Question Can someone help me with under volting my 13700k?

Feb 12, 2024
17
0
10
Ok I've been having a lot of trouble getting my CPU tuned. I have a 13700k on a gigabyte z790 aorus elite ax ddr4 with thermal grizzly aeronaught and a corsair capellix H115i AIO. I have done a lot of research on the values I should be achieving. I keep undervolting it and no temperature drops. I've gone all the down to -0.172 offset and it's still hitting 100 C during benchmarking. I really want to figure this out before I have to take everything apart to see if the mate of my cooling block to my chip isn't right. I really don't think it's my cooling system because when I stop the benchmark my temps almost instantly drop to 40 C or less. Problems I've been running into are: I had to figure out that I need to turn my AC off because when it kicks on my PC will crash during benchmark. Other things I've considered is if you research my AIO it doesn't actually come with standoffs specifically for LGA1700 so what exactly is the measurement for stand off height for my setup (I have caliper)? Another thing is I can not find the bios setup manual from gigabyte for my mobo because they use confusing terms, like what exactly do the perfdrive profiles do? I have it on spec enhance instead of optimization because they claim it runs cooler but I find people saying the contrary. My configuration is perfdrive spec enhance, undervolt protection disabled, core multipliers are 2 good cores at 54x and other 6 are 53x and E cores are 42x, power limit 253, offset is currently -0.160, load line at high. I really would appreciate some help.

Edit: to people commenting on stand offs I have researched this already at the time of the build and from what I found the stand off sizes are the same. If you know specific details of this exact model and know of an actual difference in this hardware I would like to know.
 
Last edited:
Look man, did you look it up at all? If you check the manual for the capellix H115i I'm pretty sure I used the Intel 1200/115X hardware and at the time of the build when I researched this people said they used the same ones on an LGA1700 socket and it works perfectly fine. I feel like corsair is trying to get another $8 out of me and when I bought it, it came as a CPU and AIO bundle. This build has been together for over a year and I don't have temp problems. I can play games at 300 fps and have normal operating temperatures of 60 C or less. This is only a problem specifically under a full load when benchmarking and as soon as I stop the tests my temperatures drop instantly. From the research I've done if there was a problem with contact or surface area, the temperatures wouldn't behave that way. If you have any experience lapping or de-lidding and could tell me the exact measurement my cold plate should be off of my mobo I could make the stand offs myself because I don't want to pay $8 for some stand offs that are the same as I already have. If any of my conclusions from my research are wrong, could you please explain why? I am only asking for help because I've looked all this up already.
 
Look man, did you look it up at all? If you check the manual for the capellix H115i I'm pretty sure I used the Intel 1200/115X hardware and at the time of the build when I researched this people said they used the same ones on an LGA1700 socket and it works perfectly fine. I feel like corsair is trying to get another $8 out of me and when I bought it, it came as a CPU and AIO bundle. This build has been together for over a year and I don't have temp problems. I can play games at 300 fps and have normal operating temperatures of 60 C or less. This is only a problem specifically under a full load when benchmarking and as soon as I stop the tests my temperatures drop instantly. From the research I've done if there was a problem with contact or surface area, the temperatures wouldn't behave that way. If you have any experience lapping or de-lidding and could tell me the exact measurement my cold plate should be off of my mobo I could make the stand offs myself because I don't want to pay $8 for some stand offs that are the same as I already have. If any of my conclusions from my research are wrong, could you please explain why? I am only asking for help because I've looked all this up already.
Do you know for sure with this exact product. I'm looking at the box. I'm not sure how to exactly identify it. It's a iCUE H115i elite capellix dated 2020. I looked this up and found those stand offs are the same size. I am wrong in this? If I have to take all this apart I might as well use a der8auer contact plate https://www.thermal-grizzly.com/en/cpu-contact-frame/s-tg-cf-i while I'm at it. Do you have any experience with this product and or lapping and de-lidding?

Edit: Ok the SKU is CW-9060047-WW are you 100% sure there is a difference in size from the hardware in this box and the retrofit kit because I looked this up year ago and found they are the same size. Am I wrong?
 
Last edited:
@Deafmute420

Why would Corsair create a LGA1700 retrofit kit if the new part was identical to the LGA1200 part?

The CPU height is not the same between LGA1200 and LGA1700 motherboards. Blame Intel for this change.

Have a look at the Vcore voltage that HWiNFO reports while running a consistent test like Cinebench. Make sure your undervolt is actually working correctly. What does HWiNFO report for the AC/DC loadline values?
 
  • Like
Reactions: zinkles
@Deafmute420

Why would Corsair create a LGA1700 retrofit kit if the new part was identical to the LGA1200 part?

The CPU height is not the same between LGA1200 and LGA1700 motherboards. Blame Intel for this change.

Have a look at the Vcore voltage that HWiNFO reports while running a consistent test like Cinebench. Make sure your undervolt is actually working correctly. What does HWiNFO report for the AC/DC loadline values?
Man I'm not sure, for money? Isn't that a common big business strategy, to try to sell you stuff you don't need? Why would they make a CPU AIO bundle that doesn't come with the right mounting hardware?

Well that's another problem, LLC in bios is set at high (whatever that means) their nomenclature is weird to me. My clocks are static and always running at 5300 and in HWmonitor IA offset says -0.163 and LLC/Ring offset says +0.00 is there an easy way I can send you screen shots?

Vcore is set to adaptive so why do I not have an offset?
 
Last edited by a moderator:
Well that's another problem, LLC in bios is set at high (whatever that means) their nomenclature is weird to me. My clocks are static and always running at 5300 and in HWmonitor AI offset says -0.163 and LLC/Ring offset says +0.00 is there an easy way I can send you screen shots?

Vcore is set to adaptive so why do I not have an offset?
If P-cores are always static, that means an overclock is in effect and overriding settings.
Really, the easiest way to settle this would be to go into bios:
-restore defaults, to start on a fresh slate.
-completely disable PerfDrive's features. It's just bloatware.
-set Processor Base Power to 125w, Max Turbo Power to 253w, and Turbo Duration(or Tau) to 56 ~ 128 seconds.


To post images here, you have to go through an imagehosting site, or nothing appears.
 
If P-cores are always static, that means an overclock is in effect and overriding settings.
Really, the easiest way to settle this would be to go into bios:
-restore defaults, to start on a fresh slate.
-completely disable PerfDrive's features. It's just bloatware.
-set Processor Base Power to 125w, Max Turbo Power to 253w, and Turbo Duration(or Tau) to 56 ~ 128 seconds.


To post images here, you have to go through an imagehosting site, or nothing appears.
Yeah that's what I'm trying to do now. How do I disable perfdrive?
 
You can't just turn it off? Do you have to turn off the features inside of it for it to not do anything?
Turn off Optimization, Instant 6ghz, Spec Enhance, and - well, you don't disable the E-cores.
That's why I'm asking. There is no like disable option. I have to pick one of those. Those are the different PerfDrive profiles I can pick. Optimization is the default one.

Wait, I have a question. Am I supposed to be waiting for the Extended Power Limit to kick in and throttle my power consumption down to 125w or am I supposed to be under a 100 C at 220w when its turboing up? Because I'm under 70 C at the PL1 125w limit.
 
Last edited:
PL1 125w limit
Anyone that sets PL1 to the default 125W value is not going to be able to get maximum performance out of a 13700K. Not even close.

HWmonitor IA offset says -0.163 and LLC/Ring offset says +0.00
That is why your undervolt is not working. Both the P core and the P cache need to be undervolted or else the P core undervolt request will be ignored. Either you are not setting the voltages correctly or your BIOS is not setting the voltages correctly.

Try using Intel XTU or ThrottleStop 9.6 to adjust your voltages while in Windows. You will need to disable virtualization including core isolation memory integrity if you want to be able to use software to adjust the voltage.

https://beebom.com/how-disable-virtualization-based-security-vbs-windows-11/

If you need help with ThrottleStop post some screenshots of the FIVR and TPL windows. I never use Intel XTU.

For hosting images try using,
www.imgur.com

LgFe1RU.png


When the H115 AIO was first released, LGA1700 did not exist. It only came with standoffs for LGA1200. I assumed that you were using an older AIO with the original LGA1200 standoffs on your new LGA1700 CPU. If that was the case then I would definitely be looking to find and use the correct standoffs. If you just bought this AIO then it should have shipped with the updated standoffs so no need to buy new ones.

HWmonitor
Try using HWiNFO. It reports the correct voltages. HWiNFO can also report what the AC/DC loadline values are set to.
 
I mention 125w, as that's in the cpu's specs: https://ark.intel.com/content/www/u...3700k-processor-30m-cache-up-to-5-40-ghz.html
It doesn't have to be followed to a T though; most motherboards seem to ignore it in their defaults anyway.

Wait, I have a question. Am I supposed to be waiting for the Extended Power Limit to kick in and throttle my power consumption down to 125w or am I supposed to be under a 100 C at 220w when its turboing up? Because I'm under 70 C at the PL1 125w limit.
It throttles down when Tau/Power Limit Duration runs out. But if Processor Base Power is set to match Max Turbo Power, then the cpu never needs to throttle down.


This whole thing was started by you running benchmarks on the cpu, of which that kind of temperature is the norm now.
What are the cores doing with your usual activities?
 
I mention 125w, as that's in the cpu's specs: https://ark.intel.com/content/www/u...3700k-processor-30m-cache-up-to-5-40-ghz.html
It doesn't have to be followed to a T though; most motherboards seem to ignore it in their defaults anyway.


It throttles down when Tau/Power Limit Duration runs out. But if Processor Base Power is set to match Max Turbo Power, then the cpu never needs to throttle down.


This whole thing was started by you running benchmarks on the cpu, of which that kind of temperature is the norm now.
What are the cores doing with your usual activities?
I would rather undervolt than limit my TDP. What is a good TDP? 125w is way too low, I should be around 190-210. I know Intel spec these to be running between 90-105 C. Well usually under gaming loads it is less than 20% (usually around 10%) utilization running around 40 - 60 C and I'm not sure on intensive instruction sets like password cracking. I haven't run hashcat in a while and I have CUDA cores enabled to crack with my GPU so I don't think it really even uses the CPU that much. Thank you for helping btw.
 
Anyone that sets PL1 to the default 125W value is not going to be able to get maximum performance out of a 13700K. Not even close.


That is why your undervolt is not working. Both the P core and the P cache need to be undervolted or else the P core undervolt request will be ignored. Either you are not setting the voltages correctly or your BIOS is not setting the voltages correctly.

Try using Intel XTU or ThrottleStop 9.6 to adjust your voltages while in Windows. You will need to disable virtualization including core isolation memory integrity if you want to be able to use software to adjust the voltage.

https://beebom.com/how-disable-virtualization-based-security-vbs-windows-11/

If you need help with ThrottleStop post some screenshots of the FIVR and TPL windows. I never use Intel XTU.

For hosting images try using,
www.imgur.com

LgFe1RU.png


When the H115 AIO was first released, LGA1700 did not exist. It only came with standoffs for LGA1200. I assumed that you were using an older AIO with the original LGA1200 standoffs on your new LGA1700 CPU. If that was the case then I would definitely be looking to find and use the correct standoffs. If you just bought this AIO then it should have shipped with the updated standoffs so no need to buy new ones.


Try using HWiNFO. It reports the correct voltages. HWiNFO can also report what the AC/DC loadline values are set to.
Ok after work I'll specifically use HWiNFO so we can look at those values. Ring mod in my bios is auto, should it be the same voltage offset as the Vcore? Thank you for helping btw.
 
Ok after work I'll specifically use HWiNFO so we can look at those values. Ring mod in my bios is auto, should it be the same voltage offset as the Vcore? Thank you for helping btw.

I'm not sure where exactly to go in here. One thing I noticed right off the bat though is on the voltage offsets it only shows and offset for IA and nothing else and it's also saying voltage mode in interpolative? Shouldn't it be adaptive? Oh and it shows load line is at 0.700/0.900 mOhm.
 
Last edited:
No, you can do anything you want, I just answered to uwebb because I don't think that 10% is such a huge difference, but I understand that if somebody looks for the best possible then 10% can be a big deal.

No, you can do anything you want, I just answered to uwebb because I don't think that 10% is such a huge difference, but I understand that if somebody looks for the best possible then 10% can be a big deal.
Bro I just want this tuned up right, running R23 getting over 30k pts and doing it under 90 C. That is possible, no?
 
I would rather undervolt than limit my TDP. What is a good TDP? 125w is way too low, I should be around 190-210.
The end goal of the 2 is virtually the same. I'd even say that UV is the more roundabout method and can lead to instability if too much voltage is shaved off.

Tau/Extended Boost Duration resets when tasks are complete before it ends.
If Processor Base Power = Max Turbo Power, or you're running an app that doesn't sustain a constant load on the cores, Tau will keep resetting, thus core frequencies stay high.

Most Z-series motherboards are set past the Intel spec out of the box. Essentially overclocked for you, and not a good one. No one asked the board vendors to do that, but they do it in their own little e-peen contest.
The end users who aren't aware of what's going on get shafted the most out of this.

125w is way too low... for what? You should be around 190~210w in what?
power-per-game.png

Looks pretty low for a select number of titles.

power-per-application.png

Maybe these are the kind of apps you're running on the regular? If so, you could set PBP to 190(+/-), if you want to. You don't have to follow the Intel spec.


Well usually under gaming loads it is less than 20% (usually around 10%) utilization...
Most titles only really utilize a handful of cores, and dynamically. Plus, the 13700K has 8 hyperthreaded P-cores and 8 E-cores for a total of 24... not surprising a game wouldn't make use of all those threads, save for that Ashes of the Singularity title.
 
The end goal of the 2 is virtually the same. I'd even say that UV is the more roundabout method and can lead to instability if too much voltage is shaved off.

Tau/Extended Boost Duration resets when tasks are complete before it ends.
If Processor Base Power = Max Turbo Power, or you're running an app that doesn't sustain a constant load on the cores, Tau will keep resetting, thus core frequencies stay high.

Most Z-series motherboards are set past the Intel spec out of the box. Essentially overclocked for you, and not a good one. No one asked the board vendors to do that, but they do it in their own little e-peen contest.
The end users who aren't aware of what's going on get shafted the most out of this.

125w is way too low... for what? You should be around 190~210w in what?
power-per-game.png

Looks pretty low for a select number of titles.

power-per-application.png

Maybe these are the kind of apps you're running on the regular? If so, you could set PBP to 190(+/-), if you want to. You don't have to follow the Intel spec.



Most titles only really utilize a handful of cores, and dynamically. Plus, the 13700K has 8 hyperthreaded P-cores and 8 E-cores for a total of 24... not surprising a game wouldn't make use of all those threads, save for that Ashes of the Singularity title.

I'm pretty sure my average power draw should be around 190-210w under a full load from cinebench R23 for the temperatures I want. I think uWebb429 was right about my LLC/RING. It will never go below the ring offset correct? it is set to auto in my bios so even though I'm undervolting with a Vcore offset, it is not applying the value because of that?
 
That is why your undervolt is not working. Both the P core and the P cache need to be undervolted or else the P core undervolt request will be ignored. Either you are not setting the voltages correctly or your BIOS is not setting the voltages correctly.
I think you are right about this. I now have Vcore and RING in my bios undervolted to -0.160 but still no temperature drop. HWMonitor is reflecting my undervolt values. Monitoring the CPU VCORE value in HWMonitor it is still around 1.33v at full load but I do think my VID is getting affected. I'm pretty sure I need my CPU VCORE at 1.280v

PS I am getting high 29K's pts on R23 without setting high priority
 
between 90-105 C
Intel spec is a max of 100°C. That is when thermal throttling will begin to keep the CPU at a safe temperature.

In the BIOS you could try lowering the AC/DC loadline values to 0.5 or less. That can help reduce voltage and temperatures. I use 0.25 for both. When these are not set to the same values, the reported power consumption data might not be accurate. Some motherboards use tricks like tweaked AC/DC loadline values to trick the CPU into reporting less power consumption. This can increase the amount of turbo boost available and allow it to last for a longer period of time.

Make sure Undervolt Protection is disabled in the BIOS. Use HWiNFO to double check that this is disabled.

I don't think that 10% is such a huge difference
In situations where a benchmark takes less than a minute, the 125W power limit might not be very restrictive. During a long game session where the CPU is significantly loaded, I think 125W is going to cause a loss of performance greater than 10%. I am old school. If changing a few settings can make a CPU run faster or cooler, I do it.

The default 56 second turbo power time limit is perfect for Cinebench R23 testing. Almost as if Intel chose that value specifically for this test. The high PL2 power limit allows the CPU to run at almost full speed during an entire Cinebench test. The difference in scores with the power limits removed is next to nothing.

https://www.techpowerup.com/review/intel-core-i7-13700k/6.html

I think there would be a much more significant difference when running a longer test like the 10 minute Cinebench 2024 render test.

HWMonitor is reflecting my undervolt values
HWMonitor has a minor bug where the voltages are not reported 100% correctly. They are pretty close though.

HWiNFO does not update the CPU voltages it reports in real time. If you use software to change any of the CPU voltages, you will need to restart HWiNFO so it can resample the voltage control register.

CPU VCORE
Has this dropped since you started undervolting both the core and the cache?

I am getting high 29K's pts on R23
have you increased both the PL1 and PL2 power limits or is PL1 still set to 125? Close to 31000 in R23 seems to be typical for a 13700K.
 
Intel spec is a max of 100°C. That is when thermal throttling will begin to keep the CPU at a safe temperature.

In the BIOS you could try lowering the AC/DC loadline values to 0.5 or less. That can help reduce voltage and temperatures. I use 0.25 for both. When these are not set to the same values, the reported power consumption data might not be accurate. Some motherboards use tricks like tweaked AC/DC loadline values to trick the CPU into reporting less power consumption. This can increase the amount of turbo boost available and allow it to last for a longer period of time.

Make sure Undervolt Protection is disabled in the BIOS. Use HWiNFO to double check that this is disabled.


In situations where a benchmark takes less than a minute, the 125W power limit might not be very restrictive. During a long game session where the CPU is significantly loaded, I think 125W is going to cause a loss of performance greater than 10%. I am old school. If changing a few settings can make a CPU run faster or cooler, I do it.

The default 56 second turbo power time limit is perfect for Cinebench R23 testing. Almost as if Intel chose that value specifically for this test. The high PL2 power limit allows the CPU to run at almost full speed during an entire Cinebench test. The difference in scores with the power limits removed is next to nothing.

https://www.techpowerup.com/review/intel-core-i7-13700k/6.html

I think there would be a much more significant difference when running a longer test like the 10 minute Cinebench 2024 render test.
Computerbase makes their tests with the TDP locked down and also goes far below base power, for 125 vs 253 they show a difference of 14% for the 13700k and 20% for the 13900k so yes the difference is bigger than I thought, but still not terrible, especially for the 13700k.
142W for the 13700k 9% from 253 so that would be a decent point for PL1.
 
  • Like
Reactions: uWebb429
Has this dropped since you started undervolting both the core and the cache?
No I don't think so. Max I've ever seen is 1.401v and seems to be around 1.33v most of the time. Yeah, as far as resetting HWiNFO and stuff. I'm doing this old school now with a restart, tweak setting, boot back in. Bro this is why I need help with this gigabyte board. Load Line settings are anywhere from normal to extreme. Is a lower mOhm value a steeper vdroop setting or more towards flatline? All these other mobos like ASrock have load lines labeled 1-5, mine is normal, medium, high, turbo, extreme. I wish I could find my BIOS manual because idk what gigabyte means.

Yes undervolt protection is off.
 
Last edited: