GA-X58A-UD5 Rev.2 Latency/DPC/audio issues

[Plutotype]

Hi,
this is my first post, so at first I would like to say hello to all forum members.

Recently a put together a system based on GA-X58A-UD5 Rev.2.

Last friday I received a new RME Babyface audiocard. As my future DAW software is not installed yet, wanted to test the sound in directsound playback in WMP. The install of the card is very simply. You download the newest drivers from RME, extract them somewhere in your computer and then you connect the card via USB to some stable powered USB slot. In device Manager you click Babyface on "update driver" and locate the extracted driver folder. Click ok, done. My card is installed corectly, but from the very beginning I noticed that the sound processing behaves bad. I put some wav or mp3 into the WMP, set Babyface as default output and the result?
Audio-dropouts, crackles and skiped audio frames or faster/slower playback (jus a little). Tried to play with the buffersize settings ( @48KHz:  48 - 2048 samples ) and sample rates. Higher buffer helped a little, but small crackles or audiodropouts occured from time to time.

I reinstalled the card at least 5times, tried different USB slot, did an driver upgrade for the NEC USB3.0 driver on the UD5. The audio issues were still there, so I asked at RME about the issue ( i thought its a soundcard issue ). They responded, that I have a driver/hardware problem in my computer and asked me to install the latest driver.

The issue continued when I tried to play video files with hd audio streams through well known MPC-HC player and set RME for audio playback. Before the purchase, I havent noticed the crackles or audio drop-outs, because somehow the integrated Realtek Audio chip on the Gigabyte board has hidden the DPC audio issues, which went on with the new RME Card. The crackles were immediate after starting the playback and they were not occasional. I tried to play with the buffer size in Babyface USB settings and there I have noticed USB transmission errors during playback.
http://postimage.org/image/hurcnjes/

Then I started Latency checker and LatencyMon tools to check, if there is some communication problem on the USB port. With RME set as default audiocard, I started a playback in MPC-HC with a movie in it and then I started one of the monitoring tools:

http://postimage.org/image/hrmw3xtw/
sometimes it looked like this
http://postimage.org/image/2q8frfdt0/

When I uninstalled the RME card to check, if its the RME card which causes the spikes, Realtek on the board has been set as default device, but the issues were still present. In between, I tried to disable CPU powersafe features, Turbo Boost, used GPU clock tool and MSI Afterburner ( to set GPU to fixed frequencies ), but nothing helped:
http://postimage.org/image/2q9womhlw/

To sum it up, huge DPC spikes has been notified by the LatencyMon at following drivers/sys files:

usbport.sys (USB controller or any driver which interfaces with USB)
portcls.sys (Audio port interface, so any driver which deals with sound or audio)
scsiport.sys (Anything listed under Storage Controllers or SCSI & RAID controllers in Device Manager)
ataport.sys (Anything listed under IDE/ATA ATAPI controllers in Device Manager)
hdaudiobus.sys (when using realtek audio chip )

Please take a notice, that Im using MPC-HC with madvr video renderer ( curently best of ). Using other renderers like EVR based, I couldnt reproduce these huge scsiport.sys spikes ( 2000-3999 ).
During playback with EVR rendrer, the scsiport.sys events were displayed as the highest in the Latencymon, but the DPC Count was under 500 and ISR under 250 ( execution time ).

EVR does not use DXVA, its passing everything to CPU. EVR-CP with DXVA based decoders offloads this work to the GPU processors. MadVR is different, its basically programmed like a D3D video game renderer - all work is done via GPU shaders, it does not support DXVA at all. MadVR outputs in YV12 color format only and its programmed to achieve the best results - with the costs of GPU load.

To confirm my assumption, that higher GPU load causes the red spikes in LatencyMon a did further testing with the video files and Babyface.I turned GPU-Z on and after testing I came to a conclusion, that there is a relation between GPU load and DPC spikes. EVR and EVR-CP generated approx. 0-14% GPU load. MadVR´s GPU load was between 35-43%.DPC spikes and GPU-Z looked like this:
EVR http://postimage.org/image/2ehiayb6s/
EVR-CP DXVA http://postimage.org/image/2eh1rku90/
madVR http://postimage.org/image/2efurltes/

What I try to say, is that the DPC latency issues rise with the GPU load. Im not skilled in this, but few things come off due to this:
- IRQ issue
- PCIe issue ( check my PCIe configuration - have 3 cards there )
- motherboard issue, chipset issue, bus communication
- GPU issue
-...

I dont believe RME has produced a faulty card with faulty driver. I disabled both audio audiocards ( Internal Realtek and Babyface ) and also ATI HDMI output card on my computer and then run the LatencyMon with madVR in MPC-HC:
madVR - no audio card http://postimage.org/image/ausauv38/

again same bad results on SCSIport.sys and ataport.sys

What I believe is that my system, better said, the way how my drives and PCIe cards are configured ( I have 3xPCIe cards ) behave maybe badly from the beginning. What should I do?-What kind of test do I need to do, to target the issue/cause? It is the motherboard, it is the PCIe OCZ Revodrive, it is the Perc H700 Raid Controller ( disabled - same results - tested ) or maybe the ATI 5670 card? Why does the DPC spikes show their face when RME USB card is turned on on its USB port - usbport.sys/portcls.sys ? Why does realtek getting same issues when this audio chip is on - hdaudiobus.sys?

Thanks for your opinion and please correct me if my questions are not in the right direction.

Pluto

My system

i7 980x
Win7 64bit
Gigabyte GA-X58a UD5 Rev. 2, BIOS FD ( updated a 2weeks ago via @BIOS - issue? ), the chipset driver is actual 9.1.2.1007
OCZ Revodrive (not x2), 50GB with the latest driver (for OS ) - PCIe ( card is basically a RAID controller in Raid0 )
DELL PCIe Perc H700 SAS/SATA Raid Controller with latest driver (Raid 10), using 4x WD RE4-GP 2TB
3x Samsung HD204UI
Pioneer BDR-205
Sapphire PCIe ATI Radeon HD5670 Ultimate - the issue has been tested with Catalyst 10.10e, 11.1, currently 11.1a
Seasonic Gold X-750
Fractal R3
Photos here http://www.facebook.com/photo.php?fbid=471339651826&set=a.471339546826.256461.171179716826

[GoodBytes]

This is not difficult to fix.
 > All you need to do is to re-install your SATA, motherboard, sound card and video card drivers.
 > Make sure you get the latest drivers, and that you uninstall the old one, restart your computer, and THEN you can install the latest downloaded drivers.
 

It seams that you are using Windows XP
Please do not use an outdated OS with modern computer. Windows XP doesn't know or fully supports half the technologies inside your computer.
A modern OS is highly recommended.

If you are using Windows Vista/7 with classic theme, then that is all fine.

May I recommend a far better codec pack, specially designed for Vista, and Win7?
 - Shark007 Terrible name, ugle website, but DAMN good codec pack.
It features:
 - Full thumbnail support
 - Has a nice forum with the developer full participation
 - 32 and 64-bit codecs
 - Doesn't overtire or conflict with other codecs
 - Full support on the first and excellent and light (thank you Zune team!) Windows Media Player 12 32 and 64-bit, as well as Windows Media Player. Make you feel Windows Media Player is VLC
 - Enables Windows Media Player 12 64-bit and can change all file association form the 32-bit player to the 64-bit player to use the 64-bit codecs.

Everyone I recommended it, loves this codec pack.

If you want to try it out Shark007
32-bit: http://shark007.net/win7codecs.html (install first)
64-bit: http://shark007.net/x64components.html (install second)

Be sure to uninstall your old codec pack, DivX (already included in Win7, so you don't need it), RealPlayer, and Nero (comes with codecs that conflicts with others).

Enjoy.

[Plutotype]

This is not difficult to fix.
 > All you need to do is to re-install your SATA, motherboard, sound card and video card drivers.
 > Make sure you get the latest drivers, and that you uninstall the old one, restart your computer, and THEN you can install the latest downloaded drivers.
 

It seams that you are using Windows XP
Please do not use an outdated OS with modern computer. Windows XP doesn't know or fully supports half the technologies inside your computer.
A modern OS is highly recommended.

If you are using Windows Vista/7 with classic theme, then that is all fine.

May I recommend a far better codec pack, specially designed for Vista, and Win7?
 - Shark007 Terrible name, ugle website, but DAMN good codec pack.
It features:
 - Full thumbnail support
 - Has a nice forum with the developer full participation
 - 32 and 64-bit codecs
 - Doesn't overtire or conflict with other codecs
 - Full support on the first and excellent and light (thank you Zune team!) Windows Media Player 12 32 and 64-bit, as well as Windows Media Player. Make you feel Windows Media Player is VLC
 - Enables Windows Media Player 12 64-bit and can change all file association form the 32-bit player to the 64-bit player to use the 64-bit codecs.

Everyone I recommended it, loves this codec pack.

If you want to try it out Shark007
32-bit: http://shark007.net/win7codecs.html (install first)
64-bit: http://shark007.net/x64components.html (install second)

Be sure to uninstall your old codec pack, DivX (already included in Win7, so you don't need it), RealPlayer, and Nero (comes with codecs that conflicts with others).

Enjoy.

Hi,
thanks for the answer, but it looks like you have not read my post entirely.
1. Win7 64-bit
2. All drivers up to date
3. Wont use anything else than madVR and MPC-HC for video playback ( best results )
4. The issue is clear - GPU shader load causes DPC spikes at the ports I have mentioned.
Pluto

[Dark Mantis]

Hi

I am not familiar with your particular hardware(cards) but if it is a latency problem this might help.

The problem is easilly fixed but finding the culprit can be more troublesome. Latency is caused by bad programming generally. There are various rules for how a program should access the core and if it doesn't follow these rules it can result in the core not being released for the next operation and this causes a bottleneck that backs up. The biggest headache from our perspective is that we have to find the rogue program/driver.

The best way to tackle this is to go to Task Manager and then in the Processes section stop all the processes that are not required for Windows to run. If you do this one at a time until you find the one that is causing the trouble. Then depending on what it is as to how you tackle the repair.

If none of that works you would need to go to Device Manager and disable all the devices not needed to run the system. Then you can enable them one at a time and again see which one causes the problem.

Have fun!

[Plutotype]

Dark Mantis,
thanks finally for the reply.
I have communicated here http://www.msfn.org/board/topic/140263-how-to-get-the-cause-of-high-cpu-usage-by-dpc-interrupt/page__view__findpost__p__956194 with a guru/user called Magic Andre letting him analyse ETL files using Windows performance tools kit.
All of them are still online if you look into the thread. Andre was not able to help.

Recently some things went positive regarding this issue. I have not found out the cause for th DPC latencies, but...with updated 0.976 WIN USB drivers from RME ( soundcard manufacturer ), the audio glitches went away also under high DPC latencies ( video playback ) - the WDM driver from RME tied the results with multimedia Realtek on this.
Additionally, my USB drivers from NEC/Renesas had to be updated with recent updates ( NON-VISIBLE ON THE GIGABYTE SITE  ). Since 2.0.30.0 version of the driver (from dec 2010 ) there has been recently a 2.0.34.0 update and also a device firmware update to version 3.0.2.7.0.8.
So the advice is to check all recent drivers/firmware updates directly from the device manufacturer, because sometimes Gigabyte does not care.

Thanks for your input regarding the high DPC latencies by the high GPU shader load - video playback.

Pluto

[Plutotype]

Dark Mantis,
thanks finally for the reply.
I have communicated here http://www.msfn.org/board/topic/140263-how-to-get-the-cause-of-high-cpu-usage-by-dpc-interrupt/page__view__findpost__p__956194 with a guru/user called Magic Andre letting him analyse ETL files using Windows performance tools kit.
All of them are still online if you look into the thread. Andre was not able to help.

Recently some things went positive regarding this issue. I have not found out the cause for th DPC latencies, but...with updated 0.976 WIN USB drivers from RME ( soundcard manufacturer ), the audio glitches went away also under high DPC latencies ( video playback ) - the WDM driver from RME tied the results with multimedia Realtek on this.
Additionally, my USB drivers from NEC/Renesas had to be updated with recent updates ( NON-VISIBLE ON THE GIGABYTE SITE  ). Since 2.0.30.0 version of the driver (from dec 2010 ) there has been recently a 2.0.34.0 update and also a device firmware update to version 3.0.2.7.0.8.
So the advice is to check all recent drivers/firmware updates directly from the device manufacturer, because sometimes Gigabyte does not care.

Thanks for your input regarding the high DPC latencies by the high GPU shader load - video playback.

Pluto

EDIT: Regarding USB ports, as I have mentioned, with RME Babyface updated drivers, the card has smooth playback with videos inspite of high DPC playback in the background. But although I have updated the Renesas/NEC USB 3.0 port/hub to the latest drivers, the issues were still present when the soundcard was attached to this 3.0 USB port. I had to move the cards usb cable the JMicron JMB362 E-sata/USB or the Southbridge ones, which worked flawlessly - no audio dropouts or sound artefacts. So its either problem of RMEs driver to sit on the USB 3.0 port or its the NEC/Renesas driver.

[Dark Mantis]

Yes, these things can be a dilemma. Drivers sometimes fix one issue and then cause another. Until a newer driver is released to try it is hard to say which is at fault.

[Plutotype]

Yes, these things can be a dilemma. Drivers sometimes fix one issue and then cause another. Until a newer driver is released to try it is hard to say which is at fault.

My easy logic tells me, that if the soundcard is working ok with all USB 2.0 ports that I have tried and DOES NOT WORK on USB 3.0 ports, the reason is simple - RME Babyface is USB 2.0 and not USB3.0. So works only on USB2.0. I dont believe NEC/Renesas have some bus bugs in the USB3.0 to USB2.0 bus, but as I have not any other USB 3.0 port to test it, so who knows..

[Dark Mantis]

It shouldn't make any difference as the standard is backwards compatible so regardless of whether it is USB2.0 or USB3.0 it should work.

[Plutotype]

It shouldn't make any difference as the standard is backwards compatible so regardless of whether it is USB2.0 or USB3.0 it should work.

Also RME mentiones this...but their USB driver is very special..basically its the first USB based ASIO soundcard with such low latencies.

USB Made by RME . The Babyface provides RME's unmatched USB pro audio technology for revolutionary ultra-low latencies on Mac and PC, by using two different optimized firmware versions for both operating systems. Like other RME USB 2.0* interfaces, the Babyface provides revolutionary ultra-low latencies even with multiple channels. The Babyface achieves latency values down to 48 samples on Windows and 14 samples on Mac OS X.**

* The Babyface is compatible to USB 3 chipsets.

[Dark Mantis]

Yes I am not familair with the unit itself but that would seem to bear out what I was saying.