AMD ingnots, sliced "TBread" with the crusts cut off
 

http://www.madshrimps.be/gotoartik.php?articID=84


I want to thank everyone whom contributed by printing their full sticker code, and especially thank Austin @ LowYat.net. Without such experts (enthusiasts) willing to take risks sharing their hypotheses, this would be a boring hobby.

very interesting, great article!
explains a lot how that strange AMD rating works :super:

mod: can you explain why the colors of the cpu differ? I've had a brown and a green 1700+ AIUGA through my hands, so why do the batches have different colors?

Don't shoot me for making a wild guess, but perhaps AMD doesn't "cut the crust" because of technical but economical issues. Imagine if they would try to produce CPU's from those outer crust "K" and "R" sections, and only 50% would pass the tests. Would the other 50% (which is being sold for what, 30$ a piece per 1000?) justify all the manufacturing costs involved? The "good quality" center wafre JxxxB's are already being sold for way to little money, perhaps it's just cheaper for AMD to throw those "crusts" away or recycle them.

Very interesting read, especially the first paragraph. Now I know a little bit more about microprocessor fabrication.

By the way, thanks for adjusting the blue into gray. ;)

Thank you guy's. And "shoot you" heck no, I'll thank you! Part of the reason "I put myself out there" is to learn from my mistakes. that's not to say, I start with unvalidated premises, but at some point the definition of a hypothesis (an educated guess) will be tested.

I didn't consider the economics from that perspective. In fact that's where I kinda rushed things. I couldn't find the article I read on lithography where it discussess the voltage testing, and how every piece cannot be tested due to cost. But that article would have corrborated your theory. I wonder then, at what point AMD made the decision to "bin" the remaining wafers, or crust. (God I hope I haven't coined a phrase, and it's "crust").

great work Liquid3D, thanks for sharing this info with the community :ws: !

Liquid3D, how did you come to the conclusion the "K" and "R" chips are made from cores close to the end of the wafer, and not just a low quality wafer?

Is it a logical deduction, or is it a fact?

Richba5tard I came to the conclusion based upon Austin's guide over at http://forum.lowyat.net/index.php?act=ST&f=5&t=612 the reason I give his guide so much credit, were his theory (at least in part) was (I beleive) based upon an inside source at AMD's Singapore Fab. How much, and which information is attributed to that "contact" I've no knowledge, except to say the "laser marker ID" was definately attributed to the "contact".
And from Xbit; http://www.xbitlabs.com/news/cpu/dis...322034857.html
where they claim the following on their site, which they layout in a "5 Step" process;
"The first five letters in the line deal with process information (JIUHB in our case). The first letter among five presumably informs us about the location of the core on the wafer. Closer the core was to the centre of the wafer; better overclocking potential is, as the core seems to have more potential for higher speeds. Lower letter is better, as it means that the die was closer to the centre. In general, "A" is better than "J", while K is typically worse than "J", but still has more potential than "R", which is the worst"

Out of curiosity, would you consider this plagerism? I like XBitlabs don't get me wrong, but basing this "5 Step" process to TBred ID, there's whole paragraghs cut, and pasted from Austin's Guide. That was the reason I gave so much credit to Austin at lowyat.net, without whom I'd have little to extrapolate from. My hypothesis was based on the "K -R" steppings being absent, HIS (Austin's) theory was based upon the steppings themselves.

Also jmke I followed that quote you placed in there but cannot find where this question is posited? Funny you should post that quote, because I'm almost finished with my next article which focuses on Intel the industry, and the 157nm Lithography process. That particular figure 157nm, is a measurement of the upcoming ultraviolet-light wave-length, which will image (or etch) lines in the .09 micron (90nm), .06 micron (64nm) through .04 micron (45nm) core die-shrinks. Currently DUV, Deep Unltraviolet Lithography (possibly augmented lense 248nm wave-length unltraviolet light) is used to image the .13 micron (130nm) process.

mebby it helps :)

"That particular figure 157nm, is a measurement of the upcoming ultraviolet-light wave-length, which will image (or etch) lines in the .09 micron (90nm), .06 micron (64nm) through .04 micron (45nm) core die-shrinks. Currently DUV, Deep Unltraviolet Lithography (possibly augmented lense 248nm wave-length unltraviolet light) is used to image the .13 micron (130nm) process.
"

was the answer I was looking for :=)

This is all quite amusing for someone who actually works in the industry. It's amazing how ill-informed you all are about things. Wafers are generally faster in the centre, true.
I'd be amazed if AMD binned based on wafer location to the extent you describe. As an Intel stockholder, I'd be delighted if their process was so far out of control. But I'm sure it's not.
You bet they'd sell 50% yielding edge dice - these guys are trying to make money, remember. It probably costs them $15 to assemble and test them, so if they make $30 for them, that's 50% GPM (which actually, they'd kill for).
As for why Intel chips clock faster on a similar process. Why do you think Intel has those long pipelines you all complain about so much? It's so they can clock it faster - longer pipelines means stages with shorter delays, hence faster clock. On a similar process the P4 will inevitably be able to clock faster than Atthlon. That's just physics.

Thanks for your contribution, except for the economical part of the matter, what "ill-informed" anomalies did you encounter?

Since this editorial is based upon own research and that of others it is fair to say that some parts of the puzzle could not be found without actually going to a factory and getting the raw facts and data there


I would like to see both companies work closer together to create a better and faster CPU, but thanks to capitalism this might never happen...

its really sad that we see these meager increases in speed just to get the $$$ :/

Could it be also that AMD is having problems with their Opteron scaling up and they don't wish the Xp's to become much better performers than the Athlon 64 when (if?) if comes out.

In 32 bit performance only of course!

Yes, "I" certainly am "ill-informed." However, what's more "amusing", is while I attempt to theorize (I won't even purport it's a hypothesis in your presence) experts such as yourself, sit high upon their throne of esoteric knowledge, chuckling at the "ill-informed." We enthusiasts blindly scurry down those long, dark Intel pipelines, rapidly shrinking "gate-widths" closing down on us, whilst you and your stockbroker conceal the proverbial lightswitch. Perhaps in my ignorance of the manufactirng process my assumptions were so inaccurate as to become comedian material appropriate for the next Intel Golden Parachute Dividend Event.
Why simply criticize? Why not enlighten us with those facts your "sure" of? You concur wafers are generally faster in the centre, however; you'd be amazed if AMD binned based on wafer location? How about after testing? I never claimed it was simply about wafer geography. With approximately 300 steps involved in current "Fab" methodology, perhaps there may be as many ramification detrimental to product end stages? And is it not feasable such anomalies cannot always be detected prior to incurring significant manufacturing costs? Therefore, would it not be economically astute to relable cores originally intended for upper end processors, RATHER then re-tool?

In so far as why Intel chips clock faster. This wasn't mentioned in the article, however; it's my hypothesis your divigation to it, reveals your most likely in the marketing, advertsisng, or janitorial area of the industry, not manufacturing. Your bias an an Intel stockholder, is prevelant in your regurgitation of Intel maketing fodder. In so far as the ole' clock speed theory is concerned it's self effacing to site this. Yes you are correct in that Intel architecture results in a "faster" clock rate, and it is, "just" physics. Where you insult the intelligence of enthusiasts such as myself, is the condenscending manner in which you present your argument. The formula for Bandwidth is as follows. Bandwidth = Bus Size x Clock Speed; ergo a DDR 64-bit Bus Size x 300MHz Clock Speed = 2.4GBs : Bandwidth (data). While an RDRAM 16-bit Bus Size x 800MHz Clock Speed = 1.6GBs : Bandwidth of data. So while faster, less data (information) is processed (faster) Analogous to spinning feverishly in first gear, when you have 24. Albeit a memory (or trekking) analogy, Intel has pushed speed so far down our throats were choaking on it (really fast). You have the ad-lingo ad nuaseam. Sorry to "complain". Perhaps my retort wouldn't be as nefarious, had you not been so insultive in your mannerism.

This (ideally) is a forum where idea's are shared, and sometimes challanged. Where others with more experience, assist those with less experience. Where people gain satisfaction in helping others enjoy their computing experience. To simply laugh at us "ill-informed" being an industry insider is indicative of cruelty at the expense of others. My screenname would be "Unregistered" as well, if i were to join a forum for the sole purpose of revilement. Why not enlighten us instead? Perhaps as an Intel stockholder you reprsent the atypical self centered materialist? After all you invest in a company who recently renegged on their commitment to transition to the 157nm lithography process, adversely affecting the financial stability of mutiple sub-contractors. They found they could skip two nodes utilizing a 193nm lithography process, until EUVL facilities are up and running. As much as I'd like to think Intel's first order of business is building the fastest processor, I know the first order of business is to build the least expensive, fastest processor. After all I subscribe to the Motley Fool when it comes to investing.

I'll research that, in the interim, why don't you share some of the knowledge we all covet? I promise not to stand on the corner and bum $2.5 billion and open my own Fab plant, based on what you share.

Could it also be AMD's Opteron woes, may be interconnected to Intel's skipping the 157nm process, initally one of it's largest supporters?

What "performance" are we CURRENTLY in?

Congratz with the excellent article ..

http://www.tweakers.net/nieuws/27135 ! :super:

good replies there Liquid3D, I hope "Unregistered" steps up to answer some of the questions asked!

Sry, read some comments ..

this is a c/p of an reply from liable sources,
and hard to say, but he pulls the article to pieces... :(

<Gepost door RobT maandag 26 mei 2003 - 10:02 - Score: 4

Klopt niks van dat hele verhaal met die ringen.
Zo werkt het productieproces gewoon niet.
Als de chips in het midden beter zijn dan aan de zijkant, is dit een sterke indicatie dat je productieproces niet in orde is.

* RobT is chipdesigner en hoort best nog wel eens wat van de productiekant

Dat met stofdeeltjes of scheef ingestelde lenzen geeft al aan dat ze van toeters nog blazen weten; de maskers betreffen maar een paar chips (zijn het negatief van 2 bij 2 chips of zo), dus de chips aan de rand worden op dezelfde manier belicht en geproduceert als die in het midden. Dus als er iets scheef zit aan die lenzen voor de belichting of aan de maskers in de hele constructie, dan zou dat bv de chip rechtsonder op het masker ( ! ) slecht overdragen op de wafer. Dan krijg je op de uiteindelijke wafer praktisch een schaakbordpatroon van slechte chips.
Stofdeeltjes? Sorry, maar die chips worden gemaakt in een class 10 omgeving, als het tegenwoordig niet een class 1 omgeving is (het getal geeft daarbij het aantal stofdeeltjes aan dat groter is dan zoveel micron, per kubieke meter); je kan op een wafer een paar chips hebben die kapot zijn door dergelijke onzuiverheden bij de productie, maar de helft??
In dat laatste geval zou ik het rookverbod in de waferstepper ruimte wat steviger bekrachtigen...

Mocht er iets mis zijn met je silicium, kristalroosterproblemen en zuiverheid, dan smijt je niet de helft van je chips weg, maar ga je de discussie aan met je leverancier.

Er staan zo nog veel meer onnozelheden en onjuistheden in het artikel.

De reden dat de laagstgeklokte cpu's dezelfde snelheid halen als je ze overklokt als de hoogstgeklokte, betekent gewoon dat het AMD productieproces heel goed onder controle heeft, en dus eigenlijk een heel hoge yield heeft. Dat betekent (wat hier op t.net al vaak genoeg naar voren is gekomen) dat bijna alle cpu's de hoogste snelheid halen (die is min of meer proces-, design- of thermisch begrensd). Alle cpu's hebben dus zo'n beetje dezelfde kwaliteit, halen dezelfde snelheid, dus dan kijkt amd waarschijnlijk naar het energiegebruik/warmteontwikkeling, etcetc.

Vergeet ook niet dat deze overklokkers niet per se dezelfde tests doen als AMD, of eigenlijk: per se niet.
Denk even terug aan de PIII 1.13GHz die teruggeroepen moest worden omdat Tom van THG erachter kwam dat ze geen linux kernelcompile konden doen; die cpu's waren wel door de toenmalige tests van Intel heengekomen.

Op dezelfde manier komen deze overgeklokte t-breds misschien door de tests van de overklokkers heen, maar AMD heeft bv vastgesteld dat ze niet absoluut stabiel zijn of zo. Of niet absoluut stabiel met alle goedgekeurde psu's, koelers en mobo's of zo.
>

@Dail-Up: Liquid3D doesn't speak dutch =)

sry, but don't have the time now (or this month) to translate it because of my exams..

that's why we have Systran , read the annoucement in each forum =)

edit: this translation sucks donkeyballs :^D

I'll translate it at 2 o' clock, after my test CSS

here ya go folks

"The whole story of those rings is crap, your idea of the CPU production is just wrong. If the chips in the middle are better then those on the edge, it's is a clear sign that your manufacturing process is wrong!

*RobT is chip designer and gets quite some info on the manufacturing.

That with the pieces of dust and bad lenses means that they don’t know anything about it; the wafers concern only a few chips (so, the negative of 2*2chips or so), so the chips on the edge are lighted and produced in the same way as those on the middle. So if there’s something wrong with the lenses that light it or the wafers in the whole construction, then e.g. the chip located right under the wafer would be transferred poorly on the wafer. Then you’ll finally get a wafer with a chessboard pattern of bad chips

Dust? Sorry, but those chips are made in a class 10 environment, if it isn’t already a class 1 by now (the number represents the amount of dust particles that’s bigger then 1micron, for every cubic metre); you can have a few bad chips on a wafer due to such impurities, but half of them??In that last case I’d want less people smoking in the waferstepper. …

If there might be something wrong with the silicium, crystal or impuritiets, then you’re not going to throw away half of the chips but instead have a talk with you supplier.

There are much more stupidities in that article.

The reason why the lowest clocked CPU’s (overclocked) obtain the same speed as the highest clocked, means that the manufacturing process is okay, meaning it’s got a high yield. That means (what’s already been discussed a lot) that all CPU’s obtain the highest speeds (it’s more or less limited by design or thermic reasons). All CPU’s are about the same quality, obtain the same speed, so AMD probably rates them by power usage, heat dissipation etc.

Don’t forget that these overclockers don’t do the same tests as AMD, or actually: not at all.Remember the PIII 1.13, they were called back to the factory when THG discovered they weren’t able to do Linux kernel compilation but the CPU’s passed the Intel tests anyway.

In that same way do the OC’ed T-Breds pass the overclocker’s test, but AMD found out that e.g. they absolutely weren’t stable enough or so. Or not 100% stable with all PSU’s, coolers, mainboards or so."

thanks TeuS :super:

Nice and interesting article, I am not really a modelnumber expert so it's an interesting read.

I see the things I wanted to mention mostly popped up allready. Being:
"he fact Intel is surpassing 3.06GHz using standard Deep Ultraviolet Lithography, implies the problem with AMD lie somewhere else."

That looked sort of misplaced as this is mostly architecture related and is one of the reasons why Mhz doesn't mean a thing when comparing architectures. For example the Pentium-M (Banias) Intel made is not clocked as high as a Pentium-4 M nor can they clock it as high, even though it's still Intel making them with access to the same technology. When you start to compare with other architectures (SPARC, Power4, IA64, ...) it gets even worse as the design difference and subsequent performance results are completely out of line with the Mhz rating.

Note that I can imagine how the line got in there as architecture was not you focus and thus not on your mind when you wrote that line.


The other thing was allready pasted from that RobT guy. As he mentioned it's not unlikely that these many good CPU's pop up because AMD simply has a good production process. Within the same range of cores CPU's have the same design and their speed is determined by the quality of the yield. If for example AMD sells 20% high range CPU's but have their process upto a level where 40% of their CPU's would qualify as high level than obviously half of those are going to be put in a different box.

BLMet

that's why Liquid3D published it, he claimed nowhere that all the info was 100% correct, getting this article read by lots of different people will only help pinpointing the areas of the article that could need improvement the next time around!

thanks for the comment :)

some great insights here!
http://forums.anandtech.com/messagev...readid=1056653
:ws:

lets crack the puzzle and solve this mystery of chipset fabrication :)

Thank you jmke what many critics don't realize, is, I'm playing them like a fiddle. (just kidding) I've been postulating theories since my passion for philosophy was piqued 15 years ago.. Of course I try to work with solid premesis, however; my primary intent is to coerce such reactions as to seek the TRUTH! That not to say I wrote the entire artcle with the sole intent of instigating being flamed in effigy, I put forth a concerted effort. Irregardless (I even got an email saying there was no such word :o ) the theory he calls "crap" isn't my own, and says so in the article. It's Austin, and is credited to him. When a response is emneshed with insult, it's an egotistical responce? Why would this cause so much anger?

I'll reiterate I "put myself out there" to learn. And I'm not insulted, nor terribly offended by such reactions, so long as they contain some pertinent info, were all learning.

I think what you wrote is based on other people's "facts" and added you own conclusion to that so some parts of what you stated cannot be agrued with , with you.. because you were not the one writing those facts :)

we are all here to learn

the guys over at Anandtech sure have some great comments on the article, go give it a read, they seem eager to answer some questions that might lead to everyone finding the whole truth :)

If you are just looking for a "conspiracy" or a "plot", then you have to mention this in your introduction that this was the reason why you started thinking about it. Now it's like you have found some inside information and you made an article about it.
I don't criticize the article, but you have the make that point clear.

and, eh, in my humble opinion, hardware isn't really about conspiracys and philosophy, but only Science and Economics that matter... ;)

hardware is about money. they don't much about the technical side of things, as longs as it delivers the $$$$

just to keep all the info in the same thread, here are some of the very usefull responses provided by the people over @ anandtech

jmke I think you misunderstand my retort at Anandtech. I welcome the criticism, it's the condenscending manner which I feel is inappropriate. I'm not deferring all criticisms to the following attitude; "It's all Austin's idea's, criticize him not me..."

What I'm saying is why did this person, choose not to criticize Austin's TBred guide when it was released months ago? We've all been speculating about it, and all of a sudden it's "so wrought with inaccuracies" as to be "funny," and even "unworthy" of another persons time? Well he took the time to be insultive did he not?

It's apparent to me you have a lot of respect for Anand Lai Shimpi as he does study EE at North Carolina tech, and I joined Anandtech a while ago, but didn't necessarily find it enganging. Considering your number of posts there, perhaps you haven't either? But in so far as genuflecting in the presence of someone because they claim to work at the Intel 130nm fab facility is silly. I'd had more respect for him if he had shown simple respect. Otherwise I wouldn't care if the person responding to my article was a member of Scandia Labs, and the inventor EUVL. This does not justify sarcasm and condescension. I don't care what a person's credentials are, if they cannot interact in manner befitting of their intellect. I bet he doesn't talk to his supervisor that way? I may not know everything, or anything about photolithography, comparitively, but I know a whole lot about psychology, behaviorism, displacement, projection, and where these attitudes derive from.

I also happen to know a lot about anthropology, sociology, and the sociological implications of economics. Of course we as consumers are victimized by capitalism, but what's frustrating, is we don't have to be. A business can be run profitably any number of ways. In fact what I do know about IC manufacturing, is it's impact on the environment is detrimental. That's one reason why there are so few fab facilities in the US. It's much more costly to process, and dispose of the photoresist chemicals here then almost anywhere else. Labor is another reason. Intel could remain a profitable company by relocating all it's manufacturing facilites here, that supply the US market. But when individuals (and a business is composed of people) feel a need to have five houses instead of three, retire with 500,000 shares instead of 100,000 that's called greed. Capitalism and greed however are not necessarily synonomous. In fact if Intel, and AMD chose to invest in LADI, instead of EUVL, they would not only eliminate huncreds of steps from the manufacturing process, they'd eschew the caustic chemicals, cut production times, and shrink the die size beyond EUVL. Why don't they do this? Habit, change, and complacency. So when someone who works in a fabrication facility, begins spouting economic theory, your welcome to hang on his every keystroke. But what sounds foolish to him, in my postulations, may sound foolish to me in his. And for the same reasons. I do not instantly subscribe to his version of the truth simply "because." I didn't imply there was some dep conspiracy, but hpw long have you been in these forums and had peoiple in the business clarify issues? I find their rarely doing so. It's becoming obvious I'm wrong about the "missing steppings" I never thought I was right, only that it may be a possibility. Now because of criticism, I'm learning it's most likely not possible. But that's not to say it didn't happen because neither you, nor I, nor the "expert" at Anand was either.
I can be wrong gracefully, I only wish people could be critical in a similiar fashion. I know it's about money, and that's very sad. Are you willing to embrace thsoe feelings, even momentarily? I am, and when someone so abrubtly rubs it in my face, I think that's the appropriate time to do so.

EDIT: I have to go back there, and respond again as from the quotes you just posted; this person is now, taking the time to respond in a constructive manner. And that's all I ask. What's frustrating is I'd love to add some comments to the article to avoid further misunderstanding, but once it's been printed, how do i do this?

uhm? what where?
If I were to be a "fan" of him, I would have had more then just 9 posts on his forum ;)

I just like to keep it very polite, polite = gay? ;)

I just dont want to start any dispute over there, I surely felt bad when he said "where to start , I surely cant mention all" , and was ready to answer in another way then I did.. but I also thought "whats the use"?

I'm just trying to help you out in finding answer we are looking for, not everyone seems to be helpfull right away, as in, what he stated also, lots of people have claimed certain theories to be factual while they are not at all.. he didnt read the article in depth first time around, as you have mentioned in the article that these were presumtions

I hope you dont take offence when I try to get some reactions on your writings from different corners on the web.

AMEN! I had a feeling of disgust when I read the comment @ tweakers.net

instead of just saying "I think he was wrong on those points because a.. b...c.." , no.. he had to start of with "what a load of bullcrap"

really can tick someone of, I find it very :super: you remain cool all along

I've seen the nick Liquid3D around for some time now, as I read on XS, both [M] and XS are homes away from home. what is your real home forum. www.ninjamicros.com ?

send the changes that need to be done to jmke@madshrimps.be
and I'll fix them right away! :king:

hey even RobT seems to be providing usefull info :)



more @ http://www.madshrimps.be/upload/jmke/tbread/
will translate the comments tommorow :)

I apologize jmke you have been my staunchest supporter, and you've been investing as much time in this as I have, maybe even more considering your moderating.

I really am sorry. I have nearly 100 emails daily with criticisms about spelling to punctuation. I don't mind, but example Irregardless is a word of the English language, it's in the Websters New World Dictionary, as well as others. And this is just one of many.I have replied to every email. And when wrong I promptly admit it, and even thank the person. It's becomes hurtful when people malicsiously attack my efforts reducing it to pure dirt. Which ultimately I don't mind, becuase without grains of sand there'd be no silicon. Hence there'd be nothing to even argue over, such is the cycle of life. It's just that I am human and attemtpting to be "open" leaves one vulnerable by defeifnition, ergo I 've become hypersensitive. So initially it's hurtful, and I react. I'm working on this.

However your certainly not someone who deserves that type of response. My goal for the next few days is not to apologize as much, Which means I'm going to make a concerted effort NOT to overreact, and remembering who my freinds are! And I've been taught, even if criticism comes from a malicious place, I can still learn from it. Thank's I'm working on some amendments to add at the bottom of the article.

Liquid3D you have made a extraordinary entrance at the [M] forum :ws: , your dedication to finding out the true story is something lots of people would be jealous of :)

aha! read this after I replied to your email, answers my questions :)