1. We're currently conducting systematic tests with the release version of macOS 11 (Big Sur) in order to offer full compatibility of our products as soon as possible. You can find our latest compatibility status - here


    Please note that although announced at the same time as macOS Big Sur, Apple Silicon processors are an independent release with separate compatibility concerns for NI products. Compatibility with macOS Big Sur will not yet ensure compatibility with Apple Silicon processors. At the moment, Native Instruments products are not supported on computers with Apple Silicon processors. Read more - here

    Dismiss Notice

CPU / LATENCY TSP2 problem NI say something

Discussion in 'TRAKTOR PRO / TRAKTOR SCRATCH PRO' started by leeway, Apr 4, 2011.

Thread Status:
Not open for further replies.
  1. luisfmoreira

    luisfmoreira Forum Member

    Messages:
    53
    Guys,
    I just found out something:

    On my mac, when I scratch on the DVS and get the spike thing, my processor is low on the activity monitor, but the writes on the hard drive spike ! Could this be the cause of the drop outs and the other issues? BTW anyone has any ideia of what is traktor writing when skratching?

    MBP unibody 2,66 C2D 4 GB ram 5400 RPM 70 GB free
    A8 DJ latest drivers, 2 cdj's + 2 technics 1200

    Luís
     
  2. mastermc

    mastermc NI Product Owner

    Messages:
    3,170
    what do you mean "writes" ?
     
  3. PhilL

    PhilL Moderator Moderator

    Messages:
    7,084
    Disk I/O could mean that part of your track buffer has been pushed out to the swapfile, meaning Ram reads will go look to see if the memory has been flushed to the pagefile and then written to disk. Without debugging tools it will be very difficult to know for sure. This could be a big big problem in terms of performance, but like I said it will be difficult to know for sure with debugging tools on windows I could not tell you if you could even trace this on a Mac.

    Phil
     
  4. PhilL

    PhilL Moderator Moderator

    Messages:
    7,084
    I've said it multiple times and I'll say it again!... I know many people aren't listening but here goes...

    YOU CANNOT COMPARE TSP1 LATENCY SETTINGS TO TSP2 LATENCY SETTINGS.
    They use different mechanisms and measurement points in the audio chain to obtain their values. THE NUMBERS SHOWN ARE NOT DIRECTLY COMPARABLE! ANY COMPARISON IS PURELY ANECDOTAL IN ITS NATURE.

    On the subject of observable and perceived latency your observations are interesting, I know plenty of scratch DJ's a couple of headliners too who will disagree with you here. Bottom line is this 10Msec and less will in most cases not be felt, its simply too low for brain to eye co-ordination to process it. Its like measuring the how fast a Great Dane dog walks versus a 3 Legged lizard and you you use steps per second to measure their speed! I understand you want what you had before in terms of performance. Bottom line is you can't have it, at least if you insist on using latency reported by Traktor as your measurement. Try this... Make an honest attempt to tune your system by FEEL not by an arbitrary number doled out by the computer. I.E. ignore the number and adjust it purely based on how the system feels. if it is unsatisfactory lets start tuning from there.

    I can probably dig up the study from a college back east form about 2 years ago that studied human computer latency interactions and concluded that less than 5% of the population as a whole, could reliably detect and compensate for control latencies lower than 13.5 MSecs. Sorry to say we might think we can do it but most of us cannot.

    There are probably a couple of performance issues in play here.
    1/ In TSP1 we used to have a reported setting of 2.5 or 3.5 Msec latency and godammit we want those numbers now too in TSP2. The plain fact of the matter is we measure latency in the system differently now and the numbers you get today from TSP2 CANNOT be compared to those from TSP1. In my test experience I have found that the 4 Msec latency I had before in TSP1 gives nearly identical performance to in TSP2 where latency is 8.0 -8.5 Msecs. Do not think the scale is 2X different, it isn't! The two latency scales do not run in parallel to each other nor are they flat in relation to one another.

    2/ The second issue is one of performance and performance perceptions. I have come to understand there are two actual issues here. The first is that some people are concerned / freaking out because they are observing CPU spikes in the task mangler. They are seeing spikes, but no other ill-effects. Moreover some of these spikes are highly regular in nature and the conclusion that they are caused by Traktor cannot be proven or entirely disproven. Two people I had asked to do a load test of their system without Traktor running did not report back and have not answered Email to fess-up leading me to conclude that when they put a load on their systems without Traktor running, that the same regular spikes were visible once again wihich implies that Traktor is at worst a contributor (load) to the spike issue and the polling activity on the system is at the heart of the cause.

    The second and more widely observed issue is that the load on the system is causes pops and drops. I have one verified situation where very slow motions back and forth on a track timecode control surface (TT CD or Jog-wheel) can produce extremely high loads in Traktors Audio engine and causes high CPU loads (This phenomena is not new, we have seen this issue off and on since TSP1 released. That is one case I need people to test.

    A second and increasingly obvious one is the issue of assigning two or more decks to one timecode source. Sadly I liken this at the moment to old tale of "Doctor, Doctor, My arm hurts when I do this". Doctor- "Well don't do that" It would seem there is a problem here that needs to be looked into and tweaked or fixed. I'll report the issue later today and well see where things go from there.
     
  5. PhilL

    PhilL Moderator Moderator

    Messages:
    7,084

    We have not seen your thread and anyway if you can post this message you could post your solution, Try it, you'll like the result. ;)
     
  6. mastermc

    mastermc NI Product Owner

    Messages:
    3,170
    Actually Phil you are right people just seem to be so loud about having a perfect latency. What i am concerned is about performance more than latency
     
  7. djbluephoenix

    djbluephoenix New Member

    Messages:
    9
    ok so here i go with this latency issue.

    so first i went to the knowledge base to see if there was anything there that would help me and there was.

    here is the link:
    http://www.native-instruments.com/knowledge/questions/343/Windows+XP+Tuning+Tips+for+audio+processing

    i did everything mentioned in the above knowledge base even turned the power saving options off in the usb. and turned the CPU recourses to background instead of programs.

    what i did so i can control the drivers etc. i wiped my HDD and reinstalled windows XP. after the installation was done i ONLY installed video drivers and (wireless)network drivers (usb drivers were pre-installed). after that i went to the "device drivers" section and dissabled everything with a question mark(because i didnt install the drivers yet so i could see which ones werent system drivers)

    i ran RATTV3 and it was showing that "NDIS.sys" was above the 1.0 latency range and the USB was going crazy but thats because i had my audio 10 connected.

    after searching on google it appeared that NDIS.sys is most active if a network is going so i disabled my wireless network card.

    i ran RATTV3 and it was clean across the board except for the usb drivers(because traktor audio 10 was connected)

    now before i did this HDD wipe, all my drivers were installed and i did disable everything but because the drivers were installed even if the devices were disabled the system recognized the devices.

    one little latency blip and the program drops-out.

    so you have to do a complete reinstall of the OS and not install any drivers so you don't give the system a chance to even recognize the devices...

    i did tests after the reinstall with RATTV3 and it doesn't read the devices in the log at all..

    i did some more tests(i'm on the laptop right now i use for dj'ing) and you can enable and disable the network card with no issues.. but you'll have to disable your network while using traktor 2.

    of course i said windows XP my computer is 3 or 4 years old and i've noticed even macs of this age group have problems..

    its because traktor is using ALL of the latency of the CPU and if one little thing is running (even disabled) it drops-out and gives problems.

    some older CPU's dont have a lot of crazy latency. it could be due to the slow speed of the USB plugs also in turn with the low latency of CPU's.. of course if its a bug then it can be fixed but its time for a new laptop anyway.

    the new macbook pro's USB bus speed is 10gb/sec or higher so it shouldn't give issues at all with the newer intel7 cpu's.
     
  8. chill

    chill NI Product Owner

    Messages:
    2,708
    @phil
    hmmm?
    so you think this is a tale?

    why?
    try it.

    as long as i use left deck for just deck a and right deck just for deck b, everything is fine even at 128.
    as soon as i try to control a+b via left deck i get cpu spikes and therefore pops and clicks.

    cheers
    chill
     
  9. mastermc

    mastermc NI Product Owner

    Messages:
    3,170
    I have windows vista, while your method looks interesting, its way too time consuming to reinstall everything only in hopes that it fixes things in the short run.
     
  10. Mark34

    Mark34 NI Product Owner

    Messages:
    58
    @Phill i will report what is happening on my system when cueing using TC. I am not a scratch dj so super latency is not an issue and I understand that you cannot compare latency settings. Whilst i do not get many of the sound clicks and pops i do see Trators monitor redline quite a lot. what i do not see is any increased load on the cpu it currently runs at between 20 and 30 percent with memory around the same. This does not indicate a problem with my hardware although i could be wrong. See sig for laptop I also use audio 4 dj and 8dj I do not own a new audio interface but do wonder if this is where the issue lies. What i have found is that the new audio drivers are far more stable and are much better than the previous ones. To note on the laptop i have disabled
    WIFI
    Network Card
    DVD drive
    Webcam
    Multicard reader
    ACPI
    Modem
    windows search and indexing
    There is no anti virus and windows is configured for best performance therefore there is no aero or any of the fancy windows 7 features running it is pretty much a bare bones install
    I cannot run at below 512 samples with out constant redlining but to note there is no spike on cpu it is flat and steady and not overloaded
    I cannot see this to be anything other than a software issue but am willing to be proved wrong
    Just a note TSP1 runs fine at 1.5 ms using old measurements
    Hope this helps
    Mark
     
    Last edited: Apr 16, 2011
  11. PhilL

    PhilL Moderator Moderator

    Messages:
    7,084
    @djbluephoenix, There is more to it than that but you are definitely on the right track.

    In terms of bus speed, the Mac bus speed is 480 MBits per second. The thunderbolt per may say 10Gbits per sec but its actual throughput will be below that. What matters to us though is the max speed of the slowest link, that will be the USB Port. The issue here is not so much how many bits per second can be stuffed into the pipe. What is critically important for any realtime media system is the HIGHLY REGULAR AND SUFFICIENT delivery of bits into the pipe. Consider this, if you have a consumer application that is sucking down data at the rate of 100 Mbits per second and your pipe can support 500Mbits per second BUT you only get 400 Mbits delivered every 3 seconds then the pool at the end of pipe that stores the bits while they are waiting to be processed will dry up and there will be nothing lift in the pool to be processed.

    What also complicates this is that in a sense there are multiple pumps in our system and only one power receptacle to plug the pumps into. When one pump is running the others cannot, so its a matter of making sure all the pumps are free and clear and can push the bits through as quickly as possible. If you have a blockage or a clogged filter that limits the number of bits per second that flow AND an absolute requirement to move a certain number of bits before yielding and letting another pump run, then you are going have a dry up. What we have in our case is sufficient capacity but we get hamstrung by ir-regular delivery. The root causes for ir-regular delivery are as numerous as stars in the sky but in our sky there are several REALLY prominent stars. In a real sense you should consider the ACPI and WMI Polling to be like a low pressure low volume pump that demands to be plugged in every 5 seconds and it must process bits to go to many different subsystems before it can switch off. That takes a long time and in the meantime the bit pool you really care about has gone dry.
     
  12. luisfmoreira

    luisfmoreira Forum Member

    Messages:
    53
    By Writes I mean hard drive writes, you can measure it in OSX activity monitor. Can someone with latency\cracks\pops\overweight issues verify this please? It could be a start!

    Thanks
     
  13. PhilL

    PhilL Moderator Moderator

    Messages:
    7,084
    Hence my my comment... so for the moment don't move your arm like that. It's certainly something what needs fixing but in the meantime, hold still.

    Phil
     
  14. PhilL

    PhilL Moderator Moderator

    Messages:
    7,084
    Coincidental observations in activity monitor don't guarantee that it is disk writes. Delayed writes, Caching, and several other conditions can skew or delay writes for 30 seconds or more. You may be right in your analysis but may also be way off as well.

    Phil
     
  15. Stevie

    Stevie Forum Member

    Messages:
    27
    Wow so many replies in a few days....I only read the topic start.

    In my experience , the latency problems have NOTHING to do with traktor, and I get the same thing with the cpu monitor , but if you look at the cpu with process monitor/task manager, you will also notice that this is just a false reading and doesn't mean anything.
    windows 7--latency problems due to the audio8dj drivers, to be honest I love windows 7 but my soundcard is the reason I still use XP :(.

    so windows7 and the soundcard are the suspects for me, but I notice no difference with different software.

    Just download sysinternals process explorer and look at the DPC"S you are getting, if you see a lot of red...you have the same problem . So if you use win7 I would suggest to go back to XP because that works MUCH better. I mean you wouldn't believe it I have a blazing fast computer and everything works like a dream but win7+audio8 =FAIL, couldn't even get close to low latency despite my specs.

    By the way offcourse I reported this problem , I have shown screenshots and everything but they couldn't reproduce it so appearanlty it's a problem with my system and it will never be fixed. I guess it is due to the driver in combination with my hardware+win7.....
    AAAHHH frustration boils up again I have to stop writing lol
     
  16. mastermc

    mastermc NI Product Owner

    Messages:
    3,170

    Not every machine has drivers for windows xp
     
  17. chill

    chill NI Product Owner

    Messages:
    2,708
    @phill
    ahhh ok.
    i misunderstood first.

    imho this is a bug that needs to be fixed.

    i often have to use due to space reasons just one cd player,
    setting it just to 2 decks a+b has improved it but it is still there.

    and imho it is easy to understand, when you scratch and have assinged to more than one deck, traktor regonizes it as you would scratch all assigned decks at the same time, due to this missinterpretation cpu load is exploding.

    they have to programm it in that way, that of course even if you have set several decks to one input just one is in scratch mode and the others are traded as they are, as internal decks.

    would be neak if you could explain this to ni in detail.
    i did this very precise in abug report, but the guy was not willing to trust me.

    i have already tried to split the output of my cd player and set it to input 1/2 and to 3/4 at the same time, logically this has the same impact, because when you are doing scratch movements, traktor is again interpreting it wrong.

    if i am working with only 2 decks and 2 cd players, each for one side, there is no problem at all, because i do not scratch at both decks at a time of course, doing so, which i have also tried leads to the same result.;)

    cheers
    chill
     
  18. Stevie

    Stevie Forum Member

    Messages:
    27
    So? The topic starter has win7+audiodj soundcards so I'm trying to help here by sharing my own experience , if you don't have anything productive to say please just shhhh...
     
  19. luisfmoreira

    luisfmoreira Forum Member

    Messages:
    53
    Phil the spikes coincide w heavy scratching ;)
     
  20. Mike-B

    Mike-B Forum Member

    Messages:
    25
    @phill

    I too am not that concerned about latency, though your assertion that they cannot be compared does nothing to explain the reason why.

    I am concerned with the pops and spikes in CPU that occur when I am scratching/cueing. No coincidence there. Though thankfully it seems someone else at NI has acknowledged that it is a bug - so I'm happy to wait.

    So really this thread for me is just a list of people that are experiencing the issue. As long as people have submitted bug reports directly, that is the main thing.

    What I do find concerning is the tone of your posts. I feel you are being overly dismissive and patronising... even combative. Acknowledging an issue on one the hand, then hinting that some people may be wrong or mistaken on the other is hardly helpful. I appreciate you may feel under pressure but nonetheless this is disappointing considering there are paying customers who have real issues.
     
Thread Status:
Not open for further replies.