nVidia releases hotfix driver to partially address DPC latency issues with Pascal based video cards

[zMeul]

source: https://forums.geforce.com/default/topic/951723/geforce-drivers/announcing-geforce-hotfix-driver-368-95/1/

the driver can be accessed here: http://nvidia.custhelp.com/app/answers/detail/a_id/4202

 

Quote

Today we released a new Hotfix driver that addresses the following issue:

• Fixed DPC latency bug on Pascal GPUs.

 

this issue has been reported for about a month, nVidia acknowledged it and they released a hotfix

to note: this hotfix doesn't work for multi GPU setups - nVidia still has work to do

 

Resplendence Software have updated their LatencyMon to support Windows 10: http://www.resplendence.com/latencymon

 

----

 

I see some people are confused of what DPC Latency is:

DPC (Deferred Procedure Call) is the part of Win system that deals with driver efficiency

the most common issue with DPC Latency is with audio subsystem that exhibits clicks, pops, distortion and dropouts

Edited July 26, 2016 by zMeul

[rattacko123]

46 minutes ago, zMeul said:

this issue has been reported for about a month

"hotfix"

[Michael McAllister]

Didn't appear to fix this issue for me.

 

Spoiler

_________________________________________________________________________________________________________
CONCLUSION
_________________________________________________________________________________________________________
Your system appears to be having trouble handling real-time audio and other tasks. You are likely to experience buffer underruns appearing as drop outs, clicks or pops. One problem may be related to power management, disable CPU throttling settings in Control Panel and BIOS setup. Check for BIOS updates. 
LatencyMon has been analyzing your system for  0:18:22  (h:mm:ss) on all processors.

_________________________________________________________________________________________________________
SYSTEM INFORMATION
_________________________________________________________________________________________________________
Computer name:                                        MARSONIA-PC
OS version:                                           Windows 7 Service Pack 1, 6.1, build: 7601 (x64)
Hardware:                                             ASUSTeK COMPUTER INC., P8Z77-V PRO/THUNDERBOLT
CPU:                                                  GenuineIntel Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
Logical processors:                                   8
Processor groups:                                     1
RAM:                                                  16328 MB total

_________________________________________________________________________________________________________
CPU SPEED
_________________________________________________________________________________________________________
Reported CPU speed:                                   350 MHz
Measured CPU speed:                                   1 MHz (approx.)

Note: reported execution times may be calculated based on a fixed reported CPU speed. Disable variable speed settings like Intel Speed Step and AMD Cool N Quiet in the BIOS setup for more accurate results.

WARNING: the CPU speed that was measured is only a fraction of the CPU speed reported. Your CPUs may be throttled back due to variable speed settings and thermal issues. It is suggested that you run a utility which reports your actual CPU frequency and temperature. 

_________________________________________________________________________________________________________
MEASURED INTERRUPT TO USER PROCESS LATENCIES
_________________________________________________________________________________________________________
The interrupt to process latency reflects the measured interval that a usermode process needed to respond to a hardware request from the moment the interrupt service routine started execution. This includes the scheduling and execution of a DPC routine, the signaling of an event and the waking up of a usermode thread from an idle wait state in response to that event.

Highest measured interrupt to process latency (µs):   14091.676941
Average measured interrupt to process latency (µs):   2.237491

Highest measured interrupt to DPC latency (µs):       840.246152
Average measured interrupt to DPC latency (µs):       0.885516

_________________________________________________________________________________________________________
 REPORTED ISRs
_________________________________________________________________________________________________________
Interrupt service routines are routines installed by the OS and device drivers that execute in response to a hardware interrupt signal.

Highest ISR routine execution time (µs):              78.949714
Driver with highest ISR routine execution time:       dxgkrnl.sys - DirectX Graphics Kernel, Microsoft Corporation

Highest reported total ISR routine time (%):          0.140794
Driver with highest ISR total time:                   dxgkrnl.sys - DirectX Graphics Kernel, Microsoft Corporation

Total time spent in ISRs (%)                          0.245778

ISR count (execution time <250 µs):                   5807603
ISR count (execution time 250-500 µs):                0
ISR count (execution time 500-999 µs):                0
ISR count (execution time 1000-1999 µs):              0
ISR count (execution time 2000-3999 µs):              0
ISR count (execution time >=4000 µs):                 0

_________________________________________________________________________________________________________
REPORTED DPCs
_________________________________________________________________________________________________________
DPC routines are part of the interrupt servicing dispatch mechanism and disable the possibility for a process to utilize the CPU while it is interrupted until the DPC has finished execution.

Highest DPC routine execution time (µs):              414.535143
Driver with highest DPC routine execution time:       nvlddmkm.sys - NVIDIA Windows Kernel Mode Driver, Version 368.95 , NVIDIA Corporation

Highest reported total DPC routine time (%):          0.296202
Driver with highest DPC total execution time:         USBPORT.SYS - USB 1.1 & 2.0 Port Driver, Microsoft Corporation

Total time spent in DPCs (%)                          0.547686

DPC count (execution time <250 µs):                   11550570
DPC count (execution time 250-500 µs):                0
DPC count (execution time 500-999 µs):                12
DPC count (execution time 1000-1999 µs):              0
DPC count (execution time 2000-3999 µs):              0
DPC count (execution time >=4000 µs):                 0

_________________________________________________________________________________________________________
 REPORTED HARD PAGEFAULTS
_________________________________________________________________________________________________________
Hard pagefaults are events that get triggered by making use of virtual memory that is not resident in RAM but backed by a memory mapped file on disk. The process of resolving the hard pagefault requires reading in the memory from disk while the process is interrupted and blocked from execution.

NOTE: some processes were hit by hard pagefaults. If these were programs producing audio, they are likely to interrupt the audio stream resulting in dropouts, clicks and pops. Check the Processes tab to see which programs were hit.

Process with highest pagefault count:                 nvspcaps64.exe

Total number of hard pagefaults                       1995
Hard pagefault count of hardest hit process:          782
Highest hard pagefault resolution time (µs):          126554.939143
Total time spent in hard pagefaults (%):              0.02830
Number of processes hit:                              13

_________________________________________________________________________________________________________
 PER CPU DATA
_________________________________________________________________________________________________________
CPU 0 Interrupt cycle time (s):                       71.666802
CPU 0 ISR highest execution time (µs):                78.949714
CPU 0 ISR total execution time (s):                   21.674389
CPU 0 ISR count:                                      5807603
CPU 0 DPC highest execution time (µs):                414.535143
CPU 0 DPC total execution time (s):                   45.703839
CPU 0 DPC count:                                      11012263
_________________________________________________________________________________________________________
CPU 1 Interrupt cycle time (s):                       5.170798
CPU 1 ISR highest execution time (µs):                0.0
CPU 1 ISR total execution time (s):                   0.0
CPU 1 ISR count:                                      0
CPU 1 DPC highest execution time (µs):                53.402571
CPU 1 DPC total execution time (s):                   0.668430
CPU 1 DPC count:                                      94519
_________________________________________________________________________________________________________
CPU 2 Interrupt cycle time (s):                       4.636538
CPU 2 ISR highest execution time (µs):                0.0
CPU 2 ISR total execution time (s):                   0.0
CPU 2 ISR count:                                      0
CPU 2 DPC highest execution time (µs):                47.5260
CPU 2 DPC total execution time (s):                   0.349104
CPU 2 DPC count:                                      85223
_________________________________________________________________________________________________________
CPU 3 Interrupt cycle time (s):                       4.785686
CPU 3 ISR highest execution time (µs):                0.0
CPU 3 ISR total execution time (s):                   0.0
CPU 3 ISR count:                                      0
CPU 3 DPC highest execution time (µs):                53.148857
CPU 3 DPC total execution time (s):                   0.096446
CPU 3 DPC count:                                      36963
_________________________________________________________________________________________________________
CPU 4 Interrupt cycle time (s):                       5.972615
CPU 4 ISR highest execution time (µs):                0.0
CPU 4 ISR total execution time (s):                   0.0
CPU 4 ISR count:                                      0
CPU 4 DPC highest execution time (µs):                46.60
CPU 4 DPC total execution time (s):                   0.401527
CPU 4 DPC count:                                      96595
_________________________________________________________________________________________________________
CPU 5 Interrupt cycle time (s):                       4.478912
CPU 5 ISR highest execution time (µs):                0.0
CPU 5 ISR total execution time (s):                   0.0
CPU 5 ISR count:                                      0
CPU 5 DPC highest execution time (µs):                35.286857
CPU 5 DPC total execution time (s):                   0.060398
CPU 5 DPC count:                                      18437
_________________________________________________________________________________________________________
CPU 6 Interrupt cycle time (s):                       5.402315
CPU 6 ISR highest execution time (µs):                0.0
CPU 6 ISR total execution time (s):                   0.0
CPU 6 ISR count:                                      0
CPU 6 DPC highest execution time (µs):                40.210286
CPU 6 DPC total execution time (s):                   0.921375
CPU 6 DPC count:                                      180406
_________________________________________________________________________________________________________
CPU 7 Interrupt cycle time (s):                       4.147915
CPU 7 ISR highest execution time (µs):                0.0
CPU 7 ISR total execution time (s):                   0.0
CPU 7 ISR count:                                      0
CPU 7 DPC highest execution time (µs):                37.170286
CPU 7 DPC total execution time (s):                   0.097532
CPU 7 DPC count:                                      26176
_________________________________________________________________________________________________________
 

 

7

 

[Prysin]

i find it hillarious, that whenever AMD has a major issue, that is not specific to game performance, but rather general usage. the issue is generally acknowledged and fixed in a week.

 

When Nvidia has a major issue, they spend 14 days + concocting some sort of PR statement, then another month or two fixing the error. Make me think that Nvidia doesnt take their customers seriously.

 

Also bit funny that when AMD fails, they get flamed to death just hours after bug is found. When Nvidia fails, nobody says a word and just sweeps it under the rug for as long as possible...

[Misanthrope]

4 hours ago, Prysin said:

i find it hillarious, that whenever AMD has a major issue, that is not specific to game performance, but rather general usage. the issue is generally acknowledged and fixed in a week.

 

When Nvidia has a major issue, they spend 14 days + concocting some sort of PR statement, then another month or two fixing the error. Make me think that Nvidia doesnt take their customers seriously.

 

Also bit funny that when AMD fails, they get flamed to death just hours after bug is found. When Nvidia fails, nobody says a word and just sweeps it under the rug for as long as possible...

Except for you know...this thread, your comment, comments from about 10 to 20 AMD enthusiasts that will soon come here and make a comment, etc.

[shermantanker]

What the heck is DPC latency? I have never heard of this.

[79wjd]

6 hours ago, Prysin said:

Also bit funny that when AMD fails, they get flamed to death just hours after bug is found. When Nvidia fails, nobody says a word and just sweeps it under the rug for as long as possible...

Nvidia doesnt get flamed to death? Did you forget about the 120 page long 970 thread last year?

[samcool55]

Can someone explain me what it actually fixes?

I don't have a nvidia card but some of my friends do so it would be nice to know what the issue was/is.

 

No idea what DPC is... also the fix is partially which sounds terrible tbh...

[Prysin]

1 hour ago, djdwosk97 said:

Nvidia doesnt get flamed to death? Did you forget about the 120 page long 970 thread last year?

the one and only time in 4 years i can remember such an event happen to Nvidia.

[Prysin]

3 hours ago, Misanthrope said:

Except for you know...this thread, your comment, comments from about 10 to 20 AMD enthusiasts that will soon come here and make a comment, etc.

Said the defacto Nvidia enthusiast himself.

[Misanthrope]

16 minutes ago, Prysin said:

Said the defacto Nvidia enthusiast himself.

 

2 hours ago, djdwosk97 said:

Nvidia doesnt get flamed to death? Did you forget about the 120 page long 970 thread last year?

 

[manikyath]

On 23-7-2016 at 6:10 AM, rattacko123 said:

"hotfix"

you know, gpu drivers are one of the very few pieces of software that still have significant amounts of assembly in them, and assembly devs are just the kind of people that take 4 months for something, because it's not like they have a whole lot of tools to work with

(i had a teacher that used to work as an assembly dev, and these are literally teams hired for multiple months to skin 1 byte off of a piece of software so it can be shipped on a cheaper controller, the work they do is mental, and one of the few parts of programming i have that great amount of respect for.)

 

what i'm guessing is that the dev time required for this "hotfix" took longer to locate in code than they took writing a "hotfix" for it.

[Prysin]

1 hour ago, Misanthrope said:

 

 

reading is not your forte i see.

[Misanthrope]

5 minutes ago, Prysin said:

reading is not your forte i see.

We can't all have strong points like derailing the conversation with a pointless ad hominem line like you I'm affraid.

[Prysin]

4 minutes ago, Misanthrope said:

We can't all have strong points like derailing the conversation with a pointless ad hominem line like you I'm affraid.

quality derailings are my speciality

[rattacko123]

4 hours ago, manikyath said:

 

what i'm guessing is that the dev time required for this "hotfix" took longer to locate in code than they took writing a "hotfix" for it.

Well, it took AMD about 3 days to fix their RX480 power consumption, but it took 1 month for nvidia to fix their displayport issues. For a company as large as nvidia it shouldn't take this long, they still havent fixed the stability issues of their newer drivers

[manikyath]

1 minute ago, rattacko123 said:

Well, it took AMD about 3 days to fix their RX480 power consumption, but it took 1 month for nvidia to fix their displayport issues. For a company as large as nvidia it shouldn't take this long, they still havent fixed the stability issues of their newer drivers

well, the RX480 problem was a pretty easy fix:

- pulls too much power from the slot

- should pull more power from 6pin

- bias the core to use the 6pin vrms more than the slot vrms

- send the patch out.

 

AMD can kiss their hands lucky that it's defined in software, and not hardware.

--

the speed at which a patch can be released entirely depends on luck. and to be honest i'm amazed GPU drivers work as well as they do, with the near endless amount of variables the driver developers have to account for. and sometimes you just cant find what the quack is causing a bug, like what i'm sure the DP thing ended up being. it's always something stupid, and even paying the best devs in the world doesnt make finding that one glitch any easier. 

[Tech Inquisition]

Great, another WHQL driver on the horizon for me to test...

[ManIkWeet]

7 hours ago, rattacko123 said:

Well, it took AMD about 3 days to fix their RX480 power consumption, but it took 1 month for nvidia to fix their displayport issues. For a company as large as nvidia it shouldn't take this long, they still havent fixed the stability issues of their newer drivers

Let's compare this "they are more so they must be faster" with something else:

 

I need a baby next month, so I will employ 9 females.

[Sintezza]

Yes i broaght this particular issue to the light on a few other topics.

dpc latency spikes have allways been a slight issue with Nvidia cards in general.

But with the Pascal series it was that bad, that people exaly starting to have issues with it, lots of micro stutter etc.

Nvidia finaly seems to have a half baked fix right now that might fix a problem that they basicly allready deal with for years now.

And i personaly highly doubt that this fix will be the end of their dpc latency problems.

Because high dpc latency issues are very hard to adress.

[ace_cheaply]

On 7/25/2016 at 9:48 AM, shermantanker said:

What the heck is DPC latency? I have never heard of this.

 

On 7/25/2016 at 11:23 AM, samcool55 said:

Can someone explain me what it actually fixes?

I don't have a nvidia card but some of my friends do so it would be nice to know what the issue was/is.

 

No idea what DPC is... also the fix is partially which sounds terrible tbh...

From google " Many audio problems on a computer can be caused by DPC latency. DPC stands for Deferred Procedure Call. In its simplest form it is the part of your Windows system that handles driver efficiency. If there is a driver that is taking longer than normal to process, it may prevent other drivers from being processed in time."  I only just found out about it because i was getting really bad stutter and audio hang ups in youtube.  I was advised to download latencymon to check the problem and seen this

After downloading the hotfix, reinstalling audio drivers, and uninstalling avast, I now get this, which suggests the problem is still present in part, and the "partial" fix denomination was the correct one. However, given that youtube doesn't snap, crackle and pop every thirty seconds anymore I'll count it as a win. 

 

There is a new bios for my Sabertooth z170s from three days ago, but I doubt if it solves the problem. 

 

[Notional]

On 25/7/2016 at 7:06 PM, djdwosk97 said:

Nvidia doesnt get flamed to death? Did you forget about the 120 page long 970 thread last year?

NVidia losing a class action lawsuit about the 970 VRM issue tell us it was justified. Point still stands though. NVidia gets away with murder because NVidia.

4 hours ago, ace_cheaply said:

 

From google " Many audio problems on a computer can be caused by DPC latency. DPC stands for Deferred Procedure Call. In its simplest form it is the part of your Windows system that handles driver efficiency. If there is a driver that is taking longer than normal to process, it may prevent other drivers from being processed in time."  I only just found out about it because i was getting really bad stutter and audio hang ups in youtube.  I was advised to download latencymon to check the problem and seen this

After downloading the hotfix, reinstalling audio drivers, and uninstalling avast, I now get this, which suggests the problem is still present in part, and the "partial" fix denomination was the correct one. However, given that youtube doesn't snap, crackle and pop every thirty seconds anymore I'll count it as a win. 

 

There is a new bios for my Sabertooth z170s from three days ago, but I doubt if it solves the problem. 

 

Actually it is solved. Hard pagefaults has to do with the pagefile, not the drivers. It will often go to red, so don't worry about it.

[Energycore]

On 7/22/2016 at 11:10 PM, rattacko123 said:

"hotfix"

Give Nvidia a break their drivers are so much better anyway

 

/s

[Energycore]

On 7/25/2016 at 6:40 PM, manikyath said:

well, the RX480 problem was a pretty easy fix:

- pulls too much power from the slot

- should pull more power from 6pin

- bias the core to use the 6pin vrms more than the slot vrms

- send the patch out.

 

AMD can kiss their hands lucky that it's defined in software, and not hardware.

--

the speed at which a patch can be released entirely depends on luck. and to be honest i'm amazed GPU drivers work as well as they do, with the near endless amount of variables the driver developers have to account for. and sometimes you just cant find what the quack is causing a bug, like what i'm sure the DP thing ended up being. it's always something stupid, and even paying the best devs in the world doesnt make finding that one glitch any easier. 

But increasing the number of people (number of tries) will also increase the chance of finding the solution within a certain time frame because you're getting more tries per unit of time.

[Briggsy]

On 2016-07-25 at 11:50 AM, Misanthrope said:

Except for you know...this thread, your comment, comments from about 10 to 20 AMD enthusiasts that will soon come here and make a comment, etc.

exactly. I put up with AMD's epic driver delay's for years, waiting and waiting for fixes to some fairly stupid bugs and/or missing features. Hell, its been years and AMD's mouse cursor bugs still plague people, Sometimes its a giant cursor, sometimes your cursor in windows turns into some weird i-beam looking thing. If you have a 144hz display and update AMD drivers, sometimes you end up at 60hz for no apparent reason in windows and have to change the refresh rate back manually. sometimes you get a web browser freeze up that crashes drivers and forces you to restart you computer. The last time AMD fixed something inside a week (let alone fixed at all) was that power delivery issue with the 480 because they rushed out the card before engineers could iron out all the kinks, and because it was a potential lawsuit waiting to happen.