Just some info for everyone else. FW support reached out with the request to run R23 and provide thermals.
I did so - and, the results were pretty much as everyone else said:
As seen, a temp delta of about 15 degrees, a really low score, and Core 4 was at 100c pretty much the whole time.
Biggest issue is only getting about 36 watts of package power.
They evened out more and more, and the coldest Core switched to another Core. But my highest Delta is still 9C.
@knipp30
Ouch, yep thats far off. Mine is pulling around 54w decreasing to 50w over 2 or 3 Loops and my Core Frequency is between 4.16-4.4 during cinebench R23.
Yeah, I was only getting 35.3W on mine. My RMA has been approved & getting sent an new motherboard.
Hopefully that will resolve it.
Well, I just checked and it appears to be similar to all you guys.
Biggest dT is 15 degrees Celsius. While the benchmark is running, all cores are at 4 GHz.
I’ve submitted a support request, We’ll see what happens
Pretty much same as mine - about 36W, but for 7940HS.
Waiting for replacement motherboard…
And result… less than 10k points for multicore…
Well thats the same Scores i had with the Mainboard i got replaced. You can see, that your Device pulls only 46W at max. Are you in Best Performance Mode and charged to your Set Charge Level?
and Mahlzeit nach Deutschland 
AS @ PSierra117 mentioned - please check that you have “Best Performance” as Power Mode in Settings (see below) and if it is, and you have 180W power switched on (not on battery), then your Max Power consumption for CPU should be 45W+.
Note: Looks like your max is 43W, which is low for non-battery best performance mode… I have very similar initially before replacement.
Also make sure you have latest BIOS (6.03.03 - here). Framework support also asked me to reset BIOS few times as well as reset Motherboard, but in my case nothing changed…
I received the new laptop today.
There’s a difference between Core #1 ( 81.2 C ) and Core #2 ( 99.9 C ).
The good part is that instead of 12950 points now I have 15685.
@PSierra117 out of curiosity, how is it holding up?
Any degradation/delta increase?
I’m asking because mine went from a best score of 15685 to 14971.
Somehow, I can’t get more than 15k anymore and the lowest value for the CPU Core temperature is 77.6C while the highest is 100C.
I already let the support team know, but, I’m not sure if I can go through all the tests again.
Well my modded Heattransfersystem 
had a slight degradation but its holding up fine. So instead of 16360 im at 15614 right now. I Suspect, that my PTM7950 Batch is a Fake one so its PCM6000 or something like this. I Will replace it with Genuine One when i get Home again in some Weeks.
This is a Single Run instantly after i did a 10min Loop (My X86 Utility messed up the Loop as it throttled my Thermal Limit to 70C midway through the Run so 2 Passes were off and i only scored 15035 Points in the Loop.
I did some Lapping and further modification i didn’t show here. I Also bend the 4 Mounting Legs from the CPU Cooler upwards to increase the Clamping Pressure a bit as PTM needs a fair Amount of Clamp to work perfectly. I furthermore Modified the GPU with my Fake PTM and reduced the AirGaps at the GPU Module with Duct Tape to decrease the Lost Air Mass and Funnel more through the Vents and Heatsinks. So maybe more Air gets through the CPU Heatsink too if its enough to slightly increase the Static Pressure at the GPU Module. But its all just to satisfy my Tinkering Needs as i suspects it just changes some things with a Margin of Error.
My GPU is at Maximum at 70C while Gaming in High Performance Mode.
I Really need a 240W Power Brick
Does @Alex_Uta have a 7840 or 7940 chip. 15k is about right according to the replacement mainboard results from @Snipe.
I did notice my performance degraded as I installed more programs. I think Windows often slows down a little over time. And apparently Windows 11 23H2 can cause benchmarks to go up/down as much 20% on desktop Ryzen CPUs per breaking news from Hardware Unboxed https://youtu.be/izqEZmjTfuM?si=--iBivGlZcvZKCWy. Does that apply to mobile chips? Maybe not, not sure. Does that apply to Cinebench R23? Maybe not, not sure.
In any case once you install programs like Adobe Reader, there are hidden programs to check for updates and licensing (on a free product!!!) which always run in the background and suck away a little performance.
I was able to get a better score by looking in Task Manager to see how many processes were running after boot, and carefully finding all the places where startup programs are triggered and disabling them, then rebooting and rerunning the benchmark.
Of course there are a few programs like PowerToys which I typically do want running on startup, so I add those back when I’m done benchmarking. And Adobe Reader inevitably forces itself back into startup again after a few days or weeks.
His is a 7940HS and its again behind my 7840hs. During my Test i posted today i had alot of Backgroundtasks open. Like Steam, Powertoys, X86 Universal Tuning Util etc. So nope 15k ist Not Okay.
@Micah_Lindstrom I have Ryzen™ 9 7940HS, but, now I’m curious, so, guess what? I’ll reinstall my windows 
This is the good part about using two different HDDs, one for the system and the second one for the main files.
@PSierra117 I think that yours is a lot more stable.
Do you have any photos? I was looking the other day and I can clearly see some space between the fans and the cooling system from the CPU.
For a Windows Fresh Install i use the CTT Winutil every Time right after a Reinstall for Debloat and Telemetry Delete and after every Major Update. Reduces the Background Processes to around 130 plus the Tool + UnigetUI can be used as Package Managers (like in Linux) to install and maintain any Program/Tool you need. My Current Process Count right after Restart with Tools like Logi Options, Cinebench Hwinfo etc is 194.
I do not have any Pictures by Hand.
Following up on this:
Seeing all this new discussion led me to re-check my new Mainboard’s performance, since it’s been a month since the replacement. I’m very sad to report that the situation is now somehow worse than it was before.
I replicated the same environment as my first test with the new Mainboard:
Upon running Cinebench R23 Multi-core again, the board is now scoring ~13700, with a max CPU Package Power of 37W and a core-to-core delta of 15C.
I can’t decide if I should be surprised or not. Clearly, something is going awry with Framework’s thermal solution for the FW16, but it’s not entirely clear to me what is the precise cause. It could be the Coollaboratory LM pad they’re using, but it could also be an issue with the coldplate. Either way, it seems like I’ll have to contact Framework Support once more.
Sad to hear, but not unexpected… There is pending Issue there, thats why they approve the Replacement pretty easy. But its still not resolved on Frameworks Side and also still no official Statement. Its hard to defend such an expensive Purchase.
After my Modification i am still strong where this CPU should perform.
I expanded my Support Request and had Questions about the Heatsink and LM Pad and furthermore. The Heatsink on Sale is the same as on Current Batches. They sell the Heatsink and not the Pad as the Pad is not included with the HS. Why not the Pad? Because they do not have a Replacement Guide is the official Answer. I call this BS. I doesnt add up. I really think they are Rotating behind the scenes. Either they got Screwed by the Manufacturers or they hoped it can slip. But its an Enthusiast Group they have as a Customer Group. So any Problems will Surface.
Sorry about this! 
Thanks for posting this.
Now, I’m not even sure if I should ask for a new mainboard or not.
@Chrysolite do you remember what score you had before? I’m curious about the degradation.
@Micah_Lindstrom I just reinstalled Windows and ran a test without even connecting to the internet.
So, no updates or apps installed.
@PSierra117 I wish we could get a mainboard replacement without Liquid Metal 
.
I can do even better than remember the score; I have the screenshot from the CBR23 run as well as the ranking log file.
As you can see, the average score over the course of the 10 minute run (12 passes) was 15670. The max CPU Package Power was 57.8W, with an average of 46.8W. The core-to-core delta (with Core0 being the coolest and Core3 being the hottest) was 6.8°C.
Now compare this to the test run I made today:
The performance loss is incredibly apparent from the get-go. We’ve now dropped to a 13952 average score over 11 runs (an 11% decrease from before), with CPU Package Power maxing at 37.4W (a 35% decrease) and averaging 34W (a 27% decrease). The core-to-core delta has shifted dramatically as well. Core4 now takes Core3’s place as the hottest core, and the delta is now 14.3°C (over double the prior reading).
I severely doubt that this has anything to do with potential CPU die degradation, so some element of the cooling solution is most likely degrading, thereby kneecapping the system’s maximum performance overall. Thankfully, this doesn’t affect my day-to-day tasks, so it’s not the end of the world personally, but I sincerely hope Framework resolve this sooner rather than later.
