F5F Stay Refreshed Hardware Desktop The score tends to improve with cores operating at reduced speeds...

The score tends to improve with cores operating at reduced speeds...

The score tends to improve with cores operating at reduced speeds...

O
Oskar2502
Member
138
06-13-2023, 09:53 AM
#1
I've been experimenting with overclocks and trying a Per CCD curve optimized versus shaped settings. The optimized version appears to boost max frequencies in most cores, yet it scores about 400 points lower in the multithreaded test. What's behind this difference? Optimized vs Shaped.
O
Oskar2502
06-13-2023, 09:53 AM #1

I've been experimenting with overclocks and trying a Per CCD curve optimized versus shaped settings. The optimized version appears to boost max frequencies in most cores, yet it scores about 400 points lower in the multithreaded test. What's behind this difference? Optimized vs Shaped.

A
AlmightyEag
Posting Freak
785
06-16-2023, 06:19 PM
#2
Cinebench isn't the most reliable benchmark, so a score around 400 is close to a possible error (I've noticed over 1000 points varying in runs of R23). Next, clock stretching occurs on today's CPUs—when frequency spikes too much, the system reduces speed instead of crashing. Unless you're doing extreme manual overclocking, you'll likely encounter this, so focus on performance rather than exact numbers. Also, the monitor you're using may misread clock speeds on modern processors, especially Ryzen models; it often gives incorrect data and isn't dependable. Consider Ryzen Master or HWInfo for better accuracy.
A
AlmightyEag
06-16-2023, 06:19 PM #2

Cinebench isn't the most reliable benchmark, so a score around 400 is close to a possible error (I've noticed over 1000 points varying in runs of R23). Next, clock stretching occurs on today's CPUs—when frequency spikes too much, the system reduces speed instead of crashing. Unless you're doing extreme manual overclocking, you'll likely encounter this, so focus on performance rather than exact numbers. Also, the monitor you're using may misread clock speeds on modern processors, especially Ryzen models; it often gives incorrect data and isn't dependable. Consider Ryzen Master or HWInfo for better accuracy.

T
thomasnyg
Member
58
07-07-2023, 05:29 PM
#3
Consider using a suitable benchmark or testing framework to evaluate performance.
T
thomasnyg
07-07-2023, 05:29 PM #3

Consider using a suitable benchmark or testing framework to evaluate performance.

T
track21
Junior Member
6
07-08-2023, 10:53 PM
#4
Cinebench works well but lacks high precision. Consider other CPU tests like Blender or Y-Cruncher. Choose one that matches your typical tasks, adjust settings, and run several times to smooth out variations.
T
track21
07-08-2023, 10:53 PM #4

Cinebench works well but lacks high precision. Consider other CPU tests like Blender or Y-Cruncher. Choose one that matches your typical tasks, adjust settings, and run several times to smooth out variations.