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:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    Something I found awhile back ago, couldn't exactly remember, bet you like it though.
     
    BlindBison likes this.
  2. Volteron

    Volteron Master Guru

    Messages:
    255
    Likes Received:
    37
    GPU:
    2080ti sli
    Where can I get your drivers?
     
  3. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    Just Google the respected drivers that your looking for, you will need an modified .ini in order to install the quadro 392.6x drivers, make sure you disable driver enforcement before you do, otherwise, installation will not go through.

    Also, I remember you were mentioning about liquid metal, personally I would stick with the Kyronaut, you will also need to take the necessary precautions before attempting such a thing, also do some extensive research on how liquid metal reacts to certain coolers/blocks, you will also need to protect the surrounding components before applying, personally, unless your looking into breaking records, I would stick to what you got, not worth the risk if something ever to go wrong
     
  4. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    980ti - 4-WAY SLI - CATZILLA - RANKED #19 + IMAGES - UPDATE - 7/8/22:

    Just tested 4 980ti Evga Hybrids (Modified Bios), Everything, including temps is under control.

    I will post some images of the testing system later, for now, take a gander of my score on Catzilla under 4k preset, I was unable to send the score through their server, getting error, but, I took an image showing the score.

    This was used along with my custom made bits that was specifically made for the GTX 970's in 4-Way Sli, there was some tad scaling issues with this program, mainly because, I believe, just not enough load on to the gpu's, but overall, pretty darn good.

    Just for reference, under 1 gpu mode, I get: 6786 points.

    Results under 4-Way Sli:

    4 980ti's - Catzilla - 4k preset.JPG

    Kozary, if your looking at this post, what is your 4k preset score for this program? You have to be a registered user to use the 4k preset.

    I am going to see if I can further improve the score by making some adjustments within my custom made bits.

    Also, I tried Unigine Valley 1.0 and Heaven 4.0, there is some scaling issues under 4-Way Sli, Firestrike Ultra, there is some as well, Catzilla, scaled overall the best out of the bunch, this was all used with my custom made bits.

    Testing motherboard I used for the 980ti's under 4-Way Sli, is the same one I used for the 4 Titan V's, Supermicro X10DRG-Q, in the future, I will be using the same exact mobo as my 4 GTX 970's, Gigabyte X299X Aurous Master but not anytime soon.

    Next post, Showing pics of the testing system.
     

  5. Kozary

    Kozary Master Guru

    Messages:
    292
    Likes Received:
    50
    GPU:
    2x 2080Ti NVLink
    that benchmark is junk has never worked on any of my builds and the original never had SLI. This new version it just goes blank when i sign in, probably some localization issue not being in english. I'll stick to 3dmark.
     
  6. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    GTX 980TI - 4-WAY SLI - INITIAL TESTING SYSTEM + IMAGES - UPDATE - 7/9/22:

    Here some images of my 4 980ti's testing system, Motherboard will bound to be changed in the future, when I get started with 980ti projects.

    for the radiator fans, I used my personal choice in push pull configuration: 8 Yate Loon D12BH-12's on a Lamptron Dual Fan Controller (30WX4per channel)

    Per remainder: I have 4x140mm Fans not shown on top of the Gpu's for extra cooling, they are also Yate Loon as well, same identical ones as one on my 4 GTX 970's.

    Images:

    Top down view:

    Top down view.JPG

    Angle View:

    Angle 2.JPG

    Back:

    Back.JPG

    Dual MKIII 1200W Silencer's (Classic OCZ/PC Power & Cooling):

    Dual PSU's.JPG

    Dual Lamptron Fan Controllers:

    Fan Controller.JPG

    Kozary, if you want me to test anything, let me know which ones to test out, thanks.
     
    Last edited: Jul 11, 2022
    Kozary likes this.
  7. devastator

    devastator Master Guru

    Messages:
    240
    Likes Received:
    81
    GPU:
    rx7900xt sapphire
    cant wait untill u do post when u out of videoram on ddr5 with about 8channels that be a nice read :)
     
  8. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    And what does that supposed to mean? With 8 Channels?

    Single 980ti vram is 6144MB, if all 4, still = 6144MB (Mirrored).

    I love working with limitations.
     
  9. devastator

    devastator Master Guru

    Messages:
    240
    Likes Received:
    81
    GPU:
    rx7900xt sapphire
    ram channels i like to see if they can feed gpus if the main gpu runs out ofc it cant feed all gpus as it wouldnt be enought anyway but intresting to how well it does
     
  10. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    980ti - 4-WAY SLI - DOOM 3+RESHADE+REDUX 2.0 (RC) MOD + NVIDIA INSPECTOR FORCED AA - INITIAL TEST RESULTS + IMAGES - UPDATE - 7/10/22:

    Curious about this title on this particular setup, as expected, scales just as good as the 4 GTX 970's in 4-Way Sli mode.

    I tried using the same exact Nvidia Inspector Forced AA (0x12400010 Override | 4x MSAA + 4x TrSSAA) that I used with my 4 GTX 970's, although it does work, you will have some scaling issues in certain areas & Directional view as its not putting enough load on to the 4 980ti's, So I had to up the Ante a bit, Used 4XOGSSAA+2XMSAA, now the scaling gap is drastically lessened, big difference, I also tried anything higher then 4XOGSSAA+2XMSAA, I did not see any improvement in the overall AA quality and still images does not do any justice, you have to pay absolutely close attn. while the game is in motion to really catch the difference, I didn't see any, other then reduction in performance, so for the 4 980ti's, I would go with: 0x10000019 Override | 4x OGSSAA + 2x MSAA, also, be sure to use absolute brutal settings as well as 4k (2160p) at the bare minimum as you wont get the potential scaling results as its not getting enough load to the 4 980ti's, so be sure to pack the settings well packed down.

    Images will show difference between one gpu vs 4 GPU's:

    Settings used:

    Dhewn3 + Redux 2.0 mod (RC) + Brutal Reshade settings (Same as the GTX 970's as mentioned in previous posts) - 4k (2160p) + Nvidia Inspector Forced AA: 4XOGSSAA+2XMSAA / Nvidia Inspector Forced AA: (0x12400010 Override | 4x MSAA + 4x TrSSAA)

    Reshade settings used (From previous post/attachement/MXAO @ 16 samples):


    View attachment 12135

    Loss of image quality down below will be apparent as the files were way to large to upload on here, thanks.

    1st up:

    (0x12400010 Override | 4x MSAA + 4x TrSSAA): (As a remainder, 1 GTX 970 = 19 Fps / 4 GTX 970's = 66-67 FPS:

    1 GPU:

    Doom 3 - 1 GPU -  4k - Reshade - 4XMSAA+4XTRSSAA.JPG

    4 GPU's:

    Doom 3 - 4 GPU's -  4k - Reshade - 4XMSAA+4XTRSSAA.JPG

    Time to further stress the Gpu's further with:

    Nvidia Inspector Forced AA: 4XOGSSAA+2XMSAA:

    1 GPU:

    Doom 3 - 1 GPU - 4k - reshade + 4xOGSSAA+2XMSAA.JPG

    4 GPU's:

    Doom 3 - 4 GPU's - 4k - reshade + 4xOGSSAA+2XMSAA.JPG

    As you can see, Nvidia Inspector Forced AA: 4XOGSSAA+2XMSAA gave overall better scaling results as its further putting more stress on to the Gpu's.

    Don't mind the Cpu Wattage on the 2nd image, there seems to be some kind of reading problems, totally disregard it, thanks.

    I have not conducted any long term testing as of yet, just wanted to see if can scale as well as the GTX 970's, yup, sure can, just be sure to really heavy duty reshade + @ least 4k at a bare minimum for your screen resolution as well as: Nvidia Inspector Forced AA: 4XOGSSAA+2XMSAA, don't use the other testing AA method as you won't stress the 980ti's enough in certain given areas, thanks.
     
    Last edited: Jul 10, 2022

  11. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    GTX 970 - 4-WAY - SLI - CATZILLA 4K (REGISTERED VERSION) - MY PREVIOUS, WORLD RECORD SCORE FOR THE X4 970's, BROKEN 3 TIMES - UPDATE + IMAGES - 7/11/22:

    New scores for this particular benchmark:

    Previously of mine, Current World Record holder for the X4 970's = 16835 Points, I broke my World Record, 3 times, but for some reason, I can not submit my benchmark results, having an server issue on their end it seems, I'll give it a try later, if successful, I'll update it.

    For reference:

    System Specs:

    GPU's Used: 4 X Gigabyte Windforce (With Brutal Bios Modifications) / Air Cooled, No watercooling was used on any of the gpu's.

    GPU Speeds: Absolute Insane 1589mhz on all Gpu Core's / Vram Speeds: 7840mhz (Vram Voltage was also increased).

    OS: Win 10 X64 Build 15063 (RS2-RTM) / Update Verison: 2679

    Graphic Drivers: Modified Quadro Drivers: 392.68

    CPU: Intel 10900X @ 5.1GHZ (Benchmark mode only)

    Ram: 32GB Corsair Dominator 3600mhz @ 4000mhz

    PSU: Dual MKIII 1200W Silencer's (Classic OCZ/Pc Power & Cooling Psu's)


    1st attempt:

    1.JPG

    2nd attempt:

    2.JPG

    3rd attempt:

    3.JPG

    Again, if I can submit my results, I'll revise this post with more details, thanks for reading.

    EDIT:

    Something just came up after I had to do some re-installation of windows, one of my GTX 970's, after driver install, had an yellow exclamation mark within device manager, if your into heavy bios modding, usually pertains to voltage settings, I reverted everything back to stock level, there is still an yellow exclamation mark, I had to determine which card it was, it was primary display card, which is the 1st card.

    Thankfully, I have 2 other Gigabyte Windforce G1 GTX 970 cards left, one clearly doesn't work, the other one works without any issues that I can see (If you all remember from many posts ago, this card had an Sli issue, where the screen will do massive amounts of flashing while SLI is enabled, back then, I said I routed that it was indeed infact the card itself and not the connectors, well apparently, it is functioning as we speak.

    Going to back off a bit on bios modding for awhile, not going to do another test run for awhile, at least I broke my record 3 times, well worth it to me.
     
    Last edited: Jul 12, 2022
    Kozary likes this.
  12. Kozary

    Kozary Master Guru

    Messages:
    292
    Likes Received:
    50
    GPU:
    2x 2080Ti NVLink
    Are you overclocking the bus speeds and such in the bios? Does it effect overall overclock much i heard that going to far of these can make core clocks end up lower. Right now i have slightly increased a few of those in my 980ti BIOS.

    The msi gaming gtx970s i have do not allow increased vram voltage or anything above 1.325v. These cards degraded heavily over the years pushing 220w or 150% power limit. Went from 1540mhz stable to 1460mhz and memory clock halved as well sometimes even that is unstable. One of them eventually fried the VRM actually melting part of the pcb. I just used a knife to cut it off and the card continues to work to this day. Over 1000hours of load easily at maximum possible voltages and power limits allowed within the bios.
     
  13. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    1.325V?, very very high, mine was running @ 1.275V (Gpu Core's), that is still extremely high.

    Vram Voltage: 1.262V

    Also, program: Maxwell bios editor II, under the voltage tables, the one, all the way to the bottom, next to where it says fixed, this is apparently the Vram voltage.

    1.325V, man that's alot for these card's, even under extreme watercooling.
     
  14. Kozary

    Kozary Master Guru

    Messages:
    292
    Likes Received:
    50
    GPU:
    2x 2080Ti NVLink
    Yeah i know the voltage table your talking about. It did absolutely nothing for me no difference in tdp, overclock or memory chip temps measured with a probe taped onto it. Even if you edit the bios at the end of the day its only telling the vrm what to do, you will find that on a gtx970 you can increase core voltages to 1.325v and afterburner will report 1.275 but if you measure it, it will indeed be close to 1.33v. I would say this is pretty safe honestly i only had air cooling and regularly saw temps into the high 70s not to mention the MSI cards had pathetic vrm cooling this is probably the only reason why it failed. Also about vram voltage it probably wont help a overclock if you have hynix memory. Samsung memory did benefit more from it, my 980tis with samsung can reach 8200mhz (stock volt) no problem apparently also has tighter timings.

    on a 980ti if you try and increase voltages in bios you cant even load a modern driver. You have to roll all the way back to version 340.xx or something i can't remember when they brought out the voltage restriction though, but you wont see any gains without ln2 cooling. 1.25v is plenty for maxwell. I've seen graphs of 980tis on ln2 and transient loads reach over 800w with idle at 250watts crazy stuff.

    The stuff i mentioned was on the boost states page; gpc, xbar, l2c, and SYS. As far as i understand some of those are like increasing cache speed on the CPU. There is a forum post speculating what they are but with no back to back benchmarks. I for one set them all to 1430mhz and it had no adverse effects on clock speed achieved which is 1468mhz gaming and 1520mhz benching across 4 cards no single card will go faster then 1488 stable even new.


    Question for your 980ti testing If you have edited a 980ti to run increased voltage did you have any luck with a quadro driver?
     
  15. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    Yes, tested the 980ti's with modified quadro drivers 392.68, works perfect but one problem, your modified voltages cuts down quite a bit if 4-way sli is enabled, instead of 1.249V (MAX for these card's to show up within the osd), all 4 gpu's has separate readings, whereas, if I went all out on the gtx 970's, voltage tops on all 4 gpu's will read @ 1.212V.

    In regards about:

    gpc, xbar, l2c, and SYS:

    I didn't really notice to much of a performance differences, when changing your boost clock, the max speeds on each of these parameters, it will match them all up.

    On one forum I read, best to keep the maximum portion, all matched up, otherwise, you will experience more pronounced stuttering in your desired testing applications.
     
    Last edited: Jul 12, 2022

  16. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    GTX 970 - 4-WAY SLI - BIOS - LOW VOLTAGE MODIFICATIONS + TEMP RESULTS - UPDATE - 7/15/22:

    Before I go any further, several post ago, I mentioned that I had some problems after the world record breaking attempt run with Catzilla 4k (Registered version), I managed to bring both gpu's back to life, both had bios corruption, my other primary GTX 970 that I had for years, that one is no good, I can not exactly remember if I had the original GTX 970 primary card during the Catzilla 4k run, but nontheless, the card is already showing its sign, been abused alot over the years.

    So 4-Way Sli is now back in full operation, 0 problems thus far.

    And back to the Low voltage modifications through the bios.

    I managed to under volt all 4 gpu's successfully, however within the afterburner osd, things are out of order...

    This is how I flashed the gpu's in order:

    EDIT:

    Forgot to mention earlier: All 4 gpu's are running at 1480mhz (Gpu core) at the below mentioned voltages, 1480mhz on these specific GTX 970's are my personal sweet spot.

    GPU1 @ 1.168V

    GPU2 @ 1.137V

    GPU3 @ 1.125V

    GPU4 @ 1.125V

    Anything lower on each gpu = Instability, so this is basically the lowest I can go per card.

    Now, within the osd, here's how it reads:

    Per remainder to all, if you watch my Warhammer 40k - Inquistor/Martyr video on my front page, you can see that the main primary video says GPU3 and told everyone to totally disregard it, GPU3 is infact = GPU1, so this is why the osd readings are completely way off.

    Readings through OSD:

    GPU1 @ 1.125V

    GPU2 @ 1.137V

    GPU3 @ 1.168V

    GPU4 @ 1.125V

    So basically just swap GPU3 & GPU1, that would be the correct order, again, for some reason, the motherboard that I am using (Gigabyte X299X Aorus Master), main gpu shows under Gpu3 when in reality its really Gpu 1, this I already confirmed it through GPU-z awhile back.

    Now for temp results, I ran 2 tests/back to back of Unigine Heaven 4.0 @ 4k (2160) / Max Settings + 8XAA


    Room Temp while testing: Hot 82-84 Degrees Fahrenheit (NO AC was used/Lights on, Door slightly open).

    Results within the afterburner OSD are as follows:

    (All 4 Gpu's, Fan Speed was set to very strict 100%, also my infamous spacers was used, inbetween each gpu, for fans: 4X140mm Yate loon fans (Best ones they make) on top of the Gpu's + 1 140mm yate loon fan (Their best ones)/acting like an exhaust, behind the 4 gpu's for additional cooling.)

    GPU1 @ 1.125V - 64c

    GPU2 @ 1.137V - 64c

    GPU3 @ 1.168V - 63c

    GPU4 @ 1.125V - 69c


    Thanks for reading.

    EDIT ON GPU-z / Perfcap Reason:

    Just heads up to all, within Perfcap Reason under full blown load / 4 Gpu's = No Reason, in other words, its left out completely blank.

    EDIT In a room temps of 70 degrees Fahrenheit:

    Room temps of 70 degrees Fahrenheit:

    I have 2 images, 1 showing 2 runs of Unigine Heaven 4.0 @ 4k (2160p) / Max settings + 8XAA:

    Unigine low voltage.JPG

    Just added reference on the above image results, I have seen 4th gpu high as 55c on this test.

    And the next one, this one: Catzilla 4k (Registered version), under the Raymarch test, which is puts tons of heat during this test, (I took snapshot, right before the benchmark was going to end, so the picture will be kinda dark):

    Catzilla 4k low voltage.JPG

    I think I can improve the temps on GPU4, sometimes to much spacing when using spacers can really hinder the temps, as when using to much gap inbetween each gpu, the heatsink making contact to the gpu will be out of place, thus in the end, increased temps, if to much spacing, and depending on the cards heatsink condition, your temps can really skyrocket, the moment you do your test, but it is known to cut 8c per gpu tops in reduction of temps.

    Thanks for reading.
     
    Last edited: Jul 17, 2022
  17. Kozary

    Kozary Master Guru

    Messages:
    292
    Likes Received:
    50
    GPU:
    2x 2080Ti NVLink

    Looks like you lost some clock speed?
     
  18. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    Nope, 1480mhz is the speeds I usually use, personal sweet spot, another one is 1506mhz, but far as everyday testing, speeds are @ 1480mhz.
     
  19. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    GTX 970 - 4-WAY SLI - GAME #13 - COUNTDOWN TO THE 980ti's - BALDUR'S GATE III (LATEST PATCH) + RESHADE - GAME ON HOLD INDEFINITELY UNTIL FINAL BUILD - UPDATE - 7/17/22:

    Patch 8 just got rolled out very recently, gave this a try along with my custom made bits that was made specifically for Patch #7 (Previous one), absolute no go, although the scaling is A+, The massive screen flashing, I can no longer control at the moment, in Patch #7, All you had to do is increase your shadow quality a bit to keep it under control, this patch = Nope, I also tried my resume saved game technique in hopes to keep the graphical problems under control, nope, no matter the timing = no go, also, when choosing Rogue or Thief under SLI = parts of the body is missing among other things.

    I also did some more tuning a bit = no go at the moment.

    I made a final decision to put this game on hold indefinitely as I can not keep up with the changes on every patch, it breaks my custom tuning every time.

    Once the final build is released, I will work specifically on that particular build, once its released that is and nothing else afterwards, all these updates = breaks Sli custom tuning all the time, to much changes.

    Again, Scaling = Absolute fantastic, just the patch updates = breaks my custom tuning all the time, I am not going to custom tune for every patch update, therefore I am going to wait very patiently until the final stable release build and work on that build.

    Will keep everyone posted until then, Game #13 is on indefinite hold until further notice.

    Thanks everyone for reading.

    EDIT:

    Forgot to mention that, for those who seriously wants to tune this game and want an much more easier approach, please use this version:

    Patch #5 (Build version: v4.1.1.1224125.Hotfix.3)

    Forewarned though, despite an older build, this is still for experienced Sli tuners only.

    For those who are into no holds barred/extreme tuning, might want to jump on build version after patch #5 but I would stay far away on patch #8 for the time being, thanks.

    Just one word of extreme Caution:

    Going from Patch #5 to these newer builds = Light years of difference, when it comes to tuning difficulty, so be very weary on which patch version you jump on, thanks.
     
    Last edited: Jul 18, 2022
  20. A M D BugBear

    A M D BugBear Ancient Guru

    Messages:
    4,393
    Likes Received:
    630
    GPU:
    4 GTX 970/4-Way Sli
    GTX 970 - 4-WAY SLI - HELLBLADE + RESHADE - DESERT AREA (BEGINNING) - ENGINE.INI - UPDATE - 7/17/22:

    Back on to this game, been away for a bit.

    As before, I am trying to re-work this game once again from the ground up, last area that I was working on was the desert area (Beginning of the game).

    Scaling is Excellent, with and without reshade.

    There is one particular drag spot in this area, the penalty on this, greatly depends on what settings/resolution/reshade settings your using.

    Under 1 Gpu mode, it seems doesn't affect that much performance, seems only 1 fps, or at times, it stays exactly the same, but under 4-Way Sli, without reshade, you would have to crank your resolution way up (At least 5k/2880p) + Reshade's TAA to greatly lessen the gap on your scaling loss after these very infamous drag spots, and yes, already checked under Tri-Sli operation, that to takes a similar hit as well.

    For example, @ 5120x2880p, no reshade, custom engine.ini, custom made bits, custom quality settings, I get about 22 fps under 1 gpu, right before the drag spot, after the drag spot, you may or may not lose an fps, if so, like 1 fps, under 4-Way Sli, before drag spot, I get like 86 to 87 fps range, after drag spot, goes down to 78 to 79 fps range, again, weaker the resolution and settings, bigger gap loss in your overall scaling performance, so be sure to have your resolution very high, 980ti tuners out there, this also means you have to take your resolution,etc to whole another level, personally, I would just use very heavy reshade filters and probably at least 5k(2880p) resolution along with it.

    Reshade users/tuners, depending on how heavy your settings are, you will still experience some loss in performance after this particular drag spot in question, so make sure you have brutal reshade settings turned on to make up for the scaling loss after this particular drag spot.

    Once again, I been really investigating this issue for months, I still have not find a complete solution as of yet, only performance increase in tad amounts, there is other areas, that I previously mentioned in posts way earlier, explaining that, there are other drag spots in this game, that the hit is huge, again, trying to find a workaround, other then tad boost in performance, no complete work around as of yet.

    Current Engine.ini update (Only for this area and throughout, if earlier areas, please use my earlier specified engine.ini from previous posts, thanks):

    [/Script/Engine.RendererSettings]
    r.EnableAsyncComputeTranslucencyLightingVolumeClear=1
    r.UseParallelGetDynamicMeshElementsTasks=1
    MaterialBaking.UseMaterialProxyCaching=1
    r.CompileMaterialsForShaderFormat=1
    r.Streaming.LimitPoolSizeToVRAM=1
    r.FXSystemPreRenderAfterPrepass=1
    r.PostProcessing.PreferCompute=1
    r.AlsoUseSphereForFrustumCull=1
    r.AmbientOcclusion.Compute=1
    r.DriverDetectionMethod=4
    r.OnlyStreamInTextures=1
    r.HZB.BuildUseCompute=1
    r.oneframethreadlag=2
    r.EarlyZPassMovable=1
    r.ShaderPipelines=1
    r.VirtualTexture=1
    FX.AllowGPUSorting=1
    FX.TestGPUSort=3
    r.HZBOcclusion=0 (With Reshade, if no reshade, please set it to =1, thanks)
    r.GTSyncType=1
    r.SSR.Stencil=0
    r.SSR.Quality=2
    r.EarlyZPass=2
    r.SSR.Temporal=1
    r.MotionBlur.Max=0
    r.MotionBlurQuality=0
    r.DefaultFeature.MotionBlur=0
    r.Tonemapper.GrainQuantization=0
    r.Streaming.FullyLoadUsedTextures=1

    [SystemSettings]
    r.MotionBlur.Max=0
    r.MotionBlurQuality=0
    r.DefaultFeature.MotionBlur=0
    r.Tonemapper.GrainQuantization=0
    r.DefaultFeature.AntiAliasing=0
    r.PostProcessAAQuality=0

    Also that reminds me, extremely important for non-reshade Sli tuners, please have:

    r.HZBOcclusion=

    Set to: r.HZBOcclusion=1

    Or else you will have weak scaling without Reshade.

    With reshade Sli tuners, please have it set to =0 as this will help your performance just a tad, if your running brutal reshade settings that knocks your fps way down and see some scaling loss, try this, it may help just a tad, thanks.

    One more thing about the games, built in: TAA, I would only use it under Tri-Sli operation as it has no graphical problems, only slight blur trail, you can use Reshade's TAA if you want with slight blur trail as well, if 2-Way or under 4-Way Sli, please use Reshade's: TAA as an direct alternative to the games built in TAA as you will have major problems, thanks.

    Images on this area in question later.
     
    Last edited: Jul 18, 2022

Share This Page