I have a problem with Blender on a Linux rig. I recently got access to a dual Xeon server, which is running Red Hat Enterprise. I managed to build blender 2.63.13 from source, but when I ran my first render (with BI) was disappointed - the machine was slower than my Core i7 Windows system.
Are there any performance flags or machine-related optimizations that can be done at build time? I find it hard to believe that a dual CPU server, with 16 cores, is slower than a single CPU machine. And I could really use that processing power.
I tried modifying the number of rendering threads, no improvements there.
Any suggestions are appreciated.
The i7 is a modern quad core processor, IIRC. The older Xeon, if memory serves, was designed as a server, not a compute engine. I am not surprised the i7 is faster at a task like Blender.
I agree, the i7 is newer architecture and clock for clock should be faster. But still, we are talking 4 cores vs. 16, it can't be that
Just looking at clock speed is an extremely simplistic and misleading way to compare processors. Things like cache size are very important too.
It would be interesting to compare both processors on a compute-intensive benchmark.
Blender builds are optimized but compiled for a generic x86 architecture. I don't know of compiling specifically for Xeon would give any speedup.
These Xeon processors are less than 1 year old, so their architecture is comparable with the i7. Plus, Xeons have been built with specifically for cpu-intensive tasks, so they should pack more power than an i7, which is basically a workstation cpu.
Do you have any suggestions where might I put this question to get a better chance of help? It's turned into an academic matter for me now, so I have to see it through