GDC Beast CPU @ 800MHz & Task "System Interrupts" 20-30% CPU-usage
I just received my GDC Beast Dock and fitted it with a GTX 970 (MSI). I connected it via NGFF in my WiFi-Slot, did all the preperations (installed driver, deactivated dedicated 940 MX).
The GTX970 works fine and is recognized in the system as GPU and it is connected to my external monitor without any problems.
Unfortunately, I have to other issues:
1. The Task Manager shows me, that the task "System Interruptions" uses 20-30% of my CPU. I guess this has something to do with driver issues? This leads to laggs in games, which is quite the opposit of what I wanted.
2. My CPU works at its usual frequency of up to 3,1 GHz for some minutes until it solows down to only 48 MHz, which causes my system to be even slower.
If I unplug the egpu, the task "system interruptions" is on normal level, but the CPU still drops to 48MHz after a while. The high usage for "system interruptions" is somehow plausible to me, but why is my CPU dropping to 48MHz. It is no thermal problem. I first realized this issue, when I did a FurMark bench, which worked very well for some minutes and ended in horrible laggs and low FPS.
How can I solve these two issues? Do I need to deactivate my 940MX in the UEFI? Or do I need to do a Windows reinstall (which I would prefer to avoid)?
My System is:
Acer Aspire E15 575g, Windows 10 October 2019
CPU: Intel Core i5 7200u (2,7GHz, 3,1GHz turbo boost)
dGPU: GeForce 940MX (2GB VRAM)
Thank you for your help!
after some trying to solve the issue, I can report the following:
I wanted to perform a Windows reinstall and therefore unplugged the adapter-cable and reassambled my laptop (which was unscrewed and open while doing the tests). I then startet the laptop and wanted to save some files until I realized, that somehow my CPU frequency was at its normal level again, no drop to 48MHz. Of course the high CPU usage of the task "System Interruptions" wasn't there either.
I now have two possible explenations:
1. Somehow my system may have a problem with this NGFF adapter cable (because the issue also occured with unconnected beast)
2. Maybe my laptop recognizes an opened case and saves the system by dropping the CPU frequency accordingly. I am not sure how this would save the system, but I saw a very little push-switch which seems to be pressed down by the closed laptop case.
Unfortunatelly, I didn't have time to check these two options by first testing the 48MHz issue with opened case but unplugged NGFF and then by plugging in the NGFF-cable and closing the laptop case.
Is there any possibility, that my system recognizes an opened laptop case and therfore drops the CPU frequency to prevent damages to the laptop?
I hope you can give me some helpful input.
I now checked the two possible explanations. This problem is not caused by a safety precaution, it seems to be caused by the Beast-adapter cable or more likely by the fact, that a device other than the WiFi card is connected to the PCIe slot. As long as the WiFi card is plugged in or nothing at all, the system works very well. But as soon as I plug in the connector cable for the beast, I get the mysterious CPU frequency drop to 48MHz. I already checked all advanced settings in UEFI regarding PCIe but I did't find anything which could explain this phenomenon.
As soon as I connect the beast to the cable as well, I also get the high CPU usage for the tasks "System interruptions" and "Diagnostic Policy Service".
Could anyone help me please. It seems my system recognizes the eGPU via the Beast adapter and it also seems to work. It is just that somehow the connection to the beast adapter causes CPU and task issues.
I can give another update:
I now found out, that the frequency drop (it's actually 800 MHz, not 48) is related to the power throttling which causes the CPU to run at its TDP of 100MHz with multiplier 8 = 800 MHz. Why this happens when plugged in the GDC Beast cable to the NGFF slot, I have no idea. But it seems, that by connecting this cable to my laptop, the CPU receives false information from the chipset/motherboard regarding temperature issues. This causes the CPU to throttle down to 800MHz. This problem may be solved by downloading "ThrottleStop" and desabeling "DB ProcHot".
Now my CPU runs at normal speed again, but always at Turbo Speed of 3,1 GHz, I havn't figured out yet, how to let it choose the right frequency depending on the current load.
I still don't have any idea, why the GDC Beast cable causes this problem. If anyone has an idea, please let me know.
So temporarily, there is still the issue, that when I try to connect my eGPU, the CPU usage is high because of the tasks "System interrupts" and "Diagnostic Policy Service". Does anyone have an idea why? I think it must have to do something with drivers, because of my dedicated GPU 940MX.
How can I overcome this problem?
Also wondering this; I have a Razer Core X and it does the same on my LG Gram 17. No idea how to fix it. Although I have noticed, in the "Intel Advanced Menu" in my BIOS' advanced options, there are a lot of potentially funky-looking default settings - I got the impression that LG might not've set some of them right for the laptop itself or for the best performance. So I've been meaning to dig in & google all of the options available in the Thunderbolt menu, and see if there's something I can tweak to stop the laptop from constantly interrupting.
Here's a screenshot of Process Explorer, you can clearly see System Interrupts has the highest number of Context Switches by far, which procexp's documentation says is an indication of a driver or hardware issue.
I feel pretty stupid for not being able to figure out how to post an image inline here, but here's a link:
I've been pretty suspicious of "System" and MsMpEng.exe for a while now too.
I am facing the same issue as you are. So, my spec is quite similar as yours, intel core i5-7200u with no dgpu in dell latitude 3480 laptop. I have tried gtx 1070 on NGFF cable with exp gdc beast 8.5 .
So, i have discovered my ngff pcie x1 gen3 is not that capable what i thought primarily. By checking with CUDA-Z application i have found the "host to device" bandwidth is only 700MiB/s(average). Which is not much i guess. And this bottlenecking should be the main culprit for the system interrupts and low fps/lag, I assume.