My experience with 4-way Sli thus far

Discussion in 'Videocards - NVIDIA GeForce' started by A M D BugBear, Mar 10, 2019.

  1. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    Quadro Card??? Don't own one, I am using the drivers, not the card themselves.

    &

    Sorry for all the delay, this game is holding me up big time.

    So what I did is this:

    I re-installed the OS, Fresh install.

    Soon as I got into windows, I reverted my CPU microcode to where I had it once before, and this time, I used the same updates as it was working perfectly for this game before and then reloaded my Yakuza 2 profile and guess what happens???, nope, still get those major scaling drags in those specific spots, and yes, they are REAL BAD, although the scaling is a Beast and overall Frametime seems fine, so what I am doing now is re-tuning the bits and start from scratch again, its the only way now...

    All the Unknown PREDEFINED nvidia settings is all back to default, Now focusing on the SLI compatibility bits then I'll work on the Unknown settings.

    Again, I was hoping for the final conclusion of this game, going to be longer then I originally thought since all this came up.

    However that being said, in the mean time:

    Other then the scaling drag issues in these specific spots(And they are ONLY in these spots and ONLY implies to 3-Way to 4-Way Sli mode/Even if set to very low res and All settings turned down to min, it doesn't matter), the overall scaling potential from 1 gpu to 4 gpu, during in-game cut-scenes themselves or just playing = Scales from a Dragon, in otherwords, basically perfect scaling...

    So right now, My main primary focus is to lessen the effect of those scaling drag issues, It is absolutely crazy at times.. Especially when loading up the saved game, then hitting this particular spot while heading north, that's the worst part but if you go back south and hit this spot again, its no where near as bad, going south isn't bad at all, its heading north to these spots that's bad, soon as you go back north again, the effect is still kinda bad, lose quite a bit of fps, soon as you load up the game and hit these certain spots, the initial drag is the worse, again, I am just focusing in the beginning of the game, in the city, I have not tried in other areas, no need to, If I can't smooth out in this area, no sense of checking elsewhere, soon as I get that straight then I will look into other areas, more then likely, other parts of the game will probably be the same, but right now I am focusing getting all this squared away much as possible.

    Again, Some delay here, sorry about that folks.
     
    Last edited: Oct 13, 2019
  2. venturi

    venturi Master Guru

    Messages:
    298
    Likes Received:
    124
    GPU:
    none
    meant:
    quadro driver
     
  3. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    YAKUZA KIWAMI 2 - 4-WAY SLI - UPDATE - 10/14/19:

    Update on my progress on this game.

    So far I am making vast amounts of progress, but what really sped it up is this:

    So what I did is I deleted the Yakuza 2 .exe file from the nvidia inspector profile then I put the .exe file within the BattleField 1 profile and loaded up the game, upon loading up the saved game, The initial scaling drag in this one particular spot is still there, usually when you first load up the saved game and start running, although upon observation, when going back north to these spots again, the fps is a bit higher and that's another thing I want to mention here, the scaling drag is ALOT worse when you run and come to these very specific spots when you first load up your saved game but I have manage to lessen the major initial drag effect and try to increase the overall fps when you hit these spots again while playing.

    Holy Jesus, I have seen the initial drag on this one particular spot upon loading saved game, start running before I hit this particular spot, fps went to 60+ fps to like 0 fps, inbetween was a long pause, I said wow, This game is one of its kind when it comes to 4-way sli reaction, I have NEVER seen this bad before, not like this, in all my 4-way experience, never.

    As a Reminder, this ONLY implies to 3-Way and 4-Way Sli, although under 3-Way Sli, certain specific spots are not effected, under 1 gpu and 2-Way Sli = No need to worry/0 Problems LOL.

    So what I did next is I jotted down on piece of paper the setting parameters within BF1 profile, then I deleted the Yakuza 2 profile from the BF1 profile and then put the .exe file on the correct profile(Yakuza Kiwami 2), did some other changes within the unknown PREDEFINED nvidia settings, although I am no where being finished, I am making good progress, so please check back for more results later.

    Alot of work, especially with this game, like I said before, other then the scaling drag issues in these specific spots, everything is spot on perfect.

    UPDATE:

    Getting those high MS(Frametime) spikes again during in-game intro cut-scenes, although fps seems perfect, MS is spikes up, getting that problem again.
     
    Last edited: Oct 14, 2019
  4. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    YAKUZA KIWAMI 2 - 4-WAY SLI - HIGH FRAMETIME DURING CUTSCENES - UPDATE - 10/14/19:

    Got some Staggering news here and routed the source of this problem...

    Before I start, an absolute fair warning for people who is tuning 4-Way GTX 970's with this game, if this setting is turned on then you start playing, then if you quit the game and then turn off this particular setting within nvidia inspector, the situation will be Stuck as one hard super glue and the ONLY way to correct the issue is to restore the profile through nvidia's default settings, no need to re-install the graphic drivers as I previously did, so you have to put the settings from scratch basically.

    And the moment we been waiting for:

    The cause of this HIGH/very abnormal MS(Frametime) Spikes during in-game Cut-Scenes is:

    Under unknown PREDEFINED nvidia settings:

    Go to:

    0x00C96F61 (4 Profiles)

    Set to 0x00000000

    Instead of:

    0x0000000F (Battlefield 1, Plants vs Zombies Garden Warfare 2, Star Wars: Battlefront (2015), Mirror's Edge: Catalyst).

    I repeat, if your using 4 GTX 970's in 4-Way Sli mode and your doing as a huge project like Myself and tuning this game, Under no Circumstances this particular setting shall be turned on, doing so you will be stuck with the issue, even if you turn it off after you log off the game, you MUST restore the profile to factory settings and re-input the nvidia inspector settings to where you had it before WITHOUT this setting.

    Thanks for viewing and Back to tuning this game, Hopefully should be done in few days tops, ALOT of work was spent on this game and learned great deal from it.

    ANOTHER UPDATE, PLEASE READ:

    Another thing came to my attn. here:

    I had an experience where the frametime is high during in-game cut-scenes while this particular unknown predefined nvidia settings is completely shut off, if you encounter this, make sure you have a last good working profile backed up, simply re-load your profile, it will correct the issue.

    Again, something just came up, so just heads up on it, anything else I will give heads up asap, thank you and again, that particular setting I mentioned eariler, Always have it off, cause I did checked it several times over and confirmed it is causing it, so make sure under no circumstances that setting is turned on, referring to:

    0x00C96F61 (4 Profiles)

    UPDATE AGAIN:

    Just heads up to everyone again with this issue, I turned on that setting again to absolutely make sure 100% it does effect the 4-Way Configuration and yes it does, My confirmation, but even then, if the problem arises even without this setting(VERY RARE), just re-load last good working profile that you were tuning with this game and you are all set, no need to restore factory defaults like I mentioned before, happy campin'.

    Again if your tuning this game, Always keep a Very sharp eye on that MS(FrameTime) at ALL times, Absolutely no exceptions whatsoever...

    Also, always keep your eyes on your GPU BUS usage on ALL 4 Gpu's, under 4k, Make sure its around 0%-2% range, if the game is loading up, you might see kinda high or going up and down, that's normal, no worries, talking about while your on the game itself, Always keep a close eye on this, very very important, I can not stress this enough, if you see high GPU BUS usage on any game, bouncing back and forth, like lets say 10% to 30%, its a warning signal/sign that very abnormal tuning is in effect, bits needs to be re-tune, or something else is causing the issue, if you play in this condition, chances are, very abnormal fps, stability of the game is bad, and all sorts of issues, High chance that the game might crash or hang or possibly a BSOD, so don't take the chance on it, take my word on it from my own personal experience with 4-way and also someone that has experience with it, so always keep on eye on it, thank you.

    Also to top it all of that I didn't mention eariler on this particular post, although I have found the root of cause of the Abnormal MS spikes, there might be other settings within the Unknown nvidia settings that I may have overlooked or haven't tried to see if it will effect the 4-way in general that and or to see if any of the 3 bits:

    SLI compatibility bits, SLI Broadcast Bits & SLI specific hack bits to see if it will cancel or neutralize the MS SPIKES effect, that part is still under major investigation, so I can not say for certain, but like I mentioned before, I have confirmed it with this setting so make sure you turn it off, thanks.
     
    Last edited: Oct 15, 2019

  5. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    4-WAY SLI UPDATE - USE OF NON-QUADRO DRIVERS - MODDED 436.48 DRIVERS - 10/15/19:

    This came up very unexpected, and I was very curious on how it would respond on my 4-Way setup as ALL of my games tested on this thread were using modded quadro drivers, non of them were using the normal drivers, So I decided to take a look on this and was very curious how it will respond in general and not only that, give everyone heads up what to expect.

    I have personally tested 2 games and those 2 games are:

    Yakuza Kimawi 2

    &

    Rise of tomb Raider(Using my own custom high performance SMAA profile/I have two made)

    I have tried Yakuza 2 and basically the performance is pretty much the same as the modded Quadro drivers: 431.86 give or take.

    I have checked ALL of the main in-game intro Cutscenes, and ALL is working 100% perfectly with basically spot on perfect scaling, Also MS(Frametime) is completely stable, which is another A+ here.

    During the in-game itself, The major scaling drag in those very specific spots that I am working on, also applies to this drivers so for those who's tuning Yakuza, Both Quadro and Normal drivers have this issue with 4-Way, so keep this in mind please.

    I also for my own curiosity and to give everyone heads up about the HIGH MS(FrameTime) during the intro cut-scenes that was caused by:

    Under unknown PREDEFINED nvidia settings with Modded Quadro Drivers: 431.86/431.94

    0x00C96F61 (4 Profiles): 0x0000000F (Battlefield 1, Plants vs Zombies Garden Warfare 2, Star Wars: Battlefront (2015), Mirror's Edge: Catalyst), Also applies with this as well, I already 100% confirmed it, so be forewarned about this at all times, no exceptions, Always have set to: 0x00000000<<~~~ = :).

    And this part is also Extremely important, If your using the Desktop Drivers 436.48 drivers with this setting turned on:

    0x00C96F61 (4 Profiles): 0x0000000F<~~

    The effect on the MS(FrameTime) SPIKES and its Overall Consistency During the in-game Cut-scenes is ALOT WORSE(even had my mouth wide open for long period of time, absolutely staggering) then using the modded Quadro drivers in general, so everyone, take note on this, this is very serious stuff I recommend.

    2nd game is:

    Rise of tomb Raider

    &

    Played for about good 20-30 mins or so, Absolutely 100% perfect, Scales Absolutely Phenomenal, Gawd Dam_!, and Hits very hard on scaling usage, more then Yakuza 2 in general, talking about overall scaling usage consistency, sometimes all 4 will hit 100% usage, most of the time 99%, Absolutely Brutal scaling power...

    I normally have my voltage slider to 0+ and gpu speeds set to 1468 with this game, but now its @ +50(Voltage is much higher), speed about 1555 mhz on all 4 gpu's, temps were about high 60's, and I think it was gpu 2 or 3 that ran about 74 c tops, normally its lower but the voltage is much higher this time and much higher oc then I normally run it on, but still, temps are well under control.

    As a reminder to top this all off, Anyone who's planning to do a 4-way setup or about to as a huge project, you MUST, I repeat, MUST keep the GPU and the surrounding components temps WELL under control, this is the #1 priority for a very serious 4-Way setup, especially if your using a non-watercooling setup/non AIO's like myself.

    I have seen a video I think, a guy running 4 980's in 4-way, and guess what??? Within like 5-6 mins I think, his temps on all 4 gpu's were hitting near 100C, not surprised at all and then system gave up, so again, Keeping those temps in this setup is the #1 priority, this is the most important, if not, you will be very sorry in the end.

    Check back for Yakuza 2 updates, currently still working on it and thanks for reading.
     
    Last edited: Oct 15, 2019
  6. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    URGENT UPDATE ON 4-WAY SLI - INDEPENDENT SLI CABLES - 10/16/19:

    So I wanted to do like about a week ago, use some spacers on the left side of the gpu's so the vrm's themselves can achieve better temps along with fans hitting right in front of it, I used these separate cables so I can get better exposure on the vrm's, yes they are a bit better but here comes the bad news:

    First of all, These the cables I have gotten:

    https://www.amazon.com/s?k=linkup+sli&ref=nb_sb_noss_2

    Look for the Linkup Sli Connector.

    Next, take a look at this diagram on this site:

    https://linustechtips.com/main/topic/919827-2-way-sli-with-4-way-sli-bridge/

    I have used the bottom diagram, and its basically the same as the hard connector(The positioning of the connector), the one I normally use, guess what folks??? IT DOES NOT WORK!!!

    It could be perhaps the designing of these particular cables, either way, DOESN'T WORK.

    Well first of all, Something here I have never said anything from the beginning, All my games are Custom Resolution NOT DSR, as DSR will NOT WORK with 4-Way Sli as it flickers,etc like Absolutely mad crazy and there is no way around it, it must be custom resolution in order for it to work properly, all this time I been using Custom Resolution.

    So what I did next to troubleshoot the problem, use DSR = Does not work, also native resolution = Does not work.

    And lastly I installed fresh Modded Quadro Drivers, and guess what??? It does NOT work Period.

    I might get some quick flashes upon starting up a game then it hangs.

    So next I put my bridge connector that I normally use, the hard connector, the one that came with the motherboard and guess what happens next? IT WORKS PERFECTLY.

    So if your doing 4-way with this board or in general, you might want to reconsider to use the one your motherboard gave you or use another hard one to be on the safe side, although these cables seems high quality, it does't work properly...

    Thanks for reading.
     
    Last edited: Oct 16, 2019
  7. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    CURRENT GPU TEMPS - UPDATE - 10/16/19:

    Just to keep everyone heads up on my temps:

    Game used in this test:

    Yakuza Kiwami 2

    All gpu's scaling usage is around 99%, sometimes Gpu 3 hits about 98% or slightly less.

    Ambient Temp in this testing room: slightly over 70c

    Weather:

    Humidity: 75%/Dew Point: 12c

    All 4 Gpu's were OC'ed to 1550mhz, voltage during these test on all 4 Gpu's were quite a bit higher then I normally run it at.

    Ran the game while I was stationary(idle in one area) for good 30+ mins or so

    Results are as follows:

    GPU 1: 60c-61c @ 1.218V
    GPU 2: 64c-65c @ 1.225V
    GPU 3: 64c-66c @ 1.237V
    GPU 4: 62c-63c @ 1.256V

    I had the Voltage on each card from 1st to last, lowest to highest due to temp reasons, Even @ stock Voltage, each of these 4 gpu's are not identical, all 4 are different, so I did this to keep the temps under control more effectively, Another thing I never mentioned on this thread, All 4 gpus has the same identical bios

    Game was perfectly stable throughout the test, no crash, everything was solid.

    Although I think Rise of Tomb raider hits harder on temps, I might give that a try later, but for now, these were the results.

    UPDATE:

    Forgot to mention here eariler, I am using very different spacers now, so this will help contribute to the cooling in general, yes, spacers helps ALOT!.
     
    Last edited: Oct 18, 2019
  8. Dan Longman

    Dan Longman Member

    Messages:
    29
    Likes Received:
    7
    GPU:
    2080ti watercooled
    I think you meant 70f (21c) or else I hope you get out of there!! that's deadly hot!

    but seriously though just want to thank you for this thread I thoroughly enjoy reading about your adventures on this mad endeavour!
     
    A M D BugBear likes this.
  9. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    LoL, sorry about that, I meant slightly over 70 degrees farenheit, ambient temperatures speaking.

    Also, forgot to mention this earlier, my window in my room was open during this test, also my door, the room, where my system is, was open as well.
     
    Last edited: Oct 16, 2019
  10. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    YAKUZA KIWAMI 2 - 4-WAY SLI - UPDATE - 10/17/19:

    Unknown PREDEFINED nvidia settings, tuning is all complete, working on the main bits now.

    Also, just heads up, If your running this game lower then 4k, you will see lower scaling usage, so I am working on the scaling DEPT. on this situation, there is absolutely no guarantee in this, although @ 2400p it scales EXCELLENT, I tried @ 80% scaling resolution just for kicks, scaling usage went down like more then 25% or so, so I am trying to beef up the performance while playing @ a lower resolution, again, no guarantee on this, mainly all this tuning was specifically for 4k, on all the games tested on this thread, might be special exceptions going to lower res but that is extremely rare, depending on the situation I encounter, but I will do my best for this game so that I can accommodate other resolution as well...

    Check back for more results later as this game is still:

    WORK IN PROGRESS...

    ______________________

    QUICK UPDATE ON COMPATIBILITY BITS (DX10+DX11):

    One particular hex, MUST be absolutely careful while tuning:

    0x00<<~~, This particular hex, if not tuned properly, can put the game to a complete hang during startup with a blackscreen while the music is in the background, have to do a hard reset.

    0x07<<~~ DO NOT USE this digit within this hex, for those with same setup as mine and tuning the bits, be absolutely careful, again just heads up.

    Also, there is other digits within this hex that will cut your 4-way scaling significantly, so be absolutely careful while tuning this particular hex, on top of that, you will also notice abnormal GPU BUS usage, Like I mention before, if its more then it should, you will encounter low fps, very instability in the game, Possible crash/hang or bsod, so always make sure you keep an very keen eye on that GPU BUS usage, it is a warning signal/sign, thanks for reading.

    MORE UPDATE ON COMPATIBILITY BITS (DX10+DX11):

    Just heads up again on:

    0x00<<~~

    DO NOT USE anything under 8, Use 8 and above, including the letters, if you are tuning, although 8 seems the best overall but the difference on my side is like nil pretty much, I might gain like an fps or even if that with 8, so again just heads up on it, thank you.
     
    Last edited: Oct 18, 2019

  11. gerardfraser

    gerardfraser Ancient Guru

    Messages:
    3,075
    Likes Received:
    621
    GPU:
    R9 290 Crossfire
    This really is a amazing thread,I have not read every post , I have read a few though over a period of months.Man this thread will be useful for retro gaming someday(not a dig at you),wish I had not given up on Crossfire/SLI so long ago.I have nothing to contribute other than, keep going and this information is/will be valuable to many.
     
  12. harry9x

    harry9x Member

    Messages:
    13
    Likes Received:
    0
    GPU:
    MSI GTX 980TI SLI
    @A M D BugBear First of all your contribution is out of the world.i really really appreciate your efforts. bro I want to know that what are your thoughts on Assassins Creed Origins SLI. will it be possible in near future with the help of yours? thanks a lot bro..
     
    Last edited: Oct 18, 2019
  13. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    Thanks alot.

    1st of all, my focus is not on AC: Origins at the moment, My main primary focus is Yakuza Kiwami 2, I am 100% focused on getting this darn game done.

    That being said, I have worked on this game months ago then gave up then I wanted to re-try this game since I have much more experience on it because at that time I couldn't get rid of the massive flickering, soon as I went back on it, I fixed the flickering like an instant but in turn the overall scaling potential went down significantly, but there is a particular bits that help neutralize the issue and bring the scaling up a bit, as a reminder, this game is NO WHERE being done at all, I just went back on it to see certain things could be fixed, game is going to take ALOT OF WORK.

    The bits are:

    SLI Broadcast bits:

    0x00000<<~~ this one, I believe this one is set to 3, this will help neutralize the performance drop while your flickering is gone.

    I have no profile backed up with this game cause only I went back on it for just a tad, I believed I used the act of aggression profile and I changed this particular hex within:

    SLI Compatibility bits:

    0x00000<<~~ this one, If I remembered correctly, it was set to 7, It will stop all the massive flickering issues but your scaling is reduced BIG TIME but that broadcast bits, the hex: 0x00000<<~~ this one will help bring it up a bit, don't expect awhole lot.

    Like I said before, I went back on the game to see if I can fix certain things, tuned only for a bit, and that's it, and decided to go back on it later, so as of now, I would stay away until I am done tuning.

    I will go back to AC: Origins once Yakuza 2 and Nioh(working on stabilization issue under 3-way since game is capped at 60fps/4-way=not needed), and after looking into 2 games that Venturi asked me to look into.

    But in the meantime, IF you are on normal 2-way sli configuration and nothing higher, you could try the Act of aggression profile or FARCRY 5 profile, usually these 2 kicks my 4-way sli in in some games, if it does, then I tune it from there, usually gives me some start to deal with...

    Again, thanks for viewing this thread and thanks for your reply, 4-way sli is NOT EASY, not at all, I can clearly see why no one in the world touches it, Sucker is absolutely nothing to be played around with, talking about the amount of time and everything else involved in it, no joke whatsoever, but the end results = Invaluable amounts of information acquired from the actual person who's working on the 4-way setup.

    To break it in, I would say good 6 months of every day tuning on 4-Way, just to break it in and getting somewhat used to it, every day of doing this or whenever you have time, I would say decent = 1 year of experience of doing this and nothing but tuning, if main focus is on tuning and nothing else, daily or try to go on daily, doing decent, about 1 year I would say to get somewhat decent, hell I been going a lil over 6 months on this and I am barely scratching the surface, I got LONG and I mean very very LONG ways to go.
     
    Last edited: Oct 18, 2019
  14. harry9x

    harry9x Member

    Messages:
    13
    Likes Received:
    0
    GPU:
    MSI GTX 980TI SLI
    thanks bro for quick response. and best of efforts for this mind blowing projects which will be useful to sli owners community. nvidia is just trash. it is killing its own technology.
     
  15. bobblunderton

    bobblunderton Member Guru

    Messages:
    105
    Likes Received:
    42
    GPU:
    EVGA 2070 Super 8gb
    I used to love my SLI rigs 10 years ago. I had a 7950 GX/2 (remember those in late 2006?) and a GTX 260 and then GTX 470 SLI setup, was great when new 1080p and 1200p was hard to push at good detail. There can be times where video card upgrades WILL NOT help you get more FPS with better settings though, do read on if interested.
    To the original poster:
    Until we get more games running in DirectX 12 or Vulkan at 4k, and it's something 'everyone can do' with a mainstream computer (run 4k games 60fps at 'decent' or 'medium' detail), I don't think we'll see a lot here, especially with the price of decent GPU's.
    There's a few reasons I say this, that and the price; it's as I design things for a current/next-gen 3D driving game, and make open game-worlds, I realize things. It's not details in the buildings as-per triangles, it's not details on the vehicles as-per triangles, everything hinges on one thing once the initial system requirements (VRAM for models and textures) is satisfied.
    DRAW CALLS. Click 'spoiler' to reveal a write-up on draw calls, what they are, when & how to satisfy the demand, etc.
    You could have the best SLI setup in the world with the best game-engine for it, and if you're already pegging one cpu-core (on DX11 and below), you won't go a single frame faster (though you might get a tad more consistent, it may be within margin-of-error though). If the CPU is bottle-necked on feeding one set of connected textured surfaces (linked UV's with identical texture) after the next, model after model, the GPU is sitting twiddling it's thumbs, proverbially. This is often evidenced as you'll see the in-game benchmarks at 65% use CPU and 50% use GPU or something remotely close (not maxing either without a frame-limiter in the equation). So if the example was a house, and all textures were connected that should / could be on the model of the house, then all the brick would be one draw call, all the roof texture would be one draw call (provided the roof was one consistent run, if it's split, that's a few more), all the trim on the eaves/gable ends of the roof would be one draw call, the windows and doors would be a few draw calls (unless you windows and doors all match, then it'd be two at the least, using 'batching' rendering identical surfaces at the same time), and so-on and so-forth. So while the GPU can render and fill 10,000 triangles in a frame without even maxing TDP or fan speeds out, the CPU has drown itself trying to feed it piece-meal in such a manner as this, and it can often lead to the GPU (or GPU's in this instance) loafing. This is exacerbated by some games were modeling and 3d environment is done poorly, hastily, or both (more common then you think!). When you have multiple layer textures using bump-mapping or normal mapping, specular / reflection, and shadows on it all, and then even grunge maps / variations, that's additional passes per set of like-surfaces. Now all that 'draw call' feeding the CPU must do for the graphics to be rendered on the video card, in a SINGLE core in DX11 (hence 9900k 5.0ghz processor owners, rejoice, eat your heart out, this is your day to shine!). This is why also that the 9900k reigns supreme in some things especially gaming, where for others such as content creations (me!) the Ryzen is better (disclosure: I have a Ryzen 3700x). That's getting a bit far from what I want, and let's not start another processor war thread. Long story short, the only way to fix Draw Call bottle-necking, is to FIX THE GAME or throw more GHZ / IPC at the problem with an overclock or processor/memory speed upgrade. As games go open-world more and more, and less on-rails linear, while simultaneously getting more detailed, you can see where this is going. DX11's days are numbered, but provided we have plenty of bottom-dollar talent in the industry just getting out of college on first assignment, I don't think overclocking's days (of need anyway) are certainly numbered.
    DX12 and Vulkan *BOTH* fix this problem by multi-threading draw call routines (or at-least ENABLING the developers to do this). This enables exponentially more detailed scenes in games, and as is the case, you'll often find DX12 titles happy to use more cores and a more balanced load (barring rubbish programming!) at that. Where-as by contrast, if you're seeing one core bottle-necked on a multi-threaded game engine, while the rest are loafing, chances are you might be hitting the draw call limits of your processor, where throwing the best video card made of gold that money could ever wish to buy will do NO GOOD.

    So entirely, some games will run great, and very smooth, even on meager systems with a decent video card, even if the processor is 6~8 years old and only 2.x~3.0 ghz. Others' by contrast will always run like dirt no matter WHAT you do, no matter how much you upgrade your video, even if it doesn't look like the CPU is maxed out (and this case there's no FPS limiter or clamp in place). HINT: Single-core saturation (maxed out) of draw calls doesn't always show up in-game benchmarks or FPS charts (especially true if they just show cpu-wide use %, and not per-core), but normally always shows up in task manager or your preferred hardware-watching program such as HW-INFO64 or Core-Temp.

    I hope this book of knowledge helped, but not as much as I wish developers would hurry up and use VULKAN or DX12 so we can have some more interest games / immersive game worlds.

    Thank-you to the original poster for the book of content you indeed wrote, I am sure it's helped many. If only you were around to see the sight that was getting GTAIV working with SLI 260's and 470's back almost 10 years ago when it was new - they said 'oh it never worked' ... it did on one of my systems.

    --Cheers!


    --Foot note for the site admin / general master.
    @Hilbert Hagedoorn Maybe what I wrote would shed some light into bench-marking, I don't know how deep you've dived into game engines, but I hope you get to read it sometime. If it helped, then it's small payment for using this site so many years, as it's one of the best.
    Do checkout BeamNG.Drive for bench-marking things, it can use the CPU / FPU / RAM very intensely in physics, it's a soft-body vehicle simulator. It'd be pretty awesome to have it on there with the cpu competition heating up, and it's fairly cheap! I'll pay for a license key for you in November if-needed - contact me after November 1st and I will hand you one.
     

  16. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    UPDATE ON OVERCLOCKS/GPU-Z's PERFCAP REASONS,ETC - 10/19/19:

    Went back to Maxwell bios editor II and did some changes to see if I can bring the cards out more, got great news coming along.

    First lets say this is something I have never done since I been doing this 4-Way project.

    Did some bios modding, some changes, and tested all 4 Gpu's at a staggering 1600mhz(Core), and also tried at 1.275V, which is very very very high for a 4-Way setup, I have never ran it this high, since I have Excellent cooling in store, very amazingly it was stable for like about 10 mins or so then it hanged, I had to sign off and lower the clocks just a bit.

    There was some visible artifacts @ 1600mhz, not to bad really, then I lowered it to 1595 @ 1.275v(ALL 4 GPU's), Very minimal artifacts and overall stability is excellent but ran it for only like what 15 mins or so, don't know long term, this is just heads up to everyone that it is indeed possible to run all 4 970's at a whopping 1600mhz, that is absolutely insane.

    Might lower it to 1590mhz to be safe.

    Also just heads up, I checked out GPU-Z, and under Perfcaps Reasons, it is showing as NONE while under load, very interesting.

    Also, I know this doesn't really mean anything really, not an accurate reading BUT, I checked the Power Consumption (W) within GPU-Z, only did sensor for one particular gpu, it says 216W+.

    During this short test, Remember, the voltage is ALOT higher then I ever done in the past so keep this in mind at all times please, and also bear in mind that, this is not a water-cooled/AIO system, this is using the gigabyte's own g1 windforce heatsink with alot of stuff done to it and my case and cooling.

    Temps are as follows

    GPU 1: 64c-65c
    GPU 2: 68c-70c
    GPU 3: 68c-70c
    GPU 4: 64c-65c

    I didn't check the ambient temp in my room but yes it was warm in the room during the time of the test, door was open as well. Guessing in the low 70's(Fahrenheit).

    Still hard at work with Yakuza 2, so keep your eyes peeled for it, the Dragon is coming, performance results is coming.

    UPDATE ON VOLTAGE READINGS:

    Forgot to mention earlier that the cap for the gpu voltage reading for these cards are 1.275V, although it may go higher, need to use multimeter, but yes the max readings through OSD is 1.275V.
     
    Last edited: Oct 20, 2019
    fantaskarsef likes this.
  17. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    YAKUZA KIWAMI 2 - 4-WAY SLI - UPDATE - 10/22/19:

    So the final conclusion will be up today, I just did a performance test run under 1 gpu mode and 2 gpu mode, jotted down tons of stuff, I haven't done 3-way or 4-way mode yet but will be working on it now, but just to let everyone know, from 1 gpu and 2 gpu mode, overall scaling potential and consistency = EXCELLENT!!!!, 2 thumbs up.

    Another thing I want to add here, very interesting.

    I was doing some massive tuning in this game, in the beginning, the 2nd time around while tuning, when you hit the those very minute scaling drag spots(Only in these spots), worse is when you load up your saved game for the first time upon starting up the game, when you start running into these certain specific spots, the initial drag is incredibly bad, but when you go back south, its no where as bad, then when you walk or run back north to this original specific spot, its no where as bad but the fps drop under 4-way = kinda but to so-so, after tuning, the initial drag when you start running, these specific spots, I have lessen the absolute drag upon starting up the load game.

    Now whats interesting is this, just for kicks, I restored nvidia profile default, then I just added the act of aggression under SLI Compatibility bits, no other changes were made, and guess what? There was no change in overall scaling potential nor the scaling drag issues.

    Why this was the case got my scratching my head, So I originally thought that maybe when after tuning the game and even if you did restore it the profile to default settings and just changed the SLI Compatibility bits, maybe this game didn't change a bit.

    Next, I did another testing, I completely uninstalled the Graphic drivers and re-installed the same EXACT graphic drivers during the time of the test, then quickly I re-loaded my nvidia inspector profile that took a long time to tune, because I want to rule it out, so Loaded up the game and then tested the game and guess what? NO DIFFERENCE from only using the Act of Aggression profile(while everything was set to default).

    So I was like what in the living hell is going on here, this is like the only game that does this, I believe there was another game that I tuned had similar circumstances, this is absolutely interesting to note here.

    Crazy stuff going on here, but anyways, Look for FINAL CONCLUSION to this game today, be sure to look for it.

    I will provide my custom made bits if necessary, I'll try out both ways to rule out if there is stability problems, thanks for reading.

    UPDATE:

    I knew this for quite some time, 4 Very heavy OC'ed GTX 970's in 4-way with Excellent scaling potential = Slightly faster then A 2080ti.

    For everyone's reference, here is a video running with ZOTAC 2080ti AMP edition to get a better idea how mine would compare, I would say a bit faster then this video shown however that being said, one must keep in mind that this is a recorded video, so loss on performance will be there during the recording session, so therefore the overall fps will be higher then this video shown:

    Also, the one with the major scaling drag areas(Specific Spots), Like I have mentioned before is around 0:35 of this video, pay attn. to this spot, that being said, there are other spots like this and there are other specific spots that the drag is no where near as bad, and to top it off, I can NOT get rid of it, this ONLY applies to 3-way and 4-way sli setup's, 3-way is not as bad as 4-way when it comes to these specific drag spots, some spots that affects 4-way, doesn't effect 3-way sli at all.

    Try to watch it on the actual site, much better quality.



    UPDATE ON 10/22/19:

    Alright so all my perfomance recordings have been all jotted down and their is tons and tons of information...

    There will be complete performance test on each scene recorded, from:

    MIN FPS/MAX FPS/AVG FPS/1% LOW & 0.1% LOW.

    Performance test on ALL 4 gpu's includes:

    10 in-game intro scenes(Excluding the video playback ones), City Thug Fights, 2 City Performance Tests, and 6 Spots(Stationary/Idle).

    Everything is complete, now just have to check overall stability on 4-way setup...

    There is some interesting stuff within 3-way setup, I will go over that within the final conclusion.

    Also, A particular setting, in which I always have it on high, after all of my performance test was done, I re-checked my settings, for some reason, this setting is on medium:

    Geometry Quality.

    HOWEVER, this particular setting from my close inspection, makes flat out no difference in performance, wether from low or to high, I already checked in-game cutscenes and in the game itself, then I re-checked my jotted down notes, NO DIFFERENCE, so again, sorry about that and no, I ain't doing this all over again, to much work involved, no need to re-do tests because the performance change is 0.

    Gonna take a break for awhile before I do my Long Stability Run, been on this game ALL DAY.

    Time to take a break. Check back later.
     
    Last edited: Oct 23, 2019
    fantaskarsef likes this.
  18. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    YAKUZA KIWAMI 2 - 4-WAY SLI - FINAL CONCLUSION - 10/23/19

    Here it is you all, The Dragon is out of its Den, the "MOMENT" we all been waiting for.......

    First off, let me clear some things here.

    Certain Settings needs to be shut off, Namely DOF or AKA depth of Field, otherwise, certain particular in-game cutscenes, you will see very abnormal scaling, so therefore, the setting should be off, I highly advised turning this setting on, the cost of performance hit is to much for adding so much blurfest to the game, doesn't look to bad but the performance hit for this is to much, so I strongly advised to turn it off, you can use it during the actual game play itself, but be careful, some in-game cut-scenes, you wish you had it off to begin with.

    So this Final Write-up to this game, the Graphic Settings are all set to max EXCEPT:

    Motion Blur & DOF.

    Both of those are OFF.

    AA of choice is: SMAA

    And the Resolution of choice is set to:

    3840x2400(4k/16:10 aspect ratio)

    Also, I am using the Letterbox Removal mod(to cover the left over portions on non 16:9 aspect ratio screens).

    I did long term stability test to make sure the game didn't crash or anything, ran the game for about 2 hrs or so I would say and not a single crash, Game is Rock hard Stable, longs your overclock's are in good standing, your good to go.

    I tested ALL In-game intro Cuts-Scenes and its perfectly stable, Runs extremely smooth, certain In-game cuts-scenes, you might experience some drag in the beginning of theses particular scenes, but after a few seconds, the scaling will go back to normal, these are the scenes that I was referring about when you use DOF, so MAKE SURE you have DOF OFF, otherwise these particular in-game cut-scenes will act very abnormal, Absolute fair warning.

    The Scaling drag issues in certain specific spots while playing the game, I can NOT get rid of, I am very terribly sorry, I did many things to get rid of it, no can do.

    Example of which:

    When your in the first starting the game in the city, certain spots of this game going north, the scaling will drag down big time, but when you walk back south, it doesn't hit as hard, then when you come back to these spots, chances are the overall perfomance may or may not improve at all.

    Same goes for chapter 2 and chapter 3 in the city, you will absolutely notice the scaling drags in certain spots, but if your not paying any attn. to the drag, its not all that bad, but the gap between point a to point b is more apparent then other games, like:

    Sekiro: Shadow dies twice, Dragon Quest XI to name a few

    That being said, it doesn't take it away from this game at all.

    The overall Scaling Power/Potential in this game in General = Absolute Dragon.

    The overall Scaling Potential/Power over 1 gpu in general is probably one of the best I have encountered with, despite problems with this game.

    Overall Scaling Consistency during in-game cutscenes is Excellent, just the ones DOF needs to be shut off, are the ones that have scaling issues right in the beginning of a particular scene, after a few seconds, it will quickly go back to normal so this is not an issue.

    Here comes another part, Graphic Anomalies:

    I have found a few:

    During the backyard fight with Dojima, the side most of the walls, ground area, flashes a bit, I tried Alt+Enter(window screen), then back to the game to see if it will change, nope.

    Another one is in chapter 3, when you talk to that modeling dude near the bridge, during the fight scene with 3 people, the da%n NeWsPaPerS flashes like mad crazy, again, I tried the ALT+ENTR(window mode) then back on the game to see if it will change, nope.

    And last one I found, this only pertains to 3-Way Sli, Very interesting stuff here:

    During your Energy charge, when you unleash your charge attacks, you can see a trail of energy being left there, and you can still continue to leave more trail behind, more and more as time goes by, it will stay there, even during cuts-scenes, it can overlay the cut-scenes in the area, so be very careful, however if you want to get rid of it and I already confirmed it works:

    ALT+ENTR(window screen), then back on the game = Fixes the issue.

    Next segment covers a MAJOR deal of performance test.

    Covers the following:

    1 to 4-Way Sli tests

    10 intro In-game Cutscenes(Performance tests)

    2 separate tests during in the city(Beginning)

    Thug Fights(Beginning/city)

    & 6 Spots Covered, all in random location(Stationary/Idle/FPS ONLY).

    As an Added Bonus, Under 4-Gpu mode only, I covered the following performance test as well:

    In-game cuts-scenes of: Ryuji Goda

    ______________________________________________________________________

    ALL TESTS WERE DONE @ 1590MHZ(GPU CORE)/7516MHZ(MEMORY) VOLTAGE ON GPU CORE WAS SET TO 1.275V ON ALL 4 GPU'S, SOME TESTS HERE WERE DOWN AT LOWER CLOCKS, NAMELY FROM INTRO 7 TO 10, FROM 1590MHZ TO 1585MHZ DUE TO STABILITY REASONS.

    These test are in order during the in-game intro cuts-scenes, going from one to another(EXCEPT for one particular one that was an in-game video, not actual in-game cut-scene, so I didn't include that in my tests.

    Results are as follows:

    Intro# 1:


    1 GPU: MIN FPS: 8//MAX FPS: 19//AVG FPS: 14//1% LOW: 8 FPS//0.1% LOW: 8 FPS

    2 GPU's: MIN FPS: 16//MAX FPS: 38//AVG FPS: 28//1% LOW: 17 FPS//0.1% LOW: 16 FPS

    3 GPU's: MIN FPS: 25//MAX FPS: 60//AVG FPS: 42//1% LOW: 25 FPS//0.1% LOW: 25 FPS

    4 GPU's: MIN FPS: 32//MAX FPS: 76//AVG FPS 59//1% LOW: 32 FPS//0.1% LOW: 26 FPS

    Intro# 2:

    1 GPU: MIN FPS: 15//MAX FPS: 23//AVG FPS: 18//1% LOW: 15 FPS//0.1% LOW: 14 FPS

    2 GPU's: MIN FPS: 30//MAX FPS: 46//AVG FPS: 36//1% LOW: 30 FPS//0.1% LOW: 25 FPS

    3 GPU's: MIN FPS: 44//MAX FPS: 67//AVG FPS: 52//1% LOW: 43 FPS//0.1% LOW: 38 FPS

    4 GPU's: MIN FPS: 57//MAX FPS: 89//AVG FPS: 70//1% LOW: 54 FPS//0.1% LOW: 43 FPS

    Intro# 3:

    1 GPU: MIN FPS: 12//MAX FPS: 17//AVG FPS: 15//1% LOW: 12 FPS//0.1% LOW: 12 FPS

    2 GPU's: MIN FPS: 23//MAX FPS: 35//AVG FPS: 31//1% LOW: 23 FPS//0.1% LOW: 12 FPS

    3 GPU's: MIN FPS: 38//MAX FPS: 53//AVG FPS: 46//1% LOW: 38 FPS//0.1% LOW: 31 FPS

    4 GPU's: MIN FPS: 50//MAX FPS: 70//AVG FPS: 61//1% LOW: 47 FPS//0.1% LOW: 30 FPS

    Intro# 4:

    1 GPU: MIN FPS: 16//MAX FPS: 18//AVG FPS: 17//1% LOW: 16 FPS//0.1% LOW: 11 FPS

    2 GPU's: MIN FPS: 31//MAX FPS: 36//AVG FPS: 23//1% LOW: 30 FPS//0.1% LOW: 24 FPS

    3 GPU's: MIN FPS: 47//MAX FPS: 54//AVG FPS: 50//1% LOW: 42 FPS//0.1% LOW: 38 FPS

    4 GPU's: MIN FPS: 64//MAX FPS: 68//AVG FPS: 65//1% LOW: 57 FPS//0.1% LOW: 49 FPS

    Intro# 5:

    1 GPU: MIN FPS: 12//MAX FPS: 16//AVG FPS: 14//1% LOW: 13 FPS//0.1% LOW: 12 FPS

    2 GPU's: MIN FPS: 25//MAX FPS: 32//AVG FPS: 29//1% LOW: 25 FPS//0.1% LOW: 24 FPS

    3 GPU's: MIN FPS: 36//MAX FPS: 47//AVG FPS: 43//1% LOW: 37 FPS//0.1% LOW: 35 FPS

    4 GPU's: MIN FPS: 47//MAX FPS: 68//AVG FPS: 56//1% LOW: 48 FPS//0.1% LOW: 40 FPS

    Intro# 6:

    1 GPU: MIN FPS: 13//MAX FPS: 20//AVG FPS 17//1% LOW: 13 FPS//0.1% LOW: 13 FPS

    2 GPU's: MIN FPS: 27//MAX FPS: 40//AVG FPS: 34//1% LOW: 27 FPS//0.1% LOW: 26 FPS

    3 GPU's: MIN FPS: 40//MAX FPS: 59//AVG FPS: 50//1% LOW: 40 FPS//0.1% LOW: 36 FPS

    4 GPU's: MIN FPS: 53//MAX FPS: 80//AVG FPS: 68//1% LOW: 52 FPS//0.1% LOW: 40 FPS

    Intro# 7:

    1 GPU: MIN FPS: 15//MAX FPS: 24//AVG FPS: 18//1% LOW: 16 FPS//0.1% LOW: 15 FPS

    2 GPU's: MIN FPS: 31//MAX FPS: 48//AVG FPS: 35//1% LOW: 30 FPS//0.1% LOW: 24 FPS

    3 GPU's: MIN FPS: 45//MAX FPS: 70//AVG FPS: 51//1% LOW: 42 FPS//0.1% LOW: 35 FPS

    4 GPU's: MIN FPS: 61//MAX FPS: 95//AVG FPS: 70//1% LOW: 59 FPS//0.1% LOW: 48 FPS

    Intro# 8:

    1 GPU: MIN FPS: 16//MAX FPS: 20//AVG FPS: 17//1% LOW: 16 FPS//0.1% LOW: 15 FPS

    2 GPU's: MIN FPS: 30//MAX FPS: 41//AVG FPS: 34//1% LOW: 30 FPS//0.1% LOW: 27 FPS

    3 GPU's: MIN FPS: 41//MAX FPS: 61//AVG FPS: 51//1% LOW: 39 FPS//0.1% LOW: 32 FPS

    4 GPU's: MIN FPS: 60//MAX FPS: 80//AVG FPS: 68//1% LOW: 57 FPS//0.1% LOW: 43 FPS

    Intro# 9:

    1 GPU: MIN FPS: 15//MAX FPS: 21//AVG FPS: 18//1% LOW: 15 FPS//0.1% LOW: 14 FPS

    2 GPU's: MIN FPS: 29//MAX FPS: 43//AVG FPS: 35//1% LOW: 29 FPS//0.1% LOW: 26 FPS

    3 GPU's: MIN FPS: 44//MAX FPS: 63//AVG FPS: 51//1% LOW: 43 FPS//0.1% LOW: 36 FPS

    4 GPU's: MIN FPS: 59//MAX FPS: 82//AVG FPS: 70//1% LOW: 57 FPS//0.1% LOW: 45 FPS

    Intro# 10:

    1 GPU: MIN FPS: 13//MAX FPS: 17//AVG FPS: 15//1% LOW: 13 FPS//0.1% LOW: 14 FPS

    2 GPU's: MIN FPS: 25//MAX FPS: 35//AVG FPS: 29//1% LOW: 25 FPS//0.1% LOW: 22 FPS

    3 GPU's: MIN FPS: 36//MAX FPS: 50//AVG FPS: 42//1% LOW: 35 FPS//0.1% LOW: 27 FPS

    4 GPU's: MIN FPS: 49//MAX FPS: 69//AVG FPS: 58//1% LOW: 48 FPS//0.1% LOW: 36 FPS

    ----------------------------------------------------------------------------------------------------------------------

    ADDED BONUS(4-WAY TEST ONLY):

    RYUJI GODA/IN-GAME CUTSCENE(CHAPTER 2)/FINAL:

    MIN FPS: 59//MAX FPS: 73//AVG FPS: 65//1% LOW: 54 FPS//0.1% LOW: 45 FPS

    --------------------------------------------------------------------------------------------------------------------

    YARD AREA(CEMENTRY AREA GAMEPLAY/BEGINNING):

    Results are as folllows:

    1 GPU: MIN FPS: 18//MAX FPS: 21// AVG FPS: 20//1% LOW: 18//0.1% LOW: 17 FPS

    2 GPU's: MIN FPS: 37//MAX FPS: 42//AVG FPS: 40//1% LOW: 36 FPS//0.1% LOW: 34 FPS

    3 GPU's: MIN FPS: 54//MAX FPS: 62//AVG FPS: 58//1% LOW: 52 FPS//0.1% LOW: 48 FPS

    4 GPU's: MIN FPS: 74//MAX FPS: 83//AVG FPS: 78//1% LOW: 69 FPS//0.1% LOW: 63 FPS

    ----------------------------------------------------------------------------------------------------------------------


    CITY AREA/ACTUAL GAME PLAY/BEGINNING OF THE GAME/2 SEPARATE TESTS:

    AS A REMINDER, SINCE THERE IS SCALING DRAG IN CERTAIN SPOTS WITHIN 3-WAY AND 4-WAY SLI, I MADE 100% SURE I WAS NOT EFFECTED DURING THESE TESTS, EVERY MOVE WAS MADE 100% CORRECT AND PRECISE SO ALL READINGS ARE CORRECT, AS A REMINDER AGAIN, SCALING DRAG IN THESES SPECIFIC SPOTS ONLY EFFECTS IN 3-WAY AND 4-WAY SLI.

    Results:

    City# 1

    1 GPU: MIN FPS: 15//MAX FPS: 21//AVG FPS: 19//1% LOW: 15 FPS//0.1% LOW: 13 FPS

    2 GPU's: MIN FPS: 29//MAX FPS: 42//AVG FPS: 37//1% LOW: 29 FPS//0.1% LOW: 26 FPS

    3 GPU's: MIN FPS: 43//MAX FPS: 62//AVG FPS: 55//1% LOW: 42 FPS//0.1% LOW: 32 FPS

    4 GPU's: MIN FPS: 56//MAX FPS: 82//AVG FPS: 72//1% LOW: 55 FPS//0.1% LOW: 33 FPS

    City# 2:

    Results:

    1 GPU: MIN FPS: 19//MAX FPS: 23//AVG FPS: 21//1% LOW: 19 FPS//0.1% LOW: 18 FPS

    2 GPU's: MIN FPS: 36//MAX FPS: 44//AVG FPS: 41//1% LOW: 36 FPS//0.1% LOW: 34 FPS

    3 GPU's: MIN FPS: 56//MAX FPS: 66//AVG FPS: 62//1% LOW: 54 FPS//0.1% LOW: 48 FPS

    4 GPU's: MIN FPS: 72//MAX FPS: 86//AVG FPS: 79//1% LOW: 66 FPS//0.1% LOW: 33 FPS

    ----------------------------------------------------------------------------------------------------------------------

    Thug Fights(City/Beginning of the game)/Made sure scaling drag wasn't effected during this scene, this also includes the dialog text as well before engaging in combat...

    Results are as follows:

    FINAL:

    1 GPU: MIN FPS: 14//MAX FPS: 39//AVG FPS: 17//1% LOW: 14 FPS//0.1% LOW: 14 FPS

    2 GPU's: MIN FPS: 26//MAX FPS: 52//AVG FPS: 34//1% LOW: 26 FPS//0.1% LOW: 17 FPS

    3 GPU's: MIN FPS: 42//MAX FPS: 71//AVG FPS: 51//1% LOW: 42 FPS//0.1% LOW: 26 FPS

    4 GPU's: MIN FPS: 53//MAX FPS: 82//AVG FPS: 70//1% LOW: 45 FPS//0.1% LOW: 19 FPS

    ---------------------------------------------------------------------------------------------------------------------

    IN THIS SECTION, THIS PART COVERS 6 SPECIFIC SPOTS(RANDOM BUT SAME FOR ALL 4 GPU TESTS) ALL STANDING STATIONARY(IDLE) AND ONLY FPS IS RECORED, NO NEED FOR OTHERS:


    RESULTS/FINAL:

    SPOT# 1:

    1 GPU = 20 FPS//2 GPU's = 39 FPS//3 GPU's = 59 FPS//4 GPU's = 75 FPS


    SPOT# 2:

    1 GPU = 20 FPS//2 GPU's = 39 FPS//3 GPU's = 59 FPS//4 GPU's = 76 FPS

    SPOT# 3:

    1 GPU = 23 FPS//2 GPU's = 46 FPS//3 GPU's = 68 FPS//4 GPU's = 86 FPS

    SPOT# 4:

    1 GPU = 24 FPS//2 GPU's = 49 FPS//3 GPU's = 76-77 FPS//4 GPU's = 85 FPS

    SPOT# 5:

    1 GPU = 17 FPS//2 GPU's = 35 FPS//3 GPU's = 52 FPS//4 GPU's = 68 FPS

    SPOT# 6:

    1 GPU = 19 FPS//2 GPU's = 37-38 FPS//3 GPU's = 56-57 FPS//4 GPU's = 76 FPS

    -------------------------------------------------------------------------------------------------------------------


    If anyone is interested in the custom bits here they are BUT FIRST:

    WARNING ABOUT CUSTOM BITS:

    ONLY FOR THE NVIDIA GTX 970 FAMILY+WIN 10 X64(RS2-15063/1703(RTM).

    THERE IS 2 SEPARATE PROFILES HERE, ONE WAS THE ORIGINAL TUNING PROCESS, THE OTHER ONE I HAD SOME ISSUES AND HAD TO RETUNE, THEN FIGURED OUT AFTER RE-INSTALLING THE GRAPHIC DRIVERS, ALOT OF CUSTOM MADE SETTINGS BY ME ARE NOT NEED AT ALL, BUT IF ANYONE IS INTERSTED I WILL PROVIDE THEM HERE:

    ORIGINAL TUNING BITS:

    UNDER 1 = COMPATIBILITY:

    SLI COMPATIBILITY BITS (DX10+DX11): 0x280B16F5

    SLI BROADCAST AND SPECIFIC HACK BITS = NOT USED

    COPY ENGINE BLITTER: 0x00000020 FORCE_CE_UPLOADS Force - Copy Engine Async Uploads for all uploads (ignoring location of destination surface).

    UNDER 2 - SYNC & REFRESH:

    MAXIMUM PRE-RENDERED FRAMES: USE 3D APPLICATION SETTINGS

    UNDER 5 - COMMON:

    AMBIENT OCCLUSION SETTING: Performance


    AMBIENT OCCLUSION USAGE: Enabled

    MULTI-DISPLAY/MIXED-GPU ACCELERATION: SINGLE DISPLAY PERF. MODE

    UNDER 6 - SLI:

    ANTI-ALIASING/SLI AA: DISABLED

    ALL OTHERS SET TO AUTOSELECT EXCEPT FOR:

    NVIDIA PREDEFINED # OF GPU's TO USE ON SLI RENDERING MODE DX1x: FOUR

    NVIDIA PREDEFINED SLI MODE IN DIRECT X 10: AFR

    UNDER - 8 EXTRA

    MAKE THESE CHANGES BELOW:


    ASYNC10_NVAPI_MODE: 0x00000003 ALLOW_ALL - Allow all optimizations

    CULL_BEFORE_FETCH: 0x00000006 Dynamic

    SLIMOSAIC: 0x00000340 (Battlefield 3, Battlefield 1, Battlefield 4, Dragon Age: Inquisition, Plants vs Zombies Garden Warfare 2, Star Wars: Battlefront (2015), Mirror's Edge: Catalyst).

    UNDER OTHERS:

    WHISPER MODE APPLICATION FPS: 0x0000003C

    UNDER UNKNOWN PREDEFINED NVIDIA SETTINGS, MAKE THESE CHANGES BELOW:

    0x00110999 (2 Profiles): 0x00001000 (Battlefield V, Star Wars: Battlefront II (2017))
    0x002533F7 (1 Profiles): 0x00000001 (Battlefield V)
    0x00854CAB (26 Profiles): 0x00000028 (Battlefield 4)
    0x00939833 (1 Profiles): 0x00000001 (Hitman)
    0x00D98F93 (1 Profiles): 0x00000010 (Battlefield 1)
    0x105E2A1D (268 Profiles): 0x00000004 (Battlefield V, ME: Andromeda, THE CREW 2)
    0x2030B5CD (2 Profiles): 0x00000002 (Ogre, Dawn)
    0x70092D4A (31 Profiles): 0x0000000E (Fallout 76,AC: Unity, Agents of Mayhem)
    0x7049C73C (0 Profiles): Set to: 5
    0x704D456E (0 Profiles): Set to: Some objects render at wrong depth.
    0x7051E5F5 (0 Profiles): Set to:5
    0x70B5603F (0 Profiles): Set to: d3d
    0x70CCB5F0 (27 Profiles): 0x00000400 (DRG Initiative, Kingdom Come: Deliverance)
    0x80857A28 (964 Profiles): 0x00000001 (Monster Hunter World Online, BFV)
    0x809D5F60 (399 Profiles): 0x08000001 (Battlefield 1, 1, Sid Meier's Civilization VI)

    ----------------------------------------------------------------------------------------------------------------------

    AND THE CUSTOM BITS I SPECIFICALLY USED DURING THE 2+ HOUR SESSION OF CRASH FREE:

    UNDER 1 - COMPATIBLITY

    SLI BROADCAST BITS(DX1x): 0x00008000 RESPECT _BROADCAST_ZB_RESOLVE_ON_INACTIVE_GPUS - Do not discard sync on ZB if we try to resolve inactive GPUs in broadcast.

    SLI COMPATIBILITY BITS (DX10+DX11): 0x280B18F5

    SLI SPECIFIC HACK BITS (DX1x): 0x00000400 (F1 2014, F1 2013)

    UNDER 2 - SYNC & REFRESH:

    MAXIMUM PRE-RENDERED FRAMES: USE THE 3D APPLICATION SETTING

    UNDER 5 - COMMON:

    AMBIENT OCCLUSION SETTING: PERFORMANCE

    AMBIENT OCCLUSION USAGE: ENABLED

    MULTI-DISPLAY/MIXED-GPU ACCELERATION: SINGLE DISPLAY PERF. MODE

    UNDER 6 - SLI

    SET EVERYTHING TO: AUTOSELECT

    HEAD DOWN TO UNKNOWN PREDEFINED NVIDIA SETTINGS AND MAKE THESE CHANGES BELOW:

    0x105E2A1D (268 Profiles) 0x00000004 (Battlefield V, ME: Andromeda, THE CREW 2)
    0x80857A28 (964 Profiles) 0x00000001 (Monster Hunter Online, BF V, Race Driver Grid)

    ---------------------------------------------------------------------------------------------------------------

    And to wrap things up: FINAL CONCLUSION TO THIS GAME...

    During my 2+ hr stability test, everything seems to be crash free, longs your overclocks are stable, should have absolutely no problems.

    All the in-game cutscenes seems to be running smooth and scales like one bad motha _-_-_r!

    There are certain particular cut-scenes, during the initial part, there is a slight drag, after couple of seconds, scaling will go back to normal, so 0 worries there, it is these particular cut-scenes that if DOF is enabled, you will have major problems, so always have DOF off, that being said, you can had DOF tuned on while playing the game, the performance hit for exchange of blurfest ain't worth it, even though doesn't look to shabby, but the cost of performance = forget about it.

    Other then the scaling Drag issues in certain specific spots, mainly in cities and such, and they do drag down a bit, and there is quite a bit of them and the gap between point a to point b is quite frequent, but this ONLY pertains to 3-way and 4-way sli setup.

    Other then those drag issues, Everything is EXCELLENT...

    At the end of the day, Do I recommend playing this in 4-way SLI???

    YES

    Since the in-game cut-scenes plays an very important role of this games story, and since its basically smooth, Frametime seems decent, some chokes during cut-scenes in the beginning but it will quickly go away in few seconds, in the end, Yes, I recommend 4-Way SLI MAINLY because of this reason, and despite the kinda of annoying drag issues with 3-way and 4-way, still highly recommend it.

    If you try to ignore much as possible on those drag issues, it will be a breeze, there is some you will notice though.

    Another thing that amazed me about this game compared to most of the 4-way games I tuned, the 1% LOW and 0.1% LOW's are Excellent in most cases, one of the best I have seen. A+ there.

    Another thing, DO NOT record with shadow play @ 4k recording with this game under 4-way sli with the gtx 970's, doing so, when you hit these scaling drag spots, it will HIT IT HARD, sometimes it will hang the game for like 10 seconds then come back, mainly believe the vram is being way out, so try not to use shadow play while playing for best experience...

    RECOMMENDATIONS:

    2 WAY SLI TO 4-WAY SLI = YES..

    Despite the scaling drag problems under 3-way and 4-way sli modes and yes, I put ALOT of work of trying to get rid of it or lessen the effect = X.

    Perhaps in the future I will get rid of the drag but that will be some other time, time to move on to another game.

    One hell of an experience with this game, very interesting stuff here.

    2 thumbs up??? for the most part, yes.

    Far as everything is concerned, out of 10, under 4-Way Sli, I give it an 8, could have been a perfect 10 due to those scaling drag issues, sorry about that.

    Despite these problems, 2 thumbs up, mainly because of the in-game cut-scenes, this part saved it.

    ---------------------------------------------------------------------------------------------------------------------

    To sum all up with Goodies Included:

    GRAPHIC DRIVERS USED:

    MODDED QUADRO DRIVERS: 431.86(USED DURING LONG TERM TEST)/431.94

    LONG TERM STABILITY TEST/SPEEDS USED:

    1575MHZ @ 1.275V(ALL 4 GPU's/MEMORY @ 7604MHZ EFFECTIVE)

    PCI-E SPEEDS: ALL 4 @ X16 SPEEDS/GEN 3.0


    MAX IN-GAME SYSTEM/MAIN PHYSICAL RAM USED: OVER 20GB+

    ACTUAL MONITORING/TUNING(INCLUDING OLD TUNE)/PERFORMANCE TEST CHECK: OVER 50+ HRS.

    MOTHERBOARD BIOS USED: 4001

    INTEL CHIPSET DRIVER USED: 10.1.18010.8141

    STORAGE USED: SINGLE SAMSUNG 960 PRO(512MB) @ X4 PCI-E
     
    Last edited: Oct 25, 2019
    fantaskarsef likes this.
  19. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    UPDATES ON GAMES CURRENTLY GOING TO BE WORKING SOON:

    Venturi is looking for 2 games for me to scale and yet to tried them before, 2 games:

    ECHO

    &

    The vanishing of Ethan Carter.

    I will look into these very very soon but first, need to go back to NIOH, although 4-way sli is not needed as the game is capped at 60 fps and 3-way sli is more then enough, it scales like an Animal BUT the game overall stability isn't good, it crashes here and there, so now working on the issue.

    Once that's Done, I'll post results about NIOH then begin those 2 games Venturi wants me to look into, thanks for reading.

    Be sure to always check back for future updates later, thanks for reading.
     
    fantaskarsef likes this.
  20. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,055
    Likes Received:
    284
    GPU:
    4 GTX 970-Quad Sli
    ECHO - 4-WAY SLI - FIRST UPDATE - 10/24/19:

    First time on this game, been messing with the game for a lil over an hr.

    This is for Venturi and whoever is interested in this with similar setup...

    Couple things I have cleared up

    Before I begin, I used my OWN custom made Hell Blade SLI Compatibility bits(ONLY THE BITS and nothing else) to see if I can get some head start, but everything else was unchanged, including the Unknown PREDEFINED nvidia settings.

    So this is what I did thus far:

    1st, Massive Flickering, I got rid of it like in instant by changing the following Hex within:

    SLI Compatibility Bits (DX10+DX11): 0x00000<<~~ This one and Can be Set Either 2,8, or A, anything else will Flicker and From Initial observation, All three of performs roughly the same, so either one will work, This hex Also Implies to other games I have cleared up on the flickering part, so SLI Users, you might want to take a look at this if you are experiencing Massive Flickering from the get-go, try tuning the 0x00000<<~~, try starting with a 2, then perhaps 7 but in this game's case, 7 doesn't work, every game and setup will be different.

    2nd, another one that gave me a somewhat of a booster, this also helped me in remnant: Beyond the ashes(Both this game and Remnant = UE4 engine based):

    Under:

    SLI Specific Hack Bits (DX1x): 0x00<<~~ this one and MUST be set to either 2 or 8, others will make the Screen do some massive flickering showfest, even with:

    SLI Compatibility Bits (DX10+DX11: 0x00000<<~~ 2,8 or A

    Doesn't matter, it will flicker MAD crazy but in turn you will gain more performance but again, this is just initial start on my reaction of the game.

    Game will take some time to tune, seems like its running decent, A little less then 3-Way, My gpu Bus usage is running kinda high on 1st gpu, the others is still kinda high, so its giving me a warning signal that More tuning needs to be done, no exceptions.

    MS(frametime) seems decent, games seems to be stutter free, like before, I have NOT played it for long term, this will be way later.

    And just to let everyone know, I am currently tuning this game with everything set to 0.

    In other words, under GameUserSettings:

    Within the [ScabilityGroups], I have everything like this:

    sg.ResolutionQuality=inf
    sg.ViewDistanceQuality=0
    sg.AntiAliasingQuality=0
    sg.ShadowQuality=0
    sg.PostProcessQuality=0
    sg.TextureQuality=0
    sg.EffectsQuality=0
    sg.FoliageQuality=0

    To eliminate any Unusual scaling reaction or any kind of abnormal reaction from SLI in general, so this will clear things up in the beginning.

    I will be working on this and finishing up NIOH, so be sure to check back for future updates, right here, only on GURU3D:

    ---ECHO UNDER 4-WAY SLI IS NOW UNDERGOING TUNING PROCESS/WORK IS IN PROGRESS---
     
    Last edited: Oct 24, 2019

Share This Page