Fusion crashing when rendering

Been trying to isolate an issue where my computer just simply turns off when i hit render.

It seems when i check "Enable Lighting" it causes it. Any idea why this happens?..any help would be appreciated..thanks

Here is a screenshot. When it is enabled it turns my pc off http://prntscr.com/m21d0x

Could be an overheat. How long's it been since you cleaned the dust out of your PC?

Sudden shut-downs of this sort can also occur if your power supply is going bad—the video card tries to draw its full load, and the PSU can't deliver, so it shuts down.

Could be an overheat. How long's it been since you cleaned the dust out of your PC?

Sudden shut-downs of this sort can also occur if your power supply is going bad—the video card tries to draw its full load, and the PSU can't deliver, so it shuts down.

Havent cleaned the dust out for a while...system is new from jan this year......does enabling lighting use alot more rendering power?

This is what i get rendering without lighting enabled http://prntscr.com/m21m9k

Gone up a bit halfway through http://prntscr.com/m21n8g

Enabled shadows and hasnt crashed..so it seems it is definitely lights enabled that causes it http://prntscr.com/m21p99

Could be an overheat. How long's it been since you cleaned the dust out of your PC?

Sudden shut-downs of this sort can also occur if your power supply is going bad—the video card tries to draw its full load, and the PSU can't deliver, so it shuts down.

I also tried only only half the cpu like this using the affinity option http://prntscr.com/m21r4u was something i found might help, but computer still shut down.

Havent cleaned the dust out for a while...system is new from jan this year......does enabling lighting use alot more rendering power?

Yes, lighting is going to generally be the most taxing part of a 3d scene's calculation. Turning on shadows without lighting shouldn't have any effect since shadows are generated by the lights, and so the difference would be negligible.

The CPU is fairly unlikely to be the problem. Even running full-blast most CPUs are still draw relatively little power, and generate little heat, in comparison to the GPU.

How many subdivisions does your image plane have? If it's a lot, you might try reducing its resolution to see if the crash still happens.

Also, are you overclocked? Fusion 9 doesn't work well on an OC'd system. Vito (Con-Fusion) was having trouble shortly after release because he had a brand new, finely tuned computer that crashed continually. I'm not sure what the problem was, exactly, but when he tuned it back to factory spec, the crashes stopped.

How many subdivisions does your image plane have? If it's a lot, you might try reducing its resolution to see if the crash still happens.

Only 10 http://prntscr.com/m2hpew but not sure if that is to many anyway. I tried to render on my sons Rogue which is far less powerful system and it rendered easy.

No to be honest I dont even know how to overclock a system, but I did have this custom built and it was the first one they had built like this, so not sure if some settings are not right. I use the AMD Radeon and have tried the gaming and the professional drives which can be changed in this http://prntscr.com/m2hr7b but still seems to yield the same results.

One other issue I am facing is quality. I made the image in photoshop (which I have experience in) and it is a nice smooth hires graphic 1920x1080 (tried .jpg/png/targa. I have used that to bring into fusion and in 3d then added a fastnoise etc which you can see here https://s3-us-west-2.amazonaws.com/zeep ... _final.mov But the quality is shocking, can't figure out what I am doing wrong. As far as I can tell all the settings seem ok for rendering. I am wondering if it has something to do with the camera maybe?

Render settings http://prntscr.com/m2hv47
Camera settings http://prntscr.com/m2hvi5

With only 10 subdivs, we can probably knock thermal or power problems off the list. That's a very light-weight scene that shouldn't cause problems. When I'm doing displacements I frequently raise the subdivisions to either equal to or half of the resolution of the displacement map, and even with a relatively simple scene, that can stress the GPU.

The fact that you have that RoG hardware monitoring software available to you suggests to me that the system has potentially been overclocked. And since the scene renders fine on another PC, we can eliminate any possibility of a weird malformed input image causing the problem—there's nothing wrong with the Flow itself. That leaves a hardware issue. Do you get the same crash if you set the Renderer to Software mode?

I'm afraid that's as far as I can go without being physically present. You'll probably need to find someone with experience building their own PCs and/or overclocking to help you determine if the PC's been OC'd and how to turn that off. And if it's not an overclock problem, then I'm stymied.

As for the quality issue, it's difficult to tell whether what I'm seeing in the video is all due to compression (which appears to be significant) or whether the renderer might need to be adjusted. Some things to try: Add a ChangeDepth node right after the Loader and set it to float16. In the Image tab of the Renderer3D, set the Depth to float16. That will ensure that you get no banding as a result of the 3d scene. You might also increase the subdivs of that ImagePlane to 128 or so, just to get a little bit smoother displacement. That can make a difference sometimes.

Make sure the HiQ button is on. If the image still doesn't look good in the Viewer, then we can revisit it, but if it looks nice in Fusion but look bad in the render, then the problem is the codec. Solutions depend on how the end product is going to be used.

With only 10 subdivs, we can probably knock thermal or power problems off the list. That's a very light-weight scene that shouldn't cause problems. When I'm doing displacements I frequently raise the subdivisions to either equal to or half of the resolution of the displacement map, and even with a relatively simple scene, that can stress the GPU.

The fact that you have that RoG hardware monitoring software available to you suggests to me that the system has potentially been overclocked. And since the scene renders fine on another PC, we can eliminate any possibility of a weird malformed input image causing the problem—there's nothing wrong with the Flow itself. That leaves a hardware issue. Do you get the same crash if you set the Renderer to Software mode?

I'm afraid that's as far as I can go without being physically present. You'll probably need to find someone with experience building their own PCs and/or overclocking to help you determine if the PC's been OC'd and how to turn that off. And if it's not an overclock problem, then I'm stymied.

As for the quality issue, it's difficult to tell whether what I'm seeing in the video is all due to compression (which appears to be significant) or whether the renderer might need to be adjusted. Some things to try: Add a ChangeDepth node right after the Loader and set it to float16. In the Image tab of the Renderer3D, set the Depth to float16. That will ensure that you get no banding as a result of the 3d scene. You might also increase the subdivs of that ImagePlane to 128 or so, just to get a little bit smoother displacement. That can make a difference sometimes.

Make sure the HiQ button is on. If the image still doesn't look good in the Viewer, then we can revisit it, but if it looks nice in Fusion but look bad in the render, then the problem is the codec. Solutions depend on how the end product is going to be used.

Sorry i didn't respond I have been moving house.
So totally cleaned and dusted computer when I moved.
Problem still exists but I have found that it isn't the actual renderer that is the issue I think it must be either a display codec maybe or something else, because when i just touch the slider http://prntscr.com/m3tk80 like literally move it a tiny bit it cuts my computer off instantly. I havent been able to find out if it is overclocked or not but all the settings seem to be on default as far as I can tell. It could be a hardware issue but nothing tells me anything isnt working correctly. I do know the Radeon Vega Frontier I have from what I have read uses a ton of power but I know when they built it they allowed for this and beefed up the power supply.
It is just so wierd to me why it does this.

BTW I just messaged the guy that built it and asked about it being overclocked so waiting to hear back.

Oh also tried different settings and can't get any difference in quality when rendering so stumped on that. Rendered the file from my laptop and other pc.

If I could figure out how to make the quality better at least I can render from my laptop
Maybe you could take a look at the file and see if i have the best settings? would appreciate it.
https://s3-us-west-2.amazonaws.com/zeep ... ground.rar

I've given you everything I have on the crash issue. But I do have some workflow pointers.

ChangeDepth after the Loader. Putting it after the Saver does nothing because by the time the image gets to the node there, it's already been saved. But a better idea is to set the color depth up properly to start with.

File > Preferences > [comp] > Frame Format:

depth_preferences.png (28.25 KiB) Viewed 3875 times

In the Color depth section, seen above, set the Full render (at least) to 16 bit float per channel. It doesn't actually make a huge difference here, but when it comes to doing color corrections you'll get much better results from working in floating point rather than integer.

Do the same thing in the Globals, too. In future comps, the Globals will set the defaults for new comps, but any existing comps will already have format settings that take precedence over the Globals. You can quickly check to see the depth of your current image by looking at the upper-right corner of the image, which shows the dimensions and depth of the image. For instance, 1920x1080xint8 indicates a full HD, 8-bit integer image.

http://www.bryanray.name/wordpress/anatomy-of-an-image/

Next, when you're using a FastNoise as a displacement image, it should have an even higher depth to prevent terracing:

terracing.png (466.5 KiB) Viewed 3875 times

This artifact begins to show up when the subdivisions of the image plane are quite high. I have it set to 2048. It happens because there aren't enough intermediate values, causing the displacement to "step" from one value up to the next rather than creating a smoothly interpolated gradient. This is why you should always render world position, z-depth, normals, etc in 32-bit float rather than only half-float, which is sufficient for the RGBA.

Finally, I get decent results from changing the codec from h264 to Apple ProRes 422 HQ. I think your quality issues are entirely due to the rendering codec. There's a little bit of softening coming from the displacement, but there's not really any way around that—if you stretch an image, you have to expect it to get a little blurry. You could sharpen it, but

https://we.tl/t-A1OkZ89P49

There's a link to two exports. One is h264 and the other is ProRes. You can see at a glance how much h264 munges things up.

Wow mate that's some real valuable information it is much appreciated.

I use windows and don't see the prores export option but i see BMD has a license for it, so I am assuming it is in the studio version?. I have been making sure Fusion is going to be one of my go to programs for what I do before I invest in the paid version, but that right there would sway me to if prores is an option.

I did a bit of googling and saw a comment about using DNxHD 444, so I changed all the settings in the nodes as you described and then used the DNxHD 444 as the export codec and the video came out awesome in comparison to what I have been getting. So thanks for all the detailed info I have been learning alot.

So what I do is make video tutorials mainly around crypto currencies, so mostly youtube orientated. One of the main programs I use and have been for many years is Camtaisia for screen recording. I started using Davinci because of some neat effects and options that can be created, but once you get a taste for node based motion graphics Fusion(stand alone) is where it's at, because of performance and layout .
So this particular animation I have showed you is actually a backdrop for a video tutorial. I will be overlaying all the screen recording videos over the backdrop, which poses other issues, for example you cant export prores or DNxHD from Camtaisa, only H.264 or AVI ....So what I am thinking to preserve the animation background is to export from fusion in DNxHD 444 and import into Davinci then import my screen recordings into Davinci, construct the video tutorial, then when complete export to DNxHD 444 and upload to youtube.......I could i guess import the screen recordings and do it in Fusion, but Davinci is better for handling audio and video editing in general.....what do you think?

Just ran a upload test....original file on the right is exported in DNxHD 444 played in VLC.
The one on the left is after it has been uploaded to youtube.
http://prntscr.com/m3y93j

I new they compressed but didn't think it would be that bad. I read some comments in forums that if you have the time then upload a higher quality file and it will make a difference, but that doesn't seem to be the case...

Yes, Fusion Studio is licensed for ProRes export. A year ago it was the only licensed exporter for ProRes on Windows and Linux. Adobe managed to join the party a little later than Blackmagic, but even so, Fusion is still way cheaper than Premiere, and it's well worth the cost, particularly since even if Apple yanks the license later, Fusion 9 will still have it.

YouTube's going to hose your quality any way you look at it. The best you can do is to make sure it's only getting encoded once by encoding it to YouTube's standard right out of the editor. At least that way you can be sure that what the end user sees is pretty close to what you see, even if it's not as great as you'd like.

Now, if I were setting up your workflow:
Export h264 from Camtasia.
Export tga sequences from Fusion.
Import both to Resolve and edit.
Export to h264, using whatever settings are recommended by YT

I'm not real strong on video formats because my eternal advice is "Don't use video formats." But I'm a VFX artist, and I therefore don't have to deal much with the end product, only with intermediate images.

Something that might work to improve the banding is to add a little bit of noise over the top of everything. Just enough to be barely perceptible. This is called dithering. The noise causes the compression algorithm to spread the banding across a wider area in a less uniform manner, which disguises the fact that you don't have enough colors for the gradient.

Yes, Fusion Studio is licensed for ProRes export. A year ago it was the only licensed exporter for ProRes on Windows and Linux. Adobe managed to join the party a little later than Blackmagic, but even so, Fusion is still way cheaper than Premiere, and it's well worth the cost, particularly since even if Apple yanks the license later, Fusion 9 will still have it.

YouTube's going to hose your quality any way you look at it. The best you can do is to make sure it's only getting encoded once by encoding it to YouTube's standard right out of the editor. At least that way you can be sure that what the end user sees is pretty close to what you see, even if it's not as great as you'd like.

Now, if I were setting up your workflow:
Export h264 from Camtasia.
Export tga sequences from Fusion.
Import both to Resolve and edit.
Export to h264, using whatever settings are recommended by YT

I'm not real strong on video formats because my eternal advice is "Don't use video formats." But I'm a VFX artist, and I therefore don't have to deal much with the end product, only with intermediate images.

Something that might work to improve the banding is to add a little bit of noise over the top of everything. Just enough to be barely perceptible. This is called dithering. The noise causes the compression algorithm to spread the banding across a wider area in a less uniform manner, which disguises the fact that you don't have enough colors for the gradient.

Awesome thanks for that Targa tip..and noise tip I added a bit and also a sharpen modifier on the actual image in Photoshop........

So this is what i did after a bit of research....I found out that if you can force youtube to code your videos in the VP9 format it makes the quality a tad better......so what i did was increased the size of the image resolution to my native screen size i screen record in which is 2560x1440 which makes it more than 1080p and forces the VP9 when you upload.....i exported targa and set to 50 in sequence offset, that was just a guess but seemed to work out....then imported to Davinci and exported as http://prntscr.com/m4alx8 the result when uploaded to youtube is a huge improvement http://prntscr.com/m4am49 , not sure if thats the best codec to use for export when it comes to higher than 1080p with the free version of Davinci but it seems to work well.....does Davinci Studio also have prores as an export option?

The ProRes license in Windows is only available for Fusion Studio, probably because Fusion's not a direct competitor to Final Cut, like Resolve is. Ultimately, if VP9 is what YouTube is encoding to, then that's what you should encode to. It looks like BMD hasn't yet put a VP9 encoder into Resolve for Windows, though. When they eventually do, my guess is that VP9 422 8-bit should be your codec of choice. But again, video containers are really not my forte, so I'm open to being challenged on that.

If you're trying to counteract softening from the displacement, the proper place to sharpen is after the image has been displaced. That is, you should add an Unsharp Mask after the Renderer3D to counteract the blurring that occurs when the pixels get stretched. I looks like I didn't fully finish my thought above, though: Sharpening can create an artifact called ringing if it's overdone. This usually looks like a dark outline around contrasty edges. It's particularly noticeable on vector-like graphics like this, so you'll want to use a very light touch if you sharpen here. The default control values on the Unsharp Mask are usually fine—reduce its strength using the Blend control.

The ProRes license in Windows is only available for Fusion Studio, probably because Fusion's not a direct competitor to Final Cut, like Resolve is. Ultimately, if VP9 is what YouTube is encoding to, then that's what you should encode to. It looks like BMD hasn't yet put a VP9 encoder into Resolve for Windows, though. When they eventually do, my guess is that VP9 422 8-bit should be your codec of choice. But again, video containers are really not my forte, so I'm open to being challenged on that.

If you're trying to counteract softening from the displacement, the proper place to sharpen is after the image has been displaced. That is, you should add an Unsharp Mask after the Renderer3D to counteract the blurring that occurs when the pixels get stretched. I looks like I didn't fully finish my thought above, though: Sharpening can create an artifact called ringing if it's overdone. This usually looks like a dark outline around contrasty edges. It's particularly noticeable on vector-like graphics like this, so you'll want to use a very light touch if you sharpen here. The default control values on the Unsharp Mask are usually fine—reduce its strength using the Blend control.

Thanks again for the help it is appreciated... It ended up coming out not bad at 1440x2560, I have it unlisted as you can see until they release it on their channel. Did some basic stuff i guess for most fusion users but it is good to get my feet wet and start on my fusion journey

Not bad! Certainly better than any mograph that I do. And it looks pretty clean for a YouTube video, so I think you've got your encoding workflow sorted.