Package C States above 3 are unavailable on my 9550 whilst on mains power.

So this was a conversation started on the Throttlestop guide thread, but rather than clutter Unclewebb's page I thought I would start a dedicated thread here.

I have the 6700HQ and as the thread title reads, my package C States will not go above 3 when my machine is plugged into the mains, with the package power going no lower than 2W. The second I pull the mains, the laptop is able to idle at a package C State of 8 and the power drops as low as 0.3W

Everyone else I've heard back from seems to be able to hit high package C States on battery and mains (providing they use the Intel AHCI drivers) so I'm currently baffled as to why I'm unable to do so.

It would be great to hear back from a few more people who both can and can't hit those higher states when on mains, so I can compare and contrast what drivers and settings are being used and hopefully get to the bottom of this.

 

This has potential to extend battery life and reduce thermals

Special thanks to @Che0063 and @CraftyClown who identified issues and found a simple solution. So hopefully they will draft a guide as they do more testing and have some time.

I got Package C-States working yesterday by replacing the generic M$ IDE ATA/ATAPI controller driver with Intel RST 15.8.1.1007 (latest). Now hits C8 both with mains & battery power in Windows Power Options (Balanced & High Performance).

Not a problem unique to Dell systems.

TEST ===> One way to check is via ThrottleStop (free monitoring software):

- exit all other programs

- click the button at the bottom of the panel labeled "C8" or similar

- Look at section labeled "Package C State Percentage"

--- Before, with M$ driver, my system was showing say 80% stuck no higher than C2

--- After, with updated Intel driver, my system was showing say 80% in deeper C States (e.g. C8)


POSSIBLE FIX ===> Draft rundown of basic steps that worked for me (took 3 minutes):

- Downloaded 64-bit file Intel RST (15.8.1.1007) and unzipped (name looks something like f6flpy-x64.zip)

https://downloadcenter.intel.com/download/27147/Intel-Rapid-Storage-Technology-Intel-RST-

-Device Manager>IDE ATA/ATAPI Controller

-Click on the triange to expose driver (mine was generic M$)

-Right click driver, click properties, click driver tab

-Uninstall / update manually pointing to Intel file just unzipped

-Restart

My system:
- Dell 9550 (6300HQ Samsung 850 EVO ThrottleStop)
- UEFI, legacy disabled, secure boot, AHCI)
- Did not install Intel RST during install (only used the ini for the patch yesterday noted above)

 

Yep, I'm using exactly the same driver so I don't think that's the issue.

 

If all else fails try a reinstall?

 

Yep, kind of resigned myself to that being the next logical step.

I'll have to methodically reinstall every driver and piece of software, taking notes as I go along until I work out the cause of the limitation.

 

I mean if it only happens when it's plugged in it's a power profile issue. I have no idea if mine uses higher C states when plugged in because that doesn't bother me. Have you tried playing with power profiles? Perhaps the dell power manager?

 

Well that's exactly what I presumed, however no matter what profile I use, be it with default or custom settings I get the same results. I've also looked at the Dell Power manager, but nothing there makes a difference either.

It's not a big issue I know. High C States on mains power aren't a great necessity, however my concern is more about my machine functioning as it apparently should and who knows whether this is a symptom of some other issue I have yet to discover?

 

It doesn't sound like a potential issue down the road to me. Does anyone with a 9550 get higher C states when plugged in? If it will go into them on battery (when you actually need them) but not main, it sounds like a "feature."

 

Yep, I don't disagree at all. My confusion is down to the fact I appear to be the only person this happens to. Everyone else who has reported back says they have high C states when idle no matter what power source they use.

 

And that is coming from 9550 owners? I can look at my 9560 later and let you know what I see.

 

Thanks, I'd really appreciate that. At the moment I have heard back from 9550 and 9560 owners, all of whom have the same results.

 

Ok so I've made an interesting first discovery. I've just done a full reinstall of Windows 10 using GoNz0's latest creator's update ISO (Cheers mate) and keeping the machine offline to prevent any updates, I am unable to enter high C states on battery or mains, despite having the Intel AHCI driver installed. This at least tells me there is another driver or software setting in needed to allow the required C states.

 

My 9550's Package C-States hits C8 both on mains & on battery (with new Intel driver above)

Hits C8 on Windows Power Options> Balanced & High Performance.

Hits C8 on all 4 of the Dell Command-Power Manager settings (Optimized, Cool, Quiet, Ultra Performance)

Package C-States could allow the CPU to run a bit cooler when plugged in. Also improperly acting Package C-States indicate that at least one thing is (and maybe others are) not operating correctly.

 

Well as we already know my 9560 happily does higher c states, have you reset the bios yet?

Sent from my SM-G920F using Tapatalk

 

Yes I've done a full Bios reset but it didn't affect the results.

Interestingly I have discovered I also need the Realtek SD card reader drivers installed for me to reach higher C States on battery. Without them I don't get higher C States whatever the power source. This seems to be repeatable as I can install and uninstall them and watch the ability to reach C8 come and go in Throttlestop.

I would be interested to see if anyone else can repeat this.

 

Thanks for moving this topic to a separate thread. I was playing around with my Core i5-7600 desktop computer yesterday. I have always had a problem where it does not use any of the Package C States. This is while using only the built in Intel GPU. I started by updating to the Intel AHCI driver that @GoNz0 provided a link to and still nothing. After that, I installed a Nvidia GTX 660 card which disables the Intel GPU, installed the most recent Nvidia drivers and all of a sudden Package C2 and Package C3 were working fine. For the desktop, it seems like the most recent drivers for the Intel GPU are responsible for disabling all of the package C States. I still don't see any package C6 so there might be another driver still that is causing a problem.

Here's where things get weird. I had my computer plugged in to a Kill-a-Watt meter so I could try and monitor power consumption while using ThrottleStop to switch to different package C States. With ThrottleStop set to a request value of either package C3, C6 or C7; the CPU was reporting that it was using Package C3 and power consumption was the lowest. When I switched ThrottleStop to package C8, C9 or C10; the percent of time in package C3 was still being reported at over 90% but for some reason, power consumption measured at the wall increased by 10% to 15%. I went back and forth multiple times and every time it was the same. At first I thought well maybe this is power supply related or the Kill-a-Watt is being flakey but as soon as I switched to requesting C8, power consumption went up and the package temperature of the CPU also went up which confirms that the CPU was using more power and creating more heat. Switch back to requesting C3; C3 residency time remains the same but CPU temperature instantly drops. That makes absolutely no sense but it might explain why Intel is deliberately turning off the deeper package C States. The Intel docs always have a long list of bugs that have not been resolved when a CPU ships. Some of these bugs get fixed with micro-code updates but many bugs do not. It is possible that the mobile CPUs also have a similar bug with package C States.

@CraftyClown - In the FIVR window of TS beside where it says Power Cut, it shows the micro code version that your CPU is using. Might want to post that so you can compare to others. I assume you are using the latest available bios version.

 

Always wondering when we go too far afield on the ThrottleStop thread!

Interesting results we can't really test with laptops due to batteries messing up the readings from line. Looks a bit buggy.

The Package C States are much more aggressive than the Core C States on my laptop. So the power savings even for light use may be quite limited in the real world. Package C States:

idle ~80%
internet use ~5 to ~60%
"480" youtube video ~10%

CPU microcode on my Dell XPS 9550 6300HQ is 0XBA. Same as that on @CraftyClown 's 6700HQ.

 

Package c-states are stuck at c3 max again on 9550. Not sure if that is due to BIOS or Windows update. . .

 

I have a similar problem. A few months ago I installed the AHCI driver and things started to work properly. Just checked a few days ago and C8 is never used (10W standby power consumption reported by Throttlestop). Then I reset the BIOS, and installed the Samsung NVMe 3.0 driver (was previously using Microsoft default due to the older version of Samsung causing either a rise in power consumption or DPC latency surge, maybe both, don't remember exactly). The new driver worked well (actually increased random read/write a bit but lowered sequential slightly), and C8 is used again. So I thought problem solved. Checked again today and I can't coax the processor to enter C8 even when nothing is running... Heeelp!

CPU is E3-1505m v6. Latest BIOS...

 

Hmmm... Reset BIOS to factory setting (probably did the basic reset last time). And package goes to C8 again! Will post back if anything changes.

Seems fine so far after I added back the Ubuntu boot option. Happily running at ~1W idle now. Note that this might also have a performance benefit for computationally-demanding tasks by my experience because the core can only take up the full 45W TDP (which seems to be a hard ceiling) if the package can go into higher C-states and not consume 10W no matter what. The difference is sizable if I remembered correctly. A 35W equivalent TDP for the cores will not allow full turbo boost in Prime95.

Also, just thought I mention this, I uninstalled intel thermal framework a long time ago to remove throttling. In case anyone is wondering, it doesn't seem to prevent processor from going into higher C-states (not that it should, but everything is tangentially related to us less knowledgeable...).

 

New observation. My laptop woke up from sleep this morning and refused to go into C8 yet again... Turned off Compatibility Telemetry and then left it at home all day idling but it never went into higher package C-states again. Rebooted now and it's back to normal... Very strange.

Edit: C-state stopped working again after a bit of regular browsing, outlook, foobar2000, iTunes. No background services running that I can see. Unplugged power and started working again. After I reconnected AC, I noticed that there seems to be a correlation between whether Throttlestop displays GPU temp and c-state availability. I'm using M1200 and graphics switching is set to automatic, however from what I know and the information from task manager, nothing has been running on GPU1 (GPU0 being the iGPU). Will update this once C-state stops running again; I'll disable the dGPU then and see what happens.

Edit2: Got stuck once again, set base profile in Nvidia control panel to Intel GPU hoping to disable the dGPU but to no avail. GPUz indicates minimal frequency, no utilization, and a perf cap reason "idle" on the dGPU. Tried power plans to no avail either. Finally unplugged the power again and it worked... Anyone having the same issue? Is there a way to disable the dGPU completely or report this switching bug to Dell?...

 

I'm now pretty confident that the higher package C-states are indeed disabled when dGPU is turned on (but not necessarily in use or above base frequency). Even though I know that lower C-states probably don't draw ~10W Throttlestop suggests (based on BatteryBar reports, lower and higher C-states are probably just a few W apart, if that, with higher C-states around 6W), it bothers me that Dell BIOS bugs out sometimes and don't turn off the dGPU when not in use. I searched around a bit more and there seems to be no way of disabling the dGPU short of going into BIOS (not even 100% sure if that's an option but I really can't go for that). Is there a way to report this problem to Dell so that they can maybe include it in the next BIOS update or something? Does the same problem exist on XPS 9550/9560/9570? Thanks!

 

Thanks all for the info. I have an MSI GS65, did a fresh install and turns out my CPU Package would not go into C8 power state, only C3. Installing Intel RST NVMe driver over the MS Standard NVMe driver fixed this. Seems like a general problem with the standard MS drivers.

Thanks!

 

My Dell 7577's i7 7700HQ won't go deeper than C3 Package state when my external monitor is plugged into my HDMI port and keeping my dGPU active. My individual cores all hang out in their C7 states (in the upper 90% ranges) though. However, when running only on my primary screen and dGPU asleep, my 7700HQ hits C7 Package States of 60-90+% when idling, depending on what applications I have running. I just did a major trim of Windows 1903 services since I took the May 2019 update to gain Dark Mode and it reset all my tweaks as usual. I'm back down to around 118 total processes on a fresh boot and the performance, idle, and average clock speeds are all noticeably better.