Tabula Rasa Semantics, in Microprocessor Burn-in. Part-I

CPU by KeithSuppe @ 2003-06-26

In my last article; "AMD Ingots, sliced TBread, with the crusts cut off" I proposed a hypothesis which purported to explain the propensity for "low speed" Thoroughbred-B's (1700 - 2100) to attain extraordinarily high clock-speeds. Athlon 1700's were achieving speeds as high as 2.7GHz on air cooling and minimal voltages. And in a few cases the coveted "double overclock" was achieved via LN2, and phase-change. Unfortunately, several of the premises on which my theory rested were incorrect. Here is correct version based on true facts, many hours and days of research has gone into this article. I hope you like it!

  • next

Introduction

Madshrimps (c)
3 part Article!
Part I - Part II - Part III


In my last article; "AMD Ingots, sliced TBread, with the crusts cut off" I proposed a hypothesis which purported to explain the propensity for "low speed" Thoroughbred-B's (1700 - 2100) to attain extraordinarily high clock-speeds. Athlon 1700's were achieving speeds as high as 2.7GHz on air cooling and minimal voltages. And in a few cases the coveted "double overclock" was achieved via LN2, and phase-change. Unfortunately, several of the premises on which my theory rested were incorrect.


First off, I surmized the reason for such extraordinary overclocks among "low speed" Tbred's (1700 - 2100) was due to the fact, these models were based on cores originally slated for higher speed Thoroughbreds. And although there was some validity to the assertion, its basic premise was inaccurate. I'd incorrectly hypothesized, the phenomenon was due to error's in the lithographic process (possibly lens misalignment), and that AMD had binned as "useless" many of the cores from the outer edges of the wafer. The supposed qualitative difference among inner, and outer cores from edge of the wafer, was my inspiration for the "...crusts cut off" portion of the title. After careful reconsideration I'd realized the odds of such a manufacturing catastrophy (although possible) were highly improbable. Much of my evidence for the "center wafer purity", was based on the LowYat.net "Guide to choosing an Athlon XP Thoroughbred Revision B" written by Austin (among others); which purported that outer area of the wafer produced cores, of lesser quality then the inner areas:

From our personal findings, the purity of the centre wafer will determine the high end model of a CPU. Higher end models are then called an A Thoroughbred while the lower end one is called the J Thoroughbred. AMD even had produced R Thoroughbred and K Thoroughbred for Revision A. But so far they do not exist in Revision B...For the time being we should get a JIUHB with later production date because they have a greater chance to be overclocked compared to AIUHB's with earlier production date...In General: A > J > K > R



In the months preceding the publication of the LowYat "Guide...", I'd been posting as to it's validity, and recording data from the overclocking results of TBred-B 1700, 2100 owners. The data seemed to indicate many of the assertions contained therein were empirically verifiable. I then came across an article in Overclockers.com entiled; "How to tell Them Apart", by Ed Stroglio (now edited five times); which identified two types of Tbred,s. "High end" TBreds which begin with the A-stepping, and "Low end" TBreds which begin with the J-stepping. Of course being fairly new to overclocking (I consider just over a year new) I thought it best to go by the "experts" opinion. And even wrote my article, extrapolating from their published opinions.


Another section from the Lowyat Guide, which was it's most controversial, discussed the 8th, and 9th digits in the second line of code (from the CPUID sticker) designated the processor model (or PR) from which the CPU was originally marked, or the projected speed at which it would operate. In conjunction with the Lowyat Guide, a siten by Bloodys.com released the "AthlonXP/Thunderbird processor Identifier." A software program which identified Tbred's (and other processors) which also referenced the 8th, and 9th digits in the 2nd line of code stating this was the model it was originally remarked from. Given the number of articles, programs, and empirical data I'd gathered by scanning Hardware Forums, and questioning Tbred owners, my theories were formed around these.


Perhaps my most controversial and wildly inaccurate assertion was in proclaiming the reason AMD had relabled higher-end TBred-B0's, was due to a lense misalignment in their Lithographic process. I had extrapolated this from a culmination of sources. First my ignorance of the stringent controls on the Lithogrphic process, and it's overall complexity, led me to believe there could be no other logical reason why Lower-sped Tbre'd would obtain higher Overclocks, then higher speed TBreds. And from an aspect in the Lowyat.net's; "Guide to choosing an Athlon XP Thoroughbred revision B" where it mentions four steppings A-J-K-R, in relation to their placement on the wafer, and how this effects quality.
Finally my discussing the "Center Wafer Theory" (as I've labled it) at Xtremesys, where it was brought to my attention no one had seen any K-R steppings.
This was the straw which broke my intellectual back, and all reason went out the window, as I'd thought this was the missing link which explained the replacing of low end TBred steppings with the higher end models. Whether or not these steppings exist, is not in question, because the assumption AMD's lithography process was so fraught with error as to eliminate an entire series of microchips, was non compus mentis.
  • next