The dGPU has normal Thermal Paste. It went down by 4-5C under full Load. When i set the Thermal Limit of my CPU to 65C during Gaming with Universal X86 Tuning Utility the Framework is pretty impressive how Silent it can be. BTW when Gaming the CPU only spikes to 35-40w (mostly below 30w) but for everyone affected by the Heattransfer Problem Thats Turbinesauna Territory.
2 Likes
As to your bending. I inspected mine and its hard to compare to yours. I aimed down the straight edge and i couldn’t notice any bending in the Bottom Case. My Screen is a little flexed, but not the Lower Casing. Sorry bad Focus 
1 Like
Eventually you guys could setup a “script-set” so people can replicate the exact tests you did.
This would enable a decent comparison table. We could indirectly also identify bad setup with that?
Note: We’ll need to take into account Windows and Linux OS.
You can just do a Cinebench R23 Run in Linux or in Windows, they do not defer that much. With Monitoring Tools in Linux i got less Points then in Windows. But a good Rule of Thumb is: If you hare in High Performance Mode in Linux or Windows and you are below 15k Points in CB R23 you have very likely a Throttling Issue.
Tested with Nobara 40
There were 3 scenario’s I tested.
One with only the thermal pads beneath the core.
One with thermal pads beneath the core and on top of the heat-pipes, below the top plate.
And completely stock.
After running multiple tests and reviewing the scores, i cannot say I see any significant increase or decrease in performance running the benchmark.
I see better results on a cold start (no heat soak present in the system)
I see slightly better thermals on the over al lows with the pads installed, which makes sense as there’s more surface area to dissipate heat too that is outside of the laptop body.
Having both the pads beneath and at the top also disproves the theory that it might be an issue with the mounting pressure, as with this config the cpu and heatsink/heat-pipes are basically sandwiched, with both bottom and top pressure (not a lot, but if it matter I would see some differences).
So for conclusions, please note, only 1 system was tested;
You can use this type of padding to slow the rate of system heat soaking, but I don’t see much benefit from it.
Mounting pressure seems not to be the issue.
Cause for the uneven thermals still to be determined.
side note, on a cold start I have seen it pull as high as 79 corwatt and all cores average hit ~4400mhz on full load. I have the R9.
3 Likes
I have some other idea’s but I need more information.
If you know anything about CPU details, please check:
1 Like
I Already did think about this Topic, I suspect Core 4-8 Are located in the Area nearest to the “Triangle” Edge, as those Cores are the hottest ones mostly and the Better cooled Cores are the ones more on the Edge of the Die.
As you can See on my Picture how the PTM is Spread. And where there is more PTM on the Heatsink/Core there is a “High Spot” as its Where the worst Heattransfer happens.
This Picture was taken after my First “Lapping” Attempt, but its nearly Identical to before Lapping. In The Later Lapping Runs i tried removing more Material on the Inner Area and the Other Sides, that helped a bit with better Spread Heattransfer.
3 Likes
No snark intended with the following question. It looks like those who have had their boards exchanged have seen a performance increase of 10-15 percent. Is that accurate? Are there other benefits beyond better performance at full load? I finally restarted to windows and tested my system, just curious about the pros and cons. Thanks all.
1 Like
Yeah, I’d say that’s accurate. The main benefit is CPU performance finally reaching the range expected from the 7840HS/7940HS at the power draw Framework themselves are targeting (45W sustained, somewhere between 55-65W burst peak).
However, just a bit of warning: I already had my Mainboard replaced once and, while it did perform within spec initially, it has dropped about 10% of its total performance after a little over a month AND now the core-to-core temperature delta has increased massively from 6.8°C to 20°C now. Frankly, I’m not sure what to make of this development, and clearly neither does Framework, since they’re now requesting that I send in my entire laptop for them to inspect. The cause of the thermal issues (which in turn cause the performance deficiencies) is apparently undetermined, at least to our knowledge.
2 Likes
Thank you for the reply. My test numbers weren’t great, but for my use case I’m very rarely taxing the machine, so I think I will sit tight for now, rather than go through the hassle of swapping things out and potentially winding up where I started. I appreciate the insight. Have a great day!
1 Like
The greatest Benefit if it works as intended, the Laptop is staying alot quieter during Low to Medium Use. The Cpu is only Hitting 70C at 30W instead of 100C at 30w is quiet a Jump on Noise. Why i started Testing is that i was annoyed, that the Framework was a Turbine even on low end Tasks and even Hitting 90C during Startup. which i did not expect after the Great Reviews.
3 Likes
That makes sense. I am not questioning any issues that you or others had or have, I’m just trying to understand things so that I can make an informed choice. My machine doesn’t run hot or noisy, but again, I typically do not tax it very much. I have the 7940 and 96 gb of ram. I got just over 14k in cinebench r23, so presumably that is lower than what it should be, but if the system is working for my use case, and if going through the hassle of swapping things out may have me still in the same place, I don’t know if I will pursue it. Others may make a different choice in the same circumstance, and that’s fine. I’m in waffle mode at the moment.
1 Like
Same, I started wondering what was happening with my system, fans blowing while basically it was doing nothing.
Then I stumbled on different topics in these forums and did the bench-marking and found this is not right. The board is replaced and it is hitting the numbers it should.
I am not sure if this was happening from the start as I wasn’t using the system as often as I had been more recently.
I will be on a quest getting the temps down in normal use just lower the rate for the fans to spin up, or just lower rpms in general prolonged use.
1 Like
Laptop cooling can be a bit finicky. It’s common to have certain pastes diminish in performance over a certain temperature. You can also get pump out which is when heat cycles kind of move the paste around. I wonder id putting in a pad from Thermal Grizzley would solve the issue. Carbonaut or Kryonaut.
The main reason for using Liquid metal:
- Durability: Once applied correctly, liquid metal TIMs can last longer than traditional thermal pastes because they don’t dry out or degrade over time in the same way. However, they do require careful handling due to their reactivity with some metals like aluminum.
It also has a better heat conductivity, but that doesn’t matter if it’s not making sufficient contact with the IHS and head-sink. So imho that doesn’t apply in the current implementation by framework.
I’m considering one of those Honeywell PTM7950 pads or one thermal grizzly products. To see if it improves the contact which would indirectly increase the heat dissipation and lower temps or decrease fan rpm.
For a laptop that is so easily serviceable, the downside of paste isn’t really that bad.
2 Likes
I’ve used liquid metal before and also had to repair board due to use. It definitely can mess up things due to its conductivity and messy thin nature. I haven’t taken off the heatsink or checked to see what my thermals are yet. My fan does pulse sometimes though. My thinking is carbonaut reusable pad would fit the spirit of the device. You can check out the cooling and just put back the pad. Wont pump out. Doesn’t require any intense care when installing. Maybe this is something framework could deploy via support that is cheap and mailable too.
Ill check out my thermals later. Maybe Ill be a guinea pig.
2 Likes
I’m still waiting for a reply back on my support ticket. Depending on what they say, maybe I’ll try some alternatives to the factory liquid metal. PTM-7950, IC Graphite pad, etc. We’ll see. I don’t know that mine is bad enough that I’d want them to send me a whole new board or anything. Unless they are wanting to analyze the board or something. I’d just as soon have them send me a new thermal pad to try. Then I can try some other things first, just to see how they compare.
1 Like
Mine was bad enough that they cross-shipped me a new Mainboard which did meet their expectations - for a few weeks, that is. Now they want me to send the entire laptop in for analysis and repair. Quite the troublesome issue they’ve got on their hands here…
That’s another reason why I think I might lean toward trying some thermal pads before having them spend the money to send me a new mainboard that may not be any better.
1 Like
Yeah, I’m currently on the fence as to whether or not I actually want to commit to sending it to them or just winging it with modifications of my own.
The peculiar thing about the entire situation is that the current cooling solution has demonstrated itself capable of meeting and exceeding Framework’s own stated performance target of 45W of sustained load. The big question is what the point of failure is causing the drops in sustained power output (and performance, by extension).
As @PSierra117 documented over in FW16 Batch 20 Guild, switching to what he believed at the time was PTM7950 and lapping the contact plate on the heatsink brought his board beyond Framework’s own targets. Recently, he did report that performance had dropped again to just within spec, leading him to believe that the “PTM7950” he received may actually be a different TIM altogether, so the jury’s still out on whether that’s the cause of his observed performance loss after modification.
Like I said, it’s really strange stuff going on here. Here’s hoping the FW team can figure things out.
3 Likes