I liquid metaled my FW16

Gather round, grandpa is telling the story of how he liquid metaled the FW16.

This is a continuation of this thread CPU Benchmarks and TDP but I think it’s interesting enough to warrant a new thread. TLDR at the end.
My FW16 always was thermally limited. As soon as any all core load runs, it would immidiatly hit 100°C. I thightend the screws of the heat sink a bit, which helped a little bit. But than it got worse after a couple of weeks, until it could barely sustain 3.8Ghz.
I also noticed that the temperature difference between cores was 7-10K. The air expelled by the fans was also kinda cold. So I concluded that there is something wrong with the thermal interface material.

I could reach out to support, but I didn’t buy a framework to have other people have all the fun.
So I ordered some Thermal Grizzly Conductonaut, which arrived yesterday.

This morning I went to it. Getting the mainboard out was easy enough, some connectors where a bit fiddly, but ok. Remember to BIOS disconnect the battery before.
Getting the heatsink off was not so easy. The stock liquid metal is not liquid at room temperature, so I heated up the cooler with a heat gun. Surprisingly heat pipes are very good at conducting heat away, so the whole thing was too hot to handle and I almost dropped it.
I than cleaned off the stock not-so-liquid metal with repeated use of the heat gun, cotton swaps and alcohol pads. The alcohol dissolved the glue of the foam barrier and some of the protective glue on the SMD components. I also sprayed around some liquid metal. Lucky for me it was already solid again so I could just pick it off the mainboard.

On the heatsink there are small hexagonal shapes imprinted into the copper. They look like dirt - but they won’t come off. Cleaning to thing was really annoying, it doesn’t come off, it turns solid real fast and sticks into the tiny creavase around the CPU, I never got it out of there.

Applying Conductonaut was easy, I think because everything was liquid metaled before it spreads easier than usual.

I reassembeld the laptop, fighting with myself if I trust myself that I did everything right, or listing to Linus, who says to never close the side panel, before turning on the machine for the first time.

Plugged in the power supply, the led on the side turn orange - first success, I press the power button, it lights up, second success. I look at the screen - nothing.
I contemplate all my life choices that led up to this point, I’m the stupidest person alive. I broke a 1000€ mainboard just for curiosity. I will delete my forum account. Nobody needs to know how stupid I am. Why can’t I be happy with what I have?
The screen turns on. The framework logo appears. It was just memory training.

The BIOS forgot the iGPU memory setting, one needs to turn it to normal and than back to gaming.

I run Cinebench 24. 895 points (7840HS), so about the same as before. But I don’t hit 100°C anymore and the fans are a bit quiter.

TLDR: Do not liquid metal your FW16. It’s not worth the risk. I know you gonna do it anyway. Don’t blame me.

10 Likes

So still a win :wink:

3 Likes

Terrifying.

4 Likes

If I’m not mistaken, the AMD 7000 series, at least the desktop counterparts, are designed to push as hard as they can until hitting 100C. And are designed to be able to remain at 100C indefinitely.

I would assume the reason there’s not much of a change is because the framework team have made the thermal hardware able to handle the power limits, since there’s no point to trying to get the temps down when AMD designed the CPUs that way. (At least from a performance and longevity prospective) Stupid if you ask me, but I guess Alaska appreciates AMDs new designs.

2 Likes

I don’t mind the temperature as much, but I couldn’t stomach leaving performance on the table

Don’t do it boys and girls, really don’t do it.

1 Like

Stupid question but: Generally, sometimes we use to re-paste the laptop. There should be no such need with Liquid metal?

LM can fuse into the copper, so one might to add some after time. Since the inital fusion already happend with the stock LM, I don’t expect to redo it - ever.

:rofl:
No disagreement, you’re probably right!

Not Framework related, but liquid metal:

When i got my first AMD Ryzen 7 1800x (desktop PC) i did use Thermal Grizley conductonaut liquid metal.

4 years later my PC started to show odd behaviour. In the end it turned out that the liquid metal became solid. I have absolutely no idea how that was even possible.
So yeah, repasting liquid metal is a thing

The FW 16 coldplate is not nickel-plated but raw copper?

You don’t need to use a heat gun to clean off solid old LM - add a little conductonaut to it and you can liquefy it and remove it, ideally by pulling it off with a syringe. Alcohol pads don’t dissolve LM and cotton doesn’t absorb it. If you want something to do that it’s best to use (unheated!) solder wick.

Don’t do LM if you’re scared of it. I’ve used it on laptops for years (including an intel 12th gen FW and my 13 inch AMD FW) and been fine. These laptops get daily use, tossed in backpacks, no issues.

Copper isn’t susceptible to alloying with gallium the same way aluminum is, but some grades of copper (even certain finishes) will take in a tiny bit. I’ve seen some copper coldplates completely unaffected and some discolor immediately. Good LM uses an absolute minimum of gallium to maintain the liquid state and once enough is drawn out of it the bulk can solidify at room temp. This actually won’t affect temps in a lot of cases unless it crystallizes in a funky way, like there’s uneven heatsink pressure or some other mechanical issue.

1 Like

correct

Okay, I had a look at the “Liquid Metal Deep Dive” e-mail from December again, it is mentioned there that a “vapor chamber copper plate” sits above the processor package, I also found out that the used “Coollaboratory Liquid MetalPad” is in fact gallium-free and therefore is (according to the website: Coollaboratory Liquid MetalPad – Coollaboratory) compatible with both aluminum and copper coldplates, and shouldn’t need any repaste/replacement in a FW 16’s lifetime - at least if it has been applied properly in the factory to begin with. @samokosik

@Christian_Elsner Regarding “On the heatsink there are small hexagonal shapes imprinted into the copper”, the e-mail specified this about the coldplate surface: “An etched pattern in the surface of the vapor chamber holds the liquid metal through tension”.

4 Likes

Same for the 13, ptm is very worth it though and gets you most of the perks of lm without all the risks.

Doing 45W indefinitely on the 13 after I figured out how to get around the stupid stapm limit is pretty neat though XD

1 Like

How did you get around the stapm limit? Ive been trying to figure it out on my 16 and it has been a bit frustrating.

If mine has the high heat issues, its getting a fresh coat of noctua thermal grease. I got a large tube when I build my desktop, and have plenty left. Just for games, I don’t want to risk something that could short my motherboard.

Smokeless umaf, I posted the my settings here may need adjustments for the 16 though.

What are the stock stapm limits on the 16 anyway?

Stapm throttling wasn’t really a problem with the stock paste though as it stayed under the limit anyway, not sure how that works with the 16.

1 Like

My issue is even though my system is at around 75c when boosting at 54 watts for a long period of time, stapm eventually limits me down to 45-46 watts

On the 13 stapm throttling only happens above 80C so this may be a different issue, assuming you are looking at the right sensor of course.