Page 2 of 2 FirstFirst 1 2
Results 11 to 11 of 11
  1. #11
    Super Moderator - Coderat Array Nodens PC Specs
    Nodens PC Specs
    Laptop (Model)ASUS GL702VW-GC004T
    MotherboardRampage V Extreme Edition 10
    ProcessorIntel 3930K C2
    Memory (part number)CMZ16GX3M4X2133C11
    Graphics Card #1ASUS STRIX-GTX970-DC2OC-4GD5
    MonitorASUS VG278H
    Storage #13x OCZ Vertex 3 240 (RAID5 - LSI 9361-8i)
    Storage #23x WD2003FYYS (RAID5 - LSI 9361-8i)
    CPU CoolerCorsair H100i + 4x Noctua NF-P12
    CaseCoolermaster HAF-X
    Power SupplySeasonic X-1250
    Keyboard Razer Anansi
    Mouse Logitech G502
    OS Windows 10 Pro x64
    Network RouterLinksys E4200v2 (running TomatoUSB/Toastman firmware)
    Nodens's Avatar
    Join Date
    Mar 2012
    Reputation
    266
    Posts
    4,388

    Quote Originally Posted by Reaptor View Post
    Nodens, until now I've tested DPC/ISR with the latest LatencyMon. It looks similar to your analysis. nvlddmkm.sys and dxgkrnl.sys have the highest peaks (around 400 to 500 microseconds). If I remember correctly I had almost the same result on my old system (Phenom X4 940, W7, GTX470). At this point I only would like get rid of this nvlddmkm.sys spikes and the page faults (Can I rely on this result depending page faults?), which LatencyMon is showing. Did you post your results on any Nvidia-related forum ? I'm sure this is the Nvidia driver.
    Regarding the nvidia driver my guess is that it's also related to the power saving features, specifically the dynamic clock modulation of the 6xx series. I'll get back to you as I need to do some further testing. I haven't posted information on an nvidia forum because I have a more direct way of reporting back to nvidia (I'm a registered developer with them). They are aware of the issue but the comment so far is the immaturity of the Win8 driver.
    Regarding page faults. Those are directly related to the amount of RAM available and the size of the pagefile. A page fault happens when RAM that is mapped in the virtual address space is not currently in physical RAM. In the 4GB laptop I'm using for live experimental music performance, latencymon incurs much of the page faults itself. I suggest using sysinternals process explorer as a replacement of the task manager and monitor, page fault delta and individual process page faults there.

    I have done all of the optimizations you've mentioned and tried different other things. For example I have tried to put my 1394b (my audio interface RME Fireface 800 is connected to this card) PCIe on other slots, where the card didn't work at all.
    This shouldn't have any effect (btw I use a RME Fireface UCX myself..RME hardware is amazing and the drivers are ROCK solid). You certainly know how to choose audio interfaces

    Could you please explain how the new interrupt structure works in W8 ? I've heard all devices are using only one interrupt.
    Well as far as I've found out (by Microsoft information..I'm still researching this), interrupts are passed to a single unique interrupt that calls the DPC routines at intervals. Supposedly audio has a higher priority and does not need to use a DPC but I haven't checked the WDK yet for more specifics. Even so, I assume the driver has to support this unless the kernel has provided for older drivers somehow (The only way that I can think of that would work, is by just putting the audio related DPC routine at the front of the queue but still that's a hack and also it still uses DPC). RME drivers so far do not support it for sure.

    Anyway, I'll reinstall W8 again (ohh, how I love this ...), download WPA xperf and test again - definitely for the last time - I have a lot of work to do.
    Why reinstall? Just install the Kit.

    Is this the right thing for W8 to get WPA xperf ? :
    http://msdn.microsoft.com/en-us/wind.../hh852363.aspx
    Yep. You don't need to install the entire thing. You only need the Performance kit and if you're going to use WPA yourself you should set the _NT_SYMBOL_PATH symbols variable so you can see the functions (useful in debugging, figuring out what's killing performance) . http://support.microsoft.com/kb/311503

    I'm happy that there is somebody else who also needs to have a snappy pseudo-real-time-OS Hopefully we can solve the Nvidia issues.
    I'm dabbling in experimental electronic music (algorithmic sequencing etc) where music and programming collide so yeah real-time is veeery important to me also. We'll get there. If the drivers start using the native Win8 method, spikes like these should be of no importance. I think I'm gonna hit the RME forum and see if there's any progress on that area. The brothers coding the RME drivers are absolute geniuses so if there's a short audio path to exploit I'm sure they'll bend it beyond imagination
    RAMPAGE Windows 8/7 UEFI Installation Guide - Patched OROM for TRIM in RAID - Patched UEFI GOP Updater Tool - ASUS OEM License Restorer
    There are 10 types of people in the world. Those who understand binary and those who don't!

    RealBench Developer.

Page 2 of 2 FirstFirst 1 2

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •