F5F Stay Refreshed Hardware Desktop Which OC is better for my CPU?

Which OC is better for my CPU?

Which OC is better for my CPU?

Pages (2): 1 2 Next
R
Rubiiix
Member
63
07-15-2021, 01:46 PM
#1
I've been conducting a lot of research lately. My CPU runs at 11900K using an Asus Maximus Hero XIII. I ran Blender BMW for about 19 hours with these settings: Multi: 52 AVX2/AVX512 off, LLC@4, VID 0.01, idle around 1.510 and load around 1.387. Under high power Windows mode, idle stayed at 1.510 while load was 1.394. I noticed some fluctuations and dips in performance. Reading more about this, I'm unsure if higher LLC is always bad—it can depend on voltage levels at idle. The earlier readings seemed consistent with a dip or transient. With the newer settings, I observed slightly lower idle (1.456-1.465) and load (1.394-1.403). If you have more details, feel free to share.
R
Rubiiix
07-15-2021, 01:46 PM #1

I've been conducting a lot of research lately. My CPU runs at 11900K using an Asus Maximus Hero XIII. I ran Blender BMW for about 19 hours with these settings: Multi: 52 AVX2/AVX512 off, LLC@4, VID 0.01, idle around 1.510 and load around 1.387. Under high power Windows mode, idle stayed at 1.510 while load was 1.394. I noticed some fluctuations and dips in performance. Reading more about this, I'm unsure if higher LLC is always bad—it can depend on voltage levels at idle. The earlier readings seemed consistent with a dip or transient. With the newer settings, I observed slightly lower idle (1.456-1.465) and load (1.394-1.403). If you have more details, feel free to share.

E
echigo3210
Member
180
07-16-2021, 12:00 AM
#2
You're missing the bigger picture. The performance relies more on Amp draw than just v-core adjustments. At idle, boosting single and dual cores with a higher v-core is typical, but the Amp pull remains lower than when all cores are active. When both v-core and clock speed drop, you notice the v-core reduction along with slower clocks. This stays within power limits and shouldn't cause thermal issues. An 85°C temperature under moderate voltage variation is ideal. In short, pick the voltage setting that balances speed and efficiency.
E
echigo3210
07-16-2021, 12:00 AM #2

You're missing the bigger picture. The performance relies more on Amp draw than just v-core adjustments. At idle, boosting single and dual cores with a higher v-core is typical, but the Amp pull remains lower than when all cores are active. When both v-core and clock speed drop, you notice the v-core reduction along with slower clocks. This stays within power limits and shouldn't cause thermal issues. An 85°C temperature under moderate voltage variation is ideal. In short, pick the voltage setting that balances speed and efficiency.

W
Wolf_Player24
Junior Member
40
07-16-2021, 01:24 AM
#3
If you visit LLC 8, it might not be ideal at all.
W
Wolf_Player24
07-16-2021, 01:24 AM #3

If you visit LLC 8, it might not be ideal at all.

H
hakukiki
Junior Member
34
07-16-2021, 01:02 PM
#4
Conduct stability tests using minimal LLC and v-core configurations for optimal performance. Prioritize efficiency by monitoring power usage via HWInfor64. Stable operation with reduced wattage indicates better efficiency and temperature control. Aim for a v-core voltage below 1.400V under load. The recent update is promising.
H
hakukiki
07-16-2021, 01:02 PM #4

Conduct stability tests using minimal LLC and v-core configurations for optimal performance. Prioritize efficiency by monitoring power usage via HWInfor64. Stable operation with reduced wattage indicates better efficiency and temperature control. Aim for a v-core voltage below 1.400V under load. The recent update is promising.

G
Goljat12
Member
213
07-18-2021, 11:58 AM
#5
Using the initial configurations, the core voltage stays below 1.394 under load. The idle voltage remains above 1.5 without manual adjustments. I only adjust the V/F Point table settings. For the second set of settings, the reported maximum CPU Power Package is 394, whereas earlier versions showed over 400.
G
Goljat12
07-18-2021, 11:58 AM #5

Using the initial configurations, the core voltage stays below 1.394 under load. The idle voltage remains above 1.5 without manual adjustments. I only adjust the V/F Point table settings. For the second set of settings, the reported maximum CPU Power Package is 394, whereas earlier versions showed over 400.

L
Lenicord
Junior Member
41
07-20-2021, 12:00 AM
#6
The second configuration for 400W models performs best.
L
Lenicord
07-20-2021, 12:00 AM #6

The second configuration for 400W models performs best.

C
57
07-20-2021, 02:18 AM
#7
Checked LLC@8 with -0.2 for VID at 5.2. Adjusted VID to -0.1 for 5.1 to ensure consistency. Ran a test and CPU Power Package came out to 373. Based on that, using the highest LLC setting seems optimal.
C
CamilloGamer13
07-20-2021, 02:18 AM #7

Checked LLC@8 with -0.2 for VID at 5.2. Adjusted VID to -0.1 for 5.1 to ensure consistency. Ran a test and CPU Power Package came out to 373. Based on that, using the highest LLC setting seems optimal.

L
Llabros
Senior Member
740
07-20-2021, 07:04 AM
#8
Alright, understood. Let's keep experimenting.
L
Llabros
07-20-2021, 07:04 AM #8

Alright, understood. Let's keep experimenting.

J
JbbDbb
Junior Member
1
07-21-2021, 02:42 PM
#9
I performed a 4K HDR encoding with FFmpeg using libx265 on preset Slow. Settings included AVX2 and AVX512 Offset at 0 LLC@8, VID 5.2, -0.22, adaptive voltage +0.001. No crashes or errors occurred, temperatures stayed below 88°C, with an average of 84°C. This was the first test for any AVX load, using identical parameters to SSE settings, and it appears stable so far. The AVX Voltage Guard Band is set to auto, but I still see issues on vCore: max 1.421 during idle, 1.394–1.403 during load, and peak usage reaching 1.403. CPU package power averaged 364 (max 405), which wasn’t high before bedtime. The first settings pushed the average to 413 without AVX enabled, completing a stress test in under a minute. CPU current average is 185, max 212 at 5.2GHz with AVX2 and AVX512—seems like solid optimization. The only drawback is that uncore dropped from 4.3 to 4.0. I’ll likely try another optimization later and open a thread for more details.
J
JbbDbb
07-21-2021, 02:42 PM #9

I performed a 4K HDR encoding with FFmpeg using libx265 on preset Slow. Settings included AVX2 and AVX512 Offset at 0 LLC@8, VID 5.2, -0.22, adaptive voltage +0.001. No crashes or errors occurred, temperatures stayed below 88°C, with an average of 84°C. This was the first test for any AVX load, using identical parameters to SSE settings, and it appears stable so far. The AVX Voltage Guard Band is set to auto, but I still see issues on vCore: max 1.421 during idle, 1.394–1.403 during load, and peak usage reaching 1.403. CPU package power averaged 364 (max 405), which wasn’t high before bedtime. The first settings pushed the average to 413 without AVX enabled, completing a stress test in under a minute. CPU current average is 185, max 212 at 5.2GHz with AVX2 and AVX512—seems like solid optimization. The only drawback is that uncore dropped from 4.3 to 4.0. I’ll likely try another optimization later and open a thread for more details.

P
159
07-28-2021, 02:54 PM
#10
It seems like you're praising something as a solid character build.
P
pocketchicken1
07-28-2021, 02:54 PM #10

It seems like you're praising something as a solid character build.

Pages (2): 1 2 Next