Close
Page 6 of 7 FirstFirst ... 4 5 6 7 LastLast
Showing results 51 to 60 of 68
  1. #51
    Multiboxologist MiRai's Avatar
    Join Date
    Apr 2009
    Location
    Winter Is Coming
    Posts
    6815

    Default

    Quote Originally Posted by Sam DeathWalker View Post
    Humm so 15 clients don't even take up 1G of video card ram ...


    How do you say you render 15 clients per card, there seems only one screen, are they tiled on the 2nd screen?

    If they are not tiled they are not being rendered (although there are 15 clients per card) to my way of thinking. Or if they are all in the background not visible they are also not being rendered. If that is the case your fast swaping times is just because the 580 can redraw a new screen faster then your fps. The fact you have 10fps in background would tend to indicate that you are not gaining anything by tiling in full resolution. Your card is fast enough to render a full screen in less then 1/60th of a second. Which I think should give you instant swaping no matter what.
    I don't know where this is coming from. Whether one screen is on top of the other 14 or not, they're all
    being rendered all the time as you can see in my screen shots below. Don't mind the increase in VRAM, I
    accidentally moved the camera with my mouse repeater on and I gained 300MB, at least that's what I'm
    blaming it on. The main screen is in focus so that it's at 60FPS and all windows are 1920x1080.







    Quote Originally Posted by Sam DeathWalker View Post
    I wonder what would be the result if you were in different zones, does the card use so little Video Ram because its able to share data between the 15 clients?
    No idea and that isn't on my list to test at the moment; although, such a test might give a clue as to
    whether or not textures are shared.

    Quote Originally Posted by Sam DeathWalker View Post
    How does the ram go up from when you load 1 to many is the first load a lot lot bigger then 950/15?
    I don't know but after going through Orgrimmar I think my ceiling is ~14GB since I can't seem to go any
    higher than that. I didn't look at my Resource Monitor see if Windows was holding any memory hostage in
    Standby or not. Here's a screen shot after I did a single lap in Orgrimmar before I recorded the video below.



    Quote Originally Posted by Sam DeathWalker View Post
    Or does each client just take 60MB of ram?
    From my Orgrimmar screen shot...

    1593MB / 20 = ~80MB

    However, the main is obviously taking up more than any slave so I would say it's closer to 65MB per slave
    and like 300MB for the main client.

    Quote Originally Posted by Sam DeathWalker View Post
    Do you think you can 15 box on your system (with just one 3G card) with no trouble in populated areas?
    15?! I'll give you 2 laps in Orgrimmar with 20! This was at about noon (12PM) on a Saturday -- Obviously
    not peak time.




  2. #52

    Default

    I don't belive that the gpu actually renders anything that is not visible on the display.

    Thats not to say it dosn't load the textures or set aside the memory for all clients, just that the gpu dosnt do the calculations, even if the data is available if its not going to the display. I am sure it loads the textures for all clients all the time though.

    In the first screen yes the GPU is rendering all 15 clients at the full resolution, but at 10fps.

    In the 2nd screen the GPU is only rendering 1 client but becuase your fps and effects are higher (are they?) more GPU load is being used. And in the 2nd screen more of the world is showing. Although the 2nd screen is also shown in the first I am fairly sure its at 10fps in the first screen and 60fps in the 2nd.

    My point is that in the 2nd screen with only one client showing the other 14 clients put ZERO load on the GPU.



    Well that video card is something for sure. Excellent information keep up the good work.

    300Mb for first client with most effects and 65M for addional lower alts is excellent infos.

    Maybe 2G cards are not needed, it seems that there is a lot less video information needed by the card then I had thought although the fact that there are lots of 4K read and writes would tend to support that very little video information is being used. They have made wow to be played on some really poor computers.


    Sure wish we could find out if it uses the vram that not "in use" in the gpuz display as cache.

    These people clearly are not stupid obviously if something can be done to speed up game play it probably has.

    http://developer.amd.com/documentati...ure_Topics.pdf


    That is very bad for stable performance and MultiGPU (SLI/Crossfire) scaling but it was a manageable
    solution at the time. Streaming allowed us to stay within the 32 bit limits with runtime
    data requirement sometimes exceeding the limits. On 64 bit and enough main memory
    or when using half resolution textures the texture streaming is not necessary and performance is more stable.


    Thus the conclusion is that the more Vram (and system ram) you have the better even if gpuz shows only some being used I am sure the rest is being used for cache purposes. Might have something to do with "lod" from the above artical.


    The LOD Bias affects MIP-Mapping in determinating the MIP-Level.

    MIP-Maps are collections of one and the same textures in different resolutions. To allow perspective effect and avoid underfiltering, objects in the back of a scene need to be displayed smaller and the textures in use need to have a lower resolution. The MIP-Map Detail Level globally determines if MIP-Mapping happens with the best possible resolution (and therefore best imagequality) or with best performance in mind through all MIP-Levels.

    Now LOD Bias means global displacement of the MIP-Level determination. With this value set below zero, textures of higher resolutions will more likely be applied, while with a LOD Bias greater than Zero MIP-Mapping will "prefer" textures of lower resolutions.



    gxTextureCacheSize is a CVar that affects the D3D9 API; if you're already using DirectX 10/11 mode, this CVar will not have any affect. This variable decides how much memory - in megabytes - to use for caching textures. The default value is '0', meaning the game will handle this on its own. Some players have found better performance - particularly when turning - by manually selecting a value for this. As a general rule of thumb, ~70-75% of your VRAM is best. For example, if you have a 1.5 GB GTX 480, a good value to try would be 1152 MB. The syntax would look like this:

    SET gxTextureCacheSize "1152"

    For quick reference, these are 75% values. Feel free to use more or less as depending on specific configuration settings.

    128 MB: 96
    256 MB: 192
    384 MB: 288
    512 MB: 384
    768 MB: 576
    896 MB: 672
    1 GB: 768
    1.125 GB: 864
    1.25 GB: 960
    1.5 GB: 1152
    2 GB: 1536


    Ok that answers the shared question also. 75 percent of the video ram is for texture cache, which no doubt is available to all clients.

    They want 25 percent free assuming only one client so with 10 or 20 box maybe less texture cache should be set aside according to a 80M per client rough estimate.
    Last edited by Sam DeathWalker : 08-13-2011 at 07:39 PM

    28 BoXXoR RoXXoR Website
    28 Box SOLO Nalak 4m26s! Ilevel 522! GM 970 Member Guild! Multiboxing Since Mid 2001!

  3. #53
    Multiboxologist MiRai's Avatar
    Join Date
    Apr 2009
    Location
    Winter Is Coming
    Posts
    6815

    Default

    Quote Originally Posted by Sam DeathWalker View Post
    I don't belive that the gpu actually renders anything that is not visible on the display.

    In the first screen yes the GPU is rendering all 15 clients at the full resolution, but at 10fps.

    In the 2nd screen the GPU is only rendering 1 client but becuase your fps and effects are higher (are they?) more GPU load is being used. And in the 2nd screen more of the world is showing. Although the 2nd screen is also shown in the first I am fairly sure its at 10fps in the first screen and 60fps in the 2nd.

    My point is that in the 2nd screen with only one client showing the other 14 clients put ZERO load on the GPU.
    Before I take a screen shot, I click on region 1 to make sure it's in focus and getting full attention. I do this
    every time (at least I try to remember to do this every time) because it gives a more accurate load reading as
    if someone was actually on the first client. I mentioned this before the screen shots above:

    Quote Originally Posted by Fenril View Post
    The main screen is in focus so that it's at 60FPS and all windows are 1920x1080.
    GPU-Z has a handy feature called "Always on top" so that it can stay on top of windows even when it's out of
    focus (like the Windows Task Manager).

    But for you Sam... I'm going to go the extra mile and do another test. I should have the results later today.

    Quote Originally Posted by Sam DeathWalker View Post
    300Mb for first client with most effects and 65M for addional lower alts is excellent infos.

    Maybe 2G cards are not needed, it seems that there is a lot less video information needed by the card then I had thought although the fact that there are lots of 4K read and writes would tend to support that very little video information is being used. They have made wow to be played on some really poor computers.
    Remember, this is all in DX9. If someone wants all the eye candy of DX11 with 5 or 10 clients they'll need over
    1.5GB and probably over 2GB if they really had some settings turned up.
    Last edited by MiRai : 08-13-2011 at 06:41 PM Reason: Added Last Quote/Reply

  4. #54
    Multiboxologist MiRai's Avatar
    Join Date
    Apr 2009
    Location
    Winter Is Coming
    Posts
    6815

    Default

    Here's a big-mother-of-a-picture to help show you what I'm talking about and here's what is going on in it...

    Top Monitor:
    15 tiled game windows with Client #15 (upper left window) at foreground/background FPS of 60/60. This
    ensures that even when it's in the background it's acting like it's in focus. Client #15's video settings are also
    identical to Client #1's video settings. The GPU-Z that is overlaying the 15 tiled windows is reading the GPU
    that is rendering all of those 15 clients.

    Bottom Monitor:
    14 tiled game windows that are covered up by the main window itself (Client #1). Client #1 in this particular
    screen shot is in focus so it is also getting 60 FPS just as Client #15 is. The GPU-Z that is overlaying this
    main window is reading the GPU that is rendering the main window and the 14 tiled windows behind it.

    Miscellaneous Information

    • All windows, except for 1 and 15, are capped at 10 FPS
    • Each monitor is being rendered by its own GPU
    • Each GPU is set to render 15 clients each (if this wasn't obvious from the above text)
    • Rendering in Direct X 9
    • All windows at 1920x1080
    • Windows Aero is disabled









    Now, you can clearly see that both GPU-Z's are practically identical regardless of windows being in
    view or not.
    You cannot honestly tell me that Client #1 just so happens to magically match the 15 tiled
    game windows above it. I believe this proves my point that even when a game window is not in sight, it is
    still being rendered.

    Note To Everyone And Anyone: If you happen to see a flaw in what I'm doing here please let me know or
    point it out. I am not an engineer or a someone who codes how games work so I only believe that what I'm
    showing here is indeed correct.


  5. #55

    Default

    I am coming to the conclusion that more Vram is way way better then a SSD, or even more system ram.

    Wonder if there is a 3G card thats like a 460 or 560 or something less then the 580.


    Ok as to the above. 1 and 15 are 60fps others are 10fps

    In first shot you have 2X60 and 13X10 or 120 and 130, meaning that it takes about as much gpu power to render 1 and 15 as it does to render 2-14.

    Still the total is 270 fps for all 15 clients. (can isboxer tile at different fps, I find that a bit odd).

    In the 2nd you have 1X60 for a total of 60fps

    So the 1st requires 4.5 times the gpu power (time) of the 2nd.

    We see a 1percent difference in gpu power. (18 vs. 19).


    1X = .18Y
    4.5X = .19Y

    Solve for X and Y

    3.5X = .01Y

    X = .01/3.5 Y

    X = .0028 Y

    Y = 357ish times X.

    So if your gpu is 360 or so times what you need to render 60fps (i.e. it can render 21,000 or total fps on all clients /running at 100percent gpu load/, which might be possible with that simple screen) then the 1 percent difference is explained.

    Also the gpuz thing being in only 1 percent could cut that in 1/2 depending on where exactly in the percent reading it is (i.e. could be 1.01 to 1.99 and still show 1 percent, big difference).


    The point of all this is that your 580 is so powerfull as compared to the test you are using that even the smallest difference is a lot.



    Anyway the bottem line for the reader, based on your excellent tests, and the infos I found is that money should be spent in this order:


    VRAM
    System Ram
    GPU (this depends on how high resolution and how high effects you want mostly, if you dont care then spend more on cpu)
    CPU / MB (MB is basically a reliability issue)
    SSD
    Last edited by Sam DeathWalker : 08-13-2011 at 08:20 PM

    28 BoXXoR RoXXoR Website
    28 Box SOLO Nalak 4m26s! Ilevel 522! GM 970 Member Guild! Multiboxing Since Mid 2001!

  6. #56
    Multiboxologist MiRai's Avatar
    Join Date
    Apr 2009
    Location
    Winter Is Coming
    Posts
    6815

    Default

    Quote Originally Posted by Sam DeathWalker View Post
    Ok as to the above. 1 and 15 are 60fps others are 10fps

    In first shot you have 2X60 and 13X10 or 120 and 130, meaning that it takes about as much gpu power to render 1 and 15 as it does to render 2-14.

    Still the total is 270 fps for all 15 clients. (can isboxer tile at different fps, I find that a bit odd).

    In the 2nd you have 1X60 for a total of 60fps

    So the 1st requires 4.5 times the gpu power (time) of the 2nd.

    We see a 1percent difference in gpu power. (18 vs. 19).


    1X = .18Y
    4.5X = .19Y

    Solve for X and Y

    3.5X = .01Y

    X = .01/3.5 Y

    X = .0028 Y

    Y = 357ish times X.

    So if your gpu is 400 or so times what you need to render 60fps (i.e. it can render 24,000 total fps on all clients /running at 100percent gpu load/, which might be possible with that simple screen) then the 1 percent difference is explained.
    Dammit Sam, I have no idea what all of this means. You mention 2 screen shots but, there is only 1. It's a
    double monitor screen shot and all I did was press the Print Screen button and then paste it into MSPaint --
    No hacking, chopping, or trickery was done to the above screen shot. I don't know what all the math is
    about but, my above post was to show you why I believe that clients which are not being seen by the human
    eye are still being rendered in the background by the GPU... that is all.

    As for ISBoxer tiling at different FPS, all I did was change the background FPS for that particular slot in the
    Character Set itself. You could make all 30 windows have unique foreground and background FPS if you'd
    like, there is no funny stuff going on here.

  7. #57

    Default

    No I'm not say there is any tricks at all, I'm just saying that the 1 percent difference that you show: 19 percent of full gpu load with 15 clients showing on one monitor vs. 18 percent with 1 client showing on the other monitor proves my point, even though you feel that because the are about the same it proves your point.

    If your gpu can put out 500fps total for all clients then that 1 percent is 5fps which would prove your point, as they are "about the same".

    If your gpu can put out 19,000 fps total for all clients then that 1 percent is 190fps which would prove my point. As the game is putting out 250 total fps with one gpu (60 plus 60 plus 13 X 10) showing 15 tiled clients vs. 60fps with the one client showing. Which is a difference of 190fps


    I think that on the test you used the fairly static and repeating and common to all clients log in screen that 20,000 fps over all clients for a 580 running at full load is not a stretch.


    I don't belive that they would have the gpu render screens that are not seen, considering they dont waste time rendering objects that can't be seen behind other objects.
    Last edited by Sam DeathWalker : 08-13-2011 at 10:02 PM

    28 BoXXoR RoXXoR Website
    28 Box SOLO Nalak 4m26s! Ilevel 522! GM 970 Member Guild! Multiboxing Since Mid 2001!

  8. #58
    Multiboxologist MiRai's Avatar
    Join Date
    Apr 2009
    Location
    Winter Is Coming
    Posts
    6815

    Default

    Quote Originally Posted by Sam DeathWalker View Post
    No I'm not say there is any tricks at all, I'm just saying that the 1 percent difference that you show: 19 percent of full gpu load with 15 clients showing on one monitor vs. 18 percent with 1 client showing on the other monitor proves my point, even though you feel that because the are about the same it proves your point.
    Sam, it proves my point. Client #1 and Client #15 are identical clients therefore they put out identical loads
    and take up identical amounts of VRAM. If that wasn't the case, then the bottom monitor should be reading a
    much lower number than it is, because if what you say is true, the other 14 clients aren't being rendered.
    You're telling me that 1 single client @ 60 FPS is taking up the same VRAM and putting out the same GPU load
    as 15 clients using up a total of 200 FPS at the same resolution? That's retarded...

    Quote Originally Posted by Sam DeathWalker View Post
    If your gpu can put out 19,000 fps total for all clients then that 1 percent is 190fps which would prove my point. As the game is putting out 250 total fps with one gpu (60 plus 60 plus 13 X 10) showing 15 tiled clients vs. 60fps with the one client showing. Which is a difference of 190fps
    I don't know where you're getting 60 + 60 + 130 from. Out of clients 15 through 30 only #15 is rendering at
    60 FPS. If you want to add up the FPS, the top 15 clients it would be 60 + 140 (14 clients x 10 FPS) for a total
    of 200 FPS. In your world, the bottom monitor is only rendering 60 FPS yet the VRAM and GPU load match the
    top monitor.

    Quote Originally Posted by Sam DeathWalker View Post
    I don't belive that they would have the gpu render screens that are not seen, considering they dont waste time rendering objects that can't be seen behind other objects.
    This is so ridiculous. You're basically saying that if a tree falls in the forest and no one is around... it doesn't
    make a sound. Just because you don't see the windows doesn't mean that the characters don't see the world
    around them. If what you are saying is true... then the slave characters would never be able to assist the main
    character if their screens weren't showing because they wouldn't be able to see what they should be assisting
    and would receive the error "You have no target.
    "

    EDIT: I made another video for you.

    Important Information:

    • Client #1 has higher video settings than Client #2
    • Client #1 is located in the tunnel to Orgrimmar looking at the ground
    • Client #2 is located in the Night Elf starting area
    • Client #1 is rendering at 60FPS foreground and background
    • Client #2 is rendering at 10FPS foreground and background
    • The inactive window sits behind the active window





    Now, because both clients have different video settings each window should take up different amounts of
    VRAM, however, as you can clearly see in the video... the VRAM doesn't change when I swap windows. Also,
    both clients are rendering at different framerates which affects the load on the GPU, however, as you can
    clearly see in the video... the GPU load does not deviate more than 2% from the reading at the beginning of
    the video.


    All game windows that aren't minimized, whether in the foreground or background, are rendered
    at all times.


    Is there anything else I can do for you Sam?
    Last edited by MiRai : 08-14-2011 at 04:26 AM Reason: Added More Proof / Important Info

  9. #59

    Default

    The amount of of ram isnt going to change it allocates the space.

    Maybe we are hung up a bit on the word "render". To me render means that the gpu does all the calculations needed to display a full frame. It has to determine which objects are in front of each other, the lighting, the character and world textures etc.

    Everything except rendering is always done for all clients, postion of characters, loading the textures of nearby characters, movement exptrapolatioin, etc.

    Lets say a weightlifter(gpu) can lift 1000lbs. At 1000lbs he is at 100percent(zgpu shows 100), at 500lbs he is at 50percent, at 100lbs his zgpu shows 10percent.

    You design a test that has him lift 2lbs vs. 10lbs and say that "10 lbs is FIVE times 2lbs". Surely his zgpu will show a difference of FIVE times because when he lifts 10lbs he is doing FIVE times the work of lifting 2lbs.

    Lifting 2 lbs shows his zgpu at 1 percent because the zgpu only measures in single percentages 2/1000 is rounded up to 1percent for display purposes.
    Lifting 10 lbs shows his zgpu also at 1 percent because 10/1000 is in fact 1 percent.

    The zgpu reading is the same for both (1 percent) yet one test was 5 times the other test. Thus you conclude that because the zgpu reading is 1 percent for both that he must be doing the same work (10lbs and 2lbs) 12lbs both times.

    Notice that when you ran two clients in the 2nd test you are at 12 or so percent and with 15 clients tiled you are at 19 or so percent.

    Why isn't running 15 clients using seven times as much gpu load as two clients?




    Let me ask you this, after the gpu renders the display that you see its sent to the display monitor. Where does the GPU put the rendered frame from the background when its done rendering the full screen? How many frames does it save? If you were writing the program would you render background data?


    If what you are saying is true... then the slave characters would never be able to assist the main
    character if their screens weren't showing because they wouldn't be able to see what they should be assisting
    and would receive the error "You have no target."
    You're characters arn't real people ... Position data is calculated (not rendered) for all clients at all times. Knowing that the leader is at coodenate 1021, 938 and is moving say in the X direction at 30 per second then your computer uses that informaiton to calculate his postion untill the server sends new information, thats all that is needed for position data.

    You don't have to "see" his textures to follow him, if you know his location in the world. Thats why you get character running forward when ping lag dosnt allow your computer to get server data to update the characters position. Then when the information fromt the server comes in the character snaps back to the server position.

    All the visual informationo is client side the server just sends poitional data, what the characters are and what they are wearing etc. how much damage they took and just minimal information. Your computer then takes taht minimal information and builds it into a world. The server probably dosnt even render any of the world even server side, but it does all the positional calculations and damage and all taht.
    Last edited by Sam DeathWalker : 08-14-2011 at 06:32 AM

    28 BoXXoR RoXXoR Website
    28 Box SOLO Nalak 4m26s! Ilevel 522! GM 970 Member Guild! Multiboxing Since Mid 2001!

  10. #60
    Multiboxologist MiRai's Avatar
    Join Date
    Apr 2009
    Location
    Winter Is Coming
    Posts
    6815

    Default

    Quote Originally Posted by Sam DeathWalker View Post
    Notice that when you ran two clients in the 2nd test you are at 12 or so percent and with 15 clients tiled you are at 19 or so percent.

    Why isn't running 15 clients using seven times as much gpu load as two clients?
    Because it was the login screen Sam. I could load up 150 login screens but I wouldn't be able to load 150
    characters into the game world because there's much, much more going on in the game world.


    Sam, I do not understand any of your examples or math but, how about a simple question...

    How come I can successfully FRAPS a game window that is behind another game window?

    I await your answer...

Posting Rules

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