True "merits" of high SP rating?

[Fast_N_Curious]

After a bit of research, I can see the SP score is an automated indication of potential silicon quality based solely on the voltage table, and it is USUALLY a good indicator, but it is still far from the final word on the matter. I've seen others reporting that supposedly lower-scored chips actually fared better overall.

Like, when der8auer did his 12900KS review and he compared several chips on their max stable core clock, a mere SP80 12900K did 5.2ghz @ 1.274V, while a SP88 12900KS only did the same 5.2ghz @ 1.282V.

It is my determination that SP is merely there for voltage efficiency and doesn't factor in processor scaling, hence, should not be used as a performance marker. Sure, the voltage my vary a bit between processor cores but not significantly enough to affect your overclock ceiling.

Anyone else buy into this argument or am I way out in left field?

[YoungBlade]

All of these silicon scores, like SP or the old "ASIC quality" reported by GPU-Z for some GPUs, are at best loose guidelines and at worse, next to useless. The only way to actually know the quality of silicon is to test it to death until you've attacked it from every angle to see where its strengths and weaknesses are.

 

For example, I have an old RX 580 4GB that OCs like garbage - RAM can't even get to 2000 MHz and the core can't hit 1400 MHz with stock voltages. Definitely below average. However, it actually undervolts at -60mv while maintaining the factory OC of 1386 MHz, which isn't that bad given that is technically an OC+UV. There's no way that one number is going to tell you that that sort of configuration of voltage and frequency is possible. You've just got to do trial and error until you find something that works well for you.

[Fast_N_Curious]

1 hour ago, YoungBlade said:

All of these silicon scores, like SP or the old "ASIC quality" reported by GPU-Z for some GPUs, are at best loose guidelines and at worse, next to useless. The only way to actually know the quality of silicon is to test it to death until you've attacked it from every angle to see where its strengths and weaknesses are.

 

For example, I have an old RX 580 4GB that OCs like garbage - RAM can't even get to 2000 MHz and the core can't hit 1400 MHz with stock voltages. Definitely below average. However, it actually undervolts at -60mv while maintaining the factory OC of 1386 MHz, which isn't that bad given that is technically an OC+UV. There's no way that one number is going to tell you that that sort of configuration of voltage and frequency is possible. You've just got to do trial and error until you find something that works well for you.

Thank you. I was referring to the SP rating on later gen Intel processors, I should have pointed that out. However, I assume the theory stands? 

[YoungBlade]

44 minutes ago, Fast_N_Curious said:

Thank you. I was referring to the SP rating on later gen Intel processors, I should have pointed that out. However, I assume the theory stands? 

I was just using "ASIC quality" as another example of a number that has dubious value - it's not even used anymore, because people were assigning it more value than it was worth, so it stopped being reported. I don't think that SP really means much. Maybe it has some limited value, but it can't replace actually testing overclocks/undervolts directly.

[Fendrick]

ASIC quality on an old GTX 480 was like 47% yet it could run 960mhz core clock stable on air.

 

ASIC never meant anything.