Huge performance loss due to faulty cpu clockrate on Acer Aspire 3 A315-41-R2GU - how to fix it?

JohnnyK

Hello,

I discovered a weird behaviour which I have no solution for, yet. Under full load (CPU benchmark) the CPU is clocking the cores very differently after a few seconds of runtime.
One core runs at ~3.3 GHz, two others at ~ 2.3 GHz and one at ~1.5 GHz. This results in an overall performance loss of >20%. Because one core is running at ~3.3 GHz, the cpu runs at higher voltage and gets hotter which results in the thermal throttleing of the other cores (I suppose).
I've tested this with Cinebench R15, Prime95 and AIDA64 stability test - same outcome (I've linked the images below). Which drives me nuts is that this wasn't always the case. I've seen it running stable at 2.9-3.1 GHz on all cores at a lower voltage (75°C) on all tests for a longer period of time and it never dropped into this weird behaviour. I got a maximum cinebench R15 score of 690 back then, now its only 510 - always.

Test-conditions: 20°C room temperature, 100% charged, mouse attached, WLAN off, networkcable attached, kill-a-watt-meter applied.
What I already tried:
- changing windows 10 powerplans (balanced, high, ultra),  on desktop ryzen there seems to be a additional ryzen powerplan.
- flashed between BIOS 1.08 and BIOS 1.09 multiple times (happens with both of them).
- changed the thermalpaste on the CPU (Noctua NT-H1) multiple times, checked the contact pressure and thermalpaste distribution.
- installed the newest GPU driver manually.

- rebooting several times.

Anyone with a similar experience?

Thanks for your help in advance.

JohnnyK

Sooner this day I flashed the previous BIOS (1.08) and everything worked fine after that. I updated the GPU driver again and rebooted, still no problems. Now I booted the laptop and the problem is present again. I have NO idea what this is - tomorrow I'll reinstall Windows 10.

JackE

>>> discovered a weird behaviour >>>
What date did you discover this? If around Oct 2 and if system protection was enabled, check for a major Win10update restore point in Control Panel that probably was created. Jack E/NJ

JohnnyK

JackE said:

>>> discovered a weird behaviour >>>
What date did you discover this? If around Oct 2 and if system protection was enabled, check for a major Win10update restore point in Control Panel that probably was created. Jack E/NJ

A few days after I got the laptop (2 months ago). I reinstalled windows 10 yesterday, updated it and installed the drivers which were automatically downloaded by windows - the problem still remained. Since the GPU driver provided by microsoft and the acer homepage are outdated, I installed the latest one from the AMD website for RX Vega manually (25.20.14003.1010). Magically the problem disappeared - but today it's present again with the same GPU driver. Yesterday, when everything was fine (Prime95, Aida, Cinebench R15), it scored around 650 cb with each run and drew around 40W out of the wall. Today (with faulty clockrates and higher voltage), it draws 45W (maximum) and scores around 520 cb again - so the laptop is running not only even slower, its more inefficient. I checked the taskmanager, no other task is utilizing the cpu. Im no hardware expert, but I studied informatics a few years ago and I feel like a complete ***** in this case. I can't be the only one with this problem, can I?

JackE

>>> I reinstalled windows 10 yesterday>>>

Did you re-install the original ACER-flavored factory Win10 from the hidden recovery partition using the ALT+F10 cold boot method shown in the video below? It's not the same as the Microsoft generic Win10 version. It's really the only way to rule in or rule out a hardware vs driver issue. If it starts to work again after re-installing, make sure system protection is enabled so the offending Win10update creates a restore point that you can get back to.  In fact, if you got it to work yesterday, check for a restore point now just in case a Win10update got installed while you weren't looking after the re-installation. Jack E/NJ

https://www.youtube.com/watch?v=cpzLJRAZldA

JohnnyK

JackE said:

>>> I reinstalled windows 10 yesterday>>>

Did you re-install the original ACER-flavored factory Win10 from the hidden recovery partition using the ALT+F10 cold boot method shown in the video below? It's not the same as the Microsoft generic Win10 version. It's really the only way to rule in or rule out a hardware vs driver issue. If it starts to work again after re-installing, make sure system protection is enabled so the offending Win10update creates a restore point that you can get back to.  In fact, if you got it to work yesterday, check for a restore point now just in case a Win10update got installed while you weren't looking after the re-installation. Jack E/NJ

https://www.youtube.com/watch?v=cpzLJRAZldA

Thank you for your reply. I bought the laptop with a linux distribution, so I formatted it and installed win10 - therefore I think this method does not apply. Yesterday I installed win10 and didn't connect it to the internet - the faulty clockrates were still present, even after installing all drivers again. Today I tested it for a few couple of hours and I finally found a way to remove the "bug" for some time. I have to use the laptop a few minutes on battery, shut it down, plug it back in and boot it up. Then the laptop is in "charging mode" and uses some of the power to charge the battery (depending on the system load) but it doesn'T "overboost" the CPU, so the clockrates are stable on all cores (2.6-2.8 GHz) under full load. The weird part is, when I only plug the power cable back in (when the clockrates were faulty during this session), the problem remains - so the reboot is required. I don't think that this can be fixed.

JackE

OK. Keep system protection enabled just in case.  I also suggest that you try disabling fast startup, shutting the machine completely off, fully charging the battery in the completely off mode (till the charge LED is blue). Then leave the charger connected and boot it back up again. Then see what happens going full tilt. The charger simply may not be able to keep up with battery charging plus supplying full load at the same time. Do you have the anemic 45watt charger? Or maybe its a sick 65watter. Jack E/NJ

JohnnyK

JackE said:

OK. Keep system protection enabled just in case.  I also suggest that you try disabling fast startup, shutting the machine completely off, fully charging the battery in the completely off mode (till the charge LED is blue). Then leave the charger connected and boot it back up again. Then see what happens going full tilt. The charger simply may not be able to keep up with battery charging plus supplying full load at the same time. Do you have the anemic 45watt charger? Or maybe its a sick 65watter. Jack E/NJ

Now I am even more clueless than before. As I mentioned above, the "shut down, start up" (I also called it reboot in my last post, wrong terminology) is required to get the clockrates back to normal, even if the laptop is in charging mode. Now I found out that pulling the power plug isn't necessary at all - only the "shut down, wait 10 seconds, start up" is! Normally I just clicked "reboot" cause I never had thought of it that it would make any difference, but it does. I've tested it now two times today, when the clockrates began to act faulty:

1. 5x (reboot, test -> faulty), shutdown, wait 10 sec, boot up, test -> working!

2. 2x (reboot, test -> faulty), shutdown, wait 10 sec, boot up, test -> working!

I don't think that this is a coincidence, but I'll test it again tomorrow to be sure. If you don't believe me, I already made one video and I'll make one tomorrow again, because I can't believe it myself. I just disabled fast bootup, not sure if that helps. I got the 45W charger.

JackE

Yeah, reboot used to be called a "warm boot" since all mainboard circuits remain powered with only RAM being cleared. A shutdown used to mean all mainboard circuits lose power and a subsequent startup was called a "cold boot". Nowadays, a shutdown can also mean a few  mainboard circuits remain powered and a subsequent startup is called a "fast boot" or maybe a "lukewarm boot" is more appropriate.
Disabling fast boot is a close to being truly "cold boot" like in the old days and should resolve the issue though it'll take longer to boot. Jack E/NJ     

JohnnyK

JackE said:

Yeah, reboot used to be called a "warm boot" since all mainboard circuits remain powered with only RAM being cleared. A shutdown used to mean all mainboard circuits lose power and a subsequent startup was called a "cold boot". Nowadays, a shutdown can also mean a few  mainboard circuits remain powered and a subsequent startup is called a "fast boot" or maybe a "lukewarm boot" is more appropriate.
Disabling fast boot is a close to being truly "cold boot" like in the old days and should resolve the issue though it'll take longer to boot. Jack E/NJ     

Sadly, disabling fast boot didn't help. Even more sad is that the problem occured after 15 minutes of a "cold boot", but every cold boot fixes it for an unspecified amount of time. I've put the latest version of Linux Mint on an usb-stick, installed wine and ran cinebench r15. I haven't had time to test it a lot yet, but the score was always around 620cb and it pulled ~40W (which is fine). When the clockrate-bug is active, it always draws 45W and the score drops to ~520cb. I'll keep an eye on that in the next days.

JackE

I got the 45W charger. >>> Even more sad is that the problem occured after 15 minutes of a "cold boot", but every cold boot fixes it for an unspecified amount of time.>>>When the clockrate-bug is active, it always draws 45W and the score drops to ~520cb. I'll keep an eye on that in the next days.>>>

I think you have the solution right here if my arithmetic is correct. A 45W output by the charger (minus) 45W consumed by the cpu (minus) 10W for SSD/RAM/etc (equals) less than nothing (minus 10W) left for the battery.  Jack E/NJ

JohnnyK

JackE said:

I think you have the solution right here if my arithmetic is correct. A 45W output by the charger (minus) 45W consumed by the cpu (minus) 10W for SSD/RAM/etc (equals) less than nothing (minus 10W) left for the battery.  Jack E/NJ

I'm not quite sure what you are saying. The battery is at 100% and not charging in both scenarios (idle state below 9W on medium brightness and HighPerformance). So immediately after I shut the laptop completely down, boot it back up, the bug is gone (as I said). The wattage difference is most likely due to the higher voltage for the CPU when the bug is active, because it's hovering around 1.13V to reach the clock for the 3rd core at 3.3GHz. When the bug is "fixed", it's stable at ~1.0V with clocks at ~2.7GHz on all cores.

JackE

>>>I'm not quite sure what you are saying. >>>

You said earlier "When the clockrate-bug is active, it always draws 45W ". 

Your 45 watt charger simply cannot keep up with a 45 watt CPU load plus another 10 watts minimum for the SSD/RAM/Fans running high/etc. The charger cannot deliver 55+ watts!  So the system is rapidly drawing the deficit power from the battery and likely reducing it's output voltage (depending on battery condition) at which point it the charger also tries to charge it again. It's a cascading event that can't continue for very long with the system drawing more power 55+watts than the 45watt charger can provide. Jack E/NJ

JohnnyK

JackE said:

>>>I'm not quite sure what you are saying. >>>

You said earlier "When the clockrate-bug is active, it always draws 45W ". 

Your 45 watt charger simply cannot keep up with a 45 watt CPU load plus another 10 watts minimum for the SSD/RAM/Fans running high/etc. The charger cannot deliver 55+ watts!  So the system is rapidly drawing the deficit power from the battery and likely reducing it's output voltage (depending on battery condition) at which point it the charger also tries to charge it again. It's a cascading event that can't continue for very long with the system drawing more power 55+watts than the 45watt charger can provide. Jack E/NJ

Thanks for elaborating your viewpoint. The total power supply draw is 45W under load when the clockrate-bug is present, measured with a kill-a-watt meter (as stated in my first post). After a few minutes continuous benchmarking (prime95, with bug) it is infact discharging the battery - so you are partly correct, but this is even worse. Sadly this isn't helping us in any way to resolve the problem. Why isn't the CPU lowering its voltage and distributing its clocks evenly to increase its performance and lower the power consumpotion, like it does on every cold boot. I just tested it on linux and the bug also occures as you can see in the image below. I would have limited the CPU clock via powercfg in windows like I did with my old laptop, but as I mentioned earlier, you can't, it's stays always at 1.6-1.7 GHz on all cores when in the max performance is set to 1-99%. I guess I give up at this point, thanks for the help regardless.

JackE

>>>Thanks for elaborating your viewpoint. The total power supply draw is 45W under load when the clockrate-bug is present, measured with a kill-a-watt meter (as stated in my first post). >>>it is infact discharging the battery>>>>Why isn't the CPU lowering its voltage and distributing its clocks evenly>>>

Yes, that makes sense. 45 watts is the most that the 45-watt power supply can deliver. So 45 watts is all it will draw from the AC mains wall socket if it's near 100% efficient.

The problem is that, under heavy load, all the laptop hardware together (cpu, screen, gpu, ram, fans, charging, wifi, HDD, other peripherals) apparently need more than 45watts to run properly. So it's getting the extra needed power from the battery till device voltage instabilities occur which likely accounts for why the CPU isn't behaving like you think it should. Simply put, the battery can't keep delivering this level of power for very long before the system become unstable.

Accordingly, In my opinion, the 65-watt power supply option will probably solve your issue.

Jack E/NJ       

JohnnyK

JackE said:

So it's getting the extra needed power from the battery till device voltage instabilities occur which likely accounts for why the CPU isn't behaving like you think it should. Simply put, the battery can't keep delivering this level of power for very long before the system become unstable.
Accordingly, In my opinion, the 65-watt power supply option will probably solve your issue.

True, it's definitely worth a try. I can get a 60W power supply for free (for testing purposes), I'll check it tomorrow and report back.

JackE

Good luck with it. And please post the results, good or bad. Jack E/NJ

JohnnyK

JackE said:

Good luck with it. And please post the results, good or bad. Jack E/NJ

Turns out I'm not lucky after all. I got hands on a 65W power supply and attached it. Everything is the same, except the power draw is 1W less (due to better efficency of the power supply). It seems that the maximum power draw etc. is programmed into the embedded controller which makes sense because the power supply does not have a data transfer pin (like some DELL PSUs) to communicate with the laptop. Yesterday I imported the "AMD Ryzen Balanced" powerplan (intended for Ryzen desktop cpus) which I found in a forum, it didn't help either.

Is there a way to report this issue to ACERs employees (or maybe you are JackE?) ? Since I have not found a way of reproducing this problem (besides using the PC for a while and test it occasionally), this might be difficult to detect. The laptop has so much more potential if they hadn't restricted it so badly in general. The throttling temperature of ~75°C is obviously too low and if were allowed to use ~10W more, the CPU and GPU wouldn't have to trottle it's clocks that much under load. Maybe a BIOS and EC update could fix this.

ilhamzap

Same problem.
Sadly I'm just got max 584cb at the first week, and almost 1 month now, decreased to a bit more than 500cb, now, around 450cb
I think the area around the Esc button is too hot than usual.
And I knew someone replaced his Acer product because faulty parts, then get a new one. First product only could hit 500-ish cb and the new one, a bit over 600.
Acer products for Ryzen not always in good condition before distributed?

Well, 15 default TDP is enough or a must for this laptop, if Acer set cTDP up to 25W, the cooling system is not enough, the laptop will surely melt. Gaming Laptop could get 25W because everything is specialized for it, unlike normal or slim laptops.
Yeah, 75 deg throttling temp is bad, this Ryzen max designed/ operational temp is 95 deg. My Laptop's maximum temp I ever recorded in the first week is 79 deg, and now 82 deg.

JackE

Please check for labels on the bottom of the laptop. Can you find a manufacturing date? Jack E/NJ

ilhamzap

I'm not so sure, maybe it's 2017?

"Model/Type : N17C4
51247/SDPPI/2017
(Barcode) PLD:2695"