Intel HD530 + Windows 10 Bugs #56668

New Issue

Clément Foucault · 2018-09-03T16:38:35+02:00

Clément Foucault commented

2018-09-03 16:38:35 +02:00

** ~~Bug 0~~: We have to read gl_VertexID for the vertex shader to output correct position for lighting pass (prepass seems to work).
If not, it seems the vbo data is not fetched correctly (if not at all).
This makes the object not cover the previous frame and lead to ghosting artifacts.
Workaround exists 0a26c122ed.

** ~~Bug 1~~: Environment Lighting does not work.
It's just solid black.
The texture sampling is correct. It seems to come from a problem before the diffuse convolution (encoding and decoding data is ok). Maybe that's caused by the downsampling shader. Need further investigation.
Workaround exists 2d084d4ec4.

** ~~Bug 2~~: Dot corruption (NaNs maybe) all over the mesh and seems to folow the mesh shading.
Does not appear if Subsurface scattering is ON. Does not appear if one mesh is rendered with another BSDF (other than emission). Corrupted pixels seems to follow lighting. Not cause by shadows
Enabling SSS or SSR fixes it. Enabling AO does not!
This bug affects render in a bad way (see #57331)
Caused by sampling of the maxZbuffer.
This happen even if dead codes samples it: for instance, code after a return in a function or in a function which outputs are not used.
Not sure if #ifdefs work at deleting codes. #if 0 does however.
If sampling is bypassed (bypassing Contact Shadows, AO (if not enabled uses the maxZbuffer as input), Refraction Raytrace) the glitch goes away.
Changing the texture bound to maxZbuffer (and horizonBuffer if AO is disabled) fixes it (i.e.: using the velocity buffer) but breaks the effects.
Changing maxZbuffer type to RGBA32F does not fix it and using DEPTHCOMPONENT24 breaks all raytracing effects.
Fixed by using glFlush() 117bc96ce8.

** ~~Bug 3~~: Sampling texture with the OCIO shader is sampling the texture with the wrong coordinate in the lower left quarter, and a black rectangle in the lower left quarter of the previous quarter (see picture).

Sampling with textureLod mip0 changes nothing.
Changing texture format does not change anything.
Does not appear all the time.
Enabling SSR exhibit the problem (sampling main/effect buffer).
SSRs seems to be correct but likely not (sampling the depth pyramid).
Debug MaxDepth Pyramid have the same problem (sampling the depth pyramid).
Other debug modes does not have the problem.
Disabling Mipmaps does not change anything.
AO is broken (sampling the depth pyramid the same way as OCIO).
SSS does work.
Bloom too.
Workaround exists 2d084d4ec4.

** Bug 4: Irradiance Grid have flickering dots in object engine overlays

** ~~Bug 0~~: We have to read gl_VertexID for the vertex shader to output correct position for lighting pass (prepass seems to work). If not, it seems the vbo data is not fetched correctly (if not at all). This makes the object not cover the previous frame and lead to ghosting artifacts. Workaround exists 0a26c122ed. ** ~~Bug 1~~: Environment Lighting does not work. It's just solid black. The texture sampling is correct. It seems to come from a problem before the diffuse convolution (encoding and decoding data is ok). Maybe that's caused by the downsampling shader. Need further investigation. Workaround exists 2d084d4ec4. ** ~~Bug 2~~: Dot corruption (NaNs maybe) all over the mesh and seems to folow the mesh shading. Does not appear if Subsurface scattering is ON. Does not appear if one mesh is rendered with another BSDF (other than emission). Corrupted pixels seems to follow lighting. Not cause by shadows Enabling SSS or SSR fixes it. Enabling AO does not! This bug affects render in a bad way (see #57331) Caused by sampling of the maxZbuffer. This happen even if dead codes samples it: for instance, code after a return in a function or in a function which outputs are not used. Not sure if #ifdefs work at deleting codes. #if 0 does however. If sampling is bypassed (bypassing Contact Shadows, AO (if not enabled uses the maxZbuffer as input), Refraction Raytrace) the glitch goes away. Changing the texture bound to maxZbuffer (and horizonBuffer if AO is disabled) fixes it (i.e.: using the velocity buffer) but breaks the effects. Changing maxZbuffer type to RGBA32F does not fix it and using DEPTHCOMPONENT24 breaks all raytracing effects. Fixed by using glFlush() 117bc96ce8. ** ~~Bug 3~~: Sampling texture with the OCIO shader is sampling the texture with the wrong coordinate in the lower left quarter, and a black rectangle in the lower left quarter of the previous quarter (see picture). ![40749463_246339889401771_7149057786716880896_n.png](https://archive.blender.org/developer/F4539715/40749463_246339889401771_7149057786716880896_n.png) Sampling with textureLod mip0 changes nothing. Changing texture format does not change anything. Does not appear all the time. Enabling SSR exhibit the problem (sampling main/effect buffer). SSRs seems to be correct but likely not (sampling the depth pyramid). Debug MaxDepth Pyramid have the same problem (sampling the depth pyramid). Other debug modes does not have the problem. Disabling Mipmaps does not change anything. AO is broken (sampling the depth pyramid the same way as OCIO). SSS does work. Bloom too. Workaround exists 2d084d4ec4. ** Bug 4: Irradiance Grid have flickering dots in object engine overlays

Clément Foucault self-assigned this 2018-09-03 16:38:35 +02:00

Clément Foucault commented

2018-09-03 16:38:35 +02:00

Added subscriber: @fclem

blender-admin commented

2018-09-03 16:38:35 +02:00

#57331 was marked as duplicate of this issue

Germano Cavalcante commented

2018-09-03 19:08:26 +02:00

Added subscriber: @mano-wii

Clément Foucault commented

2018-10-23 18:09:33 +02:00

Added subscribers: @MathieuDijkstra, @ChristopherAnderssarian, @lichtwerk

Mathieu Dijkstra commented

2018-11-13 18:06:52 +01:00

okay my Intel® UHD Graphics 630 seems to work 4 now on the blender-2.80-de36a2f99b3 build no crashes so far also it dos the rendering of the cube okay :)
without ssc and ssr on and i tryed every brush in sculpt mode

im happy to start on blender 2.8 :) :) :) keep up the good work

okay my Intel® UHD Graphics 630 seems to work 4 now on the blender-2.80-de36a2f99b3 build no crashes so far also it dos the rendering of the cube okay :) without ssc and ssr on and i tryed every brush in sculpt mode im happy to start on blender 2.8 :) :) :) keep up the good work

Clément Foucault commented

2019-06-14 15:20:50 +02:00

Changed status from 'Open' to: 'Resolved'

Clément Foucault closed this issue

2019-06-14 15:20:50 +02:00

Clément Foucault commented

2019-06-14 15:20:50 +02:00

To my knowledge, all these bugs were fixed. Closing.

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Intel HD530 + Windows 10 Bugs #56668