Tip:
Highlight text to annotate it
X
Good day everyone. Today we will talking about Blender Cycles performance
I had to upgrade my blender rig and i was curious,
is it better to have more CPU cores or higher clock
I wasn't sure about Blender scaling to gain speed-ups
of additional processing cores. So I did experiments
to help me pick out a new cpu. Hope it will help you too in making decision.
Enjoy. Our testing platform has
Intel Core i5 dual core processor.It is based on
the Sandy Bridge architecture and offers HyperTHreading.
The base clock speed is 2.6 GHz,
As benchmark tool
we pick latest Blender build 2.68. The operating system
is a 64-bit Windows seven with the latest service pack.
Testing scene will be a great MikePan automative model.
You can find it on Blender Artists forum - link in description.
Before we start, I made sure
that only CPU is enabled for cycles. We can check it
in user preferences - system tab, inside the compute device section.
To measure the time used to render the scene
just click render button. The results are presented in a top part of screen.
.
.
First I did exepriments with CPU multiplier
to simulate different CPU models.
We do like that to manipulate CPU speed only and nothing more.
Raise the front side bus is bascially changing everything that runs.
As we see the CPU frequency improved render efficiency so nice.
I was really suprised about this linear speed-up.
Let's start with Hyper-Threading.
The rendering time is actually faster with HT enabled
by about 30%.
Another big advantage is more responsible operating system.
While blender rendering a scene we can sitll work on another application, checking e-mail etc.
Blender has really good linear scaling.
In general, the speedup is equal to core frequency.
Dobuled frequency solo core equal to a half frequency dual core.
The number of processing cores provide
linear performance gain.
These results imply that the speedup of Blender Cycles remains steady across
the core amount and frequency.
It worth to get more cores.
thank you for watching. see you next time.