed discrepancy because of MSR usage.
Another issue, however, was performance of certain instructions. The AMD K6 had an extremely fast "LOOP" instruction. Unfortunately, some poorly written Ethernet drivers used to use this instruction for a predictable delay in speed. As such, Intel had been actually slowing down the relative clock through of this instruction over time, so that its overall speed was roughly constant so that this software continued to work. AMD did no such thing, and when their processors hit 332.5 Mhz all of a sudden these Ethernet drivers started to fail badly. Microsoft eventually issued a path, and future revisions of AMD CPUs were forced to slow down the LOOP instruction just like Intel did. I don't know if they continued this in the K8 line of CPUs.

however, free software also written to published specs ran fine, which made the AMD processors the darlings of the free software world, as they worked fine and were cheap. so AMD may have gotten a big boost from bsd, linux, etc.

I think both AMD and Intel CPUs work just fine with "free software". AMD's CPUs tend to be *cheaper* which is probably the *real* reason it was considered a darling of the free software world.

i'm really not sure if this is factual or just a nice story, but it would be nice if someone could check it out. if it's correct it probably belongs in the article as it demonstrates the way standards adhearance affects the market.

Remember, that in a sense, Intel used to *set* the standard, and that included the documentation. However, this has generally been a minor issue with AMD CPUs as their implementations always tended to be extremely close to Intel's standard. (Certainly far closer than Cyrix, or other clone makers.)
I think that what may be coming to mind was the K6 issue with Windows 95. There was a "software timing loop" issue in certain versions of Windows 95 that caused windows protection errors (and spontaneous system lockup/no boot) due to the high clock speeds used by AMD in the K6-2/350 and up processors. The relevant knowledge base article is here: http://www.microsoft.com/windows95/downloads/contents/WURecommended/S_WUServicePacks/AMDPatch/Default.asp
This was not a processor issue, it was a software problem (and therefore not particularly relevant to AMD). I suggest someone either address this item in the appropriate processor line article, or simply prune this discussion topic.
Falcomadol 15:30, 30 October 2006 (UTC)


Athlon PR rating

Is this really true?

This definition began to be applied more loosely over time as AMD struggled to compete with the ramping clock speeds of the Northwood core (>3.0 GHz). The credibility of the scheme was only saved by the arrival of the K8, where model numbers once again correlated more reasonably to actual performance.

I've generally followed the processor market fairly closely. I distinctly remember someone (either Anand or Tom I believe) claiming that one or more of the high rated Athlons did no deserve their rating. IIRC AMD responded and said that the PR rating was given based on their performance via a published set of benchmarks. If this was true, and the Athlon really did deserve whatever rating based on these published benchmarks, then this line clearly needs to go. We can make it clear that it was difficult for AMD to compete against the higher end of Intel's performance of course Nil Einne 15:55, 19 June 2006 (UTC)

Well hang on -- they were, and always have been, AMD's chosen set of benchmarks. I think that the general sentiment at the time was that AMD had been pushing it a little with the last of the K7 line. You should not automatically take the AMD corporate line over Tom or Anand. You may wish to edit this to claim that it was "disputed", but I don't think its easy to claim that its wrong.

Removed statement regarding lawsuit

I removed the following statement

The easy cooperation by these companies is an early indication of the weakness of this case, and it remains to be seen how far AMD will go when the first hearing is in 2008.

It is unsourced POV. Normally I would tag it with a citation needed but in this case, I removed it because I feel it's a bit weak. I'm not a legal expert but I see no reason to assume the cooperation gives an indication of the weakness of the case. The companies subpoened are not under any legal threat/risk at the moment as the case is against Intel. While a successful lawsuit could potentially damage any relationship they have with Intel and their cooperation could make Intel angry, it's rather difficult IMHO to say for sure what these companies would do. On the face of it, it makes most sense for them not to cooperate too much, but not to be too resistant either. Why annoy either AMD or Intel? But it depends a lot on the circumstances. The companies might not be pleased about losing their rebates etc, but they probably don't like the threats. Also, they might feel they might put themselves at greater risk from either AMD or various governments if they don't cooperate so they might see no reason to put themselves at such risk for Intel (and if Intel tried to harm or blackmail them because of the lawsuit, they could retialiate by giving this info to AMD). Some of the companies might want to send Intel a message by cooperating. Others (e.g. Microsoft) may find themselves more favouring AMD at the moment. Point being, it's rather difficult IMHO to gauge what the cooperation of these companies means. If it were Intel themselves, then we could assume if they we cooperating easily it might indicate they don't think there is any case. But since it's about third parties who are not at direct risk, from the lawsuit, it's rather difficult to say what they cooperation entails. With a reference from some legal source, preferbly one who isn't know to always favour Intel for whatever reason (although even one who does always favour Intel would be acceptable of course) Nil Einne 16:24, 19 June 2006 (UTC)

Should we add a section or a note bout the "Alchemy" processors?

Alchemy was previously produced by AMD, and then it sold all the production lines to another company, should we add something about that? --202.71.240.18 07:14, 23 June 2006 (UTC)


about the reverse HTT article

The article should keep and not to merge to this article, same as AMD Live!, Torrenza, and AMD 64, cause they are seperate techonologies developed/proposed by AMD.


There's an interesting part in zh:AMD article...

There're two timelines in the article[1].

1. The timeline of AMD, and 2. The timeline of AMD CPUs.

I think it is a nice idea to include that in the article, and add the missing items to it. And maybe arranging the sections 3-11 (or maybe 12,13 and 17?)according to the timeline? What do all of you think about that?

--202.71.240.18 09:39, 11 August 2006 (UTC)


AMD Edge

From reading the AMD support forums there is a publication called the AMD Edge. It sounds like a company newsletter focusing on PR for the technology this organization is using in their products. If anyone knows more about this, I think it deserves some mention in this article. Chapium 15:09, 1 September 2006 (UTC)

Well, I have seen a magazine named "AMD Accelerate" (Official Address, distributed by Ziff Davis Media) focusing on enterprise technologies (i.e. Opterons) but not "AMD Edge", any links to "AMD Edge"? --202.71.240.18 07:01, 2 September 2006 (UTC)

  • EDIT: "AMD Edge" has ended officially on the August 9th, 2005. It was used to promote AMD products, and also held promotions such as giving out CPUs and DDR Kits to attract people joining AMD Edge. The campaign links are still working (FX-55 Promotion and DDR Kit Promotion). However, the link for joining the program have been deleted and the login page will be redirected to the DIY page ([2]). --202.71.240.18 07:13, 2 September 2006 (UTC)

AMD64 - sentence

this sentence doesn't make sense to me, how is it supposed to be correctly written?

...the implementation of an extremely high performance point to point multiprocessor capable interconnect called HyperTransport, as part of a Direct Connect Architecture.

Shandristhe azylean 15:49, 11 September 2006 (UTC)

"Multiprocessor-capable" should be hyphenated to show intended meaning. But the sentence reads much better, and still quite gets the point across, with those two words removed. The fact that multiple processors can be used is not especially important, and makes the sentence all the harder to understand. — Aluvus t/c 17:39, 11 September 2006 (UTC)
thanks alot, aluvus, esp. for the quick response :) Shandristhe azylean 19:06, 11 September 2006 (UTC)

Wintel and Lintel

A bit OT but I felt it the best place to get the kind of views I'm after. I noticed in the Parallels Workstation page, Wintel and Lintel are used to describe the platforms. I've never really heard of Lintel before (although it's kind of obvious) and I've never really liked Wintel because it's primarily (IMHO) used by Mac users and also it doesn't really include AMD. But I'm wondering if I'm being a bit silly. What do others think? Nil Einne 20:32, 16 September 2006 (UTC)

So does that mean we should start referring to new Macs as "Mintel"? scot 18:15, 20 September 2006 (UTC)

Chartered

Most of the manufactured microprocessors are shipped from Singapore to Taiwanese and Chinese OEM/ODM manufacturing companies that build computers for Lenovo and Dell.

Is this true? AFAIK most normal AMD cores are assembled as processors in Malaysia and I believe some also in China. Does this mean Chartered are actually assembling the processors rather then just making the cores (which seems unlikely) or AMD's vendors are assembling the processors themselves (which seems unlikely to me but who knows?). Also, whatever the case, where is the testing and binning done? Nil Einne 16:08, 20 September 2006 (UTC)

Chartered owns a fab, named FAB 7 which was used to produce the cores. I believe that the chartered finish the cores and send to Malaysia for assemble and test in Suchou, China. --210.0.209.178 03:41, 27 September 2006 (UTC)

Outdated

There's a sentence in the AMD64/K8 section:

"Unlike Intel's dual-core designs, the X2 mates two cores into a single chip, rather than two chips into a single package. Intel's method may have theoretical yield advantages, but gives up some performance advantages since interprocessor communication still happens over external pins, rather than internally."

I suppose this is referring to Intel's Pentium D/Pentium Extreme, which are formed by joining two previously single-core chips into one package. However the chips based on the Intel Core Architecture were designed from the ground-up to be dual-core, so I'm assuming this sentence no longer applies (with the exception of Kentsfield which is two dual-cores put together in the same package). --QTCaptain 02:21, 14 October 2006 (UTC)

Dell servers

Brought over from an edit to the article:

COMMENT: I'm not sure if its public knowledge yet, but Dell does plan to include Opteron 2xx Series models in both the 2U and Blade space. This is in contradiction to the information above.

The above was posted by 69.169.6.22. — Aluvus t/c 03:12, 16 October 2006 (UTC)

Bias

Intel was immediately forced into a panicked internal re-design of the P6 core, which fixed many of the pipeline stalls that compromised its performance. The result was called the "Coppermine" revision. However, the rushed nature of the work put enormous pressure on Intel's manufacturing facilities and, even after it was announced, availability of the improved Coppermine chips was poor.

Can you cite a source? As for the second sentence, how was availability before Coppermine was announced? The wording is rendundant.

The X2 marks a significant step towards even greater productivity and scalability, especially for multi-threaded software applications.

This sounds like marketing material. "Productivity?" Let's keep the article relavant to AMD: the company and products.

AMD has also formed a strategic partnership with IBM, under which AMD gained silicon on insulator (SOI) manufacturing technology, and detailed advice on 90-nm implementation. IBM holds many patents on SOI technology, and Intel is reluctant to implement the process for this reason, despite the significant reductions in power consumption offered.

This statement couldn't be more wrong. Funny how there is no source given. Companies make dubious claims about their process technology. It certainly is not a fact that SOI offers better power consumption in practice than CMOS. If anything, the jury is still out.

Answer: you may not like it, but it is a fact that SOI offers better power consumption. The reason for that is that SOI reduces transistor leakage, and if less energy is leaked, less power is consumed... simple as that. The reduced leakage also enables the transitors to be packed more closely together inside the chip. In practice, manufacturers use the "liberated" space to pack EVEN MORE TRANSISTORS into the chip. The result is that the overall power consumption of an SOI processor may look similar to that of a regular processor, but the performance is quite different (more transistors = better performance). As for references, you can access the specifications of any processor that use SOI technology (I, for instance, found all the information on SOI at the APPLE website in their documentation of the PowerPC processor)

Unlike some other companies, AMD provided the technical details required for the open source BIOS project LinuxBIOS

What other companies? What technical details?

Again, innovation is key to AMD's "Virtual Gorilla" corporate strategy.

Why not include the phone number for the AMD US sales office after this sentence? This may or may not be true, but it reeks of bias and a poor encylopedia article.

Agree with everything here. I made a preliminary attempt to rewrite the K6 and K7 sections (which were particularly offensive in an autobiographical self-aggrandizement way). MORE WORK IS NEEDED. Falcomadol 16:31, 30 October 2006 (UTC)

AMD64 / K8

On May 31, 2005 AMD released its first desktop-based dual core processor family — the Athlon 64 X2. Unlike Intel's early (Pentium D and Celeron D) dual-core designs,

The Celeron D is not dual-core.

This article is in chaos!

  • There's one paragraphs mention the 80-core CPU from Intel...
  • There're places where the compeition between Intel and AMD placed in the Fab section.....
  • There're "weasel words" all over the place (please do tell me about it...)....
  • And most importantly, the section headings suck, "other productions" and the "production history" are seperated!? Aren't they the products form the SAME company, i.e. AMD!?
    • That is shall we do it the "TIMELINE" way or the good ol' "CATEGORIZED" way?
      • "Production history" and "Products", only one of the two can stay!

Post the

Notepad
 To comply with Wikipedia's quality standards, this article may need to be rewritten.
Please help improve this article. The discussion page may contain suggestions.

tag on, shall we? as well put the cleanup tags in the ATI section.

Post any thoughts below, thanks! -202.71.240.18 08:26, 10 December 2006 (UTC)

I agree with most all the sentiments expressed on this talk page. A firm supporter and devoted customer of AMD, I find that the inaccuracy, bias, and poor organization of the AMD articles hurts the company's image. I fully support a rewrite, and should I find something of value, I will outline it on this page for review.

Just as a suggestion, does anybody know of a benchmark comparison between Intel's Core 2 Duo line and AMD's Athlon 64 X2 series? AMD's website has benchmark comparisons featuring Pentium D processors, but not the new Intel dual core models. Information from such a benchmark test might be relevant in the article on the X2 AMD processors. They Call Me Dick 08:16, 16 December 2006 (UTC)

Re-Write Outline

It's pretty clear that the AMD article currently reflects the execution level of its ATi subsidiary (that is to say, it sucks). In pursuit of making the article...acceptable, I propose we follow this outline. I have not begun work on any section of the outline, (though will commence so after the holiday) - feel free to modify the outline, or take a section to write. For now, I suggest we try to keep it to one person per section, as the "I have interesting fact X about AMD, so I'll stick it in random section Y" strategy is pretty much what got us here in the first place.

For now, I recommend we keep the ATI information in a seperate article - ATI operates currently as a wholly owned subsidiary, not a division of AMD. I suggest we keep the ATI article and make it abundantly clear they are "green team" - and make appropriate lateral references in the AMD article.

I. Introduction:

   a) Brief explanation of AMD (Name, company, founded date, primary business (semiconductor manufacturing)
   b) Vital Statistics Box

II. History

   a) Founding of AMD
         i) Founders - JS III, et al - date, circumstances
         ii) Initial Products
         iii) Company philosophy and presence in the market place
   b) "Formative Era"
         i) AMD becomes a supplier of Intel "cloned" chips (reasons for doing so, and the circumstances therein)
         ii) The begging of legal troubles with Intel (brief mention of Intel transitioning from "needing" AMD to protecting their IP, not a lot of detail here
   c) Transition
         i) AMD begins to step out of the x86 "mold"
         ii) Production woes begun to stunt AMD's growth and ability to compete
         iii) The x86 Patent Settlement
   d) Techonologically Competitve
         i) Architechtural decisions that lead to technological leadership
         ii) How technological leadership became irrelevant in the face of competition
         iii) Spinoff of Spansion
   e) A Market Leader
         i) The release of the K8 signals a new direction and unification of AMD's Strategy
         ii) Intel Anti-trust suit
         iii) Purchase of ATI

III. Technology (This section is a simple breakdown of AMD's technology and pace of advancement - detail is to be found in other articles)

   a) Early Years (cloner)
   b) Undertaking Independent Designs
   c) K-era
   d) K7 and the 'marketing of technology' (K becomes Athlong)
   e) K8 - take industry leadership
   f) Opteron (Compete beyond the package - HT, ccHT)
   g) Launch of Platform Initiatives (Torrenza, the AMD Chipset business, AMD Live)
   h) Looking Ahead (K8L, Fusion, K10)

IV. Manufacturing

   a) Manufacturing Introduction
        i) Structure of AMD's manufacturing capacity
        ii) Overlook of AMD's manufacturing technology (currently 65nm 300mm SOI process) and planned process technology
   b) History
        i) Initial Production (sublet)
        ii) Desire to compete on the "grand scale" leads to pursuit of "fabs" (Jerry Sanders Quote Here)
        iii) Establishment of First Production Facilities
        iv) Pursuit of Production Technology Leadership through Partnerships (IBM)
        v) Production Future (Fab 38, Fab 36, fab 4x (luther hills))
   c) Problems
        i) Financial Proposition (Intel 100bil. market cap v. AMD 10 bil. Market cap - same number of fabs needed)
        ii) Execution Failures (Late process shrinks, fab engineers unable to execute Architecture at the same level as the competition)
        iii) Sales growth faster than Production Growth
   d) The Chartered Solution
        i) Relationship with Chartered and UMC
        ii) ATI's affect on relationship in the future

That is all I have at the moment. I hope this table doesn't come out like total snap. Sahrin 05:43, 22 December 2006 (UTC)

  • Something still not solved, what about other products (e.g. AMD Live!) and initiatives having codenames only (e.g. Torrenza, AMD Trinity etc.) away from CPU itself, though AMD is mainly a CPU making company now. Away from those, I think this is good layout. :) --202.71.240.18 08:26, 23 December 2006 (UTC)
    • P.S. Don't forget to mention AMD64 and TSMC.
    • P.S.2 I think you should provide something about "Spansion" and then its spin off. --202.71.240.18 08:29, 23 December 2006 (UTC)
      • I agree that there is some material missing, but I think, given the breadth of material in the form of other articles available (on AMD 64, on HT, on AMD Live etc.) - this article is about a public company, it should be structured more like ...an entry in the factbook for a country. In fact, I think we should add a section on a financial performance, but I am not sure if we can a) get that information or b) manage it effectively. I've made some small changes based on your feedback.Sahrin 14:43, 23 December 2006 (UTC)
  • I think that it looks good. --BenWhitey 03:00, 31 December 2006 (UTC)
  • I agree—this outline covers the matter fairly well. I believe that the information about each generation of processor should be placed in respective articles, instead of being placed in the main article. These individual processor articles are very sparse as it stands.The Squonk 06:25, 1 January 2007 (UTC)
  • Good outline, something needs done. I think the main AMD article should deal with the company, and only a brief overview of the products. A list of AMD processors should be a separate article, as should a list of its future products, chipsets, grahpics, services, etc. They all should be seperate articles that only on briefly touched upon in the main article. Nicholas 16:56, 12 January 2007 (UTC)