and scroll to the bottom you can see my findings for render threads. With a quad core setting this value to anything has almost no impact on rendering speed. Vegas fully utilizes and pegs all cores on a Quad proc to 100%, so adding additional threads has negligible results. There may be cases where having additional render threads on a quad proc is beneficial, as my tests were limited, but I'm not sure what scenarios would cause Vegas to use the additional render threads to improve performance.
I wish I could afford a computer to see where Vegas stops using all available cores by default, then I'd be able to relay that to the world. Since I have "only" 4 cores, I see all 4 of them pegged at 100% during renders, no matter how many render threads I specify.
I'm not sure what you meant about "But makes a difference with multicore PCs multi thread CPUs?"
Also: "And, RAM setting @ 1Gb. Why not more? Certainly 50-75% of 8Gb should help render times while not effecting an otherwise idol PC, shouldn't it? "
Do you mean that you've set the amount of RAM Vegas uses higher than 1GB and yet it's only using around a 1GB during renders, or is that the default value you're seeing? If it's the former, I'm not sure what to say as that sounds very odd. If it's the latter, why not adjust it yourself and see if your render times improve? :)
I'm also beginning to construct a new desktop. Currently using a dual core which I built in 2007 and feel like it's time for a rebuild. Do you mind sharing the name and model #s of the parts you used (ie case, fans, MB, Video, Audio, Memory, any other components (except monitors) you chose.
My understanding is that AVCHD is CPU intensive .. not necessarily ram. The 1GB dynamic ram is useful while doing a "dynamic ram preview" of a particular section of your timeline. But it might increase your rendering times.
The following is from Vegas online help:
------------------------------------------------------------------------------------
Select the region that contains the frames you want to cache.
From the Tools menu, choose Build Dynamic RAM Preview (or press Shift+B). Each frame in the selection is rendered.
Click anywhere in the timeline to cancel.
After rendering is finished, click Play . The cached video will play back at the frame rate specified on the Video tab of the Project Properties dialog.
If your RAM cache is not large enough to contain the entire selection, the selection is automatically adjusted to include only the cached region. You can increase the Dynamic RAM Preview max setting on the Video tab of the Preferences dialog to increase the size of the cache.
Clear the cache or turn off dynamic RAM previewing
To clear the RAM cache during playback, you can resize the Video Preview window or choose a new setting from the Video Preview Quality drop-down list.
To turn off dynamic RAM previewing, enter 0 in the Dynamic RAM preview (max) box on the Video tab of the Preferences dialog.
--------------------------------------------------------------------------
It's worth trying to understand Dynamic Ram preview - what it's for and what it does - and how that setting can adversely impact rendering times. In my experience, it can.
You may notice in that discussion that ZERO Dynamic Ram preview is NOT the best setting for rendering. I think generally that some small amount of Dynamic Ram is better than ZERO. Pay your money. Take your choice.
I've just finished building an i7 2600K PC running Win 7 Pro 64 bit 16 GB Mem.
I did some tests rendering a single clip of 1920 x 1080 50i with a duration of 2:39 and 16 frames to PAL DVD video and audio streams. (constant bit rate 9.8 Mb/s and 320 kbps audio) After some experimentation I've found that my PC renders best when set to 1 GB preview RAM and 8 rendering threads, taking 44 seconds to complete. 4 Rendering Threads took 49 seconds.
With 256 MB of preview RAM, 8 threads took 43 seconds and 4 threads took 48 seconds. I also tried 3 rendering threads with 256 MB of preview RAM and it took 54 seconds. So for either 4 or 8 threads, the difference in amount of preview RAM isn't that significant up to 1 GB but at 2 GB it takes a turn for the worse with the 8 threads taking 51 seconds and 4 threads taking 54 seconds.
From my recollection, your CPU can only process 4 simultaneous threads, but there is an advantage it seems, in using all four. In conclusion, I would try setting up with 1 GB preview RAM and 4 rendering threads. You could then do some tests of your own to fine tune the PC.
The render times depend on the cpu, yes - but also on which format you are rendering to as the different codecs will utilize your cpu and HDDs in different ways.
As a specific render job always takes xx billions of cpu cycles the speed is depending on how fast you can get these xx billions done - and that is, of course, fastest if you utilize your cpu at near 100 percent.
Best is to take a minute of your standard project and render that with different settings of 1) preview RAM (yes, it does mean something for render time also - and will lie about 500 MB to 1 GB) and max number of rendering threads (normally it means only little on quad cores but can give 1:2 improvement with more than 4 cores - see my post in this thread: http://www.sonycreativesoftware.com/forums/ShowMessage.asp?MessageID=747105&Replies=9
So why not just put 100 max render threads and 5 GB of preview RAM? -
well because too much of each wil have a significantly negative effect on the speed as it will cause Windows to spend more cpu time to manage this overhead.... so do the tests and find the best for your setup.
Following Ritsmer's suggestion, I performed the 2010 Render Test. The results were broadly in line with my earlier post on this thread, although having a bit more preview RAM was actually a help for this test. Settled on 1.5GB in the end.
Results were 155 seconds with 256MB of preview RAM and 125 seconds with 1.5 GB of preview RAM, which seems fairly typical of results for core i7 2600 CPUs that have also done the Render Test. My CPU is overclocked from the standard 3.4 GHz to 4.425 GHz but runs at about 66C (water cooled) when all cores are flat-out, so I'm pretty happy with that.