JadeRover

Yes, the display table has tons of info here is a snippet (non parsed data): Highlighted is the entire display table, notice this repeating pattern of 6 lines lines 5F10 and 5F60 starting with header "60". These are the DP_X display entries from A to F

Version 1.0 is released : Release v1.0 | Nvidia-vBIOS-Clock-Power-Tweaker · JadeRover/Nvidia-vBIOS-Clock-Power-Tweaker Editing and saving vbios is possible + checksum is fixed. Therefore I need volunteers with an external flasher to test on pascal cards these modifications !

Hello, yes there is a way to do so. The mac rumor vbios have different output table from stock vbios + documentation exists on the "DCB" table (display table) of nvidia vbios, that's the next step of the program is to read and edit these outputs. Unfortunately when testing with an RTX3000 gpu, my custom vbios did not work, the card would be bricked with error 43. Moreover, changing 2 adjacent bytes in the bios by doing +1 and -1 to the hex value (so that checksum is the same) results in a brick as well for this Turing card = meaning that Turing has more security checks then Pascal and any custom vbios bricks the card (also this was documented on mac rumors). AKA turing custom vbios won't be possible 😞

Hello, I don't think that 240Hz will work as the 7720/7520 are limited to DP1.2 signal speeds so that would be 190fps max at 1440p (according to wikipedia). What happens is that when you select that refresh rate the screen artifacts and then windows reverts to 60Hz after 15s (since you can't press the accept changes button) I don't think there are 16:9 OLED panels that support >60Hz. 16:10 they do exists, also from the vivobook line but then you would have to make sure it can actually fit in the display cover of the laptop. Go through 15.6inch OLED panels in panelook to find all the model numbers + specs

Yes, agreed, that info would be crucial. I had started a draft that would show the display table, all the info relative to "dispkay outputs" is stored in the DCB table : see here https://nvidia.github.io/open-gpu-doc/DCB/DCB-4.x-Specification.html Luckily this table has not been updated for a while so it should be compatible with all the bios from the Pascal-Lovelace era. If you want to help out you can contribute to the github. What we need is algorithms going through all the table entries and parsing the data in a user friendly format showing : DP_X = such interface + i2c controls ? (basically if it supports eDP) + enabled ?

Updated the repository -> version 0.3 Now turing cards read properly and the standalone windows exe works. @ssj92 You can now properly read the clock values of your T1000 and RTX5000

Yes that bitmap icon is killing me! It prevented me from exporting to a single exe file 😞 I would get an error

Hello everyone ! linked is my very very very early version of a pascal to ada lovelace vbios editor. as of right now it can only read bioses !! Here is the link, the details are explained in the repository https://github.com/JadeRover/Nvidia-vBIOS-Clock-Power-Tweaker Pictures :

Ok, unfortunately this will only get you so far, usually shorted capacitors/mosfets aren't visible with the naked eye. Darker areas on motherboards can be usual heatspots (around CPU VRM, battery charging circuit, etc). These can be false positives. Nothing stands out from the pictures you gave. Unknown chip is the CPU chipset btw. Ok, that's a bummer, a multimeter is a really usefull tool, even for everyday electronics (making sure wires are connected, etc). I would highy recommend getting one, even 10 dollar/euros multimeters are well worth the price in my opinion. Even with not that much knowledge, you can trace short circuits easily with one and know if your GPU is shorted, CPU vrm shorted, chipser shorted, etc... The next step would be to get the laptop to a repair shop, a good shop should have a motherboard diagonistics fee. Make sure they actually diagnose the motherboard properly, most repair shops just plug it in, say it's dead and recommend a motherboard replacement. You can read reviews/check the websites of repair shops and make sure they actually do real motherboard testing.

Do you have a multimeter ? If so you can look up guides for "check laptop power rails" and try to find the short. If I had to guess you probably have a main power rail short = would explain the no reaction to power button. What about booting the laptop without a GPU ? Do you have a spare cpu as well for testing ?

Hi, I think for dual gpu laptops it's worth a shot to tape off that PWR_LEVEL pin and see what happens. What is the performance limit reason in gpu z for the affected gpu? If it's "power" when it shouldn't be (seing 60w instead of the real 90w power limit) then it definitely can be the PWR_LEVEL signal causing this. Keep in mind that this fix doesn't give the gpu infinite power limit, you still have the power limit set in the vbios. This fix prevents the EC of the laptop telling the gpu to throttle it's power.

Hello eveyrone, Ok so this problem applies only to MXM gpus that have a "pwr" limit (as well as "thrm" limit sometimes) in GPU-Z when doing a benchmark. This is also known as the "40w power bug" Such example : Notice this is not due to hitting the power limit of the card (33% of TDP). Symptoms of the low power bug is a TDP usage around 50% = 40w if you have a TDP of 80w. I think I found the issue. It has to due with a digital signal that is sent to the GPU, the PWR_LEVEL signal : A previous fix was to block the SMB_DAT or SMB_CLK signal so that the EC chip of the laptop wouldn't be able to communicate with the MXM dGPU. This has the disadvantage of stopping all data communication between EC and MXM dGPU = no more thermal data coming in = EC usually ramps up the GPU fan to maximum. According to MXM 3.1 documentation : This signal is not mendatory and it does apply some kind of power reduction when enabled. Therefore I took a gamble and decided to cover it up and see what happens. My fix is to block off the PWR_LEVEL pin with a bit of tape. - I first put some normal paper scotch tape on my plastic desk (to get it nice and flat during cutting process) - I cut it into a thin strip with a very sharp cutter(width of MXM pin) - I removed it from my plastic desk, then applied it with tweezers. The pin is the 9th counting from the first "thin" pin, on the TOP of the MXM module aka side of the MXM gpu where there is the actual GPU die : It's okay if you cover up the 8-5th pin as well as these are reserved pins (no connection inside a laptop) I did this on my dell precision 7720 with GTX 1060M and this has so far resolved the issue, my card can now boost to 100w thanks to the modded bios. Hope this helps someone !

You are right, I think it should work afterall, using nvflashk. But I would really highly recommend a flasher anyway as the first vbios I can make will prob result in bricking the gpu unill fine tuning is done. Also, my program is coming along nicely, I can reliably find the Core clocks and mem clocks of pascal. I can also reliably get the core + mem clocks of Turing and above gpus (tested up to 4060m vbios) . But this is only using the virtual p state table. Meaning I can only edit the values of "boost" you see in gpu z. This is not the actual boost value : rtx3000 can boost to 1820mhz and that clock is not in the v p table. It must be somewhere else. This is not the case for pascal as this second "real boost" clock is in the v p table. They must have done slight changes between pascal and further generations. Because of boost 2 -> boost 3 modes. Mem overclocking is possible though, as well as power limit modification.

I can probably write the code this week, thing is I would need people to test the vbios with fixed checksum that are made, I only have a GTX1060 (in a dead precision 7720 I gotta fix) + Flashing would only work with an spi programmer, CH341a and such. modifying GPU clock + MEM clock is usefull only for quadro cards that can't overclock in software, all GTX + RTX cards have OC sliders unlocked in MSI afterburner. Power limits modification is usefull for Quadros + GTX cards I guess.

Here, look in hex for "28 4F 4C" it's the same from pascal -> blackwell The actual TDP values is the "text" are after the (F...(F -> Here it's oS...oS -> in hex it's F8 24 01 -> after little indian 01 24 F8 = 75 000(d) so 75w. (example is p3000 vbios) Def + Max value (not sure of the order though) I could probably write a small tool that can sniff out and find the power + GPU clock + MEM clock for pascal -> blackwell gpus. Shouldn't be that hard, problem is fiwing the checksum after modifications.