Dynamic RAM allocation in Vegas 14 - impact on rendering

bravof wrote on 10/14/2016, 4:20 AM

OK, here we go again: dynamic preview RAM allocation and impact on rendering times. The discussion started 6 years ago and is still unfinished.

I have 64 GB or RAM and 16 cores / 32 threads available. Ran a few tests of rendering times with different Dynamic RAM Preview settings. The findings are strange and counter intuitive, but in line with previous findings.

RAM Setting in MB --- Render times (libav H264)

200 --- 3:18 OPTIMAL SETTING

0 --- 5:30

64 --- 3:40

128 --- 3:18 (OPTIMAL AGAIN)

5000 ---- 3:47

30000 --- 3:22

At 30,000 MB I checked memory allocation: it was a linear increase during the rendering and stopped at exactly 30.8 GB. After that it does not move at all, even after rendering. Now comes a discovery: I immediately tried rendering AGAIN thinking that it had stored a lot of good stuff in memory and that it would be faster. Result:

30000 Second run --- 4:02

Surprise! It performed worse. It was probably very busy REPLACING stuff in memory while keeping the memory level constant. During the whole rendering it kept memory at 30.8 GB.

The system was not bottlenecked anywhere: processors were running at 38%, GPU at 20%, memory at 48% and disk around 0%.

In conclusion: leave your dynamic RAM Preview Max at 200 MB and don't waste time in benchmarking :)

Comments

NickHope wrote on 10/14/2016, 5:09 AM

Thanks for the testing. The whole situation might well change in the next update, as memory handling clearly has problems in version 161.

Here's a quote from the release notes that is relevant here:

Improved playback and editing performance when using a Dynamic RAM Preview cache size greater than 200 MB.

Wolfgang S. wrote on 10/14/2016, 9:14 AM

The figures suggest that one should not stay with 0 MB dynamic ram allocation. But the difference with all figures higher then 0 MB makes not such a hugh difference.

More important - the new version allows much better preview capabilities with ram allocation with more then 200 MB. For the first time it is possible to use high ram values without decreasing the playback behaviour for some type of footage. That is a clear advantage.

Eagle Six wrote on 10/14/2016, 4:14 PM

Wolfgang S. wrote on 10/14/2016, 07:14 AM

More important - the new version allows much better preview capabilities with ram allocation with more then 200 MB. For the first time it is possible to use high ram values without decreasing the playback behaviour for some type of footage. That is a clear advantage.

What type of footage and what ram values?

Best Regards......George

set wrote on 10/15/2016, 1:22 AM

Just tried on VP13.453, set back Dynamic RAM use from 0 to default 200MB, render downconvert AVCHD to DVD MPEG2, render stuck in the middle.

Wolfgang S. wrote on 10/15/2016, 1:40 AM

What type of footage and what ram values?
Best Regards......George

XAVC, ProRes, Cineform, mp4 GH4 footage.

set wrote on 10/15/2016, 1:58 AM

target render format (for above tests)?

bravof wrote on 10/15/2016, 3:51 AM

target render format (for above tests)?

If the question was for me, then H264 using libav.

Eagle Six wrote on 10/15/2016, 9:53 AM

Wolfgang S. wrote on 10/14/2016, 07:14 AM

More important - the new version allows much better preview capabilities with ram allocation with more then 200 MB. For the first time it is possible to use high ram values without decreasing the playback behaviour for some type of footage. That is a clear advantage.

What type of footage and what ram values?

Best Regards......George

Wolfgang S. wrote on 10/14/2016, 11:40 PM

What type of footage and what ram values?
Best Regards......George

XAVC, ProRes, Cineform, mp4 GH4 footage.

During a couple quick and dirty test with ProRes footage, there is no improvement in playback speed from 200 up to about 2000. When increasing above 2000, through 5000, the playback speed begins to suffer.

Best Regards......George

Wolfgang S. wrote on 10/15/2016, 9:56 AM

That was not stated that there should be an increase in preview speed.

Eagle Six wrote on 10/15/2016, 10:22 AM

Wolfgang S. wrote on 10/15/2016, 07:56 AM

That was not stated that there should be an increase in preview speed.

That's not what I said, I will explain.....there was no change in playback speed, call it performance, or call it behavior, if you like. There was a decrease in behavior, as in the playback speed when the ram went above 2000.

If there is something else you refer to by using the word 'behavior', then please explain so we are on the same page. Thank You.

Best Regards.....George

Wolfgang S. wrote on 10/15/2016, 10:50 AM

So I will explain too. With the former versions of Vegas, also with Vegas 13, an increase in the allocation of ram to the dynamic ram preview decreased preview speed expressed in fps a lot, at least for some machines, depending on the hardware but also on the type of footage.

I have seen that limit somewhere about 200 on my older i7 2700K system, where the fps in the preview went down dramatically with a higher allocation to the ram preview.

With the new feature in Vegas 14 that was improved. Preview speed in fps stays at a higher level now, even if you allocate more ram to the dynamic ram preview. I think this is an important improvement, because preview speed in fps is important to the editor.

Eagle Six wrote on 10/15/2016, 2:19 PM

Wolfgang S. wrote on 10/15/2016, 08:50 AM

So I will explain too. With the former versions of Vegas, also with Vegas 13, an increase in the allocation of ram to the dynamic ram preview decreased preview speed expressed in fps a lot, at least for some machines, depending on the hardware but also on the type of footage.

I have seen that limit somewhere about 200 on my older i7 2700K system, where the fps in the preview went down dramatically with a higher allocation to the ram preview.

With the new feature in Vegas 14 that was improved. Preview speed in fps stays at a higher level now, even if you allocate more ram to the dynamic ram preview. I think this is an important improvement, because preview speed in fps is important to the editor.

I'm not seeing that on my machine!

Best Regards......George

Wolfgang S. wrote on 10/16/2016, 2:01 AM

What do you not see on your machine?

Eagle Six wrote on 10/16/2016, 9:59 AM

What do you not see on your machine?

My time for testing on this subject is coming to an end. My original question was based on your statement, and I hoped to have received a more specific answer, which you didn't. My trial version of 14 has very limited time left, which I will use for other testing.

Your statement in response to the OP also is taking the thread off topic, which in other threads you have been critical of others for doing so, it may be best just to let this one go for now.

For me......how version 14 handles playback preview when allocating more ram, is showing no improvement, therefore it is not an advantage.

Best Regards......George

NickHope wrote on 10/16/2016, 10:47 AM

When Wolfgang wrote, "...the new version allows much better preview capabilities with ram allocation with more then 200 MB..." I believe he meant better than previous versions of Vegas. Nobody's saying that preview performance improves with more dynamic RAM, but that VP14 doesn't degrade in performance with higher values, whereas previous verisons did.

Wolfgang S. wrote on 10/16/2016, 10:48 AM

Georg, well that is up to you. I never said to allocate more ram will result in an improvement in VP14 - but compared with VP13 you have an improvement.

Eagle Six wrote on 10/16/2016, 11:33 AM

OK guys, just one more.......'improve', 'degrade' they both go hand in hand. If a feature does not make a difference in less of a 'degrade' than it has not made an 'improvement'. We can mix the words around anyway we wish, the results are what counts. The OP was in reference to Dynamic Ram Allocation -vs- render times. His results are very similar to mine......his testing is understood and for me verified.

Wolfgang, "More important - the new version allows much better preview capabilities with ram allocation with more then 200 MB. For the first time it is possible to use high ram values without decreasing the playback behaviour for some type of footage. That is a clear advantage."

Wouldn't "much better preview capabilities" be the same as "improved capabilities"?

Wouldn't "without decreasing the playback" be the same as "no improvement", if in testing there is a decrease?

Wouldn't "That is a clear advantage" be a valued advantage, if the previous statement was true? So I asked for the "some type of footage" and the RAM allocation hoping to get something specific enough to test as a verification. The response was a bit vague, but I tested what I had, with the time I had, and without something more specific did not see the advantage on my system.

Georg, well that is up to you.

And yes, of course it is up to me, that was my statement.

@ Nick Hope, Thank You Nick, your contributions are always helpful and knowledgeable.

For me, nuff said. I'll let someone else have the last word.

Best Regards.......George

Kinvermark wrote on 10/16/2016, 1:59 PM

Georg, well that is up to you. I never said to allocate more ram will result in an improvement in VP14 - but compared with VP13 you have an improvement.

So, to paraphrase - The performance DEGREDATION is LESS BAD in v14 than v13.

I found that allocating extra RAM over 200mb is still a degradation, just not as much as it once was.

I can confirm that this IS THE CASE on my machine, with GH4 and cineform 4k footage. The difference seems pretty minor, however, and I would add that the playback framerate does bounce around from test to test and from version to version.

Overall my impression is that V14 playback is better for this footage on my machine.

Wolfgang S. wrote on 10/17/2016, 8:36 AM

I have tested that on two systems now:

- on my older i7 2600K system, I have seen a really strong degredation (in preview fps) over 200mb (Vegas Pro 13)

- on my newer i7 5960X system I do not see any preview degradation (neither in VP13 nor in VP14).

The comparision is not easy, the 5960X system has significant more power. And it uses other GPUs too. What I have not done yet is to install VP14 on my older system too, to be able to compare it also there. Maybe I do that in the future, but it has no priority to me.

bravof wrote on 10/17/2016, 8:56 AM

OK... But is there any benefit in VP14 of moving beyong 200 MB in Dynamic Ram allocation?

Wolfgang S. wrote on 10/17/2016, 9:10 AM

Sure there is a benefit. You can use the dynamic ram to prerender footage, for example for effects that can not be previewed without prerendering any more. The more ram you can allocate, the more footage can be prerendered.

Kinvermark wrote on 10/17/2016, 11:49 AM

Is there a setting for RAM Preview quality?

Sometimes I find this takes too long for "quick preview" and I may as well just render/replace a cineform file (that I don't lose... unlike RAM previews)