Extreme 4-Way Sli Tuning

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:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    NEW DRIVER TESTED - QUADRO 426.32 - 4-WAY SLI - 2/26/20:

    Tried some Modded Quadro Drivers today, and looking into the files, they seem to be about a month apart then the Quadro 431.98, so looks like these are newer.

    Gave this a Whirl and as expected, 4-Way Sli works 100% perfect.

    Tried Rise of Tomb Raider..

    And it seems I notice quite a bit of performance jump with these drivers...

    Right in the beginning of Syria, one of my saved game data, before my fps is usually in the upper 90's to like 101 give or take, now hitting like 106-109'ish.

    Also in the saved game glacier area, I notice performance jump as well.

    Have not played long term with these drivers but just enough to see 4-Way works and yes it does.

    Now I can confirm that the absolute latest drivers and probably the last to use under 4-way will be these drivers, so that being said, DO NOT go any further then these drivers, if you want, try the Quadro Drivers 431.98 or the vulkan Beta 436.65 drivers, do NOT go any higher, it will NOT WORK PERIOD, that also includes 44x.xx and 45x.xx as I personally tested them, I have not tried all of the 442.xx drivers but 100% suspect it does not work, so under 4-way, I would 100% personally stay away from all 44x.xx and 45x.xx until further notice.

    Thanks for Reading.
     
    Last edited: Feb 26, 2020
  2. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    PHOENIX POINT + X-COM 2(AT A GLANCE) - 4-WAY SLI - UPDATE - 2/26/20:

    Since I am really deep into Phoenix Point, and Overall Scaling on that game is kinda not looking to good in a way, Although some stuff scales extremely well, overall scaling as of now, I would say even then, 3-way sli tops.

    I am currently working on other issues in phoenix point, this only applies in Sli mode only, does not affect 1 gpu mode, when you press F(battle view), then go back to normal view, your fps drastically goes down, 1 gpu mode does not have this issue, furthermore, I also did that resolution trick that I discovered earlier, it does not work, HOWEVER, whats very interesting is this, my fps is increased like 7-8 fps during F key(Battle view), after the resolution trick.

    There is also some performance drop, in specific areas, soon as you move one of your units around, performance goes down a bit, did resolution trick, does not work, so currently working on the issue.

    During this one particular mission, battle view mode(F key), 1 gpu mode, I get around 11 fps, now I am getting like 41-42 fps(48-49 fps/resolution trick/4 GPU's).

    So yes 4-way is working in some instances, overall = Eh, but still hard at work.

    Now that being said, I wanted to give X-Com 2 a try since very similar games, one being unreal engine 3.5/DX11 API, I figured I would have higher success of chance of scaling, Yup, from initial glance, it seems overall scaling in general, I can get away with this game ALOT easier then Phoenix Point.

    In X-Com 2, @ 4k(2400p), some stuff needs to be shut off like Screen Space Reflections, Shadow to its bottom, otherwise, you will experience Very Jerky camera movement, despite MS(Frametime) seems fine.

    Also, I have not done any hardcore tuning within the Config File, as I am leaving it alone for now, Wanted to give this game a whirl and see how it would look, initial glance, I can tell you all right now, I will have a much better overall success then Phoenix Point, by a long shot.

    Closing in about 20+ hrs into phoenix point.

    Those who are tuning this game and having the performance issue where you change your camera to battle view(F key) then back to normal, two things I discovered that will help it by a slight bit, I am no where near being done.

    First, Sli Broadcast bits (dx1x):

    0x0000000<<~~ this one, mine is set to 3

    &

    Sli Specific Hacks:

    0x0<<~~ This one as well, and its set to 5

    Normally that particular hex within the Sli Specific Hacks, does help boost performance quite a bit in some games, But usually the Digit 9 or Letter A, I never use 5, so it can accommodate other games as well, yes it does help in some games, not all.

    Final Conclusion to Phoenix Point, I am hoping before end of this week, so be sure to look for it, regardless of the outcome of the performance, I will post results, thank you..

    QUICK UPDATE ON TUNING DIFFICULTY ON PHOENIX POINT:

    One thing I would like to point out, the tuning difficulty on this game, the level:

    Very advanced level.

    I would strongly recommend doing X-com 2 first before this.

    Much much easier one is like Ghost in a tale, Rise of Tomb Raider, Hitman Absolution, Dead Space,Project cars 2 and others.

    I would NOT touch this game if your getting into 4-way at first, I would Stay far away from this game until you gather much more experience on tuning.

    Game is very advanced, So start things slowly.

    There is tons of stuff that you have to deal with while tuning under 4-way with phoenix point, Game is absolutely no joke.


     
    Last edited: Feb 27, 2020
  3. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    PHOENIX POINT - GTX 970 - 4-WAY SLI - CURRENT STATS THUS FAR - 3/2/20:

    Current Stats on this game:

    I was hoping end of last week(Very very unlikely), Now things are looking for sometime tommorrow(Hopefully).

    I am still going through both of the SLI Broadcast and Specific Hack Bits, however, last couple days, I went through Massive amounts of stuff within the Unknown ID settings within nvidia inspector, Have to scroll all the way down to Unknown nvidia settings, I have made some improvements, not awhole lot, but yes it does make some difference, I basically went through everything, certain things I did not touch cause no need to, but went through Masssive amounts of tuning within the unknown id settings...

    Now currently finishing up on tweaking Both Sli Broadcast and Specific hack bits and do my full performance report...

    For the final conclusion for this test, I will be conducting ALL 4 GPU Tests:

    Single gpu mode
    2-Way Sli
    3-Way Sli
    And lastly 4-way Sli.

    All along with: Extensive MS(Frametime) Results on all 4 GPU Test stated above, Don't usually Do MS(Frametime) during final conclusion but will this time for out of people curiosity sake.

    Also Temp Results as well, I can tell you all, The most HARD pounding on the scaling usage is when your viewing the globe, that hits incredibly hard, so my temp will be during that portion.

    Test will be conducted in 4-5 different areas, each has its own strength's and weakness in scaling.

    Far as overall scaling thus far???

    I will mention everything in the final conclusion, very interesting game we have here everyone.

    Got the next few days off from work so I will be very hard at work on this game, be sure to look for this game, I am hoping by tomorrow's time frame, way later in the day...

    A single longplay video(NOT MINE):



    SEE YOU ALL IN THE FINAL CONCLUSION...
     
    Last edited: Mar 2, 2020
    fantaskarsef likes this.
  4. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    PHOENIX POINT - GTX 970 - 4-WAY SLI - PRELUDE TO FINAL CONCLUSION UPDATE - 3/4/20:

    Alright folks, Sorry for the long A$$ delay, this isn't the real final conclusion, but a prelude, I am almost just about done jotting down TONS of performance readings however Certain things just came up, and a damning one too to boot...

    Game just very recently had a patch today, and it seems they fixed the issue that I found where using one gpu mode, while pushing the F Key(Battleview), then going back to normal screen, the normal screen will have significant impact on performance, some areas didn't had this issue before but apparently after Long investigation on this issue, I found out that it was shooting of any weapon to anything such as crates, enemies, walls, etc, right after you do that and go back to normal mode, fps goes down big time, it seems they fixed the issue, I did long time check on this issue, From TOD(Time of day/weather), Shader quality, to everything imaginable, I did all the checks all on my own time.

    Now here comes the Sli part, while messing under 2-way mode, I found out that the fps drag during the above mention situation is still there but I fixed the issue.

    For instance, lets say in Sli mode, in normal view, 46 fps, then all the sudden you push the F Key(Battleview) mode, then start shooting at a crate or an enemy, then go back, you will see the FPS went from 46 to a whopping 30 fps or under, and I have fixed this issue...

    You see, its all coming from this particular bits:

    Sli Compatibility (DX10+DX11) Bits:

    0x00000<<~~, I normally have this set to 2 as certain things if put to 1, there is some slight graphical anomalies..

    Now its set to 1 instead of 2, I had a hunch that this was the case, but didn't came around to it until very recently and my hunch was correct...

    After changing this particular hex to 1, there seems to be no reduction in performance during the above mention situation in 2-way Sli mode, during 3-way Sli mode, there is very very very little reduction in performance, however the Big daddy performance degrader is the 4-way, I probably lost almost 20 fps, that is HUGE...

    4-Way performance is actually decent in certain situations, when you do certain things and come back to what you were doing, certain time, your fps will go down, be very strongly aware of this issue.

    So those who are tuning this game, you all might want to look into this:

    0x00000<<~~ switch to one.

    Also, My Sli rendering mode of choice is: Autoselect(AFR), and not SFR(lower performance in certain locations).

    So all I needed to do is change Sli Compatibility bits to:

    0x0<<~~ This one, either 6 or 7 will work, 8 is for 4-way SLI under SFR MODE ONLY, C will work with SFR too, but for best overall performance, keep it to 6 or 7 with SLI rendering mode: AutoSelect.

    Pretty much my performance readings is almost complete, just want to jot down one particular area, Try to look for the final conclusion tonight, I am hoping before the days end, again, sorry for the delay folks.

    Also, I am closing in about 40hrs or so of tuning in this game, This game has been complete Hell to tune, Very very very stubborn game/Highly advanced game overall, man your not kiddin', one of the most challenging one to date, BAR-NONE!!!

    In the final conclusion, there will be TONS and I mean TONS things to cover, you all don't want to miss this one.

    Thanks for reading.

    UPDATE ON PROGRESS THUS FAR - 3/4/20:

    Just want to give another heads up again, lol, it came back again, that decrease in performance after pressing the f key, it is no where near as pronounced before the patch but apparently upon longer play, it has came back, this is under 1 gpu mode, went from 23 fps to 15 fps, wham, not to good, and yes. I even lowered everything to the lowest setting, even the resolution as well, it does not matter, Again, I emailed to snapshot games about the issue, yes, they are very much aware of the issue now.

    Doesn't matter through, There is one more test that I need to conduct before final conclusion is written, And regardless of the problems of this game, as promised eariler, I will conduct full performance report in the areas I have done. I am doing this one area, everything, from camera movement, battleview, everything, doing all performance monitoring, including the MS(frametime), once that's done, which won't take to long, I will work on the final conclusion, so stick around later.

    As I kinda expected, there is no way in the world I would have gotten done earlier, had to do some stuff, Last thing I did was 1hr stability and Temp check Under Absolute Brutal Load on all 4 gpu's, basically all 4 gpu's hanging on 100% usage, and came out excellent, full report on it soon.

    It won't be tonight, sorry about that, check back tomorrow, I'll do my very best to have everything up, but not tonight, no way in the hell I can get it done now.

    Thanks for reading.
     
    Last edited: Mar 5, 2020

  5. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    PHOENIX POINT - GTX 970 - 4-WAY SLI - FINAL CONCLUSION #1 - 3/5/20:

    Here it is, the moment we "ALL" been waiting for, Phoenix Point under 4-Way Sli...

    Before I start, some might want to listen to the ost while reading:



    This game overall has been probably one of the hardest game to tune BAR-NONE.

    There is So much to cover for this game, first lets start off of certain things that people who's playing this game, MUST be aware of at all times.

    As mentioned before, this applies to 1 GPU mode as well:

    While in a specific Battle Zone, when you press the F Key(Battleview), this will shift into first person mode type, soon as start firing at crates, posts, enemies and such, there might be a chance that your FPS will drop by quite a margin upon exiting the Battleview mode.

    Under close observation, I was really investigating what was actually the root of cause of all of this, I even lowered my resolution so low and all my graphic settings to bare min, I even went to disable Sli mode in general within NVCP, and ran 1 gpu mode ONLY and made main primary Physx is being ran on the primary gpu, just to see what the reaction would be, nope, still same thing.

    So Developers over at Snapshot game is very aware of this issue as we speak.

    While tuning this game, there has been Lots of problems.

    First of all, in regards about that decrease in performance while firing your weapon under Battleview then going back to default view, under closer inspection, Under 3-way and higher, In certain areas, not all, but some areas, just moving your units, one spot to another, no F-Key or anything, the the fps goes down.

    Lets say under 3-way mode, under default view in a specific zone, fps is 60, the moment you move on of your units around, I have seen it went down to upper 40's, again, the drop is significant, but certain areas far's I can tell, impacts different from one another, some areas don't effect it at all, it may take awhile to kick in, certain areas it will kick in like an instant.

    I also notice that under 3-Way and higher, the moment you push F-Key, in certain areas, there is some huge spike in MS(Frametime), so my min while doing bench recording is totally reduced, some stuff I didn't jot down cause the impact was so huge. This game has been very very very very crazy to deal with.

    Another thing that must be addressed here is the Resolution trick, I have discussed this before but this is just a re-cap:

    While tuning this game, Even at same resolution, and changing some settings, I have notice decrease in performance, let me explain here:

    Lets say under 4k, with custom settings, and your getting 25-27 fps, then you are enabling another setting that you want to tune, and it drops to 22 fps, then you want to revert your settings back to where you had it before, guess what??? Its stuck to 22, it does NOT go back to its original peak performance, this is while under 1 gpu mode, and this also pertain to Sli mode in general.

    There is a fix to this but it must be done correctly, otherwise, if you just click the esc key and click restart, the fps reduction will still be there, here is the work around:

    If and when you are experiencing this while under 1 gpu mode or tuning under Sli mode, first lower your resolution to a huge margin, if your base resolution is 2400p(4k), switch it to piss poor resolution like 1600x900, then Exit the game, do NOT go back to the main screen and switch your resolution there and the load the game, the performance reduction will STILL be present, you MUST, repeat, MUST exit out the game while this piss poor resolution is being used, Exit the game entirely then re-launch the game, go to options, under graphic settings, switch it to your desired resolution, and then load up your game, to bring your gpu potential back to its peakest, Absolutely no exceptions period here, so keep this mind at ALL times.

    And Under Sli Mode, I would strongly recommend, while under default view, do NOT zoom out, screws up the overall scaling, and keep the view to default, this ensures the peakest scaling in general and speaking about scaling usage in general, while I was tuning this game, the usage is very irrelevant, Always get a base fps from 1 gpu then work from there, the scaling usage while tuning in general is useless, you can have good scaling and still performs no more then 2 cards out of the four, so never go by that usage under Sli tuning.

    Another thing while under Sli mode, only certain areas, not all:

    While Moving your target to one of your units in first person(Battleview mode) and highlighting your units, the overall scaling goes Absolute haywire for a split second, very abnormal reaction, again, Thought the TOD(time of day) could be the cause, does not seem so, some sections you won't experience the crazy drag, some places you will.

    Another thing, FPS CAP, I worked around the issue, I read other forums where changing the graphic preset to high or very high fixes this, no it does not, And Vsync on(Bug), and yes it does but at first, it does not but there is a work around this issue.

    First if your using an OSD like Rivatuner Statics Server, make sure your Frame Rate Limiter is turned off under setup, then within nvidia inspector program, MAKE sure Vsync is switched to FORCED OFF, and in the game, MAKE sure its switch to on, I know this does not make any sense whatsoever but I have confirmed it, even with the recent patch that just came out...

    There seems to be a Vsync issue, so to uncap it, please use the above method.

    Several other things that will make the fps cap while using the above method and they are:

    Under 5 - Common:

    Multi-Display/Mixed Gpu Acceleration: MUST BE SET to single display mode/Compatibility mode, do NOT set it to 0x00000006(Quadro Gpu's) for this game, it will make the fps capped, Laser locked.

    And lastly, while tuning this game, another thing that will make this game capped(FPS) is:

    Under 8 - extra:

    ASYNC10_NVAPI_MODE: 0x00000002 ALLOW_NOBLOCKING_SLISTATE - Allow specific non-blocking optimizations for GetCurrentSliState NVAPI.

    As stated before, there was just so much to cover in this game, Absolute craziness to the MAX...

    Ok enough of that, lets fast foward the stuff that everyone has been waiting for, PERFORMANCE RESULTS..

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

    All recorded tests were done @ 4095x2559(160% Over native(1600p) resolution, as I did this for a good reason, to add more load to the gpu's over 2400p, at times 2400p wasn't scaling top notch, anything higher then 2559p, could experience some problems, so I stuck with this resolution.

    Under graphic settings, Graphic quality control slider is @ 62%, texture to medium, Shaders set to high, particle effects to high, and all other check boxes marked checked EXCEPT for Screen space reflections(cause major problems in some areas).

    Areas covered during these tests:

    Tutorial mode(Section 1), Dreamers awaken(Fort Heinlein),Fire Within(Arinna),Enter synderion(Stage Left-Nowhere),GLOBE And a scavenge area(1 gpu and 2-way sli only).

    Another important note on synderion(Stage Left-Nowhere):

    In Battleview(F-key), while viewing certain areas while zoomed out, the MS(Frametime) and scaling goes absolutely haywire, so during this particular test, I added all of the zooming and out.

    I covered some Static/Idle locations/movement patterns,etc

    MS = FRAMETIME

    Rendering mode(Autoselect/AFR)

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

    Tutorial Mode(Section 1), Static/Idle/One position:

    1 GPU: 22-23 FPS/40-58 ms

    2 GPU's: 44-48 FPS/15.6-32.5 ms

    3 GPU's: 68-73 FPS/10.7-32 ms

    4 GPU's: 70-73 FPS/10.1-16.7 ms

    Tutorial Mode(Section 1)/Battleview(F-Key)/While Targeting Enemy/Stationary Position:

    1 GPU: 26-28 FPS/21.4-47.9 ms

    2 GPU's: 49 FPS/18-24.7 ms

    3 GPU's: 37-38(Occasional Dips to 31 fps)/24.3-29.8 ms

    4 GPU's: 43-45 FPS/20.7-25.3 ms

    Tutorial Mode(Section 1)/Movement Pattern(Default view)/No zooming in or out/Results:

    1 GPU:

    MIN FPS: 13//AVG FPS: 21//MAX FPS: 25//1% LOW: 13 FPS//.1% LOW: 6//38.3 ms

    2 GPU's:

    MIN FPS: 26//AVG FPS: 40//MAX FPS: 47//1% LOW: 24 FPS//.1% LOW: 10 FPS//18.1 ms

    3 GPU's:

    MIN FPS: 40//AVG FPS: 60//MAX FPS: 69//1% LOW: 33//.1% LOW: 7 FPS//11-38+ ms

    4 GPU's:

    MIN FPS: 43//AVG FPS: 55//MAX FPS: 66//1% LOW: 20//.1% LOW: 8 FPS//10-25.1 ms

    Globe View/Zooming in and out/Movement pattern/Results:

    1 GPU:

    MIN FPS: 28//AVG FPS: 29//MAX FPS: 29//1% LOW: 25 FPS//.1% LOW: 25 FPS//27.6 ms

    2 GPU's:

    MIN FPS: 47//AVG FPS: 55//MAX FPS: 61//1% LOW: 46 FPS//.1% LOW: 15 FPS//16.3-21.5 ms

    3 GPU's:

    MIN FPS: 63//AVG FPS: 76//MAX FPS: 84//1% LOW: 43 FPS//.1% LOW: 11 FPS//7-21.5 ms

    4 GPU's:

    MIN FPS: 76//AVG FPS: 93//MAX FPS: 104//1% LOW: 39 FPS//.1% LOW: 9 FPS//9.7-25.4 ms

    ADDED BONUS - 4 GPU's(SFR/SPLIT-FRAME-RENDERING):

    MIN FPS: 79//AVG FPS: 95//MAX FPS: 104//1% LOW: 56 FPS//.1% LOW: 9 FPS//5-14.9 ms

    FIRE WITHIN-ARINNA:

    MOVEMENT PATTERN/DEFAULT VIEW/NO ZOOMING IN OR OUT:

    1 GPU:

    MIN FPS: 20//AVG FPS: 23//MAX FPS: 26//1% LOW: 19 FPS//.1% LOW: 7 FPS//38.7-51 ms

    2 GPU's:

    MIN FPS: 38//AVG FPS: 44//MAX FPS: 48//1% LOW: 31 FPS//.1% LOW: 6 FPS//21-57 ms

    3 GPU's:

    MIN FPS: 43//AVG FPS: 59//MAX FPS: 71//1% LOW: 35 FPS//.1% LOW: 7 FPS//10.8-28.3 ms

    4 GPU's:

    MIN FPS: 36//AVG FPS: 65//MAX FPS: 89//1% LOW: 35 FPS//.1% LOW: 5 FPS//8.1-30+ ms

    FIRE WITHIN-ARINNA(F-KEY/BATTLEVIEW)ZOOMING IN AND OUT/ALL MOVEMENTS:

    1 GPU:

    MIN FPS: 23//AVG FPS: 28//MAX FPS: 31//1% LOW: 21 FPS//.1% LOW: 7 FPS//31-40+ ms

    2 GPU's:

    MIN FPS: 34//AVG FPS: 49//MAX FPS: 59//1% LOW: 35 FPS//.1% LOW: 7 FPS//17.5-25.8 ms

    3 GPU's:

    MIN FPS: 38//AVG FPS: 59//MAX FPS: 88//1% LOW: 37 FPS//.1% LOW: 9 FPS//11-25+ ms

    4 GPU's:

    MIN FPS: 40//AVG FPS: 74//MAX FPS: 102//1% LOW: 37 FPS//.1% LOW: 8 FPS//6.9-21+ ms

    DREAMERS AWAKEN-FORT HEINLEIN:

    MOVEMENT PATTERN(NO ZOOMING IN OR OUT)/DEFAULT POSTION/RESULTS:

    1 GPU:

    MIN FPS: 21//AVG FPS: 24//MAX FPS 27//1% LOW: 7 FPS//.1% LOW: 5 FPS//36.5-51+ ms

    2 GPU's:

    MIN FPS: 28//AVG FPS: 36//MAX FPS: 41//1% LOW: 20 FPS//.1% LOW: 7 FPS//12-49.8 ms

    3 GPU's:

    MIN FPS: 41//AVG FPS: 63// MAX FPS: 78//1% LOW: 17 FPS//.1% LOW: 7 FPS//6.6-27.1 ms

    4 GPU's:

    MIN FPS: 45//AVG FPS: 72//MAX FPS: 94//1% LOW: 18//.1% LOW: 6 FPS//9.3-21+ ms

    DREAMERS AWAKEN-FORT HEINLEIN(F-KEY/BATTLEVIEW)/ALL MOVEMENTS+ZOOM:

    1 GPU:

    MIN FPS: 10//AVG FPS: 26//MAX FPS: 41//1.% LOW: 10//.1% LOW: 6 FPS//29.8-105 ms

    2 GPU's:

    MIN FPS: 18 FPS//AVG FPS: 47//MAX FPS: 77//1% LOW: 18 FPS//.1% LOW 10 FPS//12-57+ ms

    3 GPU's:

    MIN FPS: 27 FPS//AVG FPS: 72//MAX FPS: 102//1% LOW: 26 FPS//.1% LOW: 8 FPS//8.5-46.3 ms

    4 GPU's:

    MIN FPS: 35//AVG FPS: 83//MAX FPS: 107//1% LOW: 34 FPS//.1% LOW: 6 FPS//7.7-36 ms

    ENTER SYNDERION(STAGE LEFT-NOWHERE):

    As a remainder about this area, while your Battleview(F-Key), while zoomed out, The MS and overall scaling is Very bad, things change drastically after moving in, so during the F-Key test, All movement patterns/Zooming and out is conducted...

    ENTER SYNDERION STAGE LEFT-NOWHERE)/MOVEMENT PATTERN/DEFAULT VIEW, NO ZOOM:

    1 GPU:

    MIN FPS: 21//AVG FPS 24//MAX FPS: 27//1% LOW: 19 FPS/.1% LOW: 5 FPS//37.2-56 ms

    2 GPU's:

    MIN FPS: 29//AVG FPS: 46//MAX FPS: 52//1% LOW: 9 FPS//.1% LOW: 6 FPS//16-30+ ms

    3 GPU's:

    MIN FPS: 41//AVG FPS: 61//MAX FPS: 77//1% LOW: 10 FPS//.1% LOW: 6 FPS//7-25.8 ms

    4 GPU's:

    MIN FPS: 43//AVG FPS: 66//MAX FPS: 92//1% LOW: 7 FPS//.1% LOW: 6 FPS//10-21+ ms

    ENTER SYNDERION STAGE LEFT-NOWHERE)/F-KEY(BATTLEVIEW)/ALL MOVEMENT+ZOOM:

    1 GPU:

    MIN FPS: 21//AVG FPS: 26//MAX FPS: 34//1% LOW: 19 FPS//.1% LOW: 7 FPS//31.7-51+ ms

    2 GPU's:

    MIN FPS: 16//AVG FPS: 46//MAX FPS: 73//1% LOW: 10 FPS//.1% LOW: 7 FPS//10.9-101 ms

    3 GPU's:

    MIN FPS: 19//AVG FPS: 65//MAX FPS: 92://1% LOW: 10 FPS//.1% LOW: 7 FPS//10.9-101 ms

    4 GPU's:

    MIN FPS: 20//AVG FPS: 72//MAX FPS: 106//1% LOW: 21 FPS//.1% LOW: 8 FPS//7-25.8 ms


    SCAVENGING AREA(ONLY 1 AND 2-WAY SLI BEING USED/3&4 WAY = TO MANY PROBLEMS/JUST MOVING ONE UNIT/DECREASE OVERALL PERFORMANCE FROM 3 WAY AND UP).

    1 GPU:

    MIN FPS: 11//AVG FPS: 19//MAX FPS: 43//1% LOW: 11//.1% LOW: 5 FPS//12.1-125 ms

    2 GPU's:

    MIN FPS: 19//AVG FPS: 32//MAX FPS: 81//1% LOW: 11 FPS//.1% LOW: 4 FPS//11.9-42.3 ms

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

    This section covers 4-way under Absolute Extreme/Brutal conditions and temp check, Test was conducted during the Globe area, pretty much all gpu's were hanging around 100% usage it all times, except gpu 3(Common on all games), despite gpu 3 shows more power % utilization being used.

    About good 1 hour was conducted...

    Low 70 degrees fahrenheit Ambient Room temps/Windows Shut/Lights Off & Door to the room Wide open, Results:

    GPU #1: 61c
    GPU #2: 71-72c
    GPU #3: 72-74c
    GPU #4: 68c


    All temps were all well under control during the 1 hr brutal extreme load test, with no visible signs of graphic anomalies of any kind, no crash/bsod's or hangs, everything went excellent.

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

    Please read the next post, as it will cover the Bits themselves.

    I will do that later, please check for it.

    Also, I will be using other graphic drivers to see if certain problems in this game is changed, if any significant happens, it will be posted here.
     
    Last edited: Mar 6, 2020
  6. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    PHOENIX POINT - GTX 970 - 4-WAY SLI - FINAL CONCLUSION #2 - 3/6/20:

    If you all have not read final conclusion #1, please do so by scrolling up to the very last post, please read that through before continuing on, thank you.

    Now here comes the custom made Bits, here they are BUT FIRST:

    WARNING ABOUT MY CUSTOM MADE BITS:

    TESTED AND 100% GUARANTEED TO WORK AS HOW I DESCRIBED ON FINAL CONCLUSION #1, ONLY, REPEAT, ONLY WITH NVIDIA GTX 970 FAMILY, IT HAS NOT BEEN TESTED ON GTX 980 OR 980TI, SO PLEASE BE ABSOLUTELY AWARE OF THIS BEFORE YOU CONTINUE, ALSO ON TOP OF THAT, ONLY GUARANTEED TO WORK WITH WIN 10 X64(15063-V1703/RS2(RTM), FAILURE TO MEET THESE TWO REQUIREMENTS, I WILL GUARANTEE 0%, SO PLEASE MAKE SURE EVERYTHING IS READY TO GO BEFORE YOU CONTINUE.

    Since there is so much custom stuff in this game, I made it alot easier instead of typing all of it, and yes, there is tons of stuff.

    To make it easier on myself and to everyone, first, grab a copy of this nvidia inspector program:

    https://github.com/DeadManWalkingTO/NVidiaProfileInspectorDmW/releases/tag/v3.5.0.0

    This is what I always use, and it works perfect for what I need to do, please use this if you can, thank you.

    Next, open up the nvidia inspector profile program, on top of the program, move your cursor to export user defined settings, then on the drop down list, select the very last one, should say:

    Export all drive profiles (NVIDIA TEXT FORMAT)

    Next, open up the saved document with something like notepad, then Find the following:

    "Phoenix Point"

    Then you will come across the first line, should say:

    Profile "Phoenix Point"

    Underneath that, Delete all of it, and copy and paste this below, Make sure you do proper spacing between each line after the line: Profile "Phoenix Point", 4 spaces to be exact, The end Profile line does NOT need any spacing and here it is:

    ShowOn GeForce
    ProfileType Application
    Executable "phoenixpointwin64.exe" UserSpecified=true
    Executable "phoenixpoint.exe"
    Executable "unitycrashhandler64.exe" UserSpecified=true
    Setting ID_0x00110999 = 0x00001000 UserSpecified=true
    Setting ID_0x00120513 = 0x00000000 UserSpecified=true
    Setting ID_0x001208cb = 0x00000001 UserSpecified=true
    Setting ID_0x0017fd6d = 0x00000001 UserSpecified=true
    Setting ID_0x002c7f45 = 0x0000001e UserSpecified=true
    Setting ID_0x002ecaf2 = 0x00000001 UserSpecified=true
    Setting ID_0x0043ed70 = 0x00000003 UserSpecified=true
    Setting ID_0x005a375c = 0x99941284 UserSpecified=true
    Setting ID_0x005f5a32 = 0x00000000 UserSpecified=true
    Setting ID_0x0064b541 = 0x00000001 UserSpecified=true
    Setting ID_0x00664339 = 0x00000001 UserSpecified=true
    Setting ID_0x00667329 = 0x00000001 UserSpecified=true
    Setting ID_0x00738e8f = 0xfffffffb UserSpecified=true
    Setting ID_0x0078d9d9 = 0x00000000 UserSpecified=true
    Setting ID_0x007ba09e = 0x00000000 UserSpecified=true
    Setting ID_0x00854cab = 0x00000038 UserSpecified=true
    Setting ID_0x0085f725 = 0x25558997 UserSpecified=true
    Setting ID_0x008d624b = 0x00000000 UserSpecified=true
    Setting ID_0x008d624e = 0x00000001 UserSpecified=true
    Setting ID_0x00939833 = 0x00000001 UserSpecified=true
    Setting ID_0x0094c537 = 0x00b03007 UserSpecified=true
    Setting ID_0x0094c538 = 0x00000000 UserSpecified=true
    Setting ID_0x009cfec8 = 0x00000004 UserSpecified=true
    Setting ID_0x00a06746 = 0x50000009 UserSpecified=true
    Setting ID_0x00a06946 = 0x000000f5
    Setting ID_0x00a06946 = 0x6aba11f5 UserSpecified=true
    Setting ID_0x00a0694b = 0x0800083d UserSpecified=true
    Setting ID_0x00a25fc4 = 0x20000001 UserSpecified=true
    Setting ID_0x00a879cf = 0x08416747 UserSpecified=true
    Setting ID_0x00a89790 = 0x0000000a UserSpecified=true
    Setting ID_0x00d98f93 = 0x00000010 UserSpecified=true
    Setting ID_0x00dc68b0 = 0x00000001 UserSpecified=true
    Setting ID_0x00e32f8a = 0x00000000 UserSpecified=true
    Setting ID_0x10115c8b = 0x00000028 UserSpecified=true
    Setting ID_0x1033cec2 = 0x00000002
    Setting ID_0x1033ced1 = 0x00000000 UserSpecified=true
    Setting ID_0x1033dcd1 = 0x00000004 UserSpecified=true
    Setting ID_0x1033dcd2 = 0x00000000 UserSpecified=true
    Setting ID_0x1033dcd3 = 0x00000004
    Setting ID_0x103c2e03 = 0x60606064 UserSpecified=true
    Setting ID_0x105e2a1d = 0x00000000 UserSpecified=true
    Setting ID_0x107afc5b = 0x00000001 UserSpecified=true
    Setting ID_0x10834fee = 0x00000000 UserSpecified=true
    Setting ID_0x1094f157 = 0x00000000 UserSpecified=true
    Setting ID_0x1094f1f7 = 0x00000000 UserSpecified=true
    Setting ID_0x10a879ac = 0x00000001 UserSpecified=true
    Setting ID_0x10a879cf = 0x00000001 UserSpecified=true
    Setting ID_0x10aaa36c = 0x00000001 UserSpecified=true
    Setting ID_0x10c158ad = 0x00000000 UserSpecified=true
    Setting ID_0x10fc2d9c = 0x00000004 UserSpecified=true
    Setting ID_0x1194f158 = 0x00000000 UserSpecified=true
    Setting ID_0x200aebfc = 0x00000001 UserSpecified=true
    Setting ID_0x2089bf6c = 0x00000023 UserSpecified=true
    Setting ID_0x50166c5e = 0x00000001 UserSpecified=true
    Setting ID_0x7096eced = 0x00000000 UserSpecified=true
    Setting ID_0x709adada = 0x20000006 UserSpecified=true
    Setting ID_0x709adadb = 0x3f5e0e2d UserSpecified=true
    Setting ID_0x809d5f60 = 0x00000000 UserSpecified=true
    EndProfile

    Save the Document, then Import the saved document with your nvidia inspector, on top of the program, move cursor to:

    Import user defined settings,

    Select the 2nd item, should say:

    Import (REPLACE) all driver profiles (NVIDIA TEXT FORMAT)

    Everything should be there, including ALL the custom made bits, 8 - Extra, and ALL of the unknown or unknown PREDEFINED nvidia settings.

    People with 2-way or 3-way sli, Just change the portion where it says FOUR to its respected setting, thanks.

    FINAL CONCLUSION:

    Far as everything goes, I have put over 40+ hrs worth of effort in this game, Although there is still quite a bit of problems, even under single gpu mode, 4-way does scale pretty well overall, depending on what you are doing in the game, but if you were going to play with all seriousness and long term play, I would very strongly recommend doing 2-Way Sli over others, for couple of reasons here:

    As mention before, certain locations under 4-way sli, just moving your units, just one time, I notice your fps will be reduced quite a bit, 2-Way Sli doesn't seem to have this issue, all though there is still problems with 1 gpu mode & 2-Way Sli, but it seems under Sli mode, 2-way seems to handle it ALOT better then 3-way and 4-way.

    Or just shifting into battleview(F-Key), then going back to normal view, you will notice decreased fps overall under 3-way and 4-way, it is Way more pronounced in those modes, but not in all locations, these situations apparently only comes up in certain locations.

    Another thing, when firing your arsenal while under battleview then goes back to normal view, you will also notice decreased performance, again, 3-way and 4-way is ALOT more pronounced, single and 2-way, no where as bad.

    Another example, in the area called:

    Enter Synderion(STAGE LEFT-NOWHERE),

    While moving your camera in the normal view, 1 gpu and 2 gpu mode, the transition is alot more smoother then 3-way and 4-way, in this precise area, 3-way and 4-way while moving your camera, is ALOT more sluggish, especially while camera is zoomed all the way out, it is real bad, speaking about this area and that goes for all areas, I would not do any zooming out whatsoever, screws up the scaling, if you leave it at default position, should have no problems.

    I mean even under 1 gpu and 2-way sli mode, you can see decreased performance after firing your arsenal in battleview then coming back to normal view, doesn't happen to much but it could happen at any time, and when it does, 3-way and 4-way will do justice here, but still, game still needs work and the developers know this.

    So for just these above reason alone and yes they are pretty serious, if going to really play the game long term, I would highly advise doing 2-way, Despite still problems, some stuff 3 way and 4-way sli scales pretty good.

    So to top it all off:

    4-WAY SLI: I would only run this mode under tuning purposes only and or short term game play, despite some areas and situation, scales extremely well, you will run into problems with 1 gpu mode or 2-way sli mode, like decreased performance, this is where 3-way and 4-way will do justice, but in its decreased performance, game has lots of problems, even under 1 gpu, game is still kinda new so game is still being patched up. certain things, like I mention before, will really push it back, so until further notice, at this current stage, I would not recommend it at this time, so I will continue to work as time progress, but at this stage, I would say no, best to stay with 2-way sli cause of some situations is no where near as bad as 3-way or 4-way, you can give it a try however, its really a mixed bag situation, but again, for best overall fps stability, 2-way sli.

    Almost a year has gone by ever since I started my 4-Way Sli Project, and I can tell you all right now, this game is one of the Hardest one to tune, there is just so much problems to deal and work around with, Tuning level is most defin't at the top of the ball game, In other words: Most challenging.

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

    GRAPHIC DRIVERS USED: ALL THESE ARE MODIFIED:

    NVIDIA QUADRO DRIVERS: VERSION 431.98/426.32

    GPU CORE/MEMORY TESTED: 1 TO 4-WAY SLI MODE = 1443MHZ(CORE)/7600+(MEMORY)

    SYSTEM RAM USAGE: HIGH AS 20 GB+

    OS USED: WIN 10 x64 (15063.2108/VERSION: 1703/RS2/RTM)

    CPU UTILITY PROGRAM USED: THROTTLESTOP

    STORAGE USED: 2 SAMSUNG 950 PRO(512GB) IN STRIPPED(RAID 0) MODE.

    MOTHERBOARD USED: SUPERMICRO X10drg-q

    MOTHERBOARD BIOS USED: 3.1

    INTEL CHIPSET DRIVER USED: 10.1.18010.8141+10.1.18263.8193 update

    TOTAL TUNING/PERFORMANCE MONITORING/TESTING TIME/OVERALL STABILITY CHECKS: OVER 40+ HRS EASY.


    See you all on Technomancer & COD: Infinite Warefare, thanks for reading.

     
    Last edited: Mar 7, 2020
  7. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    UPDATE ON NEW DRIVERS BEING TESTED + PHOENIX POINT - GTX 970 - 4-WAY SLI - UPDATE - 3/6/20:

    Another heads up, I personally Tested Quadro Drivers: 432.28, Which came out last month, look at the files, dated in Jan of 2020, so these are much newer then I am using before, and yes, as what I 100% expected, due to the 43x.xx branch, these works perfectly under 4-Way Sli mode, but I only tested with my personal game of choice as if it works with this game, pretty much it will work all others, Rise of tomb raider, played for bit, Scales EXCELLENT!!!

    Far as Phoenix Point, didn't mention about these drivers with this game earlier, but the performance decrease under 4-Way Sli upon moving one of your units right off the back in certain specific areas, its still there, so be aware of it with these drivers.

    So as its stands, Use only branch 41x.xx, 42x.xx, or 43x.xx, do NOT use any of: 44x.xx or 45x.xx, as they DO NOT WORK, PERIOD...

    Also, as with older games like echo, Branch 39x.xx works great too, thanks for reading.
     
  8. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE + NEXT TIER CARD - UPDATE - 3/8/20:

    First lets go with COD: Infinite warfare...

    Earlier on, I re-tested this game after for so long, wanted to see now how it would scale, like before, an absolute beast.

    And now back on the game with full force.

    Several things here that I need to mention:

    Kinda off topic but, back in the days while testing this game @ 8k(4800) with my 1080ti, I notice that in the actual texture themselves, the actual detail of the resolution can get significantly worse when you go further in a particular spot then coming back to it, let me explain:

    Same applies to the 4-Way Sli, lets say your looking at a wall and textures seems to be where its set at, go farther a bit, then come back, you will notice now that the texture detail themselves is drastically lowered, there is a work around this issue:

    Apparently, it seems it has to do the Vram, I tried messing with the texture detail and level of detail and other things, seems like its alot better now, if for some reason, while jumping into a new zone and alot of stuff doesn't seem all that detailed, you can refresh it by changing the actual resolution itself to a lower one then going back to where you had it, but if it keeps doing it, you might want to lower both the texture and level of detail, seems to help out alot.

    Far as overall scaling, Scales like an Absolute Beast, with excellent MS(Frametime).

    And lastly rendering mode of choice:

    I have tried AFR/AFR of SFR(Combo)/and SFR.

    It seems AFR scales extremely well but it takes forever to get on the game after the saved game is done loading, man your not kiddin', next is AFR of SFR rendering mode, again it scales extremely well but the graphic anomalies is similar to SFR but worse, in SFR case, there is some flashing problems where there is lights and such, AFR of SFR is the same case but with added stuff, like this bluish/green looking graphic problem, its on the front of the screen itself, not in the background of the game itself, its in the front of the screen...

    It seems SFR scales the best out of the bunch, hops in the game right away, with ALOT less stuttering then AFR of SFR or AFR, AFR of SFR, upon loading up saved game, there is ALOT of stuttering in the beginning, after awhile it will die itself down, so no worries there, but yes, SFR, WOW, it is Super smooth like one hot knife just went through one THICK arse Butta...

    There is no way to shut off the volumetric lights themselves, lowest setting I can go is low, so no way to shut if off completely, but yes, I remember back in the days, no one could get this game to scale at all, even two way, figured hop back into the game and see how well it does now, Jesus, Absolute Brutal scaling potential...

    So working around some graphic anomaly issues, remember, this is just the beginning of the game, I have NOT venture far enough the game to know, so I will continue to give heads up about this game, so always continue to check back for the latest up to date on my 4-Way projects.

    The intro playback movie itself, it seems its not smooth at all, DQ11 had this issue as the main intro is not running in real time, its a video playback, I have not tried it with this game but I fixed it in DQ11 while tuning that game months ago, have to turn on vsync and MUST have fps cap at I believe it was 30 fps, when back on the game, just uncap it.

    And last piece, the next Tier card, thinking about this about a month ago.

    As a remainder to all, within 30 games from now(Not including the old mentioned games that this thread or old games being re-worked on, talking about New games never messed with before) I will be moving up to perhaps to 4 980 or 4 980ti's(My original plan from months ago), haven't decided yet, but from 30 games, I will be moving up, Absolutely no exceptions period...

    But before I do that, I will promise everyone, I will have Videos up on all games I have tested thus far(regardless of the performance/Going to be awhile so hang in there), and I want to test out of curiosity sake, on win 10x 64 on all current rtm build, from TH3 and there on, very curious about this and I bet many others as well...

    And just to let everyone know, using Modified Quadro Drivers: 432.28(Released in Feb of 2020)

    Thanks for reading, COD: INFINITE WARFARE is now currently in Full Swing and its now offically:

    -----WORK IN PROGRESS-----
     
    Last edited: Mar 8, 2020
  9. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE REVISTED - 1ST MAJOR UPDATE - VIDEO PLAYBACK ISSUES - 3/28/20:

    1st Major update fixed.

    About the major stuttering and possible long hang during intro playback has now been fixed.

    I tried routed the actual problem from lowering all of my settings, sync to every frame and so on.

    Apparently lowering down your resolution will indeed FIX this issue under 4-Way Sli mode.

    First I lowered down all of my in-game graphic settings and sync to every frame(60 fps) under 4k(2400p), nope, went smooth right off the back but did a complete hang, took a very long time to get rid of it, but it will come back on so no worries there, no hard reset or hard power down.

    So next I tried borderless mode while my main window screen is @ 4k(2400p), this time I tried viewing the intro playback, nope, same thing again.

    I tried lowering the resolution to 1600p(2400x1600), and its alot better but right before the end, it does a major complete hang and took forever to come back...

    Lastly, lowered down my resolution to 1200p(1920x1200), and it works perfectly, and then I tried putting all of my in-game graphic settings to where they were, re-observed the intro playback(With Vsync ON or OFF), it works perfectly...

    So those who are tuning this game under 4-way and having this major issue, trying doing the above method, it will correct the issue.

    And to top it all off, even during the intro playback and with good in-game graphic settings, like it matters anyways since its a video playback but looking at the Vram usage, Holy Moly, All out, and this is @ 1200p, so be very strongly aware of this, probably explain why about the hangs, but whats funny if it is the VRAM, I don't have an issue running this game @ 5k(5120x3200-200% from native resolution).

    I have not tried it with other in-game video playback so can not say, but will give heads up on it if something does indeed happen.

    COD: INFINITE WARFARE IS STILL CURRENTLY:

    -----WORK IN PROGRESS-----
     
  10. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - UPDATE - 3/8/20:

    For those who are trouble making this game to work for 2-way or even 3-way, I got some working bits that will work right off the back but the graphic flashing anomaly is pretty much the same as with 4-way.

    Again, its still very early in tuning but I can confirm 2-way and 3-way is indeed in-fact working but I have NOT check for long term stability so be absolutely careful.

    To operate 2-way and or 3-way sli mode, open up your nvidia inspector program, preferably the DMW version or DeamManWalking verson 3.5.0, next go to Sli Compatibility (DX10+DX11) Bits and make the following changes below BUT FIRST:

    FAIR WARNING ABOUT MY CUSTOM BITS:

    PRELIMINARY TESTING OF THESE HIGHLY EXPERIMENTAL BITS WILL BE 100% GUARANTEED TO BE WORKING, ONLY, REPEAT, ONLY FOR THE NVIDIA GTX 970 FAMILY AND NOTHING ELSE, THOSE WHO ARE WILLING TO TRY ON THE GTX 980 OR EVEN THE GTX 980TI, ABSOLUTE CAUTION IS HIGHLY ADVISED AS ITS NEVER BEEN TESTED ON THOSE, AND ONLY USE WIN 10 X64(15063.v1703/RS2-RTM).

    PRELIMINARY BITS:


    0x0A9815F5

    No need for Sli Broadcast or Specific hack bits are needed at this point, Leave all the Unknown id settings untouched.

    Just Change your Sli Rendering mode to AutoSelect, and Change the other two settings to either two or three, depending on how many gpu's your using.

    If your doing a 4-Way Project like myself, the current workings bits I am using right now is:

    0x8A9815F5

    So looking at 1 gpu peformance to 4 gpu's performance in this one area, stationary/static:

    1 gpu: 26-27 fps
    2 gpu's: 46-47 fps
    3 gpu's: 66-67 fps
    4 gpu's: 96-97 fps(AFR of SFR/Hybrid)

    And yes, I tried changing that 1 to a 2 to see all flickering went away, it did but man it slowed down everything, some games will benefit from this, some will not, some won't even make a difference, this one:

    0x00000<<~~

    Usually 1 and 7 pretty much behaves the same, same goes for A.

    Driver used: Modified Quadro Drivers: 432.28(Released on Feb/2020).

    Thanks for reading, COD: Infinite Warfare/Single player campaign is still currently:

    -----WORK IN PROGRESS-----
     
    Last edited: Mar 10, 2020

  11. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - BEYOND PERFECT SCALING RESULTS - UPDATE - 3/8/20:

    Another daming results here...

    Under AFR of SFR(Combo/Hybrid), I checked from 1 gpu and then to 4.

    I did quite a bit of in-game tuning to make sure everything is perfectly smooth, no Stuttering of any sort, and yes, since the game is capped at 125 fps, playing it at just 2400p = not utilizing all of the potential as the game caps at 125fps, so I have to increase the internal resolution as well, I am playing around 5.3K, in those areas, Texture had to set to Low, I did some testing under Textures set to normal, at a lower resolution, although it works perfectly, again, since at lower resolution, I am not fully utilizing 4-way effectively so I had to increase the resolution a bit to make better use of the overall scaling potential, so I had to lower the texture to make up for it, again, as a remainder to all, this thread is to make 4-Way to scale to its absolute best, regardless of settings, All about Scaling here, that is my main primary focus.

    But not getting into in-game settings just now, still to early, but here are the results from 1 gpu and to 4, same spot/static/no movement/stationary:

    1 GPU: 15-16 FPS

    4 GPU's: 69 FPS(AFR of SFR-HYBRID/COMBO)

    And yes, I double checked the results, this is the correct info.

    Always check back for the latest updates, and if your seeing this, just heads up if you all have not, scroll up some, you will fine some working bits that will make 2-way and 3-way sli to scale, but fair warning, bits are still to early but they do infact scale.

    GAME IS STILL:

    -----BEING FINE TUNED-----
     
    Last edited: Mar 10, 2020
  12. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE: INCOGITABLE(STAGGERING) UPDATE - GTX 970 - 4-WAY SLI - 3/11/20(REVISED):

    I deleted the original post and I am revising this section as the last one caught me off guard on the render resolution setting, and I do certainly apologize for it.


    Just achieved some Absolute Ground breaking/Earth Shattering update...

    Ok, Now I can confirm that I an hitting the 4th card, in terms of performance with no, Read this well:

    WITH NO GRAPHIC ANOMALIES, 0!!.

    How this was all done???

    First of all, this took quite some time to figure out, but going to go real snappy on this:

    First I did some changes within the SLI Broadcast (DX1x) bits:

    The one that you must change is this:

    0x00000000<<~~ Currently set to 9

    This bring all the flashing problems to a complete stop, but there is alot more to this then just that, ALOT more, read forth:

    1st of all, get into the game, switch to window mode instead of fullscreen, while in this mode, switch your Vsync to ON, then, switch your render resolution to 100%, upon loading up the game, do the following exactly, otherwise this technique will NOT WORK.

    Originally this section was done on Resident Evil 2 Remake and Bright Memory, this Technique was originally found on my own and no one else.

    So when you are resuming your saved game, watch the load bar below very very carefully, there will be a point that it will hang for a bit at the 95% mark, the moment you see it move, do ALT+ENTER to get back into full screen.

    All your flashing, graphic anomaly should be gone, now from here, you should be in fullscreen mode, switch your render resolution to 100%, turn off your Vsync, increase your render resolution to whatever it was, The down part is that, my Performance is not quite like it used to be when I had the graphic anomalies, nontheless, it is Night and Day difference.

    I checked one gpu in one static area, 5.3k, texture to low, and other settings tweaked, getting 22-23 fps, now on 4 cards, getting 72 fps roughly, I am making some difference, gonna take some time just like anything else, before I was hitting 66-67 when I discovered this, some tweaking, now getting 72 fps, so yes, making improvement, and it is infact hitting 4th card, so this is great news.

    COD: INFINITE WARFARE IS CURRENTLY UNDERGOING:

    -----MAJOR TREATMENT-----
     
    Last edited: Mar 11, 2020
  13. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE: VSYNC UPDATE - 3/11/20:

    If you look at the very last post, I totally revised it, and do apologize, I took it for granted real fast, Like before, "NEVER" EVER, go by that gpu usage, which I did, cause all of it was hitting over 90% at that time, and also my native res was @ 4k, and my osd size was the exact same as it was before, the moment I saw this, I quickly took it for granted and clapped my hands WAY WAY WAY to early, this is what I been telling people, NEVER, NEVER, NEVER go by that gpu usage, under no circumstances while you are tuning your games.

    Usually the scaling will bang up alot when you are starting to mess with the SLI BROADCAST BITS:

    0x00000000<<~~ this one.

    Notice my render resolution(SSAA) was way below 4k, the moment I went to the 1st card, it was the same, so be absolutely deadly careful about this, but I revised the post, that I can confirm now, with this thread on the line, WORKS...

    NOW, how this got all started to get this to work was the VSYNC.

    I notice when I had the Vsync turned on, when I did that RE2/Bright memory Resume game Technique, something very interesting came about, I notice that, the front most of the screen, the flashing up and down, was no longer there, all though the flashing is present within the game background itself, it seems everything is more stabilized, far as graphic anomalies is concerned..

    So this quickly prompt me do the following, for those who's tuning under 4-way, do the following, exactly in this order:

    Get on the game, make sure its in window mode, Set Vsync to ON, render res to 100% to your windows mode, then resume game, right when the loading bar goes to 95%+ mark, you should see a slight hang there before it will move again, the moment it moves, do my RE2 Remake/Bright Memory technique, right when you see that slider move again from the 95%+ mark, Quickly do ALT+ENTER, if done properly, ALL graphic anomaly that was there before is Quickly Neutralized, then from there, you should also be in fullscreen mode, then from there:

    Switch your Render Resolution(SSAA) to 100%, turn off Vsync to uncap your FPS, then switch your render resolution(SSAA) to your desired Resolution, FPS will be uncapped and All the Graphic Flashing Anomalies will non-existent, just heads up on this though:

    While you are beginning to change your settings back your settings while doing the above method, there will be a very long hang by doing so, this is normal, it make take like up to couple mins, don't worry, I can confirm no BSOD or Complete hang where you need to do a hard reset, do NOT worry about it, let it go naturally and lastly:

    Your fps will not be the same while the graphic problems was there before, but it will be slightly over 4th card performance(Tuning is WAY still to early).

    Here is my current working BITS thus far, WARNING, Game is far from finished yet and as Remainder, "ONLY" FOR GTX 970 FAMILY, WIN 10X64 (15063.V1703/RS2-RTM)..

    SLI Broadcast BITS (DX1x): 0x00000109

    SLI Compatibility BITS (DX10+DX11): 0x8CD815F5

    SLI Specific HACKS (DX1x): 0x00000060 (Crysis 2)

    As a Remainder to all, this pertains, ONLY to 4-way SLI, and please use Only the AFR of SFR mode.

    To those who wants to try 2-way or 3-way(I have not tried it yet, so I can not confirm it), Switch your:

    SLI Compatibility BITS (DX10+DX11): 0x0<<~~ leave it to 0...

    The one that MUST be changed that is needed to get rid of the massive flashing graphic anomaly is:

    SLI BroadCast BITS (DX1x): 0x00000000<<~~, MUST be set to 9(Neutralizes all graphic problems), I tried others, this works the best.

    Also, while this bits in place, be also very careful of tuning:

    SLI BroadCast BITS (DX1x):

    0x00000<<~~ this one, as certain digits, can bring up those flashing graphic anomalies again, Regardless if the above method is used, so be careful.

    Also, for best overall scaling performance, please have your AA turned off.

    I have used SMAA X1 and SMAA X2, although it seems to work, overall scaling potential seems to be lowered a bit using SMAA X2, but for tuning purposes now, I leave it to off, also, SMAA X1 seems not to have any graphic anomalies whereas SMAA X2 there is some, not to bad, but some.

    Also, screwing around with AA settings, you may experience decreased performance, for instance, if your original fps is 72, switched AA settings, says 58 or so, then back to its original setting, it will not change, usually to fix this, switch your AA settings around, toy around with it a bit to see if it will change, that or the Vsync option, it will come back eventually.

    Thanks for reading, again, if your tuning the SLI BroadCast Bits:

    0x0000000<<~~, it can make your gpu sky rocket so high but yet in turn, 1 gpu performance, so do not celebrate to quickly, like I did before, It may or may not work, depending on the digit you use and game in general, be absolutely very careful, thank you.
     
    Last edited: Mar 12, 2020
  14. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    TEMP UPDATE - DELTA FAN TRYOUT - GTX 970 - 4-WAY SLI - UPDATE: 3/12/20:

    Out of my curiosity, I had a extremely high powered and Extremely, Extremely loud 120mm Delta fan for a very long time, never really used it cause to DA_n loud, the 3 fans thats on the bench, I took the bottom one out, and replace it with this, hitting pretty much the 3 cards, I have temps results if anyone is interested.

    The four 120mm fans on top, different orientation then I normally had.

    The 120mm Delta fan:

    Product Specification:
    Model Number
    TFB1212GHE-F00
    Size 120 X 38 X 120mm
    Rated Voltage (V) 12
    Rated Speed (RPM) 4600 RPM
    Air Volume (CFM) 220.01
    Noise Level (dba) 65.0
    Bearing System Ball Bearing System
    Connector *bare (3) wire 12" tail

    Ran a Very Very Brutal test, personal choice of mine, Rise of Tomb Raider, With custom Enhanced SMAA profile, made by me, ran for good 15 mins or so:

    Ambient room temp in the upper 60 degrees fahrenheit, Windows Tightly Shut, Lights on, and door to the room, WIDE OPEN, results are as follows:

    GPU #1: 52c-53c @ 1.168V
    GPU #2: 59c-60c @ 1.175V
    GPU #3: 61c-62c @ 1.187V
    GPU #4: 63c-65c @ 1.1206V


    EXTREME RESULTS WE HAVE HERE, GPU #4 temps should be lower but the air from that fan is not really hitting the 4th card, remember, I got spacers in-between each card.

    Other then that, EXCELLENT RESULTS, EXCELLENT...


    UPDATE:

    Tried out another 120mm fan, Was using it on one 4 fans I had on top of the gpu's, but gave this one a try:

    One of my personal favorites, back in the good ol' Thermalright XP days, the 120mm Panaflo fan...

    Ambient temps about 67 Degrees Fahrenheit, Windows tightly SHUT, Lights on and the door to the room, WIDE OPEN, Results are as follows:

    GPU #1: 57c-58c @ 1.168V
    GPU #2: 71c-72c @ 1.175V
    GPU #3: 68c-69c @ 1.187V
    GPU #4: 68c-69c @ 1.1206V


    Extreme Differences, just as expected, for those who are deadly curious, these are the results, thanks again.
     
    Last edited: Mar 12, 2020
  15. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - CURRENT STATS - UPDATE - 3/15/20:

    Ok, current stats on my progress for this game thus far:

    Like before, I already 100% eliminated the flashing problem with this game, but for some reason, just yesterday, Sometimes my RE2 remake/Bright Memory Technique does not work, was working all the time, now at times it does not work, so just heads up on this, it may or may not work all the time, kinda funny, was working all the time, but will give heads up about this later.

    Actually what was interesting is this:

    Before when it was working perfectly, the elimination of the flashing problem, I wanted to try SFR mode and then AFR mode(Autoselect), to see if the flashing would come back and it did, I said to myself, ain't that stinkin' interesting, SFR of AFR rendering mode doesn't have this problem, went back on the game, it came back, was like huh, It was working perfectly, now it doesn't seem to work 100% of the time, was before, again, will give heads about it later.

    Also, When entering the game while there is no graphic anomalies/flashing,etc, I am still trying to further improve the overall 4-way scaling potential, I have gotten up pretty much up to 3-way, 3-way performance is basically perfect, some sections in the game its barely hitting the 4th card performance, so trying to further improve the overall performance.

    That being said, If one choose to play the game while certain portions of the background flashes like mad crazy, Overall scaling, like I mentioned before, Beyond perfect scaling.

    I am hoping Final conclusion will be up probably before this coming Wednesday, so be sure to look for it.

    GAME IS STILL UNDERGOING:

    ----------MAJOR SURGERY----------

    UPDATE ON FLASHING ANOMALIES:

    Ok, found the problem, Sli Broadcast Bits, as I stated before:

    0x000000<<~~ this one

    Be careful of tuning this particula hex when your Sli Broadcast is set like this:

    0x00000009<<~~ To completely neutralize all graphic problems...

    So now it looks like this currently:

    0x00000109

    So those who are tuning this game, just sli in general, if and when you get your graphic anomalies neutralized, be careful of tuning anything else, check step by step and make sure those massive flashing problem does not come back cause it will, so be absolutely careful.

    There is one graphic anomaly I have found, although minor, it is not a flickering/flashing problem, it seems it appears to be around lights and such, specific areas, while moving your camera around, it leaves star-wars light saber trail like effect, again, nothing serious, fars I can tell, its the only anomaly I found, nothing to cry about, thanks for reading.
     
    Last edited: Mar 16, 2020

  16. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - AFR/SFR RENDERING MODE UPDATE - 3/15/20:

    Ok, just to bring everyone up to speed on this and those who are tuning this game in similar fashion:

    Autoselect, AFR & SFR rendering mode = Same Exact Performance as AFR of SFR(Combo) during initial testing.

    This is while I neutralize the flashing problems in the game, and far as MS(Frametime) goes, it seems, all three is roughly the same give or take, far as stuttering wise? Initial impression, didn't see any, so all 3 modes = same give or take, very good news.

    So, those who are tuning this game, under SFR or AFR of SFR, make SURE your First Digit after the x within Sli Compatiliby bits (DX10+DX11), looks like this:

    0x8<<~~

    For AFR Rendering mode:

    0x8

    For Autoselect, looks like this:

    0x0<<~~

    And just heads up again, that Star-Wars Light Saber trail like that's left behind around certain lights while moving camera is present in all rendering mode tested above, again, nothing serious and nothing to cry about, other then that, No flashing/flickering, MS(Frametime) all modes, roughly the same and yes, all are good, low 10's to somewhere in the 14+.

    Thanks for reading.
     
    Last edited: Mar 16, 2020
    Deleted member 213629 likes this.
  17. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - QUICK CURRENT FPS RESULTS - UPDATE - 3/16/20:

    Got some quick results here while the graphic anomalies are completely neutralized...

    Just a quick recap, The setting that caused the Light Saber trail like around some lights is coming from:

    Volumetric lights

    If I switch it off, the effect is gone.

    Also, Volumetric lights is one heavy hitter in this game, performance wise, brings your fps down quite a bit.

    But going down and dirty on quick performance here:

    Base(Native)Resolution is set to 2400p(3840x2400)

    Render Resolution(SSAA) is set to 6527x4079

    With volumetric lights switched off, ambient occlusion on, Texture to low, LOD to normal, Tess = ALL.

    I did the insane render resolution to further put tremendous load on to the gpu's so it can further scale better, anything lower, overall scaling potential ability goes down, so remainder to all once again, main primary focus to this whole 4-way is scaling, to get the games to scale at their absolute best, no matter the graphic settings or anything else, All about Scaling.

    2 quick tests here, one is right in the beginning right before you jump out the window then landing down, I stop right before you jump into the pit hole.

    2nd test is in the facility in the beginning, movement pattern.

    Test #1:

    1 GPU:

    MIN FPS: 13
    AVG FPS: 24
    MAX FPS: 46

    4 GPU's:


    MIN FPS: 47
    AVG FPS: 80
    MAX FPS: 127(Game is capped @ 125, can not go any further)

    Test #2:

    1 GPU:

    MIN FPS: 17
    AVG FPS: 23
    MAX FPS: 30

    4 GPU's:

    MIN FPS: 63
    AVG FPS: 79
    MAX FPS: 108-115

    Overall, looking very good.

    Remember, this is with no graphic anomalies whatsoever.

    And as another remainder to all, Again, the insane render resolution is to further put for load to the gpu's and enhance the overall scaling ability across all 4 gpu's.

    Just for kicks, I have tried render resolution 8k(4800p) on one card, it works but 4 gpu's, anything higher then 6.5k cause the overall scaling to go haywire.

    I am also trying to improve scaling performance on lower render resolution then 6.5k, and my performance keeps about the same, so trying to further improve on it.

    Be sure to keep coming back for the on-going continuing efforts updates on my 4-way sli projects.

    GAME IS STILL UNDERGOING:

    -----MAJOR SURGERY-----
     
    Last edited: Mar 16, 2020
  18. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - PRELUDE TO FINAL CONCLUSION - UPDATE - 3/16/20:

    Warming up to the final conclusion to this game, you all don't wanna miss this one out, aiming for late wednesday or sometime into thursday.

    To warm things up even further, have a gander at these videos(NOT MINE):










    Get ready for the Showdown soon.
     
    Last edited: Mar 17, 2020
  19. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - SVG FILES/BACKUP - UPDATE - 3/18/20:

    Just heads up for those who are tuning/monitoring/performance this game under sli mode:

    It is imperative that you backup your savegame(SVG document) file, preferably to your desktop, cause lets say your doing your own performance monitoring, and if for any reason, you experience a stutter,etc, your performance monitoring will not be accurate, and if your doing movement pattern performance checking, if and when you go to another checkpoint, you will be saved to that checkpoint if your going to resume last save game, and if want to recheck your performance, your stuck, unless you restart the whole mission, so for this reason, very important, back up your SVG file, just copy and paste it inside the game folder where the actual SVG file is, saves LOTS of time.

    And another heads up, Jotting down performance is coming along very well, far from being finished, did quite a bit but no where near done, I am hoping before end of the day, if not sometime tomorrow, but hoping for today, looking very good, some stuff is rather interesting, be sure to look for it later on, and checking on the amount of time spent on this game(Revisited), well over 30+ hrs was invested, everything to this game, thanks for reading.

    ONE MORE QUICK UPDATE BEFORE FINAL CONCLUSION:

    One final thing before final conclusion will be posted, This will be, probably by far, the most jotted down on performance readings then any of the games that I posted in this thread.

    This will include, many sections and their movement patterns, static(Stationary/idle) spots, MS(Frametime), from all 4 gpu's.

    I will be doing all four 4 gpu's test so people can see performance readings from 1 gpu all the way to 4.

    Main focus is without AA, but as a added bonus in certain locations only, I will conduct From AA being off to SMAAX1 to SMAAX2.

    There will be TONS and I mean TONS of stuff, be sure not to miss any of it, very hard at work currently, should be done soon, check back later for the latest updates, currently moving closer to 40 hrs in this game, All went to tuning/performance checks/stability checks,etc, you name it.

    Thanks for reading.

    UPDATE:

    Sorry everyone for not having it done on time, I am almost done but what's taking so dang long in this game is this:

    While doing performance monitoring, certain areas, Especially: Blacksky, outside area:

    Certain times it will stutter during 3-way and 4-way, and it completely throws out my readings,
    I wanted to keep it clean and have accurate results WITHOUT any stuttering whatsoever.

    Another thing that I did not mention but wanted to mention it during the final conclusion for this game, I will do so right here:

    Right after I get on the game, the moment I adjust certain settings, there will be a great chance that the game will Completely hang for a long time, around like 1-2mins+, its not a permanent hang, just temp, it will come back, so I have to deal with this every single time I jump on the game and adjust the setting, not only that, the next thing I am going to mention also pertains from 1 gpu to 4-way sli:

    Lets say you load your saved game(check point), with 1 gpu, it says 18 fps, that fps could perhaps not be your peak, it could be perhaps 20.

    I found a way to fix the issue but it will not correct it 100%, it will take many times over to do it, its like playing the lotto but alot easier:

    You can either toy around with the Vsync option, or the post processing effect, AA.

    Under AA, you don't really need to adjust the AA option itself, you can just click the same on you have on now, to refresh the overall performance potential, keep doing it until you get your top fps performance.

    This applies to Single Gpu mode all the way to 4-way sli.

    4-way for example, it could say 45, then you toy around with the setting, next it can be 56-57, then back to 52, and so on, but if peak is 60, just keep doing it until you reach it, it could take sometime, this is the reason why I could not get the final conclusion up on time, Game is downright to the max, absolutely Insane to deal with...

    This is one of the many reason why No one in the world would touch 4-Way, Way to much to deal with, Very few people can even tolerate it, It is the Darkside of PC Gaming, absolutely no question about it.

    Going on close to 50 hrs.

    I got like a few to go then the space warfare scene to do, after that, I think I am done.

    Sorry couldn't make it tonight, I am still on the game as we speak, trying to finish it up.

    So always continue to check back right here, "ONLY" on Guru3d.com and no where else.

    Thanks for reading.
     
    Last edited: Mar 20, 2020
  20. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    3,753
    Likes Received:
    507
    GPU:
    4 GTX 970-Quad Sli
    COD: INFINITE WARFARE - GTX 970 - 4-WAY SLI - FINAL CONCLUSION #1 - 3/20/20:

    All right, here it is everyone, the moment that "ALL" the COD fans and what not, has been waiting for:

    First, if anyone wants to listen to the original OST while reading, here it is for your convenience:



    As before, Sorry for all the delays of getting the final conclusion up, Finally its all done.

    I present to you COD: Infinite Warfare, under 4-Way sli Mode, Is this even possible for this game???????, read below to find out:

    I remember playing this game while using my 2 1080ti's in the past, and judging from people's experiences in the past with this game under SLI mode, No one from what I understand, could get this game to work in SLI mode, I have not checked this for a long time to see if anyone has gotten it so don't quote me on this.

    During the initial part of this whole 4-Way Sli project, I gave this game a try, and was expected earlier, I couldn't even touch the second card...

    Fast forward like a year later of intense 4-Way tuning, Gave this a try one more last time, All I did is jumped right on to the SLI Compatibility Bits (DX10+DX11), and did some very quick Adjustments, and no other changes were done, hopped right into the game, I was really amazed to see the scaling was incredible right off the back but there was alot of graphic anomalies such as Massive amounts of flashing, Star Wars light Saber like trail that gives off on certain lights,etc.

    So this is where it gets real interesting, took sometime to figure out how to get rid of the nasty amounts of flashing, not only that, to further improve its performance after it was neutralized cause after I got it all squared away, my overall scaling potential went down quite a bit...

    It all started with Vsync believe it or not, After loading up the saved game with this setting enabled, I noticed that the front foremost of the screen, the flashing, seems it was all neutralized, there was still massive amounts of flashing in the background but seems its more stabilized, this prompt me to do the Resident Evil 2 Remake/Bright Memory Resume Technique that I personally discovered on my own time months ago, and I gave it a try with this game, it worked but there was some flashing still present, all though most of it was gone but still present, this also prompt me to do some tuning within:

    SLI Broadcast (DX1x) Bits:

    0x00000000<<~~ this one.

    After messing around with it, 9 Completely neutralized ALL graphic anomalies, not only that, took a lil bit how to get rid of that Very unique, Star Wars Light Saber like trail given off from some specific light sources in this game, to get rid of the light anomalies, you MUST turn off the volumetric lights, NO EXCEPTIONS HERE, and also, my testing that I did on my own, setting is set to OFF.

    And also just reference:

    SLI Compatibility (DX10+DX11) Bits:

    0x00000<<~~ this one, 2 will normally make everything stop flickering/flashing but in turn, massive scaling drop in most games, NOT all, but with this game, it had to be set to 1.

    So, again, for future reference here, the RE2/Bright Memory Technique, here is how to do it:

    Get into the game in windows mode, can be full screen but before you resume game, MAKE SURE the game is in windowed mode, then make sure your Native RES and Render Scale Matches up while in windowed mode, Also, MAKE SURE your Vsync is on while doing this, then resume saved game, from here, you can see the loading bar progress, bottom of the screen, once you see the bar hits 95%+ range, it will stop for a bit, right before it continues, HIT:

    ALT+ENTER

    At times, it will go so fast you will not have time to do it, so for these instances, keep repeatedly hit those 2 keys until you get into fullscreen mode, there might be a chance you might hit windowed mode again upon entering the game cause of the repeated presses, so to disrupt that, try hitting other keys and what not to stop it from going back to windows mode.

    Once in full screen mode, Everything should be all Neutralized.

    Then you can switch your Render Scale from here, but be careful, within two setting changes, you WILL experience a MASSIVE hang for up to like what it seems 2+ mins or so, So be absolutely aware of this, this also pertains to 1 gpu, cause of its very abnormally high Render Scale resolution, it will go away eventually, I have never experienced any BSOD'S, completely hanged forever, hard resets or power down, so all good.

    One thing to point out here, all though, most of my testing is without SMAA, if you have SMAA T2X2 on while doing this resume game technique, you will see a massive Imprint left on the screen and others, depending on the location, to fix this, just change your AA to OFF then back to SMAA T2X2.

    Before heading to my performance results, the one, EVERYONE has been waiting for, certain things I must point out here:

    Some of these areas while testing under SMAA T2X2, there was some stuttering in certain instances, like if your close by to certain explosion,etc, so this took a great deal of time when I was jotting down all of my performance readings, main reason why of the stuttering, settings are just to high, but I had no choice to keep my render res incredible high to keep the overall 4-way scaling at top notch level.

    Resolution of choice during this test:

    Base(Native Resolution): 2400p(3840x2400)

    Render Scale(SSAA) Resolution: 6527x4079

    People is wondering why so high of the Render Scale? Easy answer:

    To keep my 4-Way scaling WELL packed down, otherwise, if your Doing 5.3k render scale, it scales excellent for 3-Way, barely even touch the 4th card performance, so I had no choice as this is all about getting games to scale at their absolute best, no matter the changes in the graphic settings.

    Speaking about Graphic Settings, I conducted a very close observation test that may shock alot of people, but then again no:

    This particular test is about Overall Image Quality(AA/Jaggies):

    First I did 2400p with SMAA T2X2 turned on then the other setting:

    Base(Native Resolution) set to 2400p, with Render Scale of 6.5k(6527x4079) with NO, repeat, NO AA being used and guess what???

    Resolution 2400p with SMAA T2X2 provided quite a bit better overall AA image quality then the one with the beefed up render scale

    Another I personally Conducted here:

    Base(Native Resolution) set to: 2400p, Render Scale of 6.5k(6527x4079), this time with the use of SMAA X1, and guess what happens again this time:

    Still, the one with 2400p with SMAA T2X2 provided better overall AA image quality...

    Wasn't surprised, this was hrs into the game before I started this test.

    So I say to myself: Although I wanted to get back on the game to see if I can get the game to scale under 4-way, Even if it did, It is really not needed as playing @ 2400p with SMAA T2X2 will take ALOT less hit then render scale of 6.5k.

    Further observation of use of SMAA T2X2, depending which areas you are in the game, I have see maybe 2 fps to like 5 fps cost of performance while using this setting, so it will take ALOT, and I mean ALOT less hit then jacking up your Render scale to crazy Numbers, but again, this whole 4-Way Sli project is to get the game to scale to their absolute best, and there is absolutely no exceptions to this whatsoever.

    Ryse: Son of Rome was in the same situation as this.

    TAA in that game COST ALOT less performance then jacking up the SSAA to like 2x2 and TAA provided better overall AA image quality, if you go back to the earlier pages in this thread, again, SSAA in that particular game, I had no choice, to get the cards to scale at their peakest of their ability, its about games to even scale at all really, also not just Ryse: Son of Rome, other games, SLI in general, SLI does NOT like TAA, even if it did, some game like Ryse, you will see decreased performance/usage.

    One more Extremely important note here about performance, this also applies to 1 gpu(Especially using SMAA T2X2:

    Upon loading up saved game or just playing the game in general, lets say 1 gpu mode with SMAA T2X2:

    Your fps is like 18, but it should be 20, you can fix this issue by toying around with the Vsync option and the AA option, so if the AA is on SMAA T2X2, just keep clicking on the same AA setting to refresh its overall performance, fair warning though, from very close observation, it make take awhile to bring the overall Performance back to its peakest while the AA is switched to: SMAA T2X2, so be absolutely deadly careful about this at all times, sometimes under 4-way mode, it could take like 20 times or so to bring out its true potential, so just heads up about it, be very careful.

    Another very important thing about texture settings, Mine was set to low because of the special circumstances what I was undergoing, In my case, texture can set to normal, it will work BUT, the downfall is this:

    Say you are looking at a door/logo/crate, and texture seems to be fine, few steps out, come back to the same precise item you was looking at, you will clearly see that the actual texture has drastically lowered, this is the indication that the overall in-game settings are just to high and the Vram, so for this reason, mine is set to low, as it works very well under the conditions that I am running in this game.

    Again, speaking about Texture themselves, If and when you are experience Drastic decrease in texture resolution when coming back to the location you were at before, just change your render scaling resolution to a much lower one then revert back, it will fix the issue, it may come back, if this is the case, lower your texture resolution to the next lowest, then re-observe to see if will react the same way, thanks.

    And as promised, TONS and TONS of performance readings are jotted down, Let's get down to the nitty gritty to the one that everyone has been waiting for, the RESULTS themselves:

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

    This portion will cover ALL the performance test was conducted by me.

    Native Resolution set to 3840x2400

    Render Scale(SSAA) Resolution set to: 6527x4079

    Ingame graphic settings are set as follows:

    Texture level: Low
    Texture filter: High
    Shadow Map Resolution: Auto
    Shadow Caster Level: Normal
    Cache Spot Shadows: No
    Cache Sun Shadows: No
    Volumetric Lights: No
    Ambient Occlusion: Yes
    Particle Lights: High
    DOF: Off
    Motion Blur: On
    Bullet Impact: Yes
    Tesselation: All
    Particle Quality: High
    LOD: Normal
    Shader Preload: On

    Static(No movement)/Movement pattern movements/MS(Frametime)/Added bonus:

    Some areas have SMAA X1 and SMAA T2X2 as well for ones that real curious about the results.

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

    First Covering the Static(Idle/No movement), and all testing are done precise and correct for very accurate results, along with absolute precise camera positioning,etc and as before, I made sure that the overall scaling potential were at its peakest by toying around with certain settings, as failure to do this, you will not get accurate results and sometimes it will not bring out to its best potential, so I had to really make sure, potential from 1 gpu to 4 gpu's were at its best at all times, everything was absolutely carefully monitored at all times.

    As a remainder to all, ALL TEST below is WITHOUT AA, if its being used, I will specify it, thanks.

    Static: (Rising Threat/Ice Location #1):

    1 GPU: 27 FPS/36.1-37 ms
    2 GPU's: 50-52 FPS/34.7-35.9 ms
    3 GPU's: 70-71/1.5-3.9 ms
    4 GPU's: 88-91/1.6-11.1 ms

    Static: (Rising Threat/Ice Location #2):

    1 GPU: 22 FPS/44.46-2 ms
    2 GPU's: 42 FPS/34.9-43 ms
    3 GPU's: 59 FPS/4.2-34.7 ms
    4 GPU's: 72-75 FPS/3.6-11.2 ms

    Static: (Rising Threat/Inside Facility #1):

    1 GPU: 19-20 FPS/46.3-52.5 ms
    2 GPU's: 35-36 FPS/46.8/53.5 ms
    3 GPU's: 53-54 FPS/3.3-6.1 ms
    4 GPU's: 61-67 FPS/ 3.2-38+ ms

    Static: (Rising Threat/Inside Facility #2):

    1 GPU: 18 FPS/46.5-55.4 ms
    2 GPU's: 32-33 FPS/51-56 ms
    3 GPU's: 47-49 FPS/3.7-52.5 ms
    4 GPU's: 57-63 FPS/6-17.7 ms

    Static: (Black Sky/Beginning Area):

    1 GPU: 20 FPS/48.1-50.8 ms
    2 GPU's: 38 FPS/39.1-46 ms
    3 GPU's: 52-53 FPS/7.3-13.4 ms
    4 GPU's: 63-66 FPS/6.3-38.2 ms

    Static: (Black Sky/Outside Area):

    1 GPU: 24 FPS/40.5-43.5 ms
    2 GPU's: 43-44 FPS/8.4-38.9 ms
    3 GPU's: 57-59 FPS/9.4-27.2 ms
    4 GPU's: 67-70 FPS/7.2-29.3 ms

    The two test below, one is WITH AA(SMAA T2X2 and one WITHOUT AA):

    Static: Operation Blood Storm(REGROUP/NO AA):

    1 GPU: 28 FPS/34.1-36 ms
    2 GPU's: 53-54 FPS/32-35 ms
    3 GPU's: 75-77 FPS/22.9-32+ ms
    4 GPU's: 94-99 FPS/3.1-29 ms

    Static: Operation Blood Storm(REGROUP/SMAA T2X2/SAME EXACT SPOTS AS ORIGINAL):

    1 GPU: 25 FPS/39.7-41.9 ms
    2 GPU's: 46-48 FPS/37.4-38.5 ms
    3 GPU's: 62-65 FPS/30.7-41.5 ms
    4 GPU's: 84-88 FPS/3.8-31.8 ms

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

    MOVEMENT PATTERNS:

    All movement pattern was done in the same/precise manner to ensure readings are 100% accurate, did my best:


    MS(FRAMETIME)

    Some frametime recordings, only a very few, those are not recorded it will be left blank, for the time being until it's revised, thanks.

    Movement Pattern: Rising Threat/Beginning:

    1 GPU:

    MIN FPS: 14-16//AVG FPS: 19//MAX FPS: 28-30//1% LOW: 6 FPS//.1% LOW: 4 FPS//MS: 1.7-212

    2 GPU's:

    MIN FPS: 28//AVG FPS: 37//MAX FPS: 52//1% LOW: 17 FPS//.1% LOW: 6 FPS//MS:

    3 GPU's:

    MIN FPS: 41//AVG FPS: 51//MAX FPS: 77//1% LOW: 16 FPS//.1% LOW: 9 FPS//MS: 1.8-58.7

    4 GPU's:

    MIN FPS: 54//AVG FPS: 66//MAX FPS: 99//1% LOW: 17//.1% LOW: 17//MS: 1.8-

    Movement Pattern: Rising Threat/Inside Facility:

    1 GPU:

    MIN FPS: 16//AVG FPS: 22//MAX FPS: 31//1% LOW: 11 FPS//.1% LOW: 5 FPS//MS: 32.7-53+

    2 GPU's:

    MIN FPS: 30//AVG FPS: 39//MAX FPS: 56//1% LOW: 19 FPS//.1% LOW: 17 FPS//MS: 1.5-51+

    3 GPU's:

    MIN FPS: 43//AVG FPS: 57//MAX FPS: 80//1% LOW: 21 FPS//.1% LOW: 17 FPS//MS: 5-51

    4 GPU's:

    MIN FPS: 53-55//AVG FPS: 70//MAX FPS: 96//1% 26 FPS//.1% LOW: 21 FPS//MS:

    Movement Pattern: Rising Threat/Inside Facility/Mech Battle:

    1 GPU:

    MIN FPS: 10//AVG FPS: 15//MAX FPS: 26//1% LOW: 9 FPS//.1% LOW: 4 FPS//MS: 50+/125+

    2 GPU's:

    MIN FPS: 19//AVG FPS: 30//MAX FPS: 49//1% LOW: 11 FPS//.1% LOW: 4 FPS//MS: 15+/89+

    3 GPU's:

    MIN FPS: 28 FPS//AVG FPS: 42//MAX FPS 59//1% LOW: 12 FPS//.1% LOW: 11 FPS//MS: 2.7-5.1

    4 GPU's:

    MIN FPS: 39 FPS//AVG FPS: 54//MAX FPS: 75//1% LOW: 16//.1% LOW: 13 FPS//MS: 2.7-50


    Blacksky: Under Attack(Beginning Area/Shopping District):

    1 GPU:

    MIN FPS: 18//AVG FPS: 21//MAX FPS: 25//1% LOW: 18 FPS//.1% LOW: 17 FPS//MS: 40.4-55+

    2 GPU's:

    MIN FPS: 33//AVG FPS: 38//MAX FPS: 49//1% LOW: 19 FPS//.1% LOW: 18 FPS//MS: 5.0-51+

    3 GPU's:

    MIN FPS: 46//AVG FPS: 52//MAX FPS: 63//1% LOW: 22 FPS//.1% LOW: 21 FPS//MS: 7.1-41+

    4 GPU's:

    MIN FPS: 54//AVG FPS: 62//MAX FPS: 87//1%LOW: 24 FPS//.1% LOW: 20 FPS//MS: 3.0-19+


    Next test below has three independent tests by using AA, again, all 3 tests below, same Pattern movement, Except for the last one(SMAA T2X2), Very short distance as there was to much stuttering involved and throws out the performance reading, so only short distances on thoses, thanks.

    Blacksky: Under Attack(After Shopping District/Outside Area/NO AA):

    1 GPU:

    MIN FPS: 15//AVG FPS: 21//MAX FPS: 32//1% LOW: 17 FPS//.1% LOW: 15 FPS//MS: 37-66+

    2 GPU's:

    MIN FPS: 28//AVG FPS: 40//MAX FPS: 57//1% LOW: 18 FPS//.1% LOW: 16FPS//MS: 7+/58

    3 GPU's:

    MIN FPS: 41//AVG FPS: 55//MAX FPS: 80//1% LOW: 22 FPS//.1% LOW: 16 FPS//MS: 5+/58+

    4 GPU's:

    MIN FPS: 48//AVG FPS: 63//MAX FPS: 95//1% LOW: 21 FPS//.1% LOW: 15 FPS//MS: 4+/38+

    Blacksky: Under Attack(After Shopping District/Outside Area/SMAA X1/Short Distance Only):

    1 GPU:

    MIN FPS: 18//AVG FPS: 21//MAX FPS: 29//1% LOW: 18 FPS//.1% LOW: 18 FPS//MS: 34-54

    2 GPU's:

    MIN FPS: 32//AVG FPS: 40//MAX FPS: 54//1% LOW: 22 FPS//.1% LOW: 17 FPS//MS: 6.8-48

    3 GPU's:

    MIN FPS: 44//AVG FPS: 54 FPS//MAX FPS: 73//1% LOW: 23 FPS//.1% LOW: 22//MS: 3.0-43+

    4 GPU's:

    MIN FPS: 55//AVG FPS: 69//MAX FPS: 93//1% LOW: 25 FPS//.1% LOW: 21 FPS//MS: 3.0-33+


    Blacksky: Under Attack(After Shopping District/Outside Area/SMAA X2/Short Distance Only):

    1 GPU:

    MIN FPS: 17//AVG FPS: 20//MAX FPS: 27//1% LOW: 17 FPS//.1% LOW: 16 FPS//MS: 38-60

    2 GPU's:

    MIN FPS: 32//AVG FPS: 38//MAX FPS: 51//1% LOW: 19 FPS//.1% LOW: 18//MS: 11+/31+

    3 GPU's:

    MIN FPS: 43//AVG FPS: 52//MAX FPS: 71//1% LOW: 22 FPS//.1% LOW: 19 FPS//MS: 3.0-48.9

    4 GPU's:

    MIN FPS: 50//AVG FPS: 64//MAX FPS: 86//1% LOW: 24//.1% LOW: 20 FPS//MS: 3.0-40+

    Black Sky(Take the Sky/Space Warfare Scene):

    1 GPU:

    MIN FPS: 16//AVG FPS: 27 FPS//MAX FPS: 39//1% LOW: 11 FPS//.1% LOW: 5 FPS//MS: 3+/96

    2 GPU's:

    MIN FPS: 30//AVG FPS: 52//MAX FPS: 85//1% LOW: 21 FPS//.1% LOW: 17 FPS//MS: 2.7-58+

    3 GPU's:

    MIN FPS: 48//AVG FPS: 72//MAX FPS: 104//1% LOW: 21 FPS//.1% LOW: 17 FPS//MS: 2.7-58+

    4 GPU's:

    MIN FPS: 60//AVG FPS: 92//MAX FPS: 127(125 FPS CAP)/1% LOW: 33 FPS//.1% LOW: 20 FPS//MS: 1.9-20++

    Next Scene and Final Movement pattern will consist of 2 independent tests, one with AA(SMAA T2X2 and one WITHOUT AA).

    The SMAA T2X2(4 GPU's) was taken right after the Vehicle lands on the ground as it was to much stuttering in one minute area of the test, and at times, I get the shader loading screen, so drastically throws out the performance readings, thanks.

    Operation Port Amour: Civilian Terminal/In-game cutscene/Beginning/NO AA:

    1 GPU:

    MIN FPS: 14//AVG FPS: 20//MAX FPS: 30//1% LOW: 14//.1% LOW: 11 FPS//MS: 32-70

    2 GPU's:

    MIN FPS:

    MIN FPS: 26//AVG FPS: 37//MAX FPS: 55//1% LOW: 15 FPS//.1% LOW: 13 FPS//MS: 8.7-74

    3 GPU's:

    MIN FPS: 38//AVG FPS: 52//MAX FPS: 75//1% LOW: 16 FPS//.1% LOW: 15 FPS//MS: 2.5-50.6

    4 GPU's:

    MIN FPS: 49//AVG FPS: 63//MAX FPS: 94//1% LOW: 18 FPS//.1% LOW: 17 FPS//MS: 3.4-31.5

    Operation Port Amour: Civilian Terminal/In-game cutscene/Beginning/SMAA T2X2:

    1 GPU:

    MIN FPS: 10//AVG FPS: 14//MAX FPS: 17//1% LOW: 10 FPS//.1% LOW: 9 FPS//MS: 54-100.1

    2 GPU's:

    MIN FPS: 23 FPS//AVG: 32//MAX FPS: 46//1% LOW: 13 FPS//.1% LOW: 4 FPS//MS: 9.9-68

    3 GPU's:

    MIN FPS: 32//AVG FPS: 45//MAX FPS: 69//1% LOW: 13 FPS//.1% LOW: 4 FPS//MS: 9.2-72.4

    4 GPU's(AFTER VEHICLE LANDS ON GROUND DUE TO STUTTERING):

    MIN FPS: 37//AVG FPS: 53//MAX FPS: 79//1% LOW: 13 FPS//.1% LOW: 10 FPS//MS: 7.2-33+

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

    This portion will cover the Temps across all 4 GPU's, under very brutal test, about good 35-40 mins, in one spot, static mode(stationary):

    Room Temp: About 70 Degrees Fahrenheit

    Along with Windows tightly shut, LIGHTS OUT, and door to the room WIDE OPEN, results are as follows:

    GPU #1: 56c @ 1.168V
    GPU #2: 67c-68c @ 1.175V
    GPU #3: 67c @ 1.187V
    GPU #4: 64c @ 1.206V

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

    CHECK THE NEXT POST LATER FOR THE CUSTOM MADE BITS AND THE GRAND FINAL CONCLUSION TO THIS GAME, MANY THANKS FOR READING.
     
    Last edited: May 27, 2020

Share This Page