incendery

"Anyway, by trying to fix it, you'll learn many things. 🙂" very well said 🙂 this is the best way to learn

at minumum to POST you will need CPU, GPU and one stick of ram. POST stands for power on self test. the EC bios can be flashed I believe it's stored on an 8pin flash like the main bios on these machines, sourcing a file is a different story I'm not sure with that 😬 maybe someone else here with a p870 can provide one.

A small update for tonight before I can do a full writeup with more testing. The current limit modifications have been a success!! I am able to pull 130W from the CPU sustained on a heavy avx load which would previously shut the laptop down immediately. I have yet to set up my MCU to convert the fan control signal so they are both running at 100% (good cooling but loud! 😅) I will do a thorough breakdown of the process once I have done more testing but I am so far very happy with the results

thanks! It's been a while since I played with the bios modifications and there was some back and forth getting all the drivers to work properly, I've attached a bios I modified for another user here for the p775tm1g (z370) this should work with your rtx3080 out of the box. for my own machine I am using the z370 derived Dsanke bios (that person is a wizard and I am in awe at their skills) which I have tweaked with the board ID changes and also fixed the audio driver bug due to mismatched IDs P7xxTM1-0729-9900KS_BoardID7715_3080 NvdGPUDxe_audiofix.7z

I missed another component which needed changing, PR162 must be changed to 27k

check page 110~ in the schematic here (dm3 but the dm2 VRMs are essentially cloned) I'm not sure the offsets in bios do anything to affect these voltages unfortunately as the outputs appear fixed (except for vcore) and vcore/uncore are not seperately adjusted. I could be incorrect on this but it may be worth investigating 🙂 https://oem.whyopencomputing.ch/Computers/Service Manuals/2023.11.22_P775DM3_Service_manual.pdf

Life has been very busy so I haven't had much chance to work on this laptop for a long while until recently. Todays plans are to modify the current limit through PR27 which sets the OCP (144A on the DM3, 180A on the tm1g despite the same VRM configuration) changing from 30k - 38.3k there are 2 other resistors which require changing here , r177 and r162 in line with the p775tm1g schematic, I misunderstood their function before and there's a bit more going on than I first thought, the data sheet for the ncp81248 voltage regulator (close to the one in this machine) has the proper calculations for the resistor values on page 19* While the laptop is in pieces I am also converting to the modified A51M fans from Taobao, they require a PWM signal and unfortunately I did not buy the converter offered by the seller at the time so I must make my own, I am bypassing the onboard NCT3941S linear DC fan controller which takes a simple 0-3v input signal from the EC and outputs x4 the voltage to the normal fans, unfortunately the EC is not using PWM in this case which would have made my job easier! I will be using some small MCU to convert the 3v to a PWM signal to send to the fans. For now all I will be doing is attaching a breakout cable so I can tinker more with the fans externally without needing to run jumper wires to the board. More updates soon.

A strange question but what slot is your main NVME drive currently installed? I have trialed moving my drives around and putting my main drive in the slot under the main cover and not the cover for the subwoofer / SATA drives, this from some small testing seems to have improved the issue? very strange.

A small update on this project, I have pulled the laptop apart for some unrelated modifications to the fans and while it is apart I'm going to change the current limit resistor and hopefully not explode it! 😅 While I have it open I will also be modifying the Imax value through PR190 so throttlestop isn't necessary for running over 120A to the core. The changes are as follows: PR27 (OCP set) 30k -> 38.3k PR190 (IMAX) 95.3k -> 110k

Hello! this is a behaviour of the embedded controller causing the CPU to throttle when GPU load is applied, unfortunately this is seemingly only solvable with premamod EC firmware but that is practically impossible to find:( throttlestop will do fine as it continuously overrides the register change the EC makes

this sounds similar in issue to what I experienced, I believe there is an issue with one of the drivers that gets automatically pulled and it's congestion on ambush, if you put your system under load and hold a key then release do you continue to see keystrokes for a short time after releasing?

after more testing I have realized the bios change does not fix the issue 😞 I will be doing more testing with different drivers to try and diagnose it

My memory training isn't nearly as happy on the 1.05 bios! 3200mhz errors a lot when previously on xmp it was rock solid stable for months. I am stable now at 2933mhz but I wonder if the keyboard can be ported forwards to the 1.07 bios or perhaps the memory initialization backwards to the 1.05 🙂

It's been a little bit since the last update, I have been busy moving countries so not a lot of work done on the laptop 😄 I found an issue with the bios version based on 1.07, due to some problem with how the keyboard is read and buffered (I believe) there is an issue where a key held for a while will have ghost keypresses when it is released. The keyboard issue persists between seperate installs and is not present on the 1.05 version so I believe it is something with how the EC communicates with the system on differing versions. I am currently in the process of installing upgraded fans from TaoBao, I made the mistake of not ordering the DC to PWM converters from the seller so I must make my own which is taking some time, I will hopefully have more updates on this within the next month 🙂 Stay safe, stay happy.

if you look for the service manual for your model clevo there is probably a schematic you can use to help troubleshoot 🙂

I do plan on talking to the VRM at some point, I have recently moved countries so my time has been a little short and I've been using the laptop as my main PC now 🙂 I salvaged The 20mOhm shunts from some parts donor boards I had laying around at the time, nothing too specific I don't know if my machine ever supported gsync and haven't tried getting it working, using NVIDIA custom resolution settings I have the display overclocked happily to 100hz instead 🙂

Yes I did shunt mod it 🙂 I was quite conservative with it as I want it to last a good long time. I added 20mOhm shunt resistors on top of the 5mOhm already on the card so in theory (without talking to the vrm itself) it should be 180w. The performance is quite good and beats 165w dynamic boost cards in GPU benchmarks

Part 2: BIOS modifications. A late update since the last post, life has been busy 🙂 To preface this post, I am firmly a beginner with BIOS modifications and my understanding is limited 🙂 I am cobbling together information from many smart people to make this work. My goals for the BIOS modifications are: Coffee lake support. Power limit removal. Functioning audio drivers with working headphone amplifier. Native Nvidia driver installation without INF modification. 1 + 2 For Coffee lake support and power limit removal I am using Dsanke's fantastic 1.07.24 bios based on the P775tm1, while this functions mostly flawlessly it does have issues where the audio drivers no longer work and the powerful dedicated headphone amp doesn't switch on. 😞 From what I could understand in the smxdiy thread by Dsanke this has to do with the bios 1.07.24 changing from HDA to UAD support. 3 Initially I was dissapointed that this bios didn't work properly with the audio drivers and I would like to have my headphone amplifier working 😄, the symptoms of non functional amp + spdif LED always on were very similar to an issue I caused when trying to initially get native Nvidia driver support following on a Post from user SRS2236, I had changed the subsystem ID value for the intelHD audio device in error when figuring out which was for the GPU! this mismatch of IDs stopped the Realtek driver from working properly similarly to how Nvidia driver needs an INF modification. With this hunch I had a look at the boardID file in the bios provided by Dsanke and sure enough the subsystemID for intelHDaudio was set to that of the P775tm1 not the P775dm3, changing this from 7709 to the correct ID of 7705 allowed me to install a realtek UAD driver and retain working audio with a functional headphone amplifier! The only issue I was unable to resolve is that the headphone amp no longer lowers volume automatically on detecting headphones so care must be taken not to deafen yourself 🙂 Link to original post with hex values to change 4 the Nvidia driver has a set list of matching device and subsystem IDs, if your combination of IDs isn't valid then it will fail to install and require NVcleaninstall or other methods to add your combination, while this works fine it also requires extra hassle along with no automatic updates and issues with anticheat solutions. The fix for subsystem ID mismatch is to change the subsystem ID 😄. once again referring to this post for clues we can see the BoardIDPEI module is where the subsystem ID for the GPU originates, and if we can change our ID to 7715 (instead of 7705 or 7709 from p775dm3 and p775tm1 respectively) then we can trick the Nvidia driver into installing without any questions asked because 7715 is the subsystem ID for the clevo X170km which shipped with a 3080. after a bit of trial and error the values which need changing will be these 2 values (if your bios is p775dm3 derived it will be 0577, if p775tm1 derived it will be 0977) Change them to 1577 as below. E8 00 59 74 0C 48 75 10 C7 45 FC 58 15 15 77 EB DE 10 E5 1B 74 4D C7 44 24 0C 58 15 15 77 C7 44 After changing the IDs I have had 0 issues regarding GPU drivers, automatic updates etc all work flawlessly 🙂 Issues / future plans I've become aware of the CPU throttling behaviour when any GPU load is detected 😞 this isn't too problematic as throttlestop can override this behaviour but I would like to solve this. From research and my limited understanding it seems to be a function of the EC causing this throttling, when the GPU is loaded bit 38 in MSR 1A0H is flipped (Turbo mode Disable) and when load is removed it reverts back after a few seconds, Prema has to my understanding overcome this with their custom EC firmware which is not available unfortunately, I reached out to Prema but did not hear back from them (I respect their decision not to release the files due to personal or professional reasons). My plan going forward is to take some time to learn more about reverse engineering and see where that takes me 🙂 Thank you for reading!

amazing scores!! I can only dream of CPU temperatures that low 😅😅😅

intruiging, thanks for sharing!

I suspect it may not be a vbios thing at this point, it may be something with the EC and enabling the GPU.

enable above 4g decoding, that should do it 🙂

interesting! strange the tongfang vbios didn't work with the es CPUs 🤔

MXM is quite dead in the consumer space unfortunately 😞 this is a card from the last true DTR clevo with socketed CPU and GPU (X170KM)

unusual it isn't working with the QTJ1, what style mutant chip was it? one of the IHS ones? I believe I used Dsanke's 1.05 based bios when I had a QTJ2 running in mine before the heatsink swap