Rating your silicon -- art or science?

[cab]

I’ve been running these 2 in identical mobos:

Phenom II X4 965 1023FPMW wafer #1074 Default 17.0x @ 1.400 V
Phenom II X4 B55 1008FPMW wafer #0496 Default 16.0x @ 1.325 V

Based on what people have posted, my B55 will always be better that my 965 because it’s week 8 vs. 23 and because it has a lower wafer # 496 vs. 1074. But the 965 has supposedly lower leakage due to its higher default VID?

Do we have any hard evidence for these assertions or are they just random observations? Would a B55 1008FPMW wafer #1 be a collector’s item now?

[BeepBeep2]

I think most people bet on the newer week (significantly newer) being better. The "wafer #" just tells you where on the wafer the chip came from, 1074 is near outside, 0496 in between center and outside before, it was better to have one closer to the inside, but it seems with latest Bulldozer chips it doesn't matter at all. 16x @ 1.325v and 17x @ 1.400v, you can't really compare because the chips have different multipliers and one is a dual core (555) vs quad core.

No, it won't be a collectors item...I doubt any CPU will be...and what matters most is how they really overclock in practice.

[cab]

They’re both stable at 4GHz, but the B55 runs cooler and needs less voltage, but that could be due to other factors. Is anybody keeping track of week# and wafer# on the best overclocks in the sticky thread above?

What’s the reason why a “newer” week is considered better? The Phenom II X2 560 is newer, does that make it better?

[Daveburt714]

Art or Science?

I think you missed the most important variable, LUCK...

I'm no engineer, but from my experience you never really know until you get a chip and start pushing it.
Even chips with the same week code can vary significantly!

These are just my assumptions (guesses ):

1) Newer week codes may offer better performance, due to improvements in the process (Lots of variables here).
2) Theres probably some variance in the quality of the wafer the chip was imaged on (the silicon itself).
3) The skill of the operator imaging the chip.

Back in the A64 days there did seem to be some advantage to getting a chip with a low serial (wafer)#.
I actually used to put some faith in this metric until I found an 8150 with a serial of 0001 and just HAD to buy it!



That chip was actually faulty and wouldn't even blend at stock clocks!

These are my chips rated from best down:
8120, week 1135, serial# 0219, vid 1.2375. Scales well with voltage, but runs hot (needs Cold)!
8150, week 1146, serial# 0214, vid 1.3125. Temps are manageable but likes v's (good for all-in-one water?)
8120, week 1152, serial# 0052, vid 1.3750. Temps very good, but LOVES voltage (decent air chip?)

For 24/7 (stable) use they all fall in the 4.6-4.7 range with custom ambient water though.

If theres one variable thats actually important, it would seem to be the vid, and you can't know that until you actually install the chip...
Even then alot depends on your cooling.

That's why I say LUCK.

[BeepBeep2]

Because AMD and GloFo often improve upon their process as time goes on.

Phenom II 955BE C3 that is week 0945 will maybe do 4 GHz, one that is week 1045 maybe do 4.4...

For example, there was a time where it was almost impossible to get 7 GHz on Phenom II X4 and most were doing ~6.5, yet chew* purchased something like 5 of them very late in the game and all of them but one did 7-7.2 GHz.

[cab]

Here's a new one, default VID 1.45 Isn't that too high? Idling at 29C, 45C at 4GHz, I'm not pushing more because I actually use it for my work

[BeepBeep2]

Nah. That is the top end of vid and the 2012 chips seem to run extremely cool.

[cab]

Both my 965s run very stable at much lower voltage in all p-states 0.9 - 1.1 - 1.2 - 1.3
I'm doing a minus 0.15V all around with this one. So I don't understand why AMD is overvolting so much. Are they just being very conservative because of mobo voltage inconsistencies?

[FlanK3r]

It is hard to find very good FX... I have 4x FX (2x 8120 and 2x 8150). No one is 8 GHz chip ..Are around 7200-7700 MHz (7700 MHz chip died in action )

[BeepBeep2]

Both my 965s run very stable at much lower voltage in all p-states 0.9 - 1.1 - 1.2 - 1.3
I'm doing a minus 0.15V all around with this one. So I don't understand why AMD is overvolting so much. Are they just being very conservative because of mobo voltage inconsistencies?

Because of the TDP.
These later chips are extremely cool running, leads me to believe that they are not very high leakage chips.


Low leakage chips usually have higher VID.

[imamage]

Both my 965s run very stable at much lower voltage in all p-states 0.9 - 1.1 - 1.2 - 1.3
I'm doing a minus 0.15V all around with this one. So I don't understand why AMD is overvolting so much. Are they just being very conservative because of mobo voltage inconsistencies?

The last Phenom II X4 955BE I got was a late C3 , it clock nicely and hit almost up to 4.9GHz on air (Just for CPU-Z maybe)
http://www.hkepc.com/forum/attachmen...1456385ef3.jpg

I sold it to one of my friend
he have it run at 4GHz @ 1.45V pair with 4GBx4 DDR3-2000 24/7 since Sept 2011 til today (it's still rock solid and he play Battlefield 3 daily)
I kinda missed this nice CPU

[cab]

Because of the TDP.
These later chips are extremely cool running, leads me to believe that they are not very high leakage chips.


Low leakage chips usually have higher VID.

Yeah, low leakage = higher VID

..but can't they set it lower? It runs cooler with less wattage, I guess the average customer doesn't measure wattage.

I'm testing a Core i5-3570K now, we'll see if it's worth the money..