|10 October 2012||#1|
You know viewport operations of 3D Studio Max are not multi threaded at all, at least as far as I observe, when Max locks up for minutes while calculating something other than rendering, task manager reports about %8 CPU load with an 12 core system. It is %25 on a quad core and everything fits in place.
Nowadays I see task manager reporting the same %8 use but while individiual core operations are visualized in the performance tab, it looks as if 6 cores are actively running forming a total of %8??
I first thought "yes, they can't distribute a single calculation among different cores but they've at least distributed different calculations among processor cores, that's an improvement...."
However operation goes on for a very long time and the total processor usage never exceeds %8 while the use of other 5 cores don't stop (finishing what they do) after a short while.
If one job is the hard calculation, why then the other processor cores go on running like an ECG during that viewport calculation. If they also work on for that operation, then why on earth the total cpu use never exceeds %8 at all? It doesn't -even %1.
I'm confused. Only explanation I can come up with is some other application is using the CPU not 3dsMax, but it also doesn't seem so? (Only odd numbered cores are working, physical cores probably)
Have you noticed something like this in 2013?
There are times I wait for such a viewport operation for minutes and periodically wonder if there will soon be a way to put all those cores into operation to. On that case the time spent for those operations should eventually be 12 times shorter.
Furthermore I can't keep myself from thinking the fact that my nothing other that rendering times improve by upgrading our systems.
ps. I've seen some settings on my motherboard bios to individually set the core multiplier during single core use. That excited me a lot. Do any of you have information on those settings?
Single core use means lower heat dissipation and greater potential of overclocking for viewport operations.
Last edited by Byteman3D : 10 October 2012 at 02:12 PM.
|10 October 2012||#2|
Join Date: Dec 2005
when the total cpu usage doesn't excede 1 core worth of useage (in your case 8%) it means that a single threaded operation is being performed. Generally speaking, which core is used for that operation at any given moment is controlled by the operating system. This is the correct behaviour. If you start using other applications at the same time, the operating system will assign the workload across cores as efficiently as possible.
As a general note: I often see users post about multithreading and other things as if there was some switch to just implement by magic. This is the wrong assumption. In the case of multithreading it often takes explicit and complicated reworking of algorithms and rewriting of large sections of code. There are also plenty of things that generally can't be multithreaded such as file save/load or that the performance improvement would be so negligable that it would be a waste of time to implement, even on latest hardware, because there is additional work and overhead that managing multiple threads brings regardless of the operation you are trying to improve.
I hope this helps answer your concerns.
|10 October 2012||#3|
Join Date: Sep 2003
Thread automatically closed
Note that as CGTalk Members, you agree to the terms and conditions of using this website.
|Thread Closed share thread|