The performance of KeyShot Network Rendering is dependent on your CPU specification. Also, if the number of workers exceeds the number of tasks of the job, the excess workers will be left idle if no other jobs are in the job queue This will decrease the time spent on transferring the scene to all workers. To maximize the core count but minimize the worker count.The tasks assigned to each worker are generally of equal size, so if one worker is less powerful than the others it may delay the assembling of the job To have an equal amount of cores per worker.To use the same processor type for all computers.The performance of KeyShot Network Rendering is subject to network connection and configuration. The transfer time depends on network connection and file size, rendering time depends on Worker capacity and rendering settings (render passes and layers), and assembling time depends on the Manager capacity and output size (image resolution and number of frames).
Total job processing time consists of transfer, rendering, and assembling time. For still images, the tasks consist of regions of the images, and for animations, KeyShotXRs, and configurations, the tasks are whole frames. When all tasks are completed, the Manager assembles them into the finished image, video, KeyShotXR, or configuration. When a Worker finishes a task, the rendered pixels are returned to the Manager. It splits the job into tasks that are computed by the workers.
Best cpu for keyshot rendering upgrade#
And it’s always easy to upgrade your graphics card later on.The manager receives a job from the Client. GPU rendering is very exciting, but it can take keyshot a few tries to nail down brand new features like this. I think it’s safe to say - build your computer around the CPU for now as Keyshot is very good at that. But you can get a sense for it in their What’s New in Keyshot 9 webinar. I can send examples of the noise difference tomorrow if I can remember. I am running a 3 year old Lenovo laptop with a Quadro M2000M - can’t remember my CPU. The renders came out nicer even though (for some reason I don’t know) it only chugged through 30 samples (20 with denoise). Next I tried setting the GPU to render for max time of 2min 50sec to see how it compared (both with and without denoise). Tried again with GPU and denoise set to 1 - definitely better but added about 20seconds to the render time. Did the exact same shot at 128samples off the GPU and clocked in at around 45sec though it was NOTICEABLY noisey in the shadows. 128 samples off of the cpu - clocked in at 2:50sec. The CPU render looked quite close to the GPU render until it finished, and then it reverted to what you see. Keyshot does has some critical features which is why I'm trying it again, and is a bit better put together.ĬPU - 5:30 - 257 Samples (Reached 160 samples at 3:31)Įdit 4: There seems to be an issue with CPU rendering scattering mediums as well. So you can dim a light, change the colour, and get a completely new look without having to render. In maverick, I have the ability to change all the lighting settings post render. Though, perhaps I just need to change my workflow. I'm still very disappointed in the lack of post rendering editing. But I think you get what I'm feeling here.Įdit 2: OTOH, the CPU rendering in 9 seems to be vastly superior to 8's. One has had much more composition work put into it. Granted, it's not an apple to apple comparison here.
Something seems off.Įdit: Fwiw, this took about the same time to render in Maverick iirc. The whole "ray tracing" bit is a bit hard for me to believe at this point as well as the results are quite inferior to Maverick which also uses Ray-tracing, and slower. I would have to run another test, but I think I got double the samples in the same time span using CPU over GPU. Keyshot gives me the exact opposite of what I expect after using Redshift, Octane, and Maverick. Which, granted, lower spec than you, but it's not my first rodeo in GPU vs CPU rendering.
0 kommentar(er)
