I experience this on a daily basis when using Windows 10.
Is there a specific condition it occurs during? For example, like the report earlier of it being within a specific battery level range.
@nrp, I experienced it too. I’ll be doing a more comprehensive analysis in the next few days to see when it occurs and if there’s any correlation to charger, cable, and/or charge level.
So I got a USB C power meter and a wall power meter today to test. I am planning on making a deeper analysis in case some of the info can help.
I did a quick test from 50% to 88% of charge, and this is what I got.
50% - 59%: USB - 61.1 Watts Wall - 69 Watts
60% - 70%: USB - 54 Watts Wall - 60 Watts
71% - 81%: USB - 50 Watts Wall - 65 Watts
82% (Clock speed starts to dip): USB - 48 Watts Wall - 63 Watts
83%: USB - 47 Watts Wall - 59 Watts
84%: USB - 44.8 Watts Wall - 53 Watts
85%: USB - 43 Watts Wall - 47 Watts
86%: USB - 41 Watts Wall - 44 Watts
87% (Clock speed back up to normal): USB - 48 Watts Wall - 50 Watts
88%: USB - 45 Watts Wall - 52 Watts
Something weird I found with this test, the clock speed slowed down at lot later than usual (Usually it slows down at around 74%). I usually wait to charge my laptop until around 15% of battery, so maybe that could make a difference.
@nrp I am planning on making a video timelaps like @Kyle_Adams did but with the meters on the video too. Please note that the watts were fluctuating during the battery percentages but I got an average of them. Hope this can help for anything.
2 Likes
I notice that at around 9:21 (when your battery reach 74% and your cpu slow down) that your igpu shoots to 90% of usage. Could that be part of the problem, like more electricity is allocated to the onboard gpu so less to the cpu or something ? Igpu then goes back down to 20% usage right after you hit the 84% of battery and cpu speed goes back up …
1 Like
The laptop is connected to the eGPU (cf Razer Core X Chroma eGPU). But AFAIK it’s 100% charged and has been during the entire session.
The Windows task manager “Performance” tab indicated 0.18Ghz.
It happened 3 times last night when I was playing Aliens Fireteam and Cyberpunk 2077. The first 2 times it got back to normal after 1-2 minutes. The 3rd time I had to reboot.
FWIW, this has entirely stopped happening to me. I did do the suggested fix of adjusting the EMI stickers, though it stopped happening even before that.
So I was working with OBS today and my clock speed went down to 0.39GHz at 40% of battery(I was charging it since 8%). The GPU was at 88% utilization and the CPU at 5%. As soon as I stopped the OBS recording, the clock speed went back to normal.
This is the first time that the clock speed has gone down for me below ~75%.
Could something be happening with the GPU when the clock speed throttles?
1 Like
Another question, have you guys heard any “coil whine” ? I hear some of that here and there.
@Rodrigo_Luzuriaga I don’t know much about the circuitry but I’m wondering if it’s the GPU or if it’s like a power balancing issue.
Hopefully it’s able to be fixed with firmware and not a flaw in hardware which would require a redesign + replacement.
2 Likes
Does this issue affect all Framework laptops? Or only some of them?
I have yet to see this occur on my laptop. And I have cycled the battery down to 17% back to 100% a few times while working on it, pushing it hard.
So it is hard to quantify… There seems to be an issue, just getting it to repeat every time is causing difficulty in nailing the problem down for folks. It could be so many different things that only some people match up with. The more data points the better to diagnosis it.
I follow this closely with interest, but have not been able to make it happen.
1 Like
I wonder if some users experienced this issue on linux. It could add some data point on what’s going on.
@Timothy_Aponte I just did a test on Manjaro and everything seems fine. I charged from 32% to 100% and I did not see any dips in clock speed.
I installed it onto a second NVMe drive, so I wasn’t running it from a live USB.
My guess is that it is a driver issue on Windows. Hopefully the Framework team can figure out the issue and push out a new driverpack.
1 Like
I’m running Arch Linux with KDE on a fully up to date system, and I’m noticing similar behavior as mentioned above.
For example, I open up steam and set it to download a bunch of games, which causes the CPU usage to go full out while it downloads and install games as quickly as possible.
While the power cord is unplugged, everything works fine, CPU is reaching up to 3.6GHz per core as reported by powertop, then as soon as I plug the charging cable in, it takes a few seconds, and then it starts to downclock down to 400mhz on all cores. The battery is sitting at around 90% charge. I can observe the slowdown as download and install speeds drop significantly.
Some other interesting information, shortly after plugging in the power, the battery charge indicator shows that it’s “plugged in, but still discharging” for the few seconds it takes for the CPU to downclock down to 400MHz, then it switches to “plugged in, charging.” Perhaps this is tied to the battery using up all available power from the power adapter to charge up, rather than supplying it to the CPU to use for tasks?
I’m using framework’s charger, and this happens when I plug the power cord into a USB-C port on either side of the laptop.
I should also mention that when I powered on the computer starting with 100% battery, the power cable was unplugged initially.
Edit 1: some additional information, rebooting the computer while the power cord is plugged in and the battery is charging causes the CPU to be allowed to clock higher than 400MHz while downloading/installing steam games. I’m watching all 4 cores go between 850MHz and 1200MHz and the download/install appears to be going back at full speed. I think this has something to do with the CPU frequency governor and the laptop’s charging state.
Edit 2: If the framework team would like to gather any additional information, please let me know what you need me to do.
Edit 3: I did another test. The battery is still charging from the previous reboot mentioned above, I booted up a game, final fantasy 14, and all of the CPU cores appear to be slowly clocking up, starting from 1250MHz when I first launched the game at 92% charge, 1427MHz at 96% charge. 1500MHz 97%, 1600MHz 98%, 1700MHz 99%, and it seems to be stuck at 99% with the battery reporting “plugged in, still discharging” (but it’s never dropping below 99%). The CPU frequency is hovering between 1.7 - 1.8GHz on each core. My in-game performance has also gone up from the 30FPS average I had initially to 40FPS right now. I’ve shut off the game and I’m letting the battery top off. I’ve restarted the steam downloads and the CPU freqs are back up to 2.7GHz. Battery is fully charged. This looks like normal behavior again. Again i’m thinking that all of the power that the power adapter was putting out was being put toward topping off the battery and not toward powering the CPU.
Note that these are CPU core frequency stats, I haven’t looked at the iGPU stats.
2 Likes
I’ve been able to reproduce the behavior mentioned above. Steps are as follows.
Using framework power adapter, the only power management altering service I’m using is TLP (TLP - ArchWiki) with default settings
-
Starting computer with 100% battery, plugged in
-
Launching powertop on the console, and launching test game (FF14) idling in one zone staring at the same wall getting 40-42FPS in game, checking Frequency Stats, all cores except one are hovering around 1.8GHz, +/- 100MHz, with the one core sitting around 800-1000MHz
-
Confirmed battery is still reading 100%, “plugged in, but still discharging” never goes below 100%
-
Unplug power cable, 6/8 CPU cores go up to 3.1GHz, one at 2.5GHz, 1 at 1.2GHz. in-game framerate goes up to 42-46fps (weird?) powertop reports discharge rate of 39W
-
Allowing the battery to discharge to 89%. 4/8 CPU threads still at 3.1GHz, one at 2.5GHz, two at 1.2GHz
-
Plugged into power on right side of laptop (port closest to front). Battery reads “plugged in charging, about 1hr 26m to full charge.” About 30 seconds later, all CPU core freqs drop to 400MHz, battery at 90%, reads “plugged in, charging, about 13m to full charge” in-game framerate has dropped to 6-8fps
-
All CPU cores are slowly raising clockspeed, now at 500MHz across the board, approx 1m after initial “drop” to 400MHz. Some threads are scaling up faster than others, but within 100MHz of each other. In-game FPS 20
-
Battery 91%, “13m to full charge” CPU cores continue to slowly scale up now around 650MHz
-
92%, “13m to full charge” 6/8 cores at 1000MHz, 2/8 at 700MHz
-
93%, 6/8 cores 1100MHz +/- 50MHz 2/8 cores 700MHz. in-game FPS 30
-
94%, “12m to full charge” 6/8 cores 1200 MHz +/- 50MHz, 2/8 cores 750MHz. in-game FPS 31-34
-
I’ve just installed psensors while this is happening, all cores show 57C +/-2, there are other sensors that aren’t descriptive but show at most 62C and at minimum 46C, one more labeled sensor “composite” shows 46C
-
96%, “11m to full charge” 4/8 cores 1350MHz, 2/8 1250MHz, 2/8 800MHz, in-game FPS 35 +/- 1; powertop reports discharge rate of 10.4W
-
97%, “9m to full charge” 4/8 cores 1450MHz, 2/8 1350MHz, 2/8 850MHz, in-game FPS 35 +/- 1 powertop reports 9W discharge
-
98%, “7m to full” 4/8 cores 1.5GHz, 2/8 1.4GHz, 2/8 850MHz, in-game FPS 36-40, not noticing any change in any of the temperatures from psensors
-
99%, “5m to full” 4/8 cores 1.6GHz, 2/8 1.5GHz, 2/8 850MHz, in-game FPS 36-40 (leaning more toward 40 more frequently), temperatures are now going up to 63C at highest, but still staying around 61C on average
-
99%, “9hrs 53m to full”, 4/8 cores 1.7GHz, 2/8 1.6GHz, 2/8 900MHz, in-game FPS still mostly the same
-
At this point it looks like the battery isn’t going to “top off” any time soon unless I let the computer sit for a while.
One last interesting thing I noticed. When I first got the laptop, the battery health was reading 99%. It’s now reading 95% (Yikes). I haven’t fully discharged the battery, ever, the lowest it’s gone is 40% after leaving it in sleep mode for a little over half a day unplugged. I’ve cumulatively used less than 3 “full charges” of the battery since I’ve gotten it. Maybe the current battery health reading is affected by the 39W discharge then rapid recharge it was doing during the test.
Also of note, the “time to full charge” was not accurate as to how long this test actually took. From start (step 1) to finish (step 18) it was 30-40m.
Aside from the dropping to 400MHz and sticking there for a while, is all of this expected behavior? It seems undesirable to want to rapidly charge the battery, diverting the power away from the CPU, at a heavy cost to performance and potentially battery health.
Edit: after reflecting on this, I think it must be the EMI shielding issue, or some kind of short. The fact that unplugged on battery was performing better than plugged in, and the very slow charge speed despite the low clock speeds makes me think the short only happens when the Power Delivery is active, perhaps on the higher voltage since 20V is more likely to jump a gap than 5V. I’m going to double check the shielding on the ports again and get back to you.
Edit 2: I’ve reseated all of the EMI shielding stickers and I’m getting the same results. Unplugged on battery and it’s able to run full tilt no issues (3ghz all cores). As soon as I plug in the power to either side, the clock speed drops significantly then slowly crawl back up to around the 1.8ghz mark on each core once the battery tops off at 100%. Additionally, my battery health is showing it’s dropped down to 93%. What is going on?
Edit 3: I played games unplugged until the battery was drained to 51%, then plugged it back in, and everything is working normal? Clocks are fine, it’s charging fine. I here a low pitched hissing sound coming from inside the chassis (kind of reminds me of the sounds hard drives make when they’re writing data, except there’s no hard drive of course). I unplugged the power at 77% charged, the hissing stopped, plugged it back in, and it’s back again. Also, the clocks dropped back down to 400MHz and seem to be stuck there again. Rebooting does not fix the clock speed issue.
4 Likes
The fact that this isn’t happening to every single laptop implies that it’s some sort of quality control. Does anybody use a different PD charger so that we can isolate the problem?
@Shawn_Lewis , it could be happening at different levels for other users and they may not have noticed it. This is happening consistently to me at exactly 83%.
1 Like
