NVIDIA GameWorks Integration

[anonymous_user_bcf66e2c]

@WuRuiHeng;376324

@
could you tell me where I can download it ?
and…could i package project with your version.

works on the NVIDIA version, which is located : and the VXGI branch in particular is: /tree/vxgi4.8.2

[WuRuiHeng]

thank you so much!!! is a staff of nvidia!!!

[anonymous_user_7c71c8ec]

@Blakblt;376198

So apparently HDR emittance was the culprit, I just didn’t notice because after changing the setting it took about 15 seconds for the performance to improve. … It only appeared to add about 100MB to the memory, and It didn’t cap out (the memory usage fluctuated up to 1911MB at one point, 1856MB was just where it sat most of the time).

Thanks for the data. It looks like you actually are running out of video memory, although GPU-Z shows it’s not full. Just not all the data is resident in video memory, which causes massive swapping to system memory over PCI-E.
The actual difference in vidmem requirements between HDR and non-HDR emittance for MapSize=128 is close to 700 MB.

's what I get on a 4 GB GPU (Win10) in non- mode, with a very simple scene in UE4:
MapSize=64, HDR=0: 1565 MB
MapSize=64, HDR=1: 1652 MB
MapSize=128, HDR=0: 2073 MB
MapSize=128, HDR=1: 2739 MB

@Blakblt;376273

BTW, does anyone have any suggestions as how to scale GI down to mid-range hardware? My workflow doesn’t use any real-time lighting, so for low end I just turn off GI and let skylighting do the work, and for high end (say, GTX 760 and up) I can scale the GI to 4ms and still have decent quality, but what would be the best for everything in between? Would I be better off switching to LPV, or lowering the MapSize to 32 and increasing the tracing sparsity to 4?When VXGI is lowered that far it kinda looks like LPV with even lower quality, but designing levels that don’t appear different with 2 GI modes isn’t a good option either.

Isn’t MapSize=64 fast enough?
You can also try reducing the number of clipmap levels (r.VXGI.StackLevels) and scale Range accordingly.
Setting StackLevels to 1 is an interesting special case, which removes some quality issues that come from the clipmap, such as banding, but some other implicit settings aren’t optimized in the currently released version; they will be in the next version.

[anonymous_user_7c71c8ec]

@;376328

works on the NVIDIA version, which is located : and the VXGI branch in particular is: /tree/vxgi4.8.2

Right; but the latest version is in “VXGI” branch once again, not “vxgi4.8.2”. I’ll ask the repo maintainer to delete the latter branch to avoid confusion.
/tree/VXGI

[anonymous_user_bcf66e2c]

Oops my bad. Will make sure to point people to the right one in future

[anonymous_user_e6436d8c]

@;372899

I always recommend using the ZIP file. If you just see spheres flying around with no collision even in PIE then your GPU most likely doesn’t support Flex. According to the Flex documentation it requires a 660 or better GPU, my 560 TI does not support it.

So is the reason, I have the same graphics card…

[Rallii]

@;312366

Okay I have updated my merged branch to the latest 4.8 build. You can get the updated branch : https://github//UnrealEngine/tree/4.8_NVIDIA_Techs

is just the same as the preview 3 one, contains VXGI + HairWorks + FLEX + HBAO+. Now also contains 's POM material function. I will add Turbulence and WaveWorks (If I can) and also the latest LPV improvements from Lionhead

Thank you very much for , ! You are a true contributor, good sir!

[WWWWWWWWWWWWWWWWWWWW]

Unity 5.2
https://github/unity3d-jp/NVIDIAHairWorksIntegration
NVIDIA HairWorks Integration for Unity 5.2

[Blakblt]

@;376502

Thanks for the data. It looks like you actually are running out of video memory, although GPU-Z shows it’s not full. Just not all the data is resident in video memory, which causes massive swapping to system memory over PCI-E.
The actual difference in vidmem requirements between HDR and non-HDR emittance for MapSize=128 is close to 700 MB.

's what I get on a 4 GB GPU (Win10) in non- mode, with a very simple scene in UE4:
MapSize=64, HDR=0: 1565 MB
MapSize=64, HDR=1: 1652 MB
MapSize=128, HDR=0: 2073 MB
MapSize=128, HDR=1: 2739 MB

Thanks for the help !

@;376502

Isn’t MapSize=64 fast enough?
You can also try reducing the number of clipmap levels (r.VXGI.StackLevels) and scale Range accordingly.
Setting StackLevels to 1 is an interesting special case, which removes some quality issues that come from the clipmap, such as banding, but some other implicit settings aren’t optimized in the currently released version; they will be in the next version.

At point I’m not sure if MapSize6=64 is fast enough. I haven’t gotten a to test VXGI on my laptop, I’m just going off of what I’ve seen throughout thread. Scaling the range took away a lot of quality and added very little performance for me, but I haven’t messed with the clipmap because I wasn’t sure how it worked. I’ll try that next.

[Blakblt]

Ok I tried messing with the StackLevels, and it’s performance impact varied greatly per scene. One trend I noticed though was that changing StackLevels from 1 to 2 would harm the performance relatively drastically (on some occasions 5ms) but subsequent levels had less and less of an effect. In some indoor scenes I even raised the StackLevels to 6 without a noticeable performance impact. Also I would like to quickly mention that all of the data I took was taken in-editor at 1080p. I retested in a standalone window and some of those tests nearly doubled in performance. The exceptions were StoreEmittanceInHdrFormat, which ran stupidly slow no matter what I did, and MultiBounce, which took more processing power to run than it was worth (would love to see that optimized a bit, it typically doubled the GPU time!)

[CharlestonS]

Have there been any tests with VXGI in UE4 running in DX 12 on the 980 TI?

I just ordered my MSI GTX 980 TIs and i’m wondering what kind of performance to expect.

So far i have run it on dual 560TIs, a GTX 770 and a GTX 780, the 560TIs and the 770 get around 0.2 fps in the sci-fi hallway but the 780 gets over 20fps constant but from what I’ve read VXGI is targeted for 900 series running in , is that correct?

[Rallii]

Hey ,
I’ve been trying to get FleX 0.9.0 into your version but I am not having any luck. As a noob it’s no surprise though! Are you planning to update FleX in the near future?

Thanks!
Keep up the work.

You as well, ! Thank you for your hard work

[anonymous_user_bcf66e2c]

Yes I do plan to update, but am waiting on for a version of VXGI and HBAO+ before I do so. Then I will update my merged branch to 4.9 and the latest versions of whatever is available from the NVIDIA branch.

[Rallii]

@;378323

Yes I do plan to update, but am waiting on for a version of VXGI and HBAO+ before I do so. Then I will update my merged branch to 4.9 and the latest versions of whatever is available from the NVIDIA branch.

Perfect! I will patiently wait and tinker with other things Thank you, .

[Blakblt]

@;378201

Have there been any tests with VXGI in UE4 running in DX 12 on the 980 TI?

I just ordered my MSI GTX 980 TIs and i’m wondering what kind of performance to expect.

So far i have run it on dual 560TIs, a GTX 770 and a GTX 780, the 560TIs and the 770 get around 0.2 fps in the sci-fi hallway but the 780 gets over 20fps constant but from what I’ve read VXGI is targeted for 900 series running in , is that correct?

VXGI is meant for any+ card, but certain features (hdr emittance) are better with . With the right optimizations the 770 can easily hit 40 fps in the sifi hallway.

Edit: VXGI also runs significantly worse in editor. In a standalone I got 60 fps on ShooterGame with specular tracing, diffuse tracing,and a map size of 128 with static lighting disabled just by disabling hdr emittance and removing the point lights. Multi-bounce still kills me though.

[tegleg]

@;262150

Hi , I have just compiled the new version of Flex, and I am getting a strange error, where the flex balls do not collide with anything, when opening the Flex Test maps. Are the maps included outdated or am I doing something wrong, I compiled the Substance plugin at the same time, but I don’t think that should interfere.

just built from the flex branch and is what i see.
my graphics supports cuda and has the latest drivers.

anything i can try?

thanks

edit:
CUDA compute capability 3.0 or greater

mine is 2.1
oh well

[CharlestonS]

Did you run a PIE session? If so and you still see no collision then what is the exact model of your GPU? Not all Cuda supportive cards support Flex and the you’re having is exactly what happens when trying to run Flex on a non-supportive GPU.

@tegleg;379727

just built from the flex branch and is what i see.
my graphics supports cuda and has the latest drivers.

anything i can try?

thanks

edit:
CUDA compute capability 3.0 or greater

mine is 2.1
oh well

[iuhiuh]

@Blakblt;378562

VXGI is meant for any+ card, but certain features (hdr emittance) are better with . With the right optimizations the 770 can easily hit 40 fps in the sifi hallway.

Edit: VXGI also runs significantly worse in editor. In a standalone I got 60 fps on ShooterGame with specular tracing, diffuse tracing,and a map size of 128 with static lighting disabled just by disabling hdr emittance and removing the point lights. Multi-bounce still kills me though.

yea the performance is way worse in the editor than on the standalone

[CharlestonS]

I just bought and installed an MSI GTX 980 TI 6G and now i get over 90 fps constant in the VXGI SciFi Hallway so i would imagine any 900 series card would run it at playable frame rates.

It runs all of the GameWorks UE4 integrations at 90+ fps as well.

[Blakblt]

@;380002

I just bought and installed an MSI GTX 980 TI 6G and now i get over 90 fps constant in the VXGI SciFi Hallway so i would imagine any 900 series card would run it at playable frame rates.

It runs all of the GameWorks UE4 integrations at 90+ fps as well.

I was just about to test on my friend’s 980ti, so you saved me some time! To anyone that’s curious as to how well VXGI scales, in a standalone game I can get a stable 35 fps on my laptop, which has a GT 940m, and is actually using the Kepler architecture, not . The quality is also relatively high. I’m starting to think VXGI’s biggest bottleneck is actually hardware; performance seems to be quite scalable.