11980hk ES Overclocking Shenanigans, help and expertise appreciated

[TheDankKoosh]

I'm currently working with some very obscure and interesting hardware at the moment, long story short this is an ES of a 11980hk on a m-atx board.

I'm currently attempting to OC this chip to its limits, but I've hit a bit of a snag in the process and could use a few extra heads to figure it out

I'm hitting current limits in windows under all core workloads and the chip is refusing to draw more than 110w, XTU and Throttlestop ICC limits aren't changing anything, package temps under P95 SFFTS don't pass 75c

I've been messing around in the bios but I can't seem to find anything referring to current draw and possibly raising it. The bios on this bad boy is definitely dated, but fully functional when it comes to overclocking. Any help is appreciated

 

Board and CPU I'm using for reference: https://www.aliexpress.us/item/3256805120525780.html?spm=a2g0o.productlist.main.1.54d023e7R6aaQY&algo_pvid=41b6dc3d-7188-4963-9720-0b72da700757&algo_exp_id=41b6dc3d-7188-4963-9720-0b72da700757-0&pdp_ext_f={"sku_id"%3A"12000032569592621"}&pdp_npi=3%40dis!USD!388.4!190.32!!!!!%4021021b4716796295270385129d0701!12000032569592621!sea!US!2019284271&curPageLogUid=izXaVx6zvPbc 


Edit: posting screen of XTU setting under P95 load

 

[SorryBella]

18 minutes ago, TheDankKoosh said:

I'm hitting current limits in windows under all core workloads and the chip is refusing to draw more than 110w

I wouldnt be surprised if this is a BIOS limit, because 11980HK TDP up on its final products are 65W, no more no less. Whats the performance in that target TDP?

[TheDankKoosh]

3 minutes ago, SorryClaire said:

I wouldnt be surprised if this is a BIOS limit, because 11980HK TDP up on its final products are 65W, no more no less. Whats the performance in that target TDP?

It's currently scoring within a stones throw of a 11900k, but I have seen these chips do more like 6.5k multi-thread. There was a Portuguese youtuber that did some fairly comprehensive testing and overclocking on this, he didn't seem to have current throttling issues like me
 

 

 

 

[SquintyG33Rs]

34 minutes ago, SorryClaire said:

I wouldnt be surprised if this is a BIOS limit, because 11980HK TDP up on its final products are 65W, no more no less. Whats the performance in that target TDP?

it's orginally a mobile chip the TDP in spec is not the real value at all. in different laptops its configurable and here mounted on a desktop motherboard they clearly set it much higher than that.

however to answer OP the manufacturer making this Frankenstein combination job can't remove the boundaries that intel sets for the min and max of every setting. If intel sold the chip as configurable between 25 to 65W in laptops and set max 110W in turbo duration the best the bios can do usually is infinite turbo. unlike a GPU that has controllers and sensing shunt resistors externally on the board you don't have anything you can easily spoof to make it get more juice than it thinks it's getting.

[unclewebb]

@TheDankKoosh

Exit Intel XTU before using ThrottleStop. 

 

In the ThrottleStop TPL window, try setting Power Limit 4 to a value of 0.

If that does not fix the EDP throttling then you might be out of options. 

 

 

 

[TheDankKoosh]

1 hour ago, unclewebb said:

@TheDankKoosh

Exit Intel XTU before using ThrottleStop. 

 

In the ThrottleStop TPL window, try setting Power Limit 4 to a value of 0.

If that does not fix the EDP throttling then you might be out of options. 

This did not seem to do anything unfortunately, would it be possible that C-states could be the culprit?

 

1 hour ago, SquintyG33Rs said:

it's orginally a mobile chip the TDP in spec is not the real value at all. in different laptops its configurable and here mounted on a desktop motherboard they clearly set it much higher than that.

however to answer OP the manufacturer making this Frankenstein combination job can't remove the boundaries that intel sets for the min and max of every setting. If intel sold the chip as configurable between 25 to 65W in laptops and set max 110W in turbo duration the best the bios can do usually is infinite turbo. unlike a GPU that has controllers and sensing shunt resistors externally on the board you don't have anything you can easily spoof to make it get more juice than it thinks it's getting.

I'm thinking this could possibly be a C-state issue or a VRM limit that's been set into place by the manufacturer, intel's datasheet is useless for tiger lake and doesn't specify an absolute max power that can run through the chip.

Edit: After some digging in Hwinfo I've been able to rule it down to CPU TDC specifically, VRMs are plenty cool with the improved heatsinks and active cooling
Would anybody know of a way to change the TDC value outside of bios?

[SquintyG33Rs]

17 hours ago, TheDankKoosh said:

Edit: After some digging in Hwinfo I've been able to rule it down to CPU TDC specifically, VRMs are plenty cool with the improved heatsinks and active cooling
Would anybody know of a way to change the TDC value outside of bios?

that's what i'm saying i'm pretty sure that's hard enforced by the CPU itself. and you can't spoof it's reading to make it pull more

[TheDankKoosh]

2 hours ago, SquintyG33Rs said:

that's what i'm saying i'm pretty sure that's hard enforced by the CPU itself. and you can't spoof it's reading to make it pull more

I was able to do more digging last night and figured out that the motherboard defaults the CPU TDC to ~100A despite the setting technically set to unlimited, after maxing that out I was able to get the CPU to comfortably do 160w. Posting results when I get home.

 

Got a couple extra smaller issues that are impeding me now. I cannot get the cores to pull more than 1.36v under load, I suspect something is going on with the adaptive voltage/frequency curve. My best core will only do 4.92 (100.5 BCLK) at that voltage, I'm sure that I could push it to 5.1 with an extra 50-70mv

 

Secondly I'm having troubles getting memory working beyond 3200, this is partially due to VCCSA being more or less not present in BIOS, i have access to the voltage through XTU but those voltages are necessary to get higher speeds to boot at all, would there be any way for me to manipulate memory speeds and timings in windows? I know asus memtweak exists somewhere, but I don't have access to it.

[Soundbored]

Have you tried using ThrottleStop?

[TheDankKoosh]

4 hours ago, Soundbored said:

Have you tried using ThrottleStop?

Unfortunately throttlestop is a no go due to lack of per core OC and voltage controls, those are crucial for maximizing performance on newer intel chips (currently 4.7 on 6 weakest cores -20mv, 4.8 on 2nd best at -20, best 4.9 +40). Also helps to control thermals since the heatspreader isn't the greatest and high heat density. I currently have LM under the heatspreader but normal paste on top, may play with LM on top and try to throw on more voltage from the bios. Memory/VCCSA still has me stumped however.

 

[TheDankKoosh]

Current best results I was able to push out of this chip so far, any ideas or pointers would be greatly helpful as I'm replacing my 10850k with this setup

 

 

[SquintyG33Rs]

On 3/24/2023 at 9:37 PM, TheDankKoosh said:

I was able to do more digging last night and figured out that the motherboard defaults the CPU TDC to ~100A despite the setting technically set to unlimited, after maxing that out I was able to get the CPU to comfortably do 160w. Posting results when I get home.

 

Got a couple extra smaller issues that are impeding me now. I cannot get the cores to pull more than 1.36v under load, I suspect something is going on with the adaptive voltage/frequency curve. My best core will only do 4.92 (100.5 BCLK) at that voltage, I'm sure that I could push it to 5.1 with an extra 50-70mv

 

Secondly I'm having troubles getting memory working beyond 3200, this is partially due to VCCSA being more or less not present in BIOS, i have access to the voltage through XTU but those voltages are necessary to get higher speeds to boot at all, would there be any way for me to manipulate memory speeds and timings in windows? I know asus memtweak exists somewhere, but I don't have access to it.

ooo nice.

i think unless it's built into the motherboard by design the memory can't be adjusted in windows to this day. memory training needs to happen when any timing changes on current tech so i don't think it's directly a thing.

[TheDankKoosh]

10 hours ago, SquintyG33Rs said:

ooo nice.

i think unless it's built into the motherboard by design the memory can't be adjusted in windows to this day. memory training needs to happen when any timing changes on current tech so i don't think it's directly a thing.

I have been getting intimately familiar with intel XTU and I've noticed a really nice feature for newer platforms like this, voltages are retained even after reboot. So I do have VCCSA to boot higher speeds

Problem is, this only allowed me to go to 3333, I'm going to assume that the board is not well equipped to deal with high performance memory in the slightest

Tuning subs to the best of my ability I did start seeing okish latency and bandwidth, but it's far from optimal. I'm mostly relying on the IPC increase and extra cache (per core) for this to win out vs my 10850k (it definitely wins in the power consumption category).

 

[SquintyG33Rs]

On 3/26/2023 at 8:32 AM, TheDankKoosh said:

I have been getting intimately familiar with intel XTU and I've noticed a really nice feature for newer platforms like this, voltages are retained even after reboot. So I do have VCCSA to boot higher speeds

Problem is, this only allowed me to go to 3333, I'm going to assume that the board is not well equipped to deal with high performance memory in the slightest

Tuning subs to the best of my ability I did start seeing okish latency and bandwidth, but it's far from optimal. I'm mostly relying on the IPC increase and extra cache (per core) for this to win out vs my 10850k (it definitely wins in the power consumption category).

yes the clock itself is not really tied to the timings so you can move it from inside the OS. however because the timings are sensitive and you can't adjust them you will very quickly be outside of tolerance.
it's very likely that 1333 is not the limit of the controller but since you can't touch the timings it won't work past that. even if you could just give it a clock target for training on boot you could go further.

[TheDankKoosh]

On 3/30/2023 at 2:10 PM, SquintyG33Rs said:

yes the clock itself is not really tied to the timings so you can move it from inside the OS. however because the timings are sensitive and you can't adjust them you will very quickly be outside of tolerance.
it's very likely that 1333 is not the limit of the controller but since you can't touch the timings it won't work past that. even if you could just give it a clock target for training on boot you could go further.

Been a bit since I've posted, but I do have small updates with a little progress all around. I've stepped up to a BCLK 101.75 to give me an effective mem clock of 3391 with ok improvements to bandwidth and latency, this did slightly affect my core and cache clocks since I had to step down multipliers, I've found performance to be within margin of error. I'd still like to find a way to push more than 1.36v to the cores without throttling, I'm positive that I'd be able to squeeze much better single core performance on the two best cores with just a few dozen millivolts.