I have actually only ever seen drops when running a combined CPU and GPU load in a game over an extended period of time (2+ hours). I haven’t tested a synthetic combined load yet, although I might do that when I get time just to see if it’s more consistently reproducible for me. The fact that it happens more often for you is interesting.
Out of curiosity, what’s the ambient temperature in the place you use the device? My room does have an unusually high ambient temperature most of the time, close to 80 degrees F/26 degrees C. If this is a thermal issue, that might be an exacerbating factor.
Okay, I have a consistent 0.2GHz happening rn after trying to play Elden Ring for about 1 minute. It tells me that my CPU cores were under load at 100%, even though Task Manager told me it’s running at 2-3%, 0.2GHz. Also intriguing is I noticed that when the CPU dropped to 0.2 and I was still playing the game, the GPU jumped up to 100 (or maybe it was already at 100 and that’s the issue, but I can’t figure out how to get logs of graphs).
The CPU load has now (after a couple minutes) gone down to between 15-40%, but it’s still stuck at 0.2, and the GPU is active between 3-10%.
Further update: ran UserBenchmark, clocked 100% on the GPU with the CPU around 6%. Didn’t clock down to 0.2GHz, but it does say my RAM is performing way below expectation in the 16th percentile. It also says my GPU and CPU are operating as expected (if not better). I did see a drop in power consumption per the CPU Package/Core/Graphics Powers in Open Hardware Monitor, where the highest it was was 26.8W and the lowest is currently around 1.7W in the Package and Cores
@RandomUser No, it didn’t really go anywhere, because I wasn’t using a ‘supported’ system (ie. Windows or Ubuntu). Now that I am, it’s also kind of stagnant because it’s hard to replicate. They haven’t got back to me since they recommended following the advice in the thread that was linked above. Been a couple days now.
Reseat your RAM. It is possible that some jostling might have dislodged the RAM.
You may also wish to run Memtest to see if the modules might be faulty.
A system slowing down to 0.2 GHz for stability to reduce memory errors is not the expected symptoms for memory issues, but we should probably rule that out if the RAM is performing that poorly.
@Monster_user I’ve managed to run UserBenchmark when my laptop was running at 0.2 (yet again after about three minutes in Elden Ring–seems like a good benchmark game to use!). An interesting thing happened: at the very end during the GPU test, the CPU kickstarted itself back up to normal.
I can’t remember what the test told me for memory, but IIRC it was average. However, that value and the other values changed drastically during the benchmark test (obviously).
I managed a strong (/sarcasm) 0% and 1% percentile rank in SSD and memory respectively, and it couldn’t even complete the CPU testing. I got 13% percentile for the graphics card.
I’ll reseat the memory and the SSD card(s) and see what happens.
Another interesting thing to note: the CPU is running below 2.8 right now, between 2.2-2.6, when I think it normally runs just shy of 2.8 pretty consistently. I wonder if the GPU kickstart was enough to get it back to normal functioning or if it’s still lagging a bit behind.
Okay I think I found a strange pattern that happens: when I disabled the turbo step and turbo boost in BIOS, my CPU, memory, and SSD all performed below or way below expectations, but the GPU performed way above. However, when I enabled turbo step and turbo boost, suddenly it was reversed, where the CPU, memory, and SSD all performed average or far above average, while the GPU performed below average.
The GPU difference here is quite large–before it was at 87/94th percentile, whereas now it is performing at 20th percentile.
Similarly, the memory, SSD, and CPU respectively performed originally at 7/19th percentile, 36/28th percentile, and 21/26th percentile; whereas they are now performing at 43, 72, 45.
You can see that the greatest disparency was in the GPU with -67/-74 percentiles, but even the others have a massive difference.
@volcanicmaggies, What is the charge level for your battery, and did you charge it to 100% before starting those tests?
If you charge the battery to 100% before starting those tests, do you get a different result?
That seems to indicate that the system is not getting enough power to run at full load.
I’m on the Framework 60w charger. I’ve seen the battery drop while doing some gaming under Linux while plugged in.
The 90w I mentioned near the top of the thread was a typo as I was thinking about testing with a 90w, which RandomUser indicates allows an 85w max load.
@Monster_user Okay, I just changed the limit to 100, and here’s the results. Also interesting to note that my computer is currently on low battery, so it isn’t fully charged right now. When it is I’ll also try running it again.
(this is with turbo step and boost enabled)
i7: 92nd percentile
GPU: 29th percentile
SSD: 68th percentile
Memory: 38th percentile (with a note about XMP?)
@RandomUser The UserBenchmark (sorry lol) test that I carried out and mentioned to Monster_user above was carried out with the changes suggested. 3-5W is from Open Hardware. I don’t have any idea how to remove the battery and I don’t want to muck things up more than they are already…
Yes, I updated drivers for Framework’s BIOS bundle, but I have also made sure the actual graphics drivers from Intel are up to date.
@RandomUser Yes, it is running Windows 10 downloaded from Windows, the most recent version. And yes, it was 3-5W during the execution.
Here is a screen record of the values happening when the frame rates drop: fw laptop lag - YouTube
I’ll get a screen grab of the 0.2 when I can get it quickly (but it is literally in the Task Manager, CPU speed)
@RandomUser When I ran the game again (and again this is with turbo step and speed enabled) it had the same issue as the youtube video had, but I confirmed it did not go down to 0.2GHz. However, it seems like I can’t play the game right now, and I think that’s because of turbo step enabled. I am going to turn it off and replay the game and see if it goes down to 0.2. However, perhaps that’s enough: “it clearly can’t run some games and the frame rates drop but don’t stall while turbo step is enabled, but it crashes the CPU if turbo step is not enabled”
Okay from looking at that I can confirm at least that that’s the exact behavior I’m seeing, it’s just much harder for me to reproduce and doesn’t happen until several hours under load.
To be clear, this isn’t what’s causing the power throttling. Thermal throttling causes the CPU to downclock, yes, but it doesn’t change the power limit. You can test this yourself by running P95 and observing those fields in HWiNFO. You’ll see them change to “Yes”, but you won’t see your power limits drop. And if it was thermal throttling, the CPU should remain pinned at TjMAX because that limit lifts instantly when the temperature drops.
There must be something else causing the power limit decrease, probably something to do with power delivery. A battery issue could be possible, but seems unlikely to me because I haven’t seen my battery decrease in a sustained load scenario. Power adapter problems also seem unlikely because I’ve seen this happen on both the 60 W Framework charger and a 65 W USB PD adapter, both of which work fine under higher burst loads. Something to do with VRMs then? Something within the CPU itself? I don’t know honestly.
Also @volcanicmaggies there might be one more thing worth trying. I think I realized why opening Throttlestop fixed the issue for me; I had the power limits manually overridden in there. I think @RandomUser’s idea of just using a program like that to clamp the PL1 power limit at 28 W should at least alleviate the symptoms if not solve the underlying problem. You won’t lose thermal protections from doing this so it really shouldn’t be unsafe. I wouldn’t even be that opposed to just running it like that until the problem gets resolved.
Yes my bad, I should have been more clear. What I meant to say is that core temperatures are not what’s causing the PL1 to decrease here.
“Best performance (plugged in).” Based on the video, this should be what @volcanicmaggies is running too because the other two plugged in modes have lower power limits of 28 W PL1/30 W PL2 and 15 W PL1/25 W PL2, at least with the default configuration.
@OxyMagnesium@RandomUser Thank you both for taking so much time to try and help troubleshoot this. I’ve set the battery limit to 28, and Elden Ring has run okay so far, so if the problem happens again, I’ll try and get a video. As for the 200 not showing up, it hasn’t shown up since I enabled Turbo Step and Turbo Boost. So it seems like when those aren’t enabled, the CPU (because it is restricted to 2.8 when they aren’t enabled) drops to 200MHz and can’t get back out, but because with them enabled it starts at a higher CPU, it never reaches the 200MHz.
When I click the battery icon, all it says is it is not running battery saver.
@RandomUser Okay, I changed it! It was operating under Razer’s power plan since I had Razer Cortex installed. Also, I did manage to get a frame rate freeze playing full screen, but haven’t had it happen while playing in windowed mode for twenty minutes or so.