"Render threads": Understanding Threads vs. Cores

Comments

Kevin R wrote on 10/17/2010, 7:30 PM
Randy,

I'd probably choose "Custom Size" and plug in the number from "space available" (16,292 or slightly less--I've never tried using the "max") for BOTH initial and maximum. Setting both the same prevents Windows from arbitrarily deciding to re-size the swap file, which it can do occasionally.

ALSO, BE SURE to set your E drive to "No paging file" -- this is the same drive and it would be especially bad to have swap file in two places on the same drive. If your E drive is especially fast, then you MIGHT set C to None as well; HOWEVER-- there are side-effects to removing the swap file from C which is: If the E drive is ever unmounted, removed, crashed, etc. then there is no swap file anymore and the system will have trouble booting. This can be fixed in Safe Mode if it ever happens.

Is your Maxtor 120 GB drive especially fast? Moving your swap file to another drive usually isn't something that is done unless you dedicate a particularly fast volume or solid state drive to the task. If it's a regular 7200 rpm disk I don't think you'll see any large change. Also, since you already have a fair amount of RAM I'm wondering how much your swap file is utilized in the first place--have you noticed a lot of paging activity?
Kevin R wrote on 10/17/2010, 7:38 PM
Bob/farss wrote:
"My advice if looking to spend serious money would be to use a systems integrator who has a proven track record in building systems for video editing. You do pay a tad more than putting it together yourself but I found the few extra dollars worth the money"

Bob is right. You need to understand a lot of very specific issues when building a system. I have built dozens of audio/video playback systems as well as more generic systems. There is a difference.

Primarily, this is why I personally wouldn't choose an "off the shelf" solution such as the SCS or BOXX systems---because I know that I can personally make a lot of custom decisions specific to what I am building. Just thinking "RAID is the way to go" doesn't cut it. You must understand the very specific use of each volume to pick WHAT TYPE of RAID and what type of drives to populate the array with.

If you don't understand these things, then find a good integrator.

Also, I want to add that my posts above (about moving the paging file and other tips) should not be undertaken if you're not positive that you have a good understanding of WHEN and WHY they are useful and appropriate. I did not mean for anyone to just think, "hey, that sounds good" and do something to mess up their system. PLEASE make sure you understand what you are doing before you do anything.

-Kevin
Hulk wrote on 10/17/2010, 8:03 PM
Hyperthreading was developed by Intel as a means to increase the IPC (instruction per clock) efficiency of the P4 processor. In order to ramp up clock speeds Intel had to employ a 20 stage pipeline in the 1st generation P4 cores. As you can imagine a branch stall would require clearing of the entire 20 stage pipline. Quite a bit of overhead.

Hyperthreading, while only requiring about 5% increase in die space could increase efficiency (performance) by as much as 40% or even more with some well coded multithreaded applications such as TPMGEnc. In those days not many applications were multithreaded since there we few multicored processors so Hyperthreading's benefit was limited. But eventually more and more apps were coded for multithreading and the logical cores of the hyperthreaded CPU's became more meaningful.

Video applications are especially well suited for multithreading since many operations can be performed without "waiting" on the results of other operations.

A good example of this is SLI (scan line interleave) video cards. One card processes the odd lines while the other the even lines.

- Mark