MSI AfterBurner Shuttering When Fan Speed Changes

Discussion in 'MSI AfterBurner Application Development Forum' started by FDisk, Apr 18, 2010.

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

    FDisk Master Guru

    Messages:
    766
    Likes Received:
    0
    GPU:
    ASUS STRIX GTX970 OC 4GB
    First Some info of my system: HD5770 (Reference cooler), i5 750, Windows 7 x64, Catalyst 10.3, 10.4, 10.x (Doesn't matter, same effect), MSI Afterburner 1.6.0 Beta 4 (v1.5.0 has the same effect).

    OK so here is the problem I notices shuttering in all my games after installing MSI AfterBurner. To see this effect just set up fan control profile in the Afterburner options and make it increase automatically as the temperature rises.

    For example:
    40c (fan speed at 35%)
    50c (fan speed at 45%)
    60c (fan speed at 55%)
    70c (fan speed at 65%)

    Now. Every time you hear a fan speed up when it reaches the switch temperature the games shutter.
    I was going nuts with those micro shutters until I noticed that it happens every single time when I hear the fan switch speed. :bang:

    So for testing this theory what I did is just set a fan to a constant speed of 60% and ran some benchmarks and games. And what do you know, I was right!
    The micro shutters are totally gone! :)

    Set the fan profile back and the micro shutters on fan speed change are back.

    Anyone else experienced this problem?
    Can this be fixed?
     
  2. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    15,552
    Likes Received:
    2,946
    Nobody. Never. Read. Anything. Can this be fixed? Hope so. Is it so hard to peek into context help of the option you're about to use to understand what to expect after enabling it?


    Fan speed update period

    Displays and adjusts fan speed update period for user defined software automatic fan control mode.

    Hints:

    - User defined software automatic fan speed control is not absolutely free in terms of performance, some CPU time is required on each fan speed update iteration to read temperature from hardware and update fan speed if necessary. Depending on GPU and sensor model and hardware access protocols, software automatic fan control related performance hit can vary from virtually invisible to noticeable periodic performance hit causing an effect of stuttering in applications intensively using CPU, such as 3D games.
    If you are facing such effect while MSI Afterburner is running, try to temporarily disable software automatic fan speed control to identify it as a problem source by unpressing the <User define> button. Then try to decrease CPU load by increasing fan speed update period.
    Experienced users can also show performance profiler status information in the monitoring window by right clicking it and selecting the <Show status> command. Status information displays CPU usage times for different background processes including fan speed update.

    - You may disable user interface tooltips via <User interface> tab in advanced properties.


    Want it to be improved and eat less CPU time - spam nice AMD driver team and ask to optimize their fan speed control routines. But the chances to get any improvements from AMD side are close to zero.
     
  3. FDisk

    FDisk Master Guru

    Messages:
    766
    Likes Received:
    0
    GPU:
    ASUS STRIX GTX970 OC 4GB
    Thank you the the explanation. Yea I kinda missed it in the hints. Now I understand. Don't be angry.

    For now I just set it up with just two speeds at 35% speed when idle and 72% speed when above 52c. Yea, it's overclocked as hell so it gets hot. Like above 80c hot.

    So as soon as I run some game it always goes above 52c, stuttering (I know I misspelled it before :bang: ) once, but then it's all good and smooth.
    Well, maybe I just helped someone else that didn't read the hints. :grad:
     
  4. Black_ice_Spain

    Black_ice_Spain Ancient Guru

    Messages:
    4,581
    Likes Received:
    14
    GPU:
    5700XT
    you should try T-Function control if possible (i think all newer cards have it possible, at least my 4770 has it, i doubt a 5770 doesnt).

    Its the one which its only a line.

    make it 35% speed under 40ºC always and put it to raise between 40ºC and 100ºC (my option is 60-80 depending of your temps, but my GPU fan its not noisy). That will give around your current fan speed fan speeds without having problems. As far as i know its via hardware.
     

  5. FDisk

    FDisk Master Guru

    Messages:
    766
    Likes Received:
    0
    GPU:
    ASUS STRIX GTX970 OC 4GB
    I tried that once it was still stuttering every time the fan speed changed a little.
    And this way it changes it a lot. So it just made it worse.
    The best solution for this is to set it to constant speed when under load and to low speed when idle. So it wont start changing speed while gaming. That will get you killed. :)
     
  6. sideeffect

    sideeffect Master Guru

    Messages:
    323
    Likes Received:
    34
    GPU:
    Vega 56
    Why do you always have to slag off ATI? Seems every issue with with afterburner is accompanied by an ATI hate speech from you.

    The link to CPU usage doesn't make any sence either he has a quad core I5 and what games use 100% of a quad core?

    I did a quick test using Multi core optimised games
    Crysis 40 -50% CPU Usage
    L4D2 55 - 70% CPU Usage
    GTA4 60 - 75% CPU usage

    So are you saying that the ATI fan speed control routine is so unoptimised that 1 free core of 2.5+ GHZ can't handle it?
     
  7. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    15,552
    Likes Received:
    2,946
    Probably because ATI devel is a complete disaster when it comes to solving some issues and their programmers code worse than the students coming to take programming practice in my company? Like it or dislike it, it is a fact. The amount of time invested into different workarounds, added to Afterburner just to bypass issues caused by ATI's OWN drivers is already comparable to whole Afterburner project development time.
    Probably you just have to try to work yourself with ATI and NVIDIA devrels to feel the difference and to compare level of techsupport provided by both sides?
    For example, wanna know the story of voltage control implementation on the last generation hi-ends on both sides (Cypress and Fermi families)?

    Cypress family:

    We contacted AMD during implementation of voltage control on 5xxx series in a few weeks prior 5xxx announce, because we tried to implement the safest collision free I2C access layer and their own API for accessing I2C bus in ADL SDK for the Catalyst driver was not working at all. All they did was giving some flexible anwers about "probably it is locked in BIOS" or "probably it is not supported for some ASICs" and ended by sending us absolutely useless sample code for accessing DDC (which is pretty useless because VT1165 resides on alternate non-DDC I2C bus on Cypress anyway). And I2C routines are is STILL not working in AMD driver for Cypress. As a result, ANY low-level monitoring tool working with VRMs on Cypress, comes with its own I2C bus driver and have a great risk of causing VRM to enter wrong voltage state due to I2C access collision (try to google and serach for 1.6+V Cypress issues when running GPU-Z+Everest or Everest+Afterburner and so on). All that could be avoided if AMD invested at least a bit devrel resources into polishing the API they provide to developers. Most of issues reported to ATI never gets investigated and fixed. And unfortunately most of developers have the same impressions about it.

    Fermi family:

    NVIDIA themselves contacted all known hardware tool developers prior to launching Fermi and suggested everyone to rely on their own NVIDIA API for voltage control instead of direct access to VRM via I2C to provide the maximum safety to users. I faced small issue with voltage control, becaue there was no way to determine default voltage with the interfaces they provide (and as many Fermi users probably know, the chips have personal fused voltages so the hardware database approach doesn't work well) and contacted NV. Got an immediate reply, NVAPI is already being modifyed to provide this functionality. Most of issues reported to NV by their partners get investigated and fixed almost immediately after reporting.

    It doesn't make any sense only for those who never tried to use any ATI routines and have no chances to feel how slow they are.

    Yes, I'm saying that almost everything that ATI provides in ADL SDK cannot be used in realtime monitoring because it is terribly, drastically, insanely slow for realtime usage. I've already spent enough time to get rid of any ADL SDK calls in monitoring part of Afterburner because of that. Wouldn't like to do the same with fan control part, it will rely on AMD driver only (to provide maximum compatibility) and it will be as "effective" nice AMD can implement it.
    And seriously, guys, you have a good reasons to think why there are so many "TWIMTBP" logos in games. That's not a question of money, that's a question of choosing a side that can provide proper support.

    P.S. In the latest OSD server version I mentioned that new 3D OSD rendering mode gave drastical performance boost on the systems with ineffectively optimized 2D framebuffer access routines in the DirectX driver. Try to guess which driver was that?
     
  8. sideeffect

    sideeffect Master Guru

    Messages:
    323
    Likes Received:
    34
    GPU:
    Vega 56
    I can't comment on the quality of the students but The ATI driver is in good shape at the moment good enough to perform similarly to a Nvidia GPU with 40% higher transistor count.

    Yes I can understand how that would be annoying but ATI have released their own software to overclock, the CCC overclocking is working well, other tools exist like ATT perform well.

    Yes probably but to publicly critisise one company all the time while ignoring problems with the other company isn't really helpful to the end user.

    'There is a problem with the ATI fan speed control routine which needs to be fixed by ATI' is informative and the kind of answer you would expect from a developer. Insulting the company I personally don't find helpful at all and just sounds like someone with a grudge.

    No it makes no sence if you look at the evidence. The fact that the CPU load isn't even at 80% maximum while playing games with a fan profile enabled on my quad core shows that the CPU is probably not the cause of the stuttering in this case. It would be 100% load before you would notice the problem.

    I'm not arguing that it might be less efficient than the Nvidia routine and might cause issues on single or dual core systems just that in regards to the OP's issue it is unlikely to be the cause.

    Independant games run just fine on ATI cards. TWIMTP games always have some issues so perhaps the 'support' from Nvidia isn't so hot.

    Yes I read it. You forgot to mention that the Nvidia 2D driver is hardly in any better shape and was also critised for being slow on new hardware in Windows 7. ATI responded publicly stating they are improving that and have already in newer beta drivers Nvidia havn't mentioned it that i noticed.

    The impact in performance is negligable as well between the old 2D and new 3D OSD implementation. So much so that in games I tested it wasn't noticable.

    Anyway its going off topic and I don't even have the issue. I just got annoyed that you were critising someone for daring to ask for help for not reading and then pointing at an explanation that is very unlikely to be causing his problem in the first place. Even if it was it wouldn't be obvious when he has a quad core. Then making another ATI driver dig in such quick succession to ones in other posts.

    No I don't have the problem. Did you try using ATI tray tools to control the fan? Also you could modify the card BIOS with Radeon BIOS Editor to change the fan speeds without the need for 3rd party software.
     
  9. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    15,552
    Likes Received:
    2,946
    It makes no sence if you discuss the things you don't understand at all. Does the "call latency" tell you anything? It is enough to a waste 50-100ms per driver call to waste CPU time comparable with frame time. It is enough to have such ineffective just call once per second to see a stutter in the game giving stable framerate. You will NEVER see such things on CPU load graphs. Start thinking please.

    Really? Try to search this forum for stuttering issues with ATT if ATI driver is used just for software clock reading for example (this thing implemented at low-level properly takes less than a microsecond, ATI managed to provide interface doing the same 30x (!!!!!!) times slower). And if you're so stubborn, try to disable low-level monitoring part in Afterburner by adding LowLevelMonitoring=0 line to the [Settings] section of hardware profile file and feel the real power of stuttering caused by ATI driver's realtime thermal, clock and fan speed monitoring. These are the interfaces, initially designed by them for realtime usage. Beyond of my understanding how they managed to implement it so ineffectively.
    And please no need to tell me that other tools don't face ATI issues, we're fighting endless ATI related problems in cooperation with many developers, including Ray. So you're either deeply misunderstanding things or simply lying.

    One more such statement and you'll get permanent ban for spreading lie because of fanATIsm. I'm getting more that enough insults from both ATI and NVIDIA fanatics when their idol's balls are kicked. Got more tahn enough crap from NVIDIA fanboys when I declared than overclocking is broken in 196.21 WHQL, got more than enought "that's not our perfect NVIDIA drivers, that's third party crap like RivaTuner" from NVIDIA fanboys after declaring that fan control was broken in 196.75. But due to some odd reason NVIDIA prefer to intensively _monitor_ such issues and react immediately (overclocking issue was fixed in one day after reporting, driver with fan issue was pulled down almost immediately after report too) and ATI prefer to sleep ingoring the problems for ages. That's nobody's problem but their own.

    So search for another place please for holy wars, this subforum is definitively not a best place for "ATI/NVIDIA rulez just because I own ATI/NVIDIA hardware" styled fanboyism. That's the first and the last warning for you and everyone.
     
    Last edited: Apr 19, 2010
  10. FDisk

    FDisk Master Guru

    Messages:
    766
    Likes Received:
    0
    GPU:
    ASUS STRIX GTX970 OC 4GB
    I didn't mean to start a war.
     

  11. cowie

    cowie Ancient Guru

    Messages:
    13,273
    Likes Received:
    345
    GPU:
    GTX
    some of that is top ten guru all time for me man.
     
  12. TFL Replica

    TFL Replica Master Guru

    Messages:
    385
    Likes Received:
    2
    GPU:
    RTX 3060 Ti
    That was awesome. Unwinder has deep knowledge from working with both ATI and NVIDIA drivers, you'd have to be extremely dumb or a fanboy (or both in this case) to question his remarks.
     
  13. sideeffect

    sideeffect Master Guru

    Messages:
    323
    Likes Received:
    34
    GPU:
    Vega 56
    I was nothing but polite and only questioned your rational. In no way did I call you a fanboy. I only implied you were rude and patronising. Is that enough for a ban?

    Thanks thats helpful to know shame you had to get so angry about it just to explain. At least I know now to let the BIOS handle the fan speeds rather than software.
     
  14. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    15,552
    Likes Received:
    2,946
    Don't play the fool. It is enough for a ban to make ATIvsNV holy wars here and spread the lie about "publicly critisise one company all the time while ignoring problems with the other company" just to protect your idol. And both NVIDIA and ATI fanboys will treated this way. End of story, thread closed.
     
Thread Status:
Not open for further replies.

Share This Page