SuperMG3

On 5/14/2025 at 11:38 AM, GuitarG said:

I use DC screen, meaning it's LVDS driven by EDP-to-LVDS board. That board is connected by EDP from mobo.
On mobo there is 50 pin connector, housing both LVDS and EDP signals. Which one are used is determined by GPU used.

About project I think that CPU is defective, every RAM I tested threw error in memtest, meaning RAM controller in CPU is shot. And I don't see any good CPU for sale, even on marketplace....

Hello. Did you test if RTX 3000 from HP works too? If it works, it means all RTX 3000/4000/5000 work at the end. I can get the cards work on the first slot of my laptop and have better driver supports for RTX 7000 GPUs on the second slot on 3rd gen hahaha

12 minutes ago, Yarema said:

You can undervolt it in MSI afterburner. With 0,875V it should run @115W

Hello. I tried that. Runs at 138W max.

@Khenglish

@panda_zzz

Since the driver 528.49 can allow power limiting for laptops.

But the latest can't allow it on laptops, only desktops can.

Is it possible to hack the new drivers and allow power limiting? We need to find the file that allows the power limit adjustments 

4 hours ago, Khenglish said:

Since you have a programmer it's fairly safe to try other vbios. The memory strap ID on your card will likely select Micron memory settings despite whatever Techpowerup says the vbios is for. For vbios that don't match you'll likely get a brick or error 43, but you can just use your programmer to flash back.

 

The vbios chip will be 1.8V, so make sure your programmer has an adapter to step 3.3v down to 1.8v.

Yes I have a 1.8V adapter. X-vision is trying to understand how I managed to limit the power using drivers.

8 hours ago, panda_zzz said:

No one knows that. The firmware itself is encrypted and signed - two layers of protection. There was no information in the public domain about bios modification since the Turing generation.
So you have a choice to drop  tdp via msi afterburner or try different versions of the bios suitable for the chip in your video card

I use Nvidia inspector but I am limited because I can't install new drivers...

1 hour ago, Khenglish said:

You're not going to be able to reduce the TDP with a modded vBIOS. A modded vBIOS will have an invalid digital signiture and the card will refuse to run.

 

There likely are other unmodded vBIOS you can flash on your card. The memory configuration is set by a set of 3 pairs of resistors on the card. Micron, Hynix, and Samsung all have different physical programmings. vBIOS often have the same encodings between vendors, so there are likely other vBIOS that are compatible despite the card the vBIOS being intended for never having Micron vRAM. It is dangerous to experiment with flashing other vBIOS though through nvflash without having a physical programmer. nvflash will often let you flash in one direction, but refuse to let you flash back due to board ID mismatch, so you can brick the card trying this.

 

Another option is to remove and replace the current measuring shunt resistor on the card. There will be a 5mOhm resistor that you could replace with say a 6mOhm resistor for a 17% power reduction. This is a big resistor though that most soldering irons are not strong enough to remove.

Hello. I already said in the post that I know I can't flash unsigned vBIOS but I still want to know where the values are in the vbios file.

I managed to lower the TDP using the Nvidia driver bug from version 528.49 but it means I can't use the latest drivers anymore...

Anyone please?

TheDidgeridooMan aka the RTX 3060 TI modder, are you alive?

I don't understand a single crap when reading the file in hexadecimal... No way to convert hex to readable language? Like C/C++?

Even the manufacturer ignores my request... (Despite I told them I wanted to pay for a modified VBIOS )

No hardware mods needed to decrease the target TDP and to increase the thermal limit...

Hello. I managed to bend the heatpipe but didn't glue it yet because I need to do some chassis mods.

I also made a new post about the possibility to edit the 4080's vBios?

I'm a little bit late.

Actually, the vBIOS is uploaded on techpowerup. It's the "X-Vsion 4080 12GB".

Hello again.

I have an RTX 4080 MXM card. My laptop is a Clevo P570WM that can handle max to 115W of heat.

The MXM card can use up to 150W. On GPU-Z it says max 150W too. The max temperature before thermal throttling is 83C which is annoying.

I just wanted to know where I can find the value (inside the file) for the TARGET LIMIT of the TDP and the MAX THERMAL LIMIT in celsius.

I want to set the max target limit to 115W and the max thermal limit to 87C.

I can't flash any random 4080 laptop VBIOS because my card has MICRON VRAM chips. It means, every VBIOS available on the net only use Hynix/Samsung combo, no Micron available.

I know unsigned vBIOS without the correct checksum won't work. I do own a CH314A for flashing tho.

I just want to know if someone knows how to VBIOS mod the TDP limits and thermal limits. I know someone edited the vBIOS of an RTX 3060 Ti Ampère but he didn't explain how sadly...

So if any of you knows, please write it down.

Thanks.

8 hours ago, Clyde said:

Back in time, Eurocom sold a dual-fan heatsink designed for a single GPU for this model. Ask them if they can still provide you with.
If not, you can always make a DIY one. 😉

No space for one. I already use two GPUs in one system. I can just add one extra heatpipe, I tried to bend one, it's ugly but still manages to conduct the heat (hot water test).

I bought 2 more heat pipes because mine modded one can't fit since I did something wrong in the first angle.

1 hour ago, panda_zzz said:

Do you see my heatsink? There's a place to add one heat pipe, I need to curve them like the first and second heat pipes.

Do you see my heatsink? There's a place to add one heat pipe, I need to curve them like the first and second heat pipes.

EDIT: Bruh I bought brass heat pipe... (Copper and some zinc)

19 minutes ago, panda_zzz said:

 

Yes this one too. He used a rubber toy 

14 hours ago, SuperMG3 said:

I still haven't tried the 4070 (115W) on the Clevo P570WM but on my Dell Precision M6700 the temps were so high and the memory temps reached over 115C

 

Do you guys think I should add a 4th heatpipe?

 

I have 3 heat pipes. 2 for GPU and one for top VRAM/VRM

I can't send a picture because of the limit but I have a place to add a 3rd heat pipe on the copper plate of the GPU die.

I also have some margin to glue the heat pipe on the heatsink. 

I bought 3 18cm heatpipes because it's an enough length. I bought some thermal glue that works between -60C to 250C.

I did see a video on how to bend a heat pipe with bare hands, but don't you guys have something safer and can be done for free?

I still haven't tried the 4070 (115W) on the Clevo P570WM but on my Dell Precision M6700 the temps were so high and the memory temps reached over 115C

Do you guys think I should add a 4th heatpipe?

I have 3 heat pipes. 2 for GPU and one for top VRAM/VRM

12 minutes ago, ssj92 said:

I don't put liquid metal on my components anymore. 

 

PTM7950 should work well if the thermal pads are thin and allow proper contact to the core. 

 

So it seems 4080 supports eDP. 

 

Would be interesting to see it's thermal performance in M18xR2 with 3080 heatsink. 

 

If I can cool a 125w WX7100 OC in M17xR3, M18xR2 should be able to cool 4080 with 3080 heatsink in M18xR2. 

 

If that is the stock heat sink you're looking at 100w dissipation 

No all the RTX 4000 don't support eDP. 4070,4080 NO EDP!!!

I get 255FPS on FurMark at 1080p with 120W limit. Still hits the thermal throttling at the end of the test but not by too much (95-100W)

3 minutes ago, Maro97 said:

I know because you don’t do it before 

Watch the video and you’ll understand — I used to do the same as you, estimating the size of the thermal pads on GPU cards in all my laptops just by eye, without any actual measurements using a millimeter gauge. But when I applied the method shown in this video on one of the laptops I own, I realized that you must calculate the thickness of the thermal pads for each component they will cover before placing the heatsink on the GPU die.

And why mine has big gaps? If they all touch?

5 minutes ago, Maro97 said:

and in your case don’t advice to use a cooper shim on the gpu core

Eurocom told me to use a copper shim lol

4 minutes ago, Maro97 said:

Heat dissipation from the heatsink will be weak with PTM and these thermal pads, as the gap between the GPU core and the heatsink will be relatively large.

Use liquid metal along with properly sized thermal pads, because the GPU’s power is high compared to the heatsink’s performance

I don't understand the video because I can't watch it right now. I don't understand what gap do you mean? Between the GPU die and the copper heatsink?

21 minutes ago, Maro97 said:

And don’t forget this 

In the process of insulating the capacitors around the GPU chip

https://a.co/d/0SE7XoO

That's how I did it

1 minute ago, Maro97 said:

100w heatsink with 150w gpu 

I can lower to 110W using MSI Afterburner but...

It also decreases the thermal limit. From 83C to 75C. Yes it thermal throttles at 75C at 110W... Funny

I don't know. My PTM 7950 is a 20mm thick one instead of the 25mm one.

I saw some tests with liquid metal and there are no gains.

2 minutes ago, Maro97 said:

The company wasn’t wrong in what it said — the only device that has a GPU heatsink capable of handling power limits reaching up to 400W is the P870TM

For my model, that's kinda stupid to make the slot use up to 200W but only having a 100W heatsink 😕

You sure Liquid metal is better than PTM 7950? Won't it harm the chip? Any protection to add?

5 minutes ago, Maro97 said:

Use nvidia driver 528.24 the tdp pwr limit slider will activate 

Did you read my my post? I did use it. If I didn't use that driver, how am I gonna change power limits? The only thing I can't change is the thermal limit, it's grayed and no values.

Eurocom's played with the words. Now they say 200W is only for the power delivery haha