Definitive guide to TBred-B identification.

[Liquid3D]

After reading several articles, and with the supply of downgraded lower speed TBred-B's rapidly depleting, I thought it opportunistic to post this. It used to be so simple to predict the "overclockability" of CPU, simply know it's stepping, and which stepping's were the better overclockers. But now that's all changed. Even the term "downgraded" throws many off, when in fact the term is indicative of higher speed (overclocking wise) CPU's, Processors which are downgraded higher PR models. In fact the reason current 1700 TBred-B's are such great overclockers is due to their being downgraded from minimum, 2600+ models. I beleive the figure may be as high as 100% of TBred-B's are downgraded from higher models. The reason is simple; cost.
It would have cost AMD much more to "re-tool" (pardon the expression) and produce lower speed Tbred's from "fresh" wafers, then do to what they did, re-label. It's a common practice, and I doubt there would be many overclockers if it didn't exist. Imagine if all processors were made within such tight spec's as to BSOD or reach thermal shutdown at 100Mhz above default. Even worse, AMD could go the Intel route, implement a patent which monitors FSB clock speed, and slows the CPU if certain speeds are exceeded! Regardless, AMD's folly (if it can be called that) is our good fortune, however; I belive there is a finite number of these diamonds in the rough. The supply isn't limitless, and as the buying frenzy depletes the stock, identifying the best overclockers is essential. If you remember two things from this post, remember these: the 8th and 9th numbers in the second line of code designate what PR your processor was downgraded from. The seventh letter (or worse number) in the fourth line of code designates at what stage the processor was chosen from in the production cycle/s. A being first batch pick, and any number 1- 9 from re-combined batches which failed numerous QC tests (Quality Control).

So it would seem among slower speed TBred-B's it's the second line of code which contains the pertinent info. Information which can determine your CPU's potential within a few hundred MHz. I will use the venerable 1700+ as an example, as it seems to garner the most attention. It's not only the least expensive Tbred-B, but has offered up some serious clock speeds. Assuming you already know the meaning of the first line of code, here's how it works:
AXDA1700DLT3C 9391257270022.
JIUHB 0310XPMW 1999 AMD
The first seven digits in the second line of code are thought to be serial number of wafer, and the grading code of CPU core. The eigth and ninth digits highlighted in red represent the "original desgination" of the CPU. Therefore these two numbers are perhaps, the most important. The remaining four digits are thought to be the batch production code, and it's thought the lower this number is, the better i.e. 0001 would be ideal. And by that formula these last six digits would be ideal 300001. Although 30 not being an exisitng AMD CPU PR, the designation does exist. Such that 28, 29, and 30 are the most sought after, with 30 being the overclocking apotheosis.

The third line of code known as the stepping specifies the manufacturing process in reference to the wafer. These four letters are the four first letters of the TBred-B's; A > J > K > R. A being the highest quality, and indicative of those cores closest to the wafer center. A being the purist, and therefore capable of the highest speeds. So there are A TBreds and J Tbreds. however by AMDBoard article "There have not be any AIUHB Revision B's, so the next highest grade is the JIUHB", therefore all are JxxxB 1700+ Tbred-B's.

The fourth line of code: the first four digits in the fourth line of code, pertain to the manufacture date (self explanatory). The alphabetic code (in the fourth line of code) proceeding manufacture date is represents batch number per run, per date. The seventh position in lthe fourth line of code, may in fact be the second most important figure in the entire forumla; i.e. XPMW, or SP1W are important because M is thought to designate combined "lots" from different batches which failed QC tests for those specific batchs, higher letters then M, and numerals 1 - 9 and are thought to be combined, and recombined, and have undergone several production runs, after failing MANY QC tests.

If you would like a foolproof method for determining your processor's downgraded status, and all pertinent info, simply enter your CPU's codeline in the boxes provided at this site: http://mp3.zonebg.com/cpu/cpu.php

Here are the two articles from which I extrapolated the above info: http://lowyat.isentral.com/guides/athlonxp/ and http://www.xbitlabs.com/news/cpu/dis...322034857.html

EDIT: I'm not going to edit what I wrote to keep my foot in my mouth if that is where I placed it. This is to exemplify, that while I entitled this a "Definitive Guide..." it is simply an amalgamation of what I have read, and theorized to be plausible. I'm not "The Guru of 3D" I'm simply an enthusiast attempting to contribute. Apparently there are those who have questioned the author of the "fool proof" ID program I found. I simply propose a theory based on data, reaching a conclusion. If someone comes along disproving the theory not only do I NOT take it personally, I welcome the LE (Learning Experience). I susbcribe to "Theorem 5," i.e. "In 1963 Frederic Fitch published Theorem 5, claiming that if there is an unknown truth then that it is an unknown truth is itself unknowable. Clearly there are unknown truths; we are non-omniscient. So, by Theorem 5, there is an unknowable truth. This conclusion threatens any theory that entails the principle of knowability, which claims all truths are knowable:
(KP) p Kp."

[DarKcyde]

Funny, i was just looking for that guide I read a few days ago. Couldnt find it in google, and searching news page archives was taking to long.

I come back here, et voila! Figures, everything else worth looking at is here.

Originally posted by Liquid3D
and are thought to be combined, and recombined, and have undergone several production runs, after failing MANY QC tests.



[/url] [/b]

If they've failed QC tests then why have they released them? Do you have figure for how many of the chips have failed QC?

Thanks,
Craig

BTW, you should write this up as a FAQ if you have time!

Craig

[spaceman]

Heehee, the last numbers of my week 49 WP"B"W 2400+ are 280002. It kinda takes the fun out of looking, once you have a primo chip. I did find another downgraded 2800+, but not quite that sweet. It's just kind of a shame that all I use is air-cooling, a PAL 8045 w/92mm Panaflo.

[Liquid3D]

Originally posted by CCW
If they've failed QC tests then why have they released them? Do you have figure for how many of the chips have failed QC?

Thanks,
Craig

Pertaining to the third line of code which begins with the stepping, the third place following the date denotes the batch run per that particular date A being the first run etc.; i.e. AIUCB0240WPBW. From what I've extrapolated from the guides/artciles I've read there are three primary batches A, B, C, these are supposed to be the best, from the "purest" wafers, and initial runs, where all passed their Quality Control tests for their given model (speed). Then there are "combined" batches M, which may have passed most quality control, but failed to meet their 2700 rating for example, but passed for 2400. Then "re-combined" batches R, I, S, etc. these have failed some tests, and were re-processed eventuaklly meeting some lower standard. Finally digits 1 - 9, in the third place of the thrid line of code. The line immediately proceeding the date; i.e. "AIUHB0203SP1W". Numerals having failed the most QC tests, from re-combined batches.

ok cheers, so they are ok chips just lower rated for a certain speed, cool

Craig

[Fewture]

Explain to me why some guys have a 2800+ "downgraded" from a 2500+ ?
And 1700+ downgraded from a 2500+ (Barton!!).

[Liquid3D]

Originally posted by Fewture
Explain to me why some guys have a 2800+ "downgraded" from a 2500+ ?
And 1700+ downgraded from a 2500+ (Barton!!).

The extract from the AMDBoard guide may help to explain that;
"In General, assuming that:
24 = 2400+ 2.0GHz 266FSB
25 = 2500+ (should be a 2.0GHz Thoroughbred with 266FSB but it does not exist!!)
26 = 2600+ 2.083GHz 333FSB or 2.13GHz 266FSB
27 = 2700+ 2.17GHz 333FSB
28 = 2800+ 2.25GHz 333FSB
30 = 3000+ but not exist!!!

Apparently, these digits in the marking of AMD Athlon XP “Thorougbred� CPUs that describe the speed this particular CPU was designed to work at but was later “downgraded� or maybe even “upgraded� in certain cases...."

So by the guide, (which makes perfect sense since some cores are "reprocessed.") they can be Upgraded, as well as Downgraded.

EDIT: The 2500 was a typo, and was meant to read 2.0GHz

[Kurrurrin]

just as a note, they are making AIUHB tbreds. I just got one from newegg, so I would know

[Fewture]

I emailed the maker of the script

Hi,

The "Downgraded from" rating was based on observations by a malaysian overclocker. The information was NOT official and therefore not 100% guaranteed that it was true (I stated that when the script started). Recently the same person disproved that fact, according to an unofficial source from the Malaysian fab of AMD, the two characters that were believed to be "Downgraded from" rating are bing used as ID for some sort of chip tracking.

Therefore, when the last update of the identifier script was made, the "Downgraded from" was completely removed form the data returned about the given CPU.

Wednesday, April 09, 2003, 1:50:12 PM, you wrote:

SHPV> Hello,

SHPV> The cpu id script you made is causing some confusion. How did you
SHPV> get the info on the cpu's? I ask you this because the script is
SHPV> not consistent. According to the script there are 2800+ cpu's
SHPV> being downgraded from a 2500+. And 1700+ TbredB being downgraded
SHPV> from a 2500+ Barton..??

SHPV> Best regards,

SHPV> Sandor / Fewture
SHPV> Moderator GoT/DOC


--
Best regards,
Anton B.
http://www.bloodys.com/
info@bloodys.com

As I've said many times..the script is a load of crap..

[Liquid3D]

I should have known it was too good to be true. But then again, the statistical data (from two guides) is coincidentally accurate. I wouldn't be surprised if the "person" in question had to retract his/her statement in order to keep their job. I realize that seems quite nefarious, but stranger things have happened. In so far as the Barton is concerned changing the FSB isn't an issue, if however; the cache were different, it would obviously be impossible. Regardless the practice of downgrading IS an industry FACT, as it's much more cost effective to downgrade, then to "re-tool" simply to fill orders for lower clocked CPU's. Especially when there's an abundance of higher speed cores available. If you know the industry, you know it's more costly to attempt to produce many different "models" or speeds within the confines of the tight increments between models, then it is to simply mass-produce many higher clocked cores and "downgrade" them accordingly. Here's links to the two guides I extrapolated the info from (I don't purport it to be beyond challange); http://lowyat.isentral.com/guides/athlonxp/ and the following from Xbit; http://www.xbitlabs.com/news/cpu/dis...322034857.html

By the way, I can certainly see why AMD would not want such information to become public knowledge. It would obviously cost them millions of dollars, as it's already done since so many are fixating on the 1700. Why purchase a 2800, or a Barton if one can reach 2.6GHz with a $60 CPU. I'm sure that person retracted his/her original, and unsanctioned comment/s.

[afbrat]

Thanks to the thread starter "liquid3d" this is exactly the kind of info we need

[theonetheycall]

wow this thread is more than a month old.. and just to add, the 24, 25 , 26, etc.. theory has been dimissed

even so, its still an interesting read, Liquid3D done a lot of good research on steppings for us

[Tettejaak]

I did saw that some months ago, but i still didn't managed to check my barton. This is my core: axda2500kv4d f445208270100
aquca0306upaw. What am i doing wrong, the thing says: you've entered wrong data? Anyway, is doesn't really matters, but it would be nice to know.

[Liquid3D]

Originally posted by Tettejaak
I did saw that some months ago, but i still didn't managed to check my barton. This is my core: axda2500kv4d f445208270100
aquca0306upaw. What am i doing wrong, the thing says: you've entered wrong data? Anyway, is doesn't really matters, but it would be nice to know.

It's not able to detect Barton's. Thank you very much guys, but the credit really belong to LowYat.net, bloodys.com, and XbitLabs.com. I do appreciate the mention though, thanks. I'm writing a revision of this article at Madshrimps, which actually infers the 8th & 9th letter code ARE valid, after someone ast LowYat spoke to their contact in the AMD fab. In fact I'm willing to bet your Barton is an excellent overclocker, isn't it?

[antipop]

Congrat Liquid3D looks like you made some nice search job

[JeffPH]

hehe now i know my specs

[Tettejaak]

In fact I'm willing to bet your Barton is an excellent overclocker, isn't it?

Well, quite good i think He's running @10,5x230 so 2420mhz. But i'm going to try at a higher multiplier, i think the geil ram is holding me back. I'll have to do a Vmod so i can give it 3V. prolly also the chipset, but do you guys have any suggestions on a better nb cooler? Now it's the original abit cooler used with as3. Would a zalman passive nb cooler be better? Or perhaps i'll just wait about a week and try my brother's board. I've ordered him an abit nf7-s cause his old azza mobo doesn't works any more. Perhaps his board wil be better, and he doesn't needs the best board, he's using a palmino and has only pc2700 ram.

[antipop]

Originally posted by Liquid3D
The extract from the AMDBoard guide may help to explain that;
"In General, assuming that:
24 = 2400+ 2.0GHz 266FSB
25 = 2500+ (should be a 2.5GHz Thoroughbred with 266FSB but it does not exist!!)
26 = 2600+ 2.083GHz 333FSB or 2.13GHz 266FSB
27 = 2700+ 2.17GHz 333FSB
28 = 2800+ 2.25GHz 333FSB
30 = 3000+ but not exist!!!

Apparently, these digits in the marking of AMD Athlon XP “Thorougbred� CPUs that describe the speed this particular CPU was designed to work at but was later “downgraded� or maybe even “upgraded� in certain cases...."

So by the guide, (which makes perfect sense since some cores are "reprocessed.") they can be Upgraded, as well as Downgraded.

I think that whole thing about those digit is nonsense! I mean why would AMD test their cpu for 2500+ when they don't sell those(talking about tbred)? Is it for fun? How would they upgrade a chip? IMHO they test the chip at the highest clock and then go downward from it, so there is no upgrade!

One last question, it's about the place on the wafer, you say that A and J are the closest to the center of the wafer, but i've never seen a Tbred B mark with a K or an R, what do they do with the chip on the border? Do they just put them in the bin?

[Liquid3D]

antipop I don't claim to know everything about the inner workings of AMD's Fab labeling system, do you? If not I don't think you should dismiss someones theory as "nonsense" unless you have empirical evidence to contradict it. How do you know when they started production they didn't plan on introducing a 2500 model? I don't. I do know Austin's (Lowyat.net) "contact" from AMD fab Singapore stated this was an ID system, not necessarily an original PR system. Therefore the 25, could be an estimation of clock speed based upon AMD's PR? There were questions being tossed around over there about the language barrier, and perhaps what this employee describes as a "laser ID chip tracing system" could be interpreted any number of ways. It's therefore not only feasable, but plausible the 8Th and 9TH digits in 2nd line of code, could refer to the AMD PR system. Does that system have to coinside only with CPU's released to the public? I don't see why it should if it's for internal reference only. Of course it could be mere coincidence that just about every JIUHB TBred-B0 released has been capable of attatining 2200MHz at default voltage, which happens to be just higher then AMD's fastest PR released to date? But that be one hell of a coincidence? Perhaps the following update to the original guide may strengthen my position which itself is based on full codes of over 43 JIUHB 1700's (much of my research was based on Xtremesys TBred-B's);
Update:
These 2 digits are the laser marker ID and used for chip tracing according to a friend who works in AMD Singapore Fabs. But we don’t quite understand the actual meaning of “chip tracing� as from what we know, the earlier production of Athlon like Thunderbird, Palomino do not have these extra 2 digits.
Maybe those CPU with one particular code for example 24 was originally designated to be made for 2400+ model.
But due to failure of achieving that speed, it will be remarked to lower model. So whether that CPU can hit 2400+ or not with default voltage doesn't help...because it has a great chance that won't run at that speed with default voltage at all..
We should treat these 2 digits as for reference only, because not 100% of the 28 chips can run at 2800+ !!
Bear in mind that the maturity and purity of wafer also affect a lot in performance. Coding will be the 2nd consideration after confirming which core is better by your eyes... o your 6th sense..
Anyway to be safe, you should pick one Thoroughbred with 25,26,27 code instead of 24 and 28 because we have no idea which is really the best one. Judging by our experiences after reviewing more than 50pcs Thoroughbred B 1700+ with different batch and production timeframe, code 25, 26, and 27 have less “disappointment�.
From the view of statistic:
Sigma Square=difference with the average value
Mean: Average Value
Hence I have to assume that:-
IF Sigma Square(the smallest code 24, OR the biggest code 28), then disappointment OR satisfactory will be higher.
IF code = OR near to mean(25,26,and 27), the chances to get a good chip are higher, IF can get unexceptional best or worst result, you still won’t lose your confidence to buy another one again…
Since there are too many Thoroughbred 1700+ CPU out in the market, we can only give our assumption based on sampling and average from own experiences and others personal reviews, so please, don’t flame others or us if your 28 can run at super duper speed!!

EDIT: the 2500 = definition was a typo, and was meant to read 2.0GHz.
In so far as the absence of "- K - R " steppings. It's entirely feasable these cores were binned, maybe discarded. In fact my hypothesis is this is the "missing link". I purport these outer parameter cores, were intended to fill the orders for all "lower speed" Tbreds. Any number of fabrication anomalies could have resulted in their destruction. There are repetitive steps in the photolithographic process, in which ulrtaviolet imaging and etching repeat to form the core's three dimensional microcircuitry. Any number of accidents, undetectable until actual voltage tests could be run, would result in discarded batches. In fact given the consistancy only the outer core steppings are abscent, this may be evidnce of lense misalignment. You probably know this, but many may not know the meaning of the .13 micron designation. This measurement describes the "gate width" or width across the pathways etched (imaged) in the microchips core, along which electrons travel. The smaller the gate width, the faster the speed, and lower the voltages the processor can operate at (although amperes increase). As the die process shrinks the chance for error grows.

Why then would almost every Tbred-B 1700, and 2100 sold to date be capable of 40% to 105% overclocks? Yet none of the higher PR models are attaining anything like this. In fact given the Austin's TBred guide, the speeds which the higher PR Tbreds are overclocking corroborate almost perfectly? If AMD has been able to produce, and reproduce consistantly processors capable of 40% - 105% overclocks on lower speed TBred's, why can't they reproduce this formula with the higher PR's? I maintain the reason is clearly they didn't intend to sell these cores as remarked lower speeds. The speeds consistantly attained with 1700, and 2100's are perfectly suited to meet the PR of the highest 256K cache 333FSB TBred's available. Albeit not the best overclocking processor ever released, a company doesn't profit selling CPU's which cater to that 1% of total sales, the overclocker. They profit selling CPU's which meet carefully (cost controlled) researched manufacturing spec's.

[afbrat]

is there a guide like this for the 333 series 2500?
I have one and would like to see info like this for it or can I use this info for mine too??
I ask because my second line has 14 digits not 13
AXDA25000KV4D
95073555260798
AQXDA 0313MPMW
according to your info this was originally designated a 2600 but was recombined thus the M... although the "A" in the 3rd line means it is good quality??
I have had this one up to 2200 MHZ @ 200 X 11

all of the 2500s are 166FSB (333)

[antipop]

Originally posted by Liquid3D
[b]antipop I don't claim to know everything about the inner workings of AMD's Fab labeling system, do you? If not I don't think you should dismiss someones theory as "nonsense" unless you have empirical evidence to contradict it.

I might have wrongly exprimed myself, you might be right about the 2 digits as well as you might be wrong! It's just an hypothesis, i admit that there's strong proof on both side of the question and we can discuss that matter for a long time!
I know that you don't know everything about how amd names its cpu. But i was wondering some question out loud. I know that a fabric of cpu cost a lot of investment and last a short time. So you have to get your investement back in the shortest time possible! That's why it would sound logical that amd would produced as much cpu as available one a single wafer. And they would also use the border, if the lens was misaligned don't you think they would have pointed it at the beggining and corrected it?