Another request for further advice on mV Boost. When I set mV Boost to 100, HWiNFO64 reports (+)100 for IA Voltage and CLR Offsets.

[dc_IV]

This is likely by design and I am just not viewing it with the correct context, but when I have -50.8 mV undervolt for CPU Core and CPU P Cache, and 100 set for mV Boost, it displays as 0.100 V in HWiNFO64. I guess that fact that it is showing an offset is good for knowing that my undervolt was set, but it is confusing to me.

This is with mV Boost @ 800MHz NOT set.

This is with mV Boost set to 100

 

[Bogus5]

I understand that in the BIOS you also have the ability to lower CPU voltages. The BIOS and ThrottleStop settings add up, so if you lower the core voltage in the BIOS to -50 mV and in the TS to -50 mV, it's a total of -100 mV.

The "mv Boost @ 800 MHz" function only adds the set voltage value when the CPU is seemingly idle, at least that's how I understood it.

 

[dc_IV]

... it's a total of -100 mV.

In my case, HWiNFO64 is showing the offets as positive values, so that is why it is confusing, at least to me.

 

[unclewebb]

it displays as 0.100 V in HWiNFO64

The ThrottleStop monitoring table displays the same -0.0508 mV core and P cache offset whether mV Boost is set to 0 or 100. I do not know why HWiNFO is reporting the mV Boost value as the core and cache offset values. Try posting your question on the HWiNFO forum. The developer of HWiNFO might know something that I do not know.

The mV Boost feature was originally developed on a 10850K. It has not been tested by me on anything newer. Some 13th Gen users have reported that using mV Boost has helped them to stabilize a larger core and cache undervolt compared to having mV Boost set to 0. Based on that, mV Boost seems to still work correctly on the newer CPUs.

To show what this feature does, I set the Speed Shift Min and Max values both to 8 to lock my CPU at 800 MHz.

Idle VID = 0.6897
mV Boost = 0

Idle VID = 0.7797
mV Boost = 100

In this test, a mV Boost request of +100 mV results in a VID increase of +90 mV.

Hopefully someone with a 13th Gen CPU can try repeating this test.

 

[dc_IV]

Hopefully someone with a 13th Gen CPU can try repeating this test.

EDIT! LOL, I found I could set Max Cache Ratio to 8, and the numbers show closer to 100 mV. I did see a spike to 1.3xxx once while looking at it for a bit.

EDIT2: I also deleted my .ini since I tweaked so much, I did not want a weird setting, but thank you @unclewebb for suggesting this test, I feel much more confident, and YES 100 mV allowed me to get to -145 on CPU Core and CPU P Cache and stable Cinebench scoring. Using -145 allowed me to get 30642 on the 10 minute throttle test, and no Throttlestop was 26449. That's a 13.7% lower result versus a cooler running system, or at least less time a throttle.

So I tried this, but I could not get a steady FID of 8.00. It was bouncing around between cached ratios. I set Speed Shift in TPL to Min and Max of 8. I forced a FIVR cache ratio of 8 as well, but it was a bouncer! I was able to see enough though that mV Boost at 0 was a VID of around 0.7332, and with 100, it was around an average of 0.8200, so for an i9-13900HX, it seems to be working!

mV Boost 0

mV Boost 100: it was difficult to get a good cell phone picture, and the other numbers I saw in included 0.8337 and 0.8330 and also less than 0.8200 so I will say it is rough 0.8200 average.

 

[unclewebb]

Thanks @dc_IV for your testing.

I noticed one thing about your settings. For most Intel CPUs, if you want to undervolt the Intel GPU, you need to undervolt both the Intel GPU and the iGPU Unslice the same amount. If you do not undervolt both, your request will likely be ignored.

 

[dc_IV]

Thanks @dc_IV for your testing.

I noticed one thing about your settings. For most Intel CPUs, if you want to undervolt the Intel GPU, you need to undervolt both the Intel GPU and the iGPU Unslice the same amount. If you do not undervolt both, your request will likely be ignored.

Ah, good to know. I actually have a BIOS controlled iGPU/dGPU MUX and I set Hybrid Graphics to Disabled in the BIOS, so I don't think I am actually ever using the on die UHD 770 GPU, so maybe it is moot in that respect.

 

[Bogus5]

Ah, good to know. I actually have a BIOS controlled iGPU/dGPU MUX and I set Hybrid Graphics to Disabled in the BIOS, so I don't think I am actually ever using the on die UHD 770 GPU, so maybe it is moot in that respect.

It's very strange how different results in the Cinebench R23 test can be obtained on the same CPU with very similar ThrottleStop settings. With a 150 mV reduction of the core voltage, the CB result is around 33500 (10 minutes test), and at -170 mV around 35000 (also 10 minutes). What brand of computer do you have?

On the Asus G16 series (2023) laptop, the CPU power was permanently reduced after using GHelper, resulting in much lower results in the Cinebench R23 test. Only restoring the system to the state before using GHelper helped.

 

[dc_IV]

What brand of computer do you have?

I have an Alienware m16 R1 with the 4080 and i9-13900HX. Does your Asus have the H or HX in it? I wonder if there is a huge difference or not, and I don't know if I'll ever take advantage of my HX being unlocked. I don't know if my system would hit 35000, that seems crazy high.

Do you know your PL4 value? On the m16 R1, it is 330. I wonder if your PL4 is higher, or if the Tri-fan cooling on yours allows for a nice undervolt and amazing Cinebench result.

 

[Bogus5]

Exactly, it is a G16-G614JI model with an i9-13980HX CPU, so we have identical processors. DELL unfortunately throttles its laptops, even the high-end ones, and those brands that are the most powerful are, in order: MSI, Lenovo, Asus. The case culture on the Asus laptop isn't very high, but it certainly helps to have a third fan on the motherboard chip.