Seeing this post led me to check NotebookCheck’s numbers on the 7840HS to compare them with mine and, unfortunately, my unit is also performing well under spec.
I’m scoring ~14200 in R23 Multi when the NC minimum is 15515 and median is 16194. Single-core is also suffering, with ny unit scoring 1660 whereas the NC minimum is 1718 with a median of 1775. My CPU Package Power spikes to 47W but averages 40W.
Setting Windows 11 to Best Performance did not seem to affect scores substantially. I will be starting an RMA request with Framework tomorrow.
Careful, you can’t compare against NotebookCheck since they have the slightly faster 7940HS chip (which I got too).
Comparing against PSierra and Snipe who have your same chip, it appears your multicore score is a bit too low. But I don’t think I’ve seen anyone share R23 single-core results for a 7840HS chip so nothing to compare there. I would guess your single-core result matches expectations. Maybe try Cinebench 2024 as well, since LTT shared their scores for that with your same chip the 7840HS.
I highly suggest checking your CPU Package Power as well. Per , both chips should consume about the same amount of power. My first mainboard’s chip would consume only about 28W during benchmarks while plugged into the 180W charger, at full 100% battery, with Windows set to “Best performance”. My replacement starts out at about 49W and gradually drops to 46W as the computer heats up during the benchmark.
As far as I know, cooling is the only reason for the chip to not draw 45W or more (besides power modes of course; it draws about 16W during my benchmarks on battery with Windows set to “Best power efficiency”, which is expected).
Well funny enough, i did 2 CBR23 Runs today to check against Notebookcheck Framework16 and 7840HS Average.
Mine was also at ~14300 Multicore and around 1700~ Singlecore.
14300 is greatly underspec, Above 16000 is the Average 7840HS on Notebookcheck. There must be an underlying Problem with QS or maybe Batch 20.
The New Mainboard has the same Manufacturing Date, but the whole Assembly looks revised, there are new Foam Pads on the Heatsink.
And the best of all, i get about 1000 Points more in CB23 MC and the Temps are much more even overall!
Sounds like your R23 multi-core score now enters the range expected for the 7840HS, certainly good news. I actually just opened a support ticket this afternoon, we’ll see what the say.
Well certainly there is some Problems with either the Cooling System or the “Revolotionary Liquid Metal”
As there was still a Thermaldelta and it didn’t feel quite right, even with my new Mainboard.
I decided to remove the Heatsink and take a look. The Heatsink looks, ehm funny, it has hexagonal Bumbs. I Suspect them to be the Reason of the High Tempdelta on the FW16. As those Bumbs has to be filled by Thermal Interface Material. Possibly those are there to help with the surface Tension to keep the LM in Place.
Yes i know, the LM Pad is still not orderable. I Decided to remove all the PTM LM and install PTM 7950. Well, The Thermaldelta got reduced and i got even more CB R23 Points after the initial Burnin. And then today i removed the Heatpipe again and took a look how the PTM 7950 was spread.
Same but more Uniform.
I have decided to “Lap” the Hexagonal Surface to reduce the Bumps and reapply PTM 7950. Now if i ever want to get Liquid Metal under there again i have to get a new Heatsink, as it was coated Copper. And bare Copper without Coating hates Liquid Metal.BUT now i get nearly 15900 Points in CB R23 and the Thermals are more even across all Cores. With the Original Assembly it was 15400.
Edit:
Well i did a new Run right now and i got even more Points, So the PTM 7950 is still in its Burnin Cycle. The CPU Power goes up to 61w and reduces to 51W during an Single Run.
First off, props to you for having the nerve to remove the heatsink and give this a look. Second, I figured PTM7950 would make for a great TIM for something like the FW16, so I will definitely consider making use of it if/when I get a new Mainboard.
And speaking of the RMA, I’ll keep y’all posted on how the RMA process goes for me.
I received my first response from Framework to my original support query. Here’s what they sent:
Thanks for contacting Framework Support. We’re sorry to hear that you are having an issue with your Framework laptop.
We apologize for the delay, as we’re having a high ticket volume.
In order to determine the best course of action, may we ask you:
When did the issue start?
Did you make any changes to your laptop before the issue occurred? (hardware/software)
Have you tried updating the BIOS? If not, kindly try to update it and see if there are any changes.
Try to set the BIOS settings to default and see if there are any changes.
Access the BIOS by tapping F2 or F12 during the boot-up.
Once in the BIOS, please press F9 to load the default setup.
Please press F10 to save and exit.
Does the issue occur when the laptop is connected to the charger? Or battery? Or both?
We noticed that your order does not include SSD and RAM. For documentation purposes, can you share with us the model and specs of your SDD and RAM?
Can you visually inspect the fan(s) and look for any debris or obstructions that may be blocking the airflow? If possible, kindly try to clean the fan using compressed air.
Are there any hubs, docking stations, printers, or other devices plugged into your laptop? If yes, kindly share with us the exact model.
Can you try to remove all the expansion cards and see if there are any changes?
We are looking forward to your response. Thank you.
Regards,
Framework Support
Here’s my reply:
Hey there,
I can’t exactly pinpoint when the issue started, since I had simply been using the laptop without benchmarking it and had no indication that anything was necessarily wrong outside of what seemed to be overly aggressive fan curves and some odd player lag in Counter-Strike 2, though I attributed that to perhaps a quirk with the game itself or a network latency issue.
The laptop has been in the exact same hardware configuration since I received it, outside of swapping out a USB-C expansion card for an HDMI expansion card every once in a while.
The BIOS is currently on version 3.03, the latest stable revision released by y’all.
After setting the BIOS to defaults and returning to Windows, I ran Cinebench R23 again with only HWiNFO and Covenant Eyes running in the background (to match the parameters of my original test) and saw multi-core scores drop about 300 points and single-core dropped 5 points, as well as observing the same anomalous core-to-core thermal deltas.
The issue occurs when the laptop is connected to the Framework 180W USB-C charger. I didn’t test under battery in order to give the laptop its full power budget during benchmarking.
My SSD of choice is a TeamGroup MP44L 2TB (TM8FPK002T0C101) and my RAM kit of choice is G.Skill Ripjaws DDR5 2x32GB 5600MT/s CL40 (F5-5600S4040A32GX2-RS)
Upon removing the Graphics Module to visually inspect the fans, all I could see was a small amount of dust as expected from regular use. The fins on the mainboard heatsink were practically impeccable and had no obstructions.
The only device plugged into the laptop during the original round of testing (as well as gaming) is my Logitech G502 LIGHTSPEED’s wireless receiver.
I removed all the expansion cards and plugged in the charger to the left rear Expansion Card port and re-ran Cinebench R23 with the same parameters as the other two tests and once again observed lower scores than the original test and the same core-to-core thermal delta I mentioned in my first message.
Looking forward to y’all’s instructions on how to proceed from here!
It seems like they’re sending out a generalized first response to get common issues resolved and out of the way before going in-depth on a given support ticket. We’ll see how everything shakes out.
Yeah as i told, you must get through the first Question Lap, before they send you to the experts, but for me it was 5 Days from initial Contact, until i was notified i will get a replacement.
I Just did a Cinebench24 Run and you can see my old Score and the new Score with my new Mainboard and modified Cooling Solution. over 70Points more is massive in cb24
Even my Single Core Score was massivly affected before. 102 vs 80 Points. Thats Wild!
Wish you all the Best, i hope you get alot better Scores with the RMA’ed Board. My Modification, is far off the Chart for the most Users, but i was pretty comfortable not to screw it up, and even when it would have worsened, i would have wait to replace the Liquid Metal Pad (When available). My Current suspicion is, that the Manufacturer of the Pad, has a Quality Problem OR its not enough Liquid Metal, to fill the hexagonal Dips. Because even without lapping the Surface for more than a quarter of a Hair (didnt wipe out the hexagons) improved the Thermals even further from the Pure PTM 7950. I would never have bothered doing it, if the PTM7950 hadn’t improved the Performance from the Liquid Metal.
Hey Guys i want to Update you as my fellow Batch 20’ers
I did further Modification on my Replacementboard.
further Lapping worked and gave some extra Points (more than 200)
All CBR23 Runs i posted have been Single Runs and the CB24 Runs 10 min Loops.
I will perform another Step, i ordered, Knife Sharpening Stones with 400/1000/3000/8000 Grit
To get it as flat and shiny as an Mirror. But as “rough” as it looks, it gave me better Systemperformance by 2-4% just by this Step. Overall i improved the Performance about 6% on my Replacement Mainboard (CBR23 nearly 1000 Points extra), just by manually modifying the Mating Area and replacing the Liquid Metal with PTM7950 (which should be vice versa)
And i am even beating Micas 7940hs with my 7840hs right now.
So when the Sharpening Stones arrive i will rip it apart again and do some more “manufacturing”
Thanks for the info. I also had very bad performance and decided to remove the heatsink and apply standard thermal paste (with slight increase in performance). What do you mean by “lap” ?
I have used a File to file down this elevated plate with the hexagonal holes/dents. Today i will get my knife sharpening stone to smooth it out and maybe fully remove the plate so the Die can directly touch the Vapor Chamber.
If you buy PTM7950 you will get more performance as with standard thermal Paste.
You can put “lapping” in Google translate if you are not native english. I am german native. Thats why i mix up the upper and lowercase all the time
Yep, that was also a Reason for me to stop at the current Point. If the Heatsink was sold seperatly i would have bought one and would try to remove the former etched Plate, to remove another Point of Heat transfer Problems. To have the Die directly on the VaporChamber. Currently it has to gap 3 Points of Failure to get the Heat into the Heatpipe.
i love it, when an idea works out. 
.