Jump to content

X299 ACPI latency issue (Asus Rampage VI Apex -- EC slowdown)


Silent_Scone
 Share

Recommended Posts

Hi,

Using build v5.95.4569 on Windows 17074 Insider Preview. I was having this issue on earlier builds of AIDA, so renewed my maintenance period but to no avail.

When low-level sensor support is enabled, AIDA and ACPI will put a heavy overhead on the system as one thread is loaded.This is easily replicated by enabling and disabling the option. Can someone please look into this?


Many thanks

 

DxDiag.txt

Link to comment
Share on other sites

Please right-click on the bottom status bar of AIDA64 main window --> Sensor Debug --> Sensor Profiling Dump. Copy-paste the full results into this topic, or attach the results as a TXT file to your post. You may need to enable status bar in AIDA64 / main menu / View first.


Thanks,
Fiery

Link to comment
Share on other sites

  • 2 weeks later...
On ‎2018‎. ‎03‎. ‎12‎. at 6:00 PM, Silent_Scone said:

Hi Fiery, thanks for the response. Attached the dump. Unfortunately, I wouldn't be able to tell you when this started as I'd only noticed it very recently.

 

 

sensorprof.txt

Thank you.  For me it looks as if the Aquaero monitoring module causes the latency issues. Can you please try to disable the Aquaero module and check if it makes it any better?  You can do that by right-clicking on the bottom status bar of AIDA64 main window --> HWMon Modules --> Aquaero/b]. You may need to enable status bar in AIDA64 / main menu / View first.

Link to comment
Share on other sites

On 3/26/2018 at 4:21 PM, Fiery said:

Thank you.  For me it looks as if the Aquaero monitoring module causes the latency issues. Can you please try to disable the Aquaero module and check if it makes it any better?  You can do that by right-clicking on the bottom status bar of AIDA64 main window --> HWMon Modules --> Aquaero/b]. You may need to enable status bar in AIDA64 / main menu / View first.

Having tried this, the issue is still there (have tried the latest build posted Monday, too). Only when AIDA is launched does the load get placed on the system

Link to comment
Share on other sites

On 3/26/2018 at 4:21 PM, Fiery said:

Thank you.  For me it looks as if the Aquaero monitoring module causes the latency issues. Can you please try to disable the Aquaero module and check if it makes it any better?  You can do that by right-clicking on the bottom status bar of AIDA64 main window --> HWMon Modules --> Aquaero/b]. You may need to enable status bar in AIDA64 / main menu / View first.

Hi,

 

The load subsides when Motherboard is unticked from HWMon Modules.

Link to comment
Share on other sites

On ‎2018‎. ‎03‎. ‎31‎. at 8:58 AM, Silent_Scone said:

Hi,

 

The load subsides when Motherboard is unticked from HWMon Modules.

Thank you for your test runs.  Please right-click on the bottom status bar of AIDA64 main window --> Sensor Debug --> ISA Sensor Dump. Copy-paste the full results into this topic, or attach the results as a TXT file to your post.

Also right-click on the bottom status bar of AIDA64 main window --> Sensor Debug --> Embedded Controller Dump. Copy-paste the full results into this topic, or attach the results as a TXT file to your post.


Thanks,
Fiery

Link to comment
Share on other sites

 

Hi Fiery, please let me know if you find anything.

On 4/1/2018 at 9:27 AM, Fiery said:

Thank you for your test runs.  Please right-click on the bottom status bar of AIDA64 main window --> Sensor Debug --> ISA Sensor Dump. Copy-paste the full results into this topic, or attach the results as a TXT file to your post.

Also right-click on the bottom status bar of AIDA64 main window --> Sensor Debug --> Embedded Controller Dump. Copy-paste the full results into this topic, or attach the results as a TXT file to your post.


Thanks,
Fiery

3

 

ecdump.txt

Link to comment
Share on other sites

On ‎2018‎. ‎04‎. ‎02‎. at 12:46 PM, Silent_Scone said:

 

No problem, attached.

ecdump.txt

Thank you.  We're investigating this issue.  Apparently something makes the EC register readout slow down tremendously on certain Asus motherboards, but only intermittently.  Sometimes it's lightning fast (like 0.2 millisec), but in other cases it can be 1000 times slower (over 200 millisec).

Link to comment
Share on other sites

On 09/04/2018 at 5:58 PM, Fiery said:

Thank you.  We're investigating this issue.  Apparently something makes the EC register readout slow down tremendously on certain Asus motherboards, but only intermittently.  Sometimes it's lightning fast (like 0.2 millisec), but in other cases it can be 1000 times slower (over 200 millisec).

Thank you, a few users report that disabling Thunderbolt3 from within the BIOS resolves the issue, but not all cases are experiencing this issue due to AIDA being open. I've yet to try replicating that but might help.

Link to comment
Share on other sites

4 hours ago, Silent_Scone said:

Thank you, a few users report that disabling Thunderbolt3 from within the BIOS resolves the issue, but not all cases are experiencing this issue due to AIDA being open. I've yet to try replicating that but might help.

Thank you for the info.  We'll do relevant test runs tonight as well.  Meanwhile, we've added more debug information to the Embedded Controller dump.  Please post an updated such dump using the latest AIDA64 beta build:

https://www.aida64.com/downloads/latesta64xebeta

Thank you in advance!

Link to comment
Share on other sites

19 hours ago, Fiery said:

Thank you for the info.  We'll do relevant test runs tonight as well.  Meanwhile, we've added more debug information to the Embedded Controller dump.  Please post an updated such dump using the latest AIDA64 beta build:

https://www.aida64.com/downloads/latesta64xebeta

Thank you in advance!

 

 

No problem, have downloaded the beta and dumped the EC again. To replicate the issue with TB3 I'm fairly sure you will need the TB3 extension card, too. Unfortunately, I do not have this to hand. The problem is possibly related, however, AIDA isn't required to reproduce the interrupt on ACPI in the instances reported.

ecdump.txt

Link to comment
Share on other sites

1 hour ago, Silent_Scone said:

No problem, have downloaded the beta and dumped the EC again. To replicate the issue with TB3 I'm fairly sure you will need the TB3 extension card, too. Unfortunately, I do not have this to hand. The problem is possibly related, however, AIDA isn't required to reproduce the interrupt on ACPI in the instances reported.

ecdump.txt

Thank you!  With TB3 disabled we can still experience the same issue on our own Asus Prime X299-A motherboard based test system.  In your dump you can see the anomaly as:

Time #3 = 0.230 / 0.272 / 0.269 / 0.225 / 0.258 / 0.241 / 0.266 / 0.230 / 0.230 / 0.301 / 0.218 / 262.524 / 0.338 / 0.443 / 0.230 / 0.246 msec

AIDA64 reads 16 registers from the EC there, and one case out of 16 it slows down tremendously.  The chances of that happening seems to be similar on our Prime X299-A motherboard.  So typically once per every 10 or 20 EC register accesses gets a penalty that way.  We'll contact Asus to find out what can we do on our end to avoid the penalties.

 

Link to comment
Share on other sites

21 hours ago, Fiery said:

Thank you!  With TB3 disabled we can still experience the same issue on our own Asus Prime X299-A motherboard based test system.  In your dump you can see the anomaly as:

Time #3 = 0.230 / 0.272 / 0.269 / 0.225 / 0.258 / 0.241 / 0.266 / 0.230 / 0.230 / 0.301 / 0.218 / 262.524 / 0.338 / 0.443 / 0.230 / 0.246 msec

AIDA64 reads 16 registers from the EC there, and one case out of 16 it slows down tremendously.  The chances of that happening seems to be similar on our Prime X299-A motherboard.  So typically once per every 10 or 20 EC register accesses gets a penalty that way.  We'll contact Asus to find out what can we do on our end to avoid the penalties.

 

Glad you're able to reproduce the problem! I wish I could tell you when exactly it began, as I've been using AIDA in conjunction with this board for some time.

Link to comment
Share on other sites

  • 2 weeks later...
On 4/12/2018 at 9:25 AM, Fiery said:

Thank you!  With TB3 disabled we can still experience the same issue on our own Asus Prime X299-A motherboard based test system.  In your dump you can see the anomaly as:

Time #3 = 0.230 / 0.272 / 0.269 / 0.225 / 0.258 / 0.241 / 0.266 / 0.230 / 0.230 / 0.301 / 0.218 / 262.524 / 0.338 / 0.443 / 0.230 / 0.246 msec

AIDA64 reads 16 registers from the EC there, and one case out of 16 it slows down tremendously.  The chances of that happening seems to be similar on our Prime X299-A motherboard.  So typically once per every 10 or 20 EC register accesses gets a penalty that way.  We'll contact Asus to find out what can we do on our end to avoid the penalties.

 

 
4

Hi Fiery, is there any chance of an update on this issue? :)

Link to comment
Share on other sites

1 hour ago, Silent_Scone said:

Hi Fiery, is there any chance of an update on this issue? :)

We've sent a request to Asus to investigate this issue, but they haven't completed their efforts to replicate the anomaly and dig down deeper to find out what exactly causes the slowdowns.  I'll let you know once we know more about this issue.  I'm afraid at this point we have no tools to get around the problem or to fix it on our end. 

Link to comment
Share on other sites

8 minutes ago, Fiery said:

We've sent a request to Asus to investigate this issue, but they haven't completed their efforts to replicate the anomaly and dig down deeper to find out what exactly causes the slowdowns.  I'll let you know once we know more about this issue.  I'm afraid at this point we have no tools to get around the problem or to fix it on our end. 

 

No problem, thanks for the update

Link to comment
Share on other sites

  • 2 weeks later...

Hi Fiery, thanks for the above link. A test BIOS was released for X299-A to a particular user that allegedly remedies similar problems with high latency on ACPI (not strictly AIDA).

If you do not have it yet I will see if I can grab it for replication. I do not have this board here to test, unfortunately.

Link to comment
Share on other sites

On ‎2018‎. ‎05‎. ‎15‎. at 11:51 AM, Silent_Scone said:

Hi Fiery, thanks for the above link. A test BIOS was released for X299-A to a particular user that allegedly remedies similar problems with high latency on ACPI (not strictly AIDA).

If you do not have it yet I will see if I can grab it for replication. I do not have this board here to test, unfortunately.

We'd rather wait for a public BIOS release for Prime X299-A.  It's an essential machine in our test fleet (due to AVX-512 support and many CPU cores), so we'd try not to mess it up with a test BIOS :) 

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
 Share

×
×
  • Create New...