Looking for a good undervolt guide i7 13700KF and Asus Tuf B660m-Plus

[SaiKaesu]

I own 13700 KF and Asus TUF B660m-Plus D4 Mobo
I want to try undervolt. I tried a few times but I didn't get the results I wanted. 13700KF draws too many volts under load and the temperatures are rising. I got the most stable cpu voltage around 1,280V.
How can i make UV on my mobo.
is there any guide?

Stock settings hit 100* and than TT
C23 Score: 27576 to much low

Actual UV settings;
XMP: Enabled (Profile 1)
Switch microcode: 0x104 microcode
Intel Adaptive Boost tec: Enabled
LLC: Level 3
Actual VRM Core Voltage: Auto
Global Core SVID Voltage: “Offset” - .125
Cache Core SVID Voltage: “Offset” -.125
C State: Enabled.
C State function: Auto.

is it nice?

 

[Deleted member 232534]

1. set "turbo boost short power max" to 140 W
2. set "turbo boost power max" to 140W

3. set "core voltage offset" to -0.050
4. stability stress test - if it passes the stability test lower the "core voltage offset" to -0.060 and repeat

-the temperatures should drop with every decrease of "core voltage offset"

-once you determine the lowest "core voltage offset" you can modify the watt limit to whatever you like and enter the settings to bios

-stability test is at least prime95 preferably aida64. Cinebench is not enought to test the stablity

 

[SaiKaesu]

1. set "turbo boost short power max" to 140 W
2. set "turbo boost power max" to 140W

3. set "core voltage offset" to -0.050
4. stability stress test - if it passes the stability test lower the "core voltage offset" to -0.060 and repeat

-the temperatures should drop with every decrease of "core voltage offset"

-once you determine the lowest "core voltage offset" you can modify the watt limit to whatever you like and enter the settings to bios

-stability test is at least prime95 preferably aida64. Cinebench is not enought to test the stablity

View attachment 309670

140W Set limit to much low right?

Right, that makes sense. We are making 140w because we look at the heat values we give with offset. I'm raising the watt limit after setting the lowest temperature right? Thanks i’ll try.

 

[GerKNG]

everything auto, adaptive offset -X mv until unstable. then slowly lowering powerlimits until you're satisfied with the efficiency.

 

[chrcoluk]

Screenshot doesnt show your vcore, SA VID isnt vcore, and if was your vcore 1.2v for a load vcore is quite low.

Please show the what is actually labelled as vcore from hwinfo64, its in the motherboard readings.

 

[ir_cow]

Set your PL1/PL2 limits and call it a day
Waaaay easier vs undervolting.

 

[SaiKaesu]

Screenshot doesnt show your vcore, SA VID isnt vcore, and if was your vcore 1.2v for a load vcore is quite low.

Please show the what is actually labelled as vcore from hwinfo64, its in the motherboard readings.

Actual UV settings;
XMP: Enabled (Profile 1)
Switch microcode: 0x104 microcode
Intel Adaptive Boost tec: Enabled
LLC: Level 3
Actual VRM Core Voltage: Auto
Global Core SVID Voltage: “Offset” - .180
Cache Core SVID Voltage: “Offset” -.180
C State: Enabled.
C State function: C8

Set your PL1/PL2 limits and call it a day
Waaaay easier vs undervolting.

Is just PL1/PL2 enough?

everything auto, adaptive offset -X mv until unstable. then slowly lowering powerlimits until you're satisfied with the efficiency.

every day i try something

Cooler DeepCool AK620

 

[chrcoluk]

Ok post the vcore whilst the benchmark is running, not when its finished, as you just posted the idle vcore, not the load vcore.

But ultimately you have tried quite an aggressive undervolt, and if its unstable you need to reduce it.

 

[SaiKaesu]

Ok post the vcore whilst the benchmark is running, not when its finished, as you just posted the idle vcore, not the load vcore.

But ultimately you have tried quite an aggressive undervolt, and if its unstable you need to reduce it.

Ok, I'll post while the benchmark is running.
If doesn’t work UV. What values should I decrease? SVIDs? or PL1-PL2?

 

[chrcoluk]

Pl1 and Pl2 are the power limit, they are a safe way to reduce your power consumption and heat. But potentially with the loss of peak performance on heavily threaded workloads (probably wont affect games).

Undervolt reduces power consumption without a loss of performance, but is running the CPU out of spec so might cause instability.

For now try Pl1 and Pl2 at 175w, remove the SVID offset and see how things look.

If you not happy with that then either reduce further or then try the undervolt, but I think -180 is too high, start at maybe -50.

 

[SaiKaesu]

Pl1 and Pl2 are the power limit, they are a safe way to reduce your power consumption and heat. But potentially with the loss of peak performance on heavily threaded workloads (probably wont affect games).

Undervolt reduces power consumption without a loss of performance, but is running the CPU out of spec so might cause instability.

For now try Pl1 and Pl2 at 175w, remove the SVID offset and see how things look.

If you not happy with that then either reduce further or then try the undervolt, but I think -180 is too high, start at maybe -50.

okay, first I'll make the PL1-PL2 175W and then remove the offsets.

Images in the current settings.

 

[SaiKaesu]

okay, first I'll make the PL1-PL2 175W and then remove the offsets.

Images in the current settings.

@chrcoluk what do u think current UV settings? Bechmark score?

Core clocks very low under load..

 

[chrcoluk]

@chrcoluk what do u think current UV settings? Bechmark score?

I not commented as I asked you to do the PL without UV first.

The UV we dont know if it is stable.

If you trying to work out why score is below what you think it should be hwinfo64 (and XTU) will report if any throttling and the reason why.

On hwinfo64 I think you better off turning on the log, and starting the log when run benchmark then stopping at end, then attach the log here, as then we see everything, not just snapshots of data.

 

[SaiKaesu]

I not commented as I asked you to do the PL without UV first.

The UV we dont know if it is stable.

If you trying to work out why score is below what you think it should be hwinfo64 (and XTU) will report if any throttling and the reason why.

On hwinfo64 I think you better off turning on the log, and starting the log when run benchmark then stopping at end, then attach the log here, as then we see everything, not just snapshots of data.

I tried set PL1-2 175W, remove the SVID offset. Temp is not bad peek 85s average 75s but P-Core 4,6GHz E-Core 3,6 GHz. And then tried -.050 offset looks like same results. Core voltage 1.18-1.24V

 

[mahanddeem]

1.29v vcore max voltage is good (for an average chip)
I kind of have the same max vcore (1.305v) with almost everything on stock (but SVID Behavior on Best case senario) on Asus Z790 Prime and my 13700K. And power limits 253w max.

You might need a better cooler. I use Noctua D15 max gaming temp (as well as my daily PC use) is around 75c

Your CB R23 is somehow low. 29k to 30k is about normal for stock.

 

[chrcoluk]

1.29v vcore max voltage is good (for an average chip)
I kind of have the same max vcore (1.305v) with almost everything on stock (but SVID Behavior on Best case senario) on Asus Z790 Prime and my 13700K. And power limits 253w max.

You might need a better cooler. I use Noctua D15 max gaming temp (as well as my daily PC use) is around 75c

Your CB R23 is somehow low. 29k to 30k is about normal for stock.

Yeah indications are the OP has at least an ok chip as the vcore is not high, even on my fixed bios without an UV my vcore on certian loads still hits 1.34v (max clocks all core under light load when no vdroop). But it is difficult to determine extent of vcore range based on chip quality due to the variances caused by bios as well.

 

[SaiKaesu]

1.29v vcore max voltage is good (for an average chip)
I kind of have the same max vcore (1.305v) with almost everything on stock (but SVID Behavior on Best case senario) on Asus Z790 Prime and my 13700K. And power limits 253w max.

You might need a better cooler. I use Noctua D15 max gaming temp (as well as my daily PC use) is around 75c

Your CB R23 is somehow low. 29k to 30k is about normal for stock.

AK620 And Nh-D15 almost the same isn't it?

Yeah indications are the OP has at least an ok chip as the vcore is not high, even on my fixed bios without an UV my vcore on certian loads still hits 1.34v (max clocks all core under light load when no vdroop). But it is difficult to determine extent of vcore range based on chip quality due to the variances caused by bios as well.

The frequencies don't work exactly at every value I make UV. TT on stock settings. frequencies drop. It goes up to 253Watts. When I make PL1-2 200W and apply a little UV, I can get frequencies close to P-E Core 5287/4188 MHz. But I can never quite keep the frequencies high. I'm talking about SVID offsets in the scenario where I'm currently applying an aggressive SVID "-.15 or -.18". Only then it consumes 200Watt and I can get good results. In terms of heat, frequency and voltage.. I made the SVID's -.150, although I don't quite like it. I set PL1-2 200w and quit. I don't know exactly how to make a UV and how to benefit from it.

 

[chrcoluk]

If your only concern right now is frequencies and nothing else, set PL1/2 253w. No UV, Enable hwinfo logging, run cinebench multi core run. Then check log if any throttles are triggered. If they are triggered for power limit can raise the PL further until it is satisfied, not necessarily a good day to day config, but useful for this test.

The chips can throttle for a multitude of reasons, hwinfo will tell you which.

If you got no experience undervolting and are not prepared to do lots of testing, then ir_cow's advise is probably the best in my opinion, set PL1/2 to a level your cooler can handle and at wattage you can stomach, for all lightly threaded loads it will still run at max performance, and also for mainstream multi threaded workloads like FPS games, it will only potentially throttle at heavy workloads like encoding or heavy AVX.

 

[SaiKaesu]

If your only concern right now is frequencies and nothing else, set PL1/2 253w. No UV, Enable hwinfo logging, run cinebench multi core run. Then check log if any throttles are triggered. If they are triggered for power limit can raise the PL further until it is satisfied, not necessarily a good day to day config, but useful for this test.

The chips can throttle for a multitude of reasons, hwinfo will tell you which.

If you got no experience undervolting and are not prepared to do lots of testing, then ir_cow's advise is probably the best in my opinion, set PL1/2 to a level your cooler can handle and at wattage you can stomach, for all lightly threaded loads it will still run at max performance, and also for mainstream multi threaded workloads like FPS games, it will only potentially throttle at heavy workloads like encoding or heavy AVX.

I just wanted to undervolt for games. High performance, a cold CPU and quiet fans
In short, a cooler cpu without losing performance.

When I set only PL1-2 253w and leave the rest of the settings stock, the cpu gets very hot in games.