Hi Vlado, I’m having similar issues and I think it has something to do with Raycaster memory settings for Hyperthreading. I’ll try explain as best I can… (without rambling too much!) Vray SP3 and Sp4…
We have scene with something like 30 unique Vray Proxy meshes in the scene, with something like 3000 vray proxies within the scene. The proxies are all bits of vegetation, so all pretty heavy, it’s almost 1gb of files so quite heavy. We’ve done a few projects like this and not had any issues. Rendering usage is normally almost 100% all the time.
We initially thought our slow-down was because of loading over our network, so I wrote a script which runs pre-render which collects all the .vrmesh files and copies them locally, so there’s definetely no slow-down because of network pulling.
It seems to only be affecting our newest render blades, Our other blades are 4 cores and 8 core. We’ve recently upgraded our render farm and we now have a nice farm full of Intel Xeon e5530, 2 physical cores, 4 processing cores, hyperthreaded, so that means we get 16 Render Buckets

Vray normally has no issues using all 16cores except when it comes to VrayProxies where normally the CPU drops to 50% or lower… By looking at the task manager you can see either, 50% of the Cores doing all the work and the other 50% doing nothing, or you’ll see all the CPUs working but maybe only a couple above 50% usage and most below 25% usage, which adds up to 50% Total usage.
Ram doesn’t seem to be a factor either, the Physical ram is mostly being used but the Page file is still hardly used, We upped one of the blades up to 24gb of RAM and it still had exactly the same issue.
I’ve done some experimentation and by tweaking the Raycaster settings I can get Vray to use about 90-95%, so it’s rendering each bucket, but rendering each bucket much slower…
I haven’t done any proper benchmarking tests yet but it’s on my todo list, and if you’d find it useful for solving this issue I can make it a priority as we could really do with a solution to this.
@Stan, Try these settings and see if they improve your rendering…
Max Tree Depth 40
Min leaf size 0.0
Face/level coef. 2.0
Dynamic Memoery limit 3000mb
That should get your CPUs using 100% of their power but you might notice that buckets take longer to render, it’s just a battle of seeing is it worth each bucket rendering slower and having all the buckets rendering the image…
I shall hopefully look into this more soon and will report back..