One Blend file gives very different render results on two different computers

I have one Blender File that I am bouncing (via oneDrive) between two computers. So being the same file, I reason that with Cycles being physically correct, I should get the same render result on both machines. However, as per the attached samples, I am clearly getting different lighting levels.!

Renders are in cycles, GPU with experimental features enabled, and the lights are Point Lights with IES profiles (Texture Coordinate and Mapping Nodes) and Blackbody node added to the emission node (pretty standard setup from examples I have seen)


Acer Predator Triton 900 (Nvidia GeForce RTX 2080 Max-Q 8Gig GDDR6)


MSI G5 73VR GRF (NVidia GeForce® GTX 1060 with 6GB GDDR5)

Why am I getting quite different lighting levels?

Would like to be educated as to what I am doing, possibly incorrectly, in my setup that the file renders the lights at such different levels.

Thanks in advance

have you updated the videocard drivers on both computers?

hey @thinsoldier, yes I have - through gForce Experience on both computers earlier this week if I recall.

:frowning: hmm… if you cranked up the light values on the good render do you get a similar result as the bad render or is it very different?

Also are you certain all files necessary were copied to the other computer? Are there any files necessary for the IES stuff?

1 Like

Okay, no I have not tried to adjust the settings to match the output on the two machines. My goal is not to adjust the settings to get the same result. Rather I am trying to understand why the two renders (at least from the lighting aspect) are so different.

To give some background, the MSI belongs to a friend who loaned it to me long term (he upgraded and the machine was gathering dust) while the ACER is a machine he had for a tech review (so must be returned - so I am not investing too much into the machine). So I am not overly concerned (per se) that I am not getting matching results. My concern is probably more along the lines of “which of the two renders is the ‘correct’ render”.

Regarding the IES proiles.
Just opened the file, took a look at the node set-up and from what I see the IES profiles have been packed into the blend file (showing //…\…11.ies for example). But beyond that the profiles are also located in the same oneDrive, so, at least in theory, they should be available on both machines.

Bit of background that may or not have an impact, I am getting a couple of warnings about some textures not being found - so cannot pack them. But from what I can see they are not the culprits, the rest of the texturing seems to be “close enough” between the two - allowing for the different light levels. And I am guessing that even the HDRI is rendering with a slightly different value. Probably incorrect terminology, but the ACER (RTX graphics) seems to be producing a more saturated render than the GTX of the MSI.

When I started learning 3D they said only CPU rendering give us the same result on machines.

I’m certain I get the same results between cpu and gpu so I should get the same results on any gpu.

my theory was that if the IES files were missing, a 1000 strength light that is knocked down significantly by the IES details would be at full 1000 strength and look very different.

If you disable all lights, add a sun light and render, do you still get different results?

@lehuan5062

Yes, indeed, I am pretty sure it must be down the the graphics cards. But my concern is still, that if Cycles is physically correct, then surely the render should actually be the same regardless of CPU or GPU? And the only difference should be the speed of the render?

Although this might also explain, why, on my little Dell Inspiron with basic NVidia card, I get a much poorer quality render.

Okay tried with CPU and still the same problem.

Good theory, and yes, would be an obvious solution.

But the fact that Blender does not seem to be complaining about not finding the IES (and I could be wrong here) means to me that it has found the IES profiles and using them correctly.

do you have this same problem with any other scene?

could you render this scene and see if there is any difference?

If there is no difference, add an IES light to it and render and see if there is a difference.

If there is no difference, turn of all lights in your scene and see if there is a difference.

If there is no difference, triple check your IES lights.

@thinsoldier

I have not tested other scenes. As I mentioned in an earlier post, prior to this I have been bumbling along with a very modest Dell Inspiron, which allowed me to model (even large scenes) but ground to a halt when I attempted to render. So this scene is the closest I have gotten to “Photo Realism” yet.

And I don’t usually have access to two quite powerful machines, so would ordinarily not have noticed. In fact looking back at the renders I attempted on my Dell - the quality was quite shocking relatively speaking, though that was at a much earlier stage in the scene where I had not yet added lights (think I would have broken the Dell or the blend file).

Am busy disabling all the lights in the scene, doing a “save as” and will try the render just with HDR and then will try it again with just a sun light. See how they turn out.

Just a thougt, maybe they use different engines, check both for CUDA / OPTIX.

I know this might be a long shot, but on the MSI I am using the Windows Store version of Blender 2.92 while on the ACER I am using the download from blender.org (also 2.92)

But I cannot think there would be a vast difference?

Think they are both on CUDA, but will take a look now and check

Busy with a saveas to see if disabling the lights will make a difference.

Cycles should give the same result on any machine as long as you use the exact same version/build of Blender and the exact same .blend file (preferably with packed textures). Also similar Blender configurations with the same color management and addons.
Also, to ensure the same results, on both machines it should be rendered with GPU only or GPU+CPU. If you use CPU only in some of them, the result may vary in a few cases as in some scenes with volumetrics.

This is my thinking - same blend file. Which I understand contains all the settings for the file. So blender itself should not be adjust settings when I open the file. Gosh I hope I am not doing something obviously stupid.

I have no volumetrics in play, and prior to this thread have not changed settings at all. Hence my confusion at this apparent weirdness

i’m still betting its a path or missing file problem

For what it is worth, unsure if perhaps this was just due to differing screens on the two machines (the ACER has 4k UHD). Thought perhaps just an effect of screen quality, so did the two renders and opened them in Affinity Photo. And they are clearly quite different

And are you using the exact same version/build of Blender on both machines?

Have you done tests using CUDA only on the two machines and other test using Optix only on the two machines? (sorry if you already mentioned this somewhere in the thread)