AnsweredAssumed Answered

R9 380 Core Clock Dropping - Windows 10 FCU 17.11.1

Question asked by dreamc on Nov 7, 2017
Latest reply on Nov 21, 2017 by alleyviper

Like • Show 0 Likes0
Comment • 12

Recently I updated to WIndows 10 1709 (from 1703) and started noticing some strange behavior - from time to time for a second FPS drops to approximately 20-30 (still going smooth with uniform delays between frames) for a second and then returns to normal.

This affects most applications using GPU - games and even Firefox (it uses hardware acceleration after all, however it is more slow to recover from this state).

When under high load this is also coupled with core clock going down, as you can see from the attached screenshot.

I tried updating Crimson Drivers, though I had almost the latest ones. This didn't help.

Tried cleaning drivers with DDU and reinstalling them - no good.

GPU temperature does not go higher than 65. CPU load is normal. Tried with my AV disabled.

Does anyone else have this problem?

And what can I do in this situation?

Describe your system:

AMD Graphics Card
- MSI R9 380
Desktop or Laptop System
- Desktop
Operating System
- Windows 10 64bit 1709
Driver version installed
- Radeon Software Crimson Edition 17.11.1
Display Devices
- DELL U2414H, 60 Hz over DisplayPort
- Samsung UE55MU6195 UHD TV over HDMI
Motherboard + Bios Revision
- ASUS Z97-AR, Bios Version 2801
CPU/APU
- Intel Core i5 4670
Power Supply Unit Make, Model & Wattage
- Corsair 650W (do not remember the exact model)
RAM
- 2x 8GB in dual channel

Message was edited by: Matt B We have updated the title of this discussion with relevant details to better describe your issue.

Attachments

i188^cimgpsh_orig.png286.3 KB

No one else has this question

Outcomes

kingfish Nov 7, 2017 9:29 AM

Power Limit? That would be my guess...

AMD graphics performance

Actions

dreamc @ kingfish on Nov 7, 2017 9:50 AM

All the settings are default (and even before updating to 1709 I had all on default, however there was no such problem).

And I hope GPU does not try to lower it power state under High load...

Actions

kingfish @ dreamc on Nov 7, 2017 10:02 AM

It will definitely throttle the card.

Actions

dreamc @ kingfish on Nov 7, 2017 12:12 PM

"Definitely throttling the card" under load MUST be considered a not desired behavior and fixed.

This did not affect me in any way before 1709 update. Testing it right now.

Actions

dreamc @ kingfish on Nov 7, 2017 8:05 PM

Unfortunately, this didn't help.

I attached a couple of screenshots, where you can see some spikes in core clock.

Is there anything else to check?

Attachments

2.PNG163.2 KB
1.PNG188.7 KB

Actions

alleyviper @ dreamc on Nov 8, 2017 7:32 AM

Please try to raise power limit to something like 20% (or above, if that's not max). This is to check if there's any change to those clocks while gaming, in case it's simply max available TDP kicking at default 0%.

Also, expand the setting under "power limit" (arrow at the right) to check what lies there or change the interface to a more modern one. Just in case it's not a temp target that is set very low.

Actions

dreamc @ alleyviper on Nov 8, 2017 8:22 PM

Where do I set the power limit? Crimson WattMan allows setting only _specific_ voltage for every state without specifying limits.

And... well... I'm not sure this is the right way to do it... As even browser (as I already stated) can "drop FPS" for a couple of seconds.

Actions

alleyviper @ dreamc on Nov 14, 2017 8:21 AM

My previous response was related to the MSI Afterburner screenshots that seem gone now. On them Power Limit was at 0%, I wasn't suggesting to change voltages. So I told you to raise power limit to ~20% and test gaming again to see if the card will keep higher clocks.

EDIT: Nvm about the screenshots, they're on your post. I couldn't seem them due to forum view.

Actions

dreamc @ alleyviper on Nov 9, 2017 4:32 PM

Oh... I didn't get it at first, as I do not use Afterburner for anything except for graphs (and those screenshots are still there). Basically I do not over/under clock my hardware at all.

However, I decided to test what you said and it didn't help - still periodic lags. Attached a screenshot.

Attachments

3.PNG155.4 KB

Actions

dreamc Nov 14, 2017 8:14 AM

I found the culprit of this one.

The moment I disconnect the second display (in this case the UHD TV), everything comes back to normal - no more UI lags, no more strange core clock drops, and even memory is now automatically downclocked when under low load (when the TV is plugged in, the memory clock always stays on 14xx mhz).

Why is this the problem?

I can't imagine this GPU to not be able to handle showing a static desktop. Or is it Windows 10 issues?

Is there anything I can do to make this work?

Actions

dreamc @ dreamc on Nov 19, 2017 11:06 AM

I'm still looking for any solutions or who should I contact regarding this issue.

Any help is appreciated.

Actions

alleyviper @ dreamc on Nov 21, 2017 6:06 AM

To mitigate the problem, have you tried to use alternative connections to check if for any reason the card changes behavior? I've seen many reports of stuck high memory clocks in the past when a secondary monitor is connected with a decent resolution, but never about lowered clocks. Maybe try to connect the hdmi TV to a DP via adapter on your card to see if anything changes, and do something similar with your monitor to check for combinations even if temporarily (like connecting it to a DP or DVI (or even dvi-i>vga port if its using hdmi). Also, you could try requesting the last vbios from its manufacturer. Good luck!

Actions

12 Replies

kingfish Nov 7, 2017 9:29 AM
Mark Correct
Correct Answer
Power Limit? That would be my guess...
AMD graphics performance
Like • Show 0 Likes0
Actions
- dreamc @ kingfish on Nov 7, 2017 9:50 AM
  Mark Correct
  Correct Answer
  All the settings are default (and even before updating to 1709 I had all on default, however there was no such problem).
  And I hope GPU does not try to lower it power state under High load...
  Like • Show 0 Likes0
  Actions
  - kingfish @ dreamc on Nov 7, 2017 10:02 AM
    Mark Correct
    Correct Answer
    It will definitely throttle the card.
    Like • Show 0 Likes0
    Actions
    - dreamc @ kingfish on Nov 7, 2017 12:12 PM
      Mark Correct
      Correct Answer
      "Definitely throttling the card" under load MUST be considered a not desired behavior and fixed.
      This did not affect me in any way before 1709 update. Testing it right now.
      Like • Show 0 Likes0
      Actions
    - dreamc @ kingfish on Nov 7, 2017 8:05 PM
      Mark Correct
      Correct Answer
      Unfortunately, this didn't help.
      I attached a couple of screenshots, where you can see some spikes in core clock.
      
      Is there anything else to check?
      Attachments
      2.PNG163.2 KB
      1.PNG188.7 KB
      Like • Show 0 Likes0
      Actions
      - alleyviper @ dreamc on Nov 8, 2017 7:32 AM
        Mark Correct
        Correct Answer
        Please try to raise power limit to something like 20% (or above, if that's not max). This is to check if there's any change to those clocks while gaming, in case it's simply max available TDP kicking at default 0%.
        Also, expand the setting under "power limit" (arrow at the right) to check what lies there or change the interface to a more modern one. Just in case it's not a temp target that is set very low.
        Like • Show 0 Likes0
        Actions
        dreamc @ alleyviper on Nov 8, 2017 8:22 PM
        Mark Correct
        Correct Answer
        Where do I set the power limit? Crimson WattMan allows setting only _specific_ voltage for every state without specifying limits.
        And... well... I'm not sure this is the right way to do it... As even browser (as I already stated) can "drop FPS" for a couple of seconds.
        Like • Show 0 Likes0
        Actions
        alleyviper @ dreamc on Nov 14, 2017 8:21 AM
        Mark Correct
        Correct Answer
        My previous response was related to the MSI Afterburner screenshots that seem gone now. On them Power Limit was at 0%, I wasn't suggesting to change voltages. So I told you to raise power limit to ~20% and test gaming again to see if the card will keep higher clocks.
        
        EDIT: Nvm about the screenshots, they're on your post. I couldn't seem them due to forum view.
        Like • Show 0 Likes0
        Actions
        dreamc @ alleyviper on Nov 9, 2017 4:32 PM
        Mark Correct
        Correct Answer
        Oh... I didn't get it at first, as I do not use Afterburner for anything except for graphs (and those screenshots are still there). Basically I do not over/under clock my hardware at all.
        
        However, I decided to test what you said and it didn't help - still periodic lags. Attached a screenshot.
        Attachments
        3.PNG155.4 KB
        Like • Show 0 Likes0
        Actions
dreamc Nov 14, 2017 8:14 AM
Mark Correct
Correct Answer
I found the culprit of this one.
The moment I disconnect the second display (in this case the UHD TV), everything comes back to normal - no more UI lags, no more strange core clock drops, and even memory is now automatically downclocked when under low load (when the TV is plugged in, the memory clock always stays on 14xx mhz).

Why is this the problem?
I can't imagine this GPU to not be able to handle showing a static desktop. Or is it Windows 10 issues?
Is there anything I can do to make this work?
Like • Show 0 Likes0
Actions
- dreamc @ dreamc on Nov 19, 2017 11:06 AM
  Mark Correct
  Correct Answer
  I'm still looking for any solutions or who should I contact regarding this issue.
  Any help is appreciated.
  Like • Show 0 Likes0
  Actions
- alleyviper @ dreamc on Nov 21, 2017 6:06 AM
  Mark Correct
  Correct Answer
  To mitigate the problem, have you tried to use alternative connections to check if for any reason the card changes behavior? I've seen many reports of stuck high memory clocks in the past when a secondary monitor is connected with a decent resolution, but never about lowered clocks. Maybe try to connect the hdmi TV to a DP via adapter on your card to see if anything changes, and do something similar with your monitor to check for combinations even if temporarily (like connecting it to a DP or DVI (or even dvi-i>vga port if its using hdmi). Also, you could try requesting the last vbios from its manufacturer. Good luck!
  Like • Show 0 Likes0
  Actions

R9 380 Core Clock Dropping - Windows 10 FCU 17.11.1

Attachments

Outcomes

Attachments

Attachments

Related Content

Recommended Content