PDA

View Full Version : Can recent TBred's be "ID' as to their overclocking potential?


Liquid3D
04-14-2003, 08:17 AM
I started this thread to investigate the validity of the following program from Bloodys.com; http://mp3.zonebg.com/cpu/cpu.php
Of course this in itself wouldn't warrant such a thread, and it was my subsequent reading of the following articles from two sites I hold in high esteem which then motivated me to post; http://lowyat.isentral.com/guides/athlonxp/ and; http://www.xbitlabs.com/news/cpu/di...0322034857.html

This pertains to recent to .13 micron core's in Tbred-B's bearing a PR at or below model XP 2100, and more specifically XP 1700's. Whether or not the 8th and 9th numbers in the 2nd line of code actually denote the original model from which your CPU (or any relabled lower speed TBred) was downgraded from, is thought to be a fallacy. There may be those who wish to quash this system of AMD code interpretation, claiming it's completely baseless, and simply "rumored" to be true.

Where I became slightly confounded is in the accuracy of the data. The pure number of exceptional overclock's prima facie corroborate the theory? Emprical evidence is but the first step in any sound argument. So I shall begin with a Hypothetical Syllogism;
Premise A) If 1700 TBred-B, then high overclock.
Premise B) If high overclock, then 8th, 9th places in 2nd line of code are 26 or greater.
Conclusion) If 1700 TBred-B, then 8th, 9th places in 2nd line of code are 26 or greater.
The argument is valid, whether the premise's are true will most likely wring out this thread. Onto the rest of the ID process.
The first letter of the stepping (which is the first place in the 3rd line of code) supposedly denotes the proximity of the core to the wafer center, A being the closest, then J, K, and R being the furthest from center (highlighted below).
AXDA2100DUT3C 9397647280017 AIUHB0301SP1W 1999 AMD Recently TBred-B's have either been A or J. And certainly one would concur the 3rd alphabetic (or number 1-9) also in the third line of code (following the date) has validity. A being the first production batch, for that production run, B being second, until C, and M denoting combined batches. If there is a number in that place 1-9 then this denotes re-combined batches which have falied Q&C, and been re-processed, as in the example above.
I've yet to see a Tbred-B 1700, or 2100 with an A, B, or C in the third alphabetic (following the date) of the third line of code. My 2400 has a B in that position, and my friend's 2600 has an A there. Most lower speed's TBred lately have M denoting a combined batch. In so far as the 2nd line of code, the last four digits (following the magic 8, 9th) are indicative of the CPU number in the batch, lower being better. Such that in the example above, the processor is 17th from that re-processed, re-combined, production batch. Perhaps 90% of this "code" is true except the significance of the 8th, and 9th places in the second line of code? However...

Over at ExcaliperPC they sell TBred-B's based on stepping, and date, and have included the following disclaimer; "please also do not ask us to look for 27 or 28 in the last 5 digits. We see that is just a rumor and we don't support it" regardless they got it wrong. It's the last six digits, but that's beside the point. This can stir more suspicion, then it deters. If they said nothing it be one thing, but any opportunity to sell a chip a vendor can take advantage of, they usually do. Perhaps AMD released an internal memo, requiring reseller's not to release, or advertise based on this "code"? Most of these relabled TBred's 8th, and 9th numbers are 26,27,28, even 29, and 30. So why not read the dang numbers? It may be as simple as time, althoug to estimate the stepping, you would have to read most. Anyone buying a Tbred-B 1700 or 2100 is going to consider a 2600 a great find? And I'm sure many will agree, the reason these 1700 JIUHB etc. are all reaching 2.3GHz to 2.7GHz is because they are in fact relabled from the wafer slated for higher speed TBred production. I don't need to iterate here it's common practice, and cost effective (given certain conditions) to relabel.
I would think it's logical for the person whom originally released this "inside info" to retract their statement. Because just as these relabled Tbred's are going to be short lived, so is that person's career. It's not that AMD would fear everyone rushing to buy TBred-B 1700's instead of $300 2800's or better, because the overclocker market is so miniscule perhaps 1% of total revenue. And I don't belive it's AMD's fear, unscrupulous dealers will "remark" the CPU's selling them as 2600's (although it's feasable). I purport the cost to AMD would come in the guise of Industrial Psychology. The Lion's share of AMD sales comes from OEM vendors, and such. A market where security, dependability, and reliability are the attributes which are most sought after. And businesses will pay more for it, ss is evident in Intel's earnings. And with Intel putting into place yet another security feature (patent 6,535,988 detects FSB overclocking, slowing the processor) they become even more attractive.
So if anyone has anything they would like add I'd be grateful, and I'd be especially appreciative if any TBred owners of late (past four months) could submit their entire code no matter the model. Thank you.

dropadrop
04-14-2003, 09:10 AM
The reason excalibur stopped posting the last digits of the code, was that even the person who made the script admitted it doesn't have anything to do with being downgraded from some certain speed. Still, his script had allready spread around, and everyone and his dog where looking for processors with 28 as the "magic digits". If you look at the results, those numbers don't really affect the scaling of the cpu's.

Liquid3D
04-15-2003, 08:44 AM
That's the conundrum, the data does support the theory. That's why I posted this, in hopes more people would be willing to post the entire code of their CPU, and the overclocked speed. I know the person retracted his claim, that's what it says in my post, however; of the 50 or so results I've glimpsed here and there across Hardware Forums, every overclock is indicative of the "relabled" code, or 8th, and 9th places in 2nd line. The only other disproof besides the person retracting his/her story (which I've theororized above) is that the CPU's are overclocking beyond their rated spec's, and that's the point of overclocking, finding the manufacturers true "ceiling". Irregardless, recent lower-speed TBred's are/were cores intended for much higher PR models, as their overclocking prowess would indicate. And as you must know the practice is quite common, as it's cost effective in many (in THIS case) cases, to simply repeat the lithographic process on an exceptional wafer batch. I doubt any Manufacturer would think "Oh we better get some lower quality wafers now, to fill the few hundred thousand lower speed CPU's we need." That's just not cost effective. If you've struck on an exceptional core yeild, it would simply be repeated until superflous, or redundant. Either way the 2.3GHz average clock speed of these 1700/2100 JIUHB/AIUHB (respectively) exceeds anything AMD is currently making in 256k cache. Even the upcoming 3000+/400FSB, with 512k cache is only clocked at 2200MHz.
Of course I'm simply speculating, and I know when speculating, I make a spec of u, and lating.

GVCryan
04-15-2003, 06:58 PM
it is a neat theory, but I tested his script with several chips, and a TbredA (week 0241)came up as downgraded from a 2700 (can't be done).

2 tbredb's came downgraded from 2500+'s.

1 TbredB came from a 3000+ batch(oddly enough the worst clocker (just over 2.3 at 2.1 volts)

1 TbredB came from a 2900+ (doesn't exist)

Also, if you look around the internet, you will find that the 2500 and 2600 downgrades are basically the best.

Another interseting point, is the default voltages. None of these higher chips are 1.5v, and it is a fairly large amount of work to change the default voltage (permantently). Not just remarking.

spaceman
04-15-2003, 07:26 PM
Has anyone looked harder into the theory, that the DLT3C 1700's might be remarked mobile chips? That would explain the lower vcore, that's for sure. I tried to download the white papers for AMD's mobile chips the other day, to try to figure it out, but they are only available on a CD now. :rolleyes: I wasn't THAT interested, or maybe I was just being cheap. ;)

Liquid3D
04-15-2003, 09:46 PM
In response to GVCryan's findings (and with all due respect), my intent was to label this post "Can recent TBred-B's be ID..." and in the introductory paragragh I do specify TBred-B's. In so far as the contradictory markings, the author presents (in my opinion) an acceptable (albeit incomplete) explanation;
"Believe it or not, those CPU with "26", "27" or even "28" tested to be stable at that particular speed with default voltage 1.65v, or even 1.525v which is fantastic !! That's why they run much cooler at extreme voltages, and can be overclocked to high speeds up to 3000+ (2.4GHz).

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...."
It's true no 2500 exists, however they went from 2600 to 2700, so their propensity for keeping PR on an even model number basis with the TBred core, hasn't been adhered to. Why not a 2500 TBred? Thre's no reason a change couldn't have been made from a pure marketing perspective. With the upcoming relase of Barton (at that time), as not to confound the PR concept acroos two product lines?
And in further support of the cost issue X-bit makes the following assesment;
"Looking at the situation from AMD’s standpoint we can assume that the yield of higher-end Thoroughbred B0 chips is quite good and the company should have no problems with lowering the prices on these products. On the other hand, we see that the company needs to remark the fast CPUs to fulfill the demand on cheaper models. It is not good for the company, I assume, since it may mean that AMD has problems with selling the fast chips rapidly, but still hopes for increase of demand on such microprocessors later and does not want to slash the prices on them now. On the other hand, if the company substantially cuts the prices on higher-end models, it has to sell the slower processors at extremely low prices what may cause losses.
I disagree with the concept it's "...not good for the company..." as I feel it's always cost effective to remain with the same facility, repeating the same conditions if their yeilding the exceptional results overclockers have been benefitting from. However, I also wanted to reiterate my theory was based on articles, and performance of Tbred-B cores only. Both quotes above were taken from the articles linked to in the original post.