CODES (abridged)

Gautam: said:

Austin was where I first got this info from, btw. What first gave me a heads up about the third letter being significant were the GameVe DLT3C's. Unlike the ones from Excal and SVC, these were nothing spectacular at all. They did about as well, if not worse as a DUT3C, maxing out at 2.2-2.3 on air, 200mhz or more below the DLT3C's. The only thing that looked odd in their stepping codes were the third letter- TPXW. I have seen TPMW's, and they do just as well as any other DLT3C. But these X's were clearly worse. There was a huge thread on amdmb.com about them, and all of them did rather poorly in comparison to other DLT3C's.

One of the best Barton steppings was the AQUCA XPAW. Most that you see these days are AQXCA XPMW's, and they do not seem to do as well. IF you look on the Vr-zone lists of the best overclockers of both Bartons and T-Breds, the front pages are dominated by processors of several stock speeds, but almost all having their third letter as A's, B's and C's. I'll do some more searching, but for now that's the best I've got.

Basically, I've been hunting down the 4 letter codes, and I can't draw any conclusions about any of the letters in it from own findings except for the third letters. TPMW's, UPMW's, VPMW's, WPMW's and XPMW's all perform the same, but an XPAW's seem to be better than XPMW's within the same processor type. All 1700+'s are xPMW's, so we can't say anything for sure about them, but within Barton 2500+'s, xPAW's appear to be better than xPMW's. The difference between A and M appears to be negligible at most, though, but the difference between an M and an S or X is rather large.

Click to expand...

Gautam said:

The reason that the 1700+'s are M's is becayse they are lower in wafer integrity. Wafer integrity does not always equal overclockability. In the 1700+'s case, the lower quality, for example, translates into its "dislike" for high voltages. Hitechjb1 made a post supporting this; I'll dig it up.

Click to expand...

OC Detective said:

Gautam, your mention of voltage set off a trigger in my head and I went and scoured the vr-zone database to come up with an alternative theory to what this third letter means, this involved recording data on cpu type, manufacture date and "third letter" of over 80 Tbreds and 25 Bartons (who said get a life OCD?).
My belief is not so much a dislike for higher voltages is exhibited in higher alpha's such as "M" rather that they can operate at default speed at a LOWER core voltage.
My rationale for this is because my research shows that those cpus assigned lower alphas such as A, B and C are typically from higher speed cpus such as XP2400 and above and older XP2200 Tbred A's. When it comes to XP2100's again older models have lower values such as A whereas newer ones from 2003 exhibit letters such as M, N and R. With the XP1700's virtually all of them (the DLT3C's anyway) this year have been M's however last year it was a mixture of M's and lower letters such as D. The DUT3C's XP1700's seem to be mostly lower letters such as A although there are some M's which indicate to me that AMD realised that they could run at lower voltage hence the introduction of the DLT3C!!.
With the Bartons it is even more clear cut the earlier models up to around 0310 have lower alphas whereas those from 0312 exhibit the letter M or higher. This would indicate to me not that wafer integrity is decreasing but that AMD have improved the process as production matures and the voltage needed to run at default has reduced. Indeed perhaps AMD will introduce Bartons to run a lower voltages in the near future lol.
The glaring problem with this is why does the vr-zone have the A,B's and C ranked higher? - thats simple it because the people who are early adopters are also the pioneers who push the envelope with extreme cooling techniques and want to test the newest products straight away.
The best evaluation would be to look at the middle slice of results and compare air cooling figures only and you will see a clearer picture.

Click to expand...

Gautam said:

Okay, I think I've had a eureka moment here. I've looked over my previous evidence, and I think I've got an explanation that all makes sense.

http://www.amdforums.com/showthread...ve&pagenumber=8

If you look closely, you'll notice that these do very well at below 1.7v or so, but once you get higher, things begin to fall apart. Several needed 1.8v to reach only 2.3ghz. At lower voltages, they do better, but begin to do progressively worse at higher ones. So, at below 1.7v, the TPXW's do as well, if not better than any, but at higher ones, do a lot worse than most. My friend has one of these, in fact, and an SK6+. It can't even do 2.3ghz, no matter what. If it were an M or higher, then it would have been able to clock higher as more voltage was put in. The diminishing returns were too heavy on this one; it can't take the higher temperatures caused by the voltage. Now, my UPMW, which I have tested under an SK6+ myself went about 100mhz higher than his with identical cooling and voltage; 2369mhz.

This could explain this:

Originally posted by hitechjb1:
Theoretically, the 1700+ DLT3C's have certain properties that can take lower Vcore to do same CPU frequency as other higher Vcore rated ones. Its max stable Vcore would be lower than the 2100+ and 2500+ by about 100-150 mV. It is due to process variation giving transistors of lower threshold voltage (probably you call it "lower" wafer quality). There are good and bad consequence to it. It can run faster at lower Vcore (about 100 mV lower), but would create more heat due to leakage at same higher Vcore, which stops it from going further (due to diminishing return on frequency at around 50C) at higher Vcore like other siblings. Actually, if these 1700+ DLT3C are cooled well at very low temperature that heat due to leakage becomes non-issue, it should perform equally well or maybe even better than the 2100+ and 2500+ due to low threshold, faster transistors (my conjecture, not proved). Eventually, for a given chip, the eventual outcome depends on the tradeoff between the lower threshold (faster) and the leakage (heat), and how the chip was born at silicon, and have to test the chip/stepping on a case by case.

Click to expand...

So as we can clearly see, having higher wafer quality does not always equate to higher overclockability. In some cases it can, in some cases it can hurt. OC Detective hypothesizes that those with lower third letters need a higher stock voltage. If they need higher voltages to sustain lower speeds, then in theory, they should end up being able to sustain higher voltages overall, and perhaps, end up ahead of those with lower third letters in some cases. To take this a step further, this can potentially explain why the M's that are most common do so well. They are a sort of compromise between needing high voltages and not being able to sustain them. With this in mind, it makes perfect sense that these are the overclocker's choice. But are they the choice for all overclocks? The highest overclocks in the VR-Zone database have lower third letters across the board. They all use extreme forms of cooling in which a heavily overvolted processor can thrive, provided that in can sustain these voltages. So when pushed to the limits, these higher wafer integrity processors with lower letters do the best, but with lower forms of cooling can potentially do the worst. With the mid-level cooling that most of us have, the middle is what works the best. I may be thinking far out, but you have to admit, it does make plenty of sense .

Click to expand...

Originally posted by hitechjb1
Why the 1700+ can run so fast at low Vcore?

The Tbred B 1700+ DLT3C is based on the same 0.13 micron bulk silicon process as all the other model 8 (Tbred A and Tbred B) from XP 1600+ to 2800+ (recently 3000+). (BTW, Tbred B has one more metal layer than Tbred A, both are 0.13u.)

The hammers (Opteron, Athlon 64) are based on 0.13 mircro SOI process, will go to 0.09 mircro eventually.

The Tbred B 1700/1800+ have the same transistor count, same L1, L2 cache size, same number of metal layers, same chip dimensions, ... as the 1.6 and 1.65 V rated Tbred B.

Side track: Same for Barton, which is also based on the 0.13 micron process. But it is a different chip, different transistor counts, chip dimension and has bigger L2 cache of 512KB instead of 256KB in Tbred.

I think the reason why the Tbred B 1700+ DLT3C can work at rated 1.5V and can be clocked at simliar highest clock frequency (if not better) as all the other 1.6V and 1.65V rated Tbred B is due to the following:

Its transistors have lower threshold characteristics due to process variation which produces transistors with shorter channel length. Shorter channel means lower transistor threshold, runs faster, draws larger leakage current and higher active current (hence higher active power). According to AMD spec, the 1.5V 1700+ has higher rated current than the 1.6V 1700+ (about 7% more).

Threshold voltage of a transistor is the gate voltage above which the transistor will conduct current orders of magnitude higher from source to drain compared to that below the threshold. Chips with lower threshold transistors can perform equally well with a lower supply voltage (Vcore) as those with higher threshold, because the transistors can conduct at a lower gate voltage.

This is normal for a given silicon process (say 0.13u) to have such variation that some transistors in certain chip die have shorter channel length (less than 0.13u) or some have longer channel length. Those that have shorter channel length have faster intrinsic speed and can run as fast when smaller Vcore is applied (pros). On the other hand (cons), due to the lower threshold voltage which draws higher leakage current and generates more heat at the same higher Vcore, these chips can run as fast at a low Vcore as the higher Vcore rate chips, but they will max out at a lower Vcore compared to the higher Vcore rated siblings.

The 1700+ has a run-away current at a lower Vcore compared to the 2100+. Run-away current refers to the leakage current and the heat generated positively feeding each other resulting instability.

The final oc success of the Tbred B 1700+/1800+ DLT3C is a race between its natural, born, intrinsic characteristics, the balance and tradeoff between the smaller channel length, lower transistor threshold, hence faster, and the opposing, negative behaviour of higher leakage current, and heat generated.

Click to expand...