Has the issues been fixed yet with machines with more than 64 cores? Can I safely run VRay on a single machine with 120 cores (HT) and use all cores without having to do any workarounds or hacks to make it work?
This machine might be used as a DR slave. The workstations will typically have 12 cores only
No; the problem is that the affinity mask for a processor in the standard Win64 API is 64-bit. For more cores, a different API needs to be used, which involves the so-called processor groups. So we need to add extra code in V-Ray to make sure that we can spawn threads on all cores. We did implement this, but there were some issues which we need to iron out (f.e. if the user manually sets processor affinity from the task manager, weird things started happening).
The question is: would one 12 nodes renderfarm as the fuchsundvogel one render as fast as one huge HP computer. If you count the Ghz, it’s similar, but I think you have time loss with the 12nodes of the renderfarm.
can anyone confirm ?
the HP is around 50k, the 12nodes renderfarm around 6/7k
Tasks that are multi-threaded but not shared as LC will be slower.
Task single threaded will compute faster on the small desktop nodes as the clock speed is faster.
Network is one other factor.
I would rather go for a bunch 5820k with 32Gb of ram or maybe the 5960x if you need 64Gb.
You could build a node with a 5960x OC at 4.3 Ghtz with 64Gb of Ram for $1600.
You could have 30 of those for $50k.
Small nodes : 30x 12 core x 4.3Ghtz = 1032 Hgtz
Big node : 4x 15 core x 2.8 = 168 Hgtz
That’s 6 times faster for the same $, sur you need to add the prince of a couple of licenses, but still…
if you go for 5820k with 32Gb of ram it’s even more.
Unless you really need a huge amount of ram : 64Gb+ or do simulations like FFX, realflow, etc you don’t want to go with high expensive nodes, specially for rendering.
I am planning to get 72 Threads for rendering (Maya vray)
If I use Linux is there still a restriction with core utilization or is this only a Windows problem at this stage?
Can Linux use beyond 64 Threads at 100% CPU utilization without some DR slave hacks??