Hi, I had the same issue but I just disassembled the heatsink based on the iFixIt teardown. For future reference if anyone runs into this thread, the foil between the heatsink and the chip was not taken off. I scrubbed down the PTM since I figure it wouldn’t work well after cycling into foil and applied normal thermal paste, now the system functions normally (although no PTM sucks of course…)
3 Likes
I can confirm, ive wasted 2 weeks of my life doing back and forth with framework support asking me to do all kinds of test and to wait as my ticket is being escalated. I saw this post and guess what
5 Likes
So many questions. Beggars belief!
Uff. That is steep. I would not have expected that level of messing up. Sure, errors can occur in production but aren’t they testing equipment before sending it away? Such a film would be obvious at a simple startup of the system with some Live OS, that has the temperatures readings already on the deskop or something. That way also booting issues could be filtered out.
But it makes sense. I would have been surprised if a PTM could fail so dramatically in heat transfer just because of a slightly non-optimal seating of the heatsink.
3 Likes
Yeeah it kind of feels like there was no burn-in test of any kind after assembly, either that or it’s somewhat inadequate. I understand that Framework is not the biggest company but the amount of tinkering required with my mobo was a bit overkill. Maybe they just leave the foil for all standalone mobos and it wasn’t communicated well? I’m not sure.
1 Like
Framework isn’t actually producing anything. Their Taiwan partner is. I forgot their name but it is a really big OEM. Maybe Framework needs to step up the game at enforcing quality control there, or need to pay for more quality control.
Is it Compal?
1 Like
We are looking into this miss. Please open a support ticket, if you don’t already have one, if any adverse affects have arisen from this.
1 Like
Will you both please shoot me a DM with the email address you used to contact support? I want to make sure I have both of your tickets. Thanks in advance.
I got an email from support earlier today that offered me another PTM kit, which was nice! I’m happy to see that we’re being taken care of.
2 Likes
For me (Batch 13) temps shoot up to 100 degrees seconds after a heavy load is introduced. They stay that way, HWiNFO reports sustained CPU Package Power in the range of 100W. When back to idle, the temps fall quickly to 30-ish. Throughout the test the heatsink is at best warm in touch.
Got to say, I don’t love this. I read the thread and it seems things are not as bad as for OP, but still, not what I expected.
Should I simply report this to the support or is this within normal parameters?
Definitely report this to the customer support. This sounds to me like something is wrong with the PTM pad or the heat sink mounting. Possibly the protective film has not been removed as in the cases above (which makes you wonder if there is any quality control at the end of assembly)
Compromised cooling like that will seriously cripple the APU’s performance and isn’t an acceptable state.
If you are told to reseat the heatsink, demand a free replacement PTM pad. If they deliver a faulty product, that is the least Framework can do (and from what I have heard they are also doing it)
1 Like
Run some more tests and it is getting weird.
Here’s the benchmark, and the full results are here. I picked a few of the benchmarks they were using and… I got same-ish results. 100% CPU load, around 100W draw and nice temps of ~70°C.
However, those are not the tests I run yesterday. Yesterday I run this particular browser-based test from PCMod.pl (Firefox) I stumbled upon yesterday. If you want to try it yourself, just press Start Benchmark - looking at the JS it seems to be an ad-hoc test with matrix multiplication. No GPU load. I get 50% CPU load, 100W draw and ~100°C. I simply don’t get it.
Here are the readings I just took to convince myself I am not crazy (LibreHardwareMonitor, consistent with HWiNFO):
What is going on here?
I’m very surprised at how difficult it is to source the Honeywell PTM7958 thermal pad used by Framework. Makes me wonder if there’s a difference between 7950 and 7958?
Is there another pad that is comparable?
My board got up into the high 90s when first booting into Windows. The replacement is at least 10 degrees cooler under the same operation
Different loads heat up different parts of the CPU and there are differences on heat dissipation, could be related to Uneven CPU thermals!
1.3k replies 
Thanks, seems like a possible explanation.
For more context, even if I set the fan RPM to 100% and still the cursed test brings the APU temperature close to 100°C. I’d love to know if I am the only one suffering from this.
By the way, I noticed the fan was very lazy to start spinning with default settings. So now I am sporting “Cpu Fan Sensor” mode set to “CPU” rather than the default “CPU Average” and now it reacts more quickly. Are there any resources explaining what those settings actually mean?
Did some more benchmarks. It all boils down to the power per core. Benchmarks with 100% CPU Load draw ~6.5W per core, whereas the 50% CPU Load one has 8 cores going at close to 13 W. Here’s how it looks on a graph:
It starts making sense now, but I am still not 100% sure whether temperature for the second case is pathological or not.
What determines which cores are loaded by which type of request? I would have thought this is at the kernal level? If at the kernal level, are these temps replicatable across different OS or are they OS dependent?
So, Thermal Grizzly make an alternative - reviewed here. The reviewer noted it’s a little more tricky to use correctly than the Honeywell PTM7950