Is there a guide on how to troubleshoot xperf outputs?

M

MarceloLagos

Member
Joined
Feb 16, 2016
Posts
5
I'm having sound crackling and stutters. I've been reading the forums and running xperf and it seems to always point to the DirectX driver or hal.dll driver.
I've updated drivers, disabled/modified BIOS settings regarding the CPU (not overclocked) , USB, etc.
But I can't figure out the xperf output. Is there a guide I could use to understand it?

Update:

Just in case:
ASUS Maximus VIII HERO
i5 6600
2x8GB G.Skill DDR4-3200
SSD Samsung EVO 850 and a 1TB WD disk.
SoundBlaster Z and EVGA Nvidia 970

CPU.7z - Google Driv
 
Thanks for the reply.
It's a very new Windows 7 installation without any security suite nor antivirus.

Not sure what to make of from observing hal.dll as the one showing the problem... is it hardware?
 
If you've got the .ETL trace uploaded somewhere (after zipping, of course) I can take a look and show you what I find.
 
Hi all -

Marcelo - any luck yet?

Nice specs on your system. Do the pops & clicks mainly happen when gaming? Especially when playing lower-end games? Or?

[cluberti - I think Marcelo gave a link to his file on Google Drive - it's highlighted in blue]
 
Hi guys, I think I found a big clue:

I built the pc using an Asus Xonar card first and thinking it was the card I went for the SB Z but the issue remained, so it wasn't the card.
It is very easy for me to reproduce: start a youtube music video and then open the CPU.etl. Stuttering starts as soon as wpa finishes loading the file and whenever I start selecting curves or zooming in.
Weird thing is I have no problems in high end games like Battlefield 4. Only on "lower-end" games (like bejeweled) the problem appears. And I can feel the stuttering in the mouse cursor when it happens (it jumps a bit).

Anyhow, tonight I tried moving the card to other PCI slots and there were no changes... until I tried the PCIx8 slot (usually used for a second video card).
Now I can't reproduce the stutter unless I really stress the system (disk access, CPU, etc.), which should be normal I believe.

Now I'm thinking it's maybe a Motherboard problem? The short PCI slots should be good enough for a sound card?
(I also tested the BIOS settings for the PCI ports moving them from Auto to PCI Gen 3 speed, etc.)

At least it's more stable now, but I wonder if it's still a problem just masked by using a PCIx8 port and I also wonder if it will affect the first PCI performance (video card) because it will go from x16 to x8 speed.

Any suggestion is welcomed.

Thanks!
 
Hi again


I'm no expert at gaming systems, but it seems like it would be nice to let the 970 have a full strength x16 slot. Have you tried using any of the "sync" settings in the nVidia drivers? I've heard gamers talk about V-Sync / Advanced (or Adaptive?) V-Sync / and a new one called G-Sync ... Seems like it might help if the cause of your stutters and pops are because your card is able to play video at such high rates - paradoxically, the monitor can't keep up, and the video stuttering (so I've heard) can affect the audio stream as well. Have a look at the nVidia website - especially for feature for your card. You could check for anything new with your sound device as well (but it looks like you already have the latest drivers).

We can also see if jcgriff or cluberti find anything in your trace log.

You could try setting Bios back to defaults, if you've changed a lot of settings ... just as a test.
 
Thanks for the suggestions OldGrayGary. Although I should clarify the following:
- When using the video card at full capacity (like with 3d games) there is no stuttering at all. Everything is smooth.
- I've cleared BIOS settings several times, and even tested using lower speeds for the memory.
- The issue mostly happens on normal 2D Windows desktop, while playing a low-end game (windowed game) like Bejeweled 3, watching a YouTube video, etc.

I believe for a single card using x8 vs x16 on PCI 3.0 shouldn't be much of a difference, but that is not my main concern. I would like to know if it's the Motherboard the one that fails to handle the PCI3.0 at 16x, or if it's the video card, or something else.
I already changed the sound card to test, but buying another 970 or another VIII Hero without being certain which one it's at fault is kind of expensive. I've been thinking on buying the 970 so if it's not the culprit at least I would have SLI after replacing the motherboard...

I wish there was another way to test this without having to buy another card or motherboard.
 
Actually, the popping problems that those "sync" utilities (or features) are supposed to solve are for both the very high and very low ends of the ranges. That's why I'm thinking they are worth a go. That's why I thought of them when you mentioned that it happens when watching a YouTube video. That's about as low-end as it gets. Can't hurt!

I'm sorry that I'm not much experienced in the Windows performance trace logs - I've never had to use them (in all these years!), but I can see why they'd be helpful for high-performance systems. [The high-performance systems I worked on for 25 years were mainframes :) ] But when I took a look at your etl file in EventViewer, it looked to me like the same message over & over --- (or very close to the same message .. what changed was a "payload" value)

If any of the hardware wizards are reading this: in the Processing Error Data section, ErrorCode 15003 appears in every single line of the log. The log is given the name "CPU" when extracted.

Maybe the trace experts will have some ideas to share with us.

[I wouldn't buy another motherboard or card just yet ... unless they'd send you a free one under warranty :) ]
 
Hi again .... apparently, it looks like a tool called the "Windows Performance Analyzer" (which quickly gets referred to as WPA in the articles) is handy for viewing and analyzing the trace logs. I'm thinking that perhaps you might have it already - if you downloaded the Windows Assessment and Deployment Kit ("Windows ADK", for short). A brief intro that I skimmed says it can read the .etl trace files.

If you want to read up on that tool a bit, here's the link
Windows Performance Analyzer - Windows 10 hardware de

It is, as the title suggests, aimed at developers rather than end users. ... but if it proves useful as a diagnostic - I guess I might have to have a look (teach an old, old dog a new trick)
 
Thank you Gary. That is what I was looking for, although it seems that is not easy to grasp...
And yes, I was using wpa to analyze the output of xperf but I need more knowledge to correctly interpret the results.

On a side note, I explained my tests in a support ticket with Asus and they immediately offered a RMA... I guess the MB is more likely to be at fault than the graphics card...
 
Nice to hear that Asus still lives up to their reputation for good customer service.

If all goes well with a replacement, I suppose our Windows ADK & WPA educations won't be quite as immediately pressing :) ... Though it does look like an interesting set of tools to explore (...from the reading I looked at earlier, much of it looks to be aimed at large company-wide roll-outs of Windows 10 ... but it's also possible that their might be tools here and there that a consumer or enthusiast might benefit from ...)

Hope it works out, then you get to enjoy a speedy non-stuttering computer at last!
 
In looking at the DPC/ISR timeframe from your trace, and zooming in on one of the (very obvious) peaks on the DPC usage by module/stack graph (seen below the Timeline graph in the screenshot below), we can see that this behavior is driven primarily by some sort of function in the nvidia kernel driver (nvlddmkm.sys), and a corresponding call to the directx kernel module (dxgkrnl.sys). There's also an ndis.sys call that happens (likely due to chrome.exe being open), but that isn't contributing at all (it's what I would normally see as expected behavior with a browser open):
directx_nvidia.jpg
The entire DPC is attributed to DirectX (the giant flattop peak in the bottom graph), but when you break it down you can see that it's something to do with the nvidia kernel driver driving it. This is actually not entirely a surprising behavior, and we see it regularly when accelerated apps run in a window (say, Chrome, or a windowed game) that don't occur when full-screen. I have the same Asus motherboard, using onboard sound, but I am using an Asus Strix GTX970 card and I have no issues at all with my setup. I hate to say it, but this is one of the problems with DiY systems, namely hardware and driver compat. This unfortunately happens regularly, where some sort of network or audio behavior can trigger it, but less common is a video trigger (as we see here). I'd guess the placement of the card will indeed matter, but only Asus could really tell you what the optimal PCI-E slot is for the card you have, and then if anything needs to be adjusted in the UEFI OS to make sure it works appropriately.
 

Has Sysnative Forums helped you? Please consider donating to help us support the site!

Back
Top