F5F Stay Refreshed Hardware Desktop Assistance with overclocking Need tips on boosting performance? Let me know how I can help.

Assistance with overclocking Need tips on boosting performance? Let me know how I can help.

Assistance with overclocking Need tips on boosting performance? Let me know how I can help.

J
71
11-30-2025, 01:27 PM
#1
Hey there, I’m trying to figure out how to update my system specs. I’ve got some knowledge but I’m nervous about making mistakes. I already built a car from the frame up, so I’m familiar with the numbers and OCPX terms. If you’re open to it, feel free to drop me a Discord—I’d really appreciate your help! Thanks, Ed#6101
J
JesperSlagroom
11-30-2025, 01:27 PM #1

Hey there, I’m trying to figure out how to update my system specs. I’ve got some knowledge but I’m nervous about making mistakes. I already built a car from the frame up, so I’m familiar with the numbers and OCPX terms. If you’re open to it, feel free to drop me a Discord—I’d really appreciate your help! Thanks, Ed#6101

A
54
11-30-2025, 01:49 PM
#2
Adjusting performance on a 9900K involves mainly managing voltage and clock speed. The key is balancing these factors to achieve optimal results. Personally, I’d start with a decent LLC configuration—likely level 2 or 3 in Tweakers Paradise. Begin with basic settings: VCore at 1.3V, core ratio at 45, and uncore/cache/ring at 40. In Windows, perform a stress test for a short period. If stable and temperatures are acceptable, tweak the core ratio in the BIOS. If stability is lacking but temps are fine, raise voltage slightly by 0.025V. If temperatures remain stable, lower voltage by the same amount. Should both temperature and stability be off, reduce the core ratio multiplier and adjust voltage accordingly. Continue this process until you reach a satisfactory balance. For testing reliability, run stress tests like Linpack Xtreme or Prime95 Small FFTs for about 30 minutes initially, then extend to 2 hours if needed. Various tools such as OCCT can also be useful depending on your needs.
A
Alixander_Hill
11-30-2025, 01:49 PM #2

Adjusting performance on a 9900K involves mainly managing voltage and clock speed. The key is balancing these factors to achieve optimal results. Personally, I’d start with a decent LLC configuration—likely level 2 or 3 in Tweakers Paradise. Begin with basic settings: VCore at 1.3V, core ratio at 45, and uncore/cache/ring at 40. In Windows, perform a stress test for a short period. If stable and temperatures are acceptable, tweak the core ratio in the BIOS. If stability is lacking but temps are fine, raise voltage slightly by 0.025V. If temperatures remain stable, lower voltage by the same amount. Should both temperature and stability be off, reduce the core ratio multiplier and adjust voltage accordingly. Continue this process until you reach a satisfactory balance. For testing reliability, run stress tests like Linpack Xtreme or Prime95 Small FFTs for about 30 minutes initially, then extend to 2 hours if needed. Various tools such as OCCT can also be useful depending on your needs.

W
WeedlessT
Junior Member
18
11-30-2025, 03:05 PM
#3
On dime boards, the BIOS includes an auto overclock option. For instance, you can choose 9900k at 5.0ghz. It might be a method to begin gradually and assess the results.
W
WeedlessT
11-30-2025, 03:05 PM #3

On dime boards, the BIOS includes an auto overclock option. For instance, you can choose 9900k at 5.0ghz. It might be a method to begin gradually and assess the results.

X
xanderzone317
Posting Freak
957
11-30-2025, 03:33 PM
#4
X
xanderzone317
11-30-2025, 03:33 PM #4

B
B4LLS
Member
65
12-01-2025, 05:56 AM
#5
I've experienced more problems with underclocking on AMD compared to Intel. Underclocking AMD leads to either extreme slowdowns (high clock speeds but poor performance) or crashes when using the curve optimizer during idle states that are nearly impossible to fix. On Intel, you simply tweak one setting assuming a Z-series board, and it works fine—though B-series boards have some limitations. Overall, undervolting feels more like an adjustment than a true change, unlike overclocking which is quite different.
B
B4LLS
12-01-2025, 05:56 AM #5

I've experienced more problems with underclocking on AMD compared to Intel. Underclocking AMD leads to either extreme slowdowns (high clock speeds but poor performance) or crashes when using the curve optimizer during idle states that are nearly impossible to fix. On Intel, you simply tweak one setting assuming a Z-series board, and it works fine—though B-series boards have some limitations. Overall, undervolting feels more like an adjustment than a true change, unlike overclocking which is quite different.

M
MaxBuddyRoo
Member
95
12-01-2025, 07:03 AM
#6
it's about what you're focusing on too, though. It seems to be mostly temperature readings. For instance, my 3600* running at -200V with a voltage offset didn't have clock stretching—it just got boosted longer, which improved performance because temperatures were lower. The same goes for my 5800X3D, GTX 1060, GTX 1070, and RTX 3070 (my Intel i5U crashes under any offset settings). *Interesting: at stock it would draw around 65W as advertised, but with -200V/+200Hz it went up to about 110W.*
M
MaxBuddyRoo
12-01-2025, 07:03 AM #6

it's about what you're focusing on too, though. It seems to be mostly temperature readings. For instance, my 3600* running at -200V with a voltage offset didn't have clock stretching—it just got boosted longer, which improved performance because temperatures were lower. The same goes for my 5800X3D, GTX 1060, GTX 1070, and RTX 3070 (my Intel i5U crashes under any offset settings). *Interesting: at stock it would draw around 65W as advertised, but with -200V/+200Hz it went up to about 110W.*

B
Brady_Gaming
Junior Member
14
12-02-2025, 06:40 AM
#7
When attempting to undervolt with offsets, the outcome varies depending on the situation. Without an offset, the result is a significant drop—around -200mV. A -300mV offset also leads to similar issues. The main clocks continue advancing, yet overall performance declines. In some AGESA versions, using a positive voltage offset can even improve the score. All Intel processors maintain consistent performance and stability under these conditions, unlike the 5900X which experiences unpredictable crashes when stress tests are applied. It seems the chips either function correctly or fail entirely if voltages aren't precisely set.
B
Brady_Gaming
12-02-2025, 06:40 AM #7

When attempting to undervolt with offsets, the outcome varies depending on the situation. Without an offset, the result is a significant drop—around -200mV. A -300mV offset also leads to similar issues. The main clocks continue advancing, yet overall performance declines. In some AGESA versions, using a positive voltage offset can even improve the score. All Intel processors maintain consistent performance and stability under these conditions, unlike the 5900X which experiences unpredictable crashes when stress tests are applied. It seems the chips either function correctly or fail entirely if voltages aren't precisely set.

A
Andreasx345
Member
178
12-02-2025, 03:28 PM
#8
that's what I mentioned doesn't occur... it never increased, consistently around 4.1 to 4.2. The issue is that it took longer to boost, which you can only notice because benchmarks are rising, like 500 points for instance. Personally, I think that's quite low, but others seem to be satisfied with the changes. I mainly focus on overall lower temperatures too. Sorry, just about voltage offset—clocks shouldn't rise since it's locked by a multiplier (around 42x...). So there must be another factor at play here.
A
Andreasx345
12-02-2025, 03:28 PM #8

that's what I mentioned doesn't occur... it never increased, consistently around 4.1 to 4.2. The issue is that it took longer to boost, which you can only notice because benchmarks are rising, like 500 points for instance. Personally, I think that's quite low, but others seem to be satisfied with the changes. I mainly focus on overall lower temperatures too. Sorry, just about voltage offset—clocks shouldn't rise since it's locked by a multiplier (around 42x...). So there must be another factor at play here.