Isotropix Forums

PBR setup — Hadrware & Software

General Discussion about Isotropix and CG related topics

PBR setup — Hadrware & Software

Unread postby hometask » Sun Jun 04, 2017 8:14 pm

It is time for the truth and understanding to surface, common technical sense and bad grammar to emerge.

ABSTRACT
So "speed performance" of render engine (how fast do I get the final result) = a) my hardware (and its setup to OS) + b) how "fast" render engine's code handles all the data and all the settings.

MAIN — HADRWARE

Q1. In a simple form: How does all my hardware affects "speed performance"?

— CPU: Core count vs Core Clock vs L1(2,3) cache size. AVX? AVX2? TFlops?
— RAM: Size (C.O. — more the better), Channels (C.O. — more the better), Clock? (note to compare to GDDR) vs Timing (Lag)? ECC impact?
— ROM: Should I go for some price-biting sata, pci-e or strait to the gates of sas-raid-enterprise-me-to-the-ground high-rotating setup?


MAIN — HARDWARE's settings

Q2. Should I mess with the BIOS?

This one concerns all the settings that different vendors offers. They are usually described as meaningless one-liners so I could easily throw my board's manual in a trashcan.
While I want to read: "Lib X after switching on Parameter Y skyrockets".

MAIN — OS and Dependencies.

Q3. What is the best Software setup (OS, omg-.net, bless-my-python) to run "blazingly" fast, what tweaks and variables should I use?

Yes, I can see in my Windows Task Manager that all my CPU are 100% busy, but please, tell me how to check my settings "under the hood". Is it 1% more efficient if switch to server OS, because some libraries tends to run faster then? I'll consider. Is there any performance gain on *nix or command-save-us-all-line? I'll test it.

MAIN — Application defines it all.
For now I'd like to reserve this for questions. Primarily because it deals with obj vs alembic packing, material shaders etc. Since 3.5 did a switch towards "Arnolds mats", "ground-up re-written" I tend to wait and see.

BOTTOM LINE

This topic and 3 large questions came from years of learning and years of trials and errors. Clarisse iFX is unique in a way of CG workflow or any workflow. By the time this app reaches it's authentic maturity I hope I will rise to "master" it. And I would expect to understand 1) "speed performance" and — "scalability". Should I go beyond 2 socket setup to 4s or 8s or this new i999-core will bring me to my knees? That raises more questions.

I'm expecting pure numbers, vs's, links, conclusions for instructions, settings, tweaks, safe tips etc.

Thank you in advance. I will monitor this thread and I think all participants (Artists, Enthusiasts, Hard Core Programmers and System Engineers) will be at least interested by it. To dare you to control.


UPDATE 1.

Ok, so I've found elephant in a server room. Q1 is omitted.
Probably all possible systems are here http://ranker.sisoftware.net/top_device_all.php
In a nutshell, newer socket — better, more sockets — better.
Last edited by hometask on Sat Jun 10, 2017 5:42 pm, edited 1 time in total.
hometask
 
Posts: 3
Joined: Wed Apr 19, 2017 4:52 am

Re: Best PBR rendering setup — Hadrware & Software

Unread postby isoyann » Wed Jun 07, 2017 12:04 pm

Hi,

I'll only answer to the Q1.. other "Artists, Enthusiasts, Hard Core Programmers and System Engineers" may have tested the other aspects and give you numbers:

Q1. In a simple form: How does all my hardware affects "speed performance"?

The more core you have, the better. Currently the best performance are achieved by combining the highest frequency clock and maximum core number.

Ram size : you just need to make sure than your project fit in memory. If that's the case, more ram doesn't affect the speed. ECC impact should be minimal..

Thanks
User avatar
isoyann
 
Posts: 601
Joined: Mon Jan 28, 2013 12:11 pm


Return to General Discussion