Realtime with EDIUS 4.60 & 5

[wags]

Is it just me or does it take more power to drive EDIUS 5.

I seem to be able to get more real time tracks going with 4.60 than 5 now?

To prove it I rolled my system back from an image with 4.60, got clips up and applied 3D or pips to them.

Next I un installed 4.60 and installed 5 and tried it again with the exact same effects - clips and found I have less realtime performance with multiply tracks using 5.

[STORMDAVE]

You're not alone.

Can you do a test with Multicam to see if performance dropped there too? I think I'm seeing the same thing.

[wags]

Hi Dave.
I tried multicam on 8 x Canopus 1440/1080 HQ clips and it seemed to play back fine although could not see V2 and V3..edit...oops had then off....ha ha?
It plays only with 5 realtime.

With 4.60 I could have 9 Canopus HQ pips playing back in realtime, now struggles 6 with version 5.

With 3D Picture in Picture I used to get 3 playing back realtime now it struggles with just two.

Could this be a bug?

[STORMDAVE]

Could be? I hope it gets fixed.
Thank you for trying, I thought I was the only one :)

[kwshaw1]

This sounds like bad news for those of us who are just barely getting the real-time performance we need with Edius 4. If anyone finds performance optimizations for version 5 please post them here.

[Philip]

I have been thinking about this thread and thought I would jump in with some possible explanations. First let me say I used to be a programmer but never did the kind of stuff that would build an editing suite but some concepts remain universal no matter what language or app you are using.

Each time a new release has been rolled out a few users have complained about RT being reduced. We complain a little and then before the next release is out we have upgraded our computers and RT is back to good again. I remember when we went from 2.? to V3. Holy crap...lots of complaints about reduced RT.

If you think about how a computer works my explanation of reduced RT is this. To create RT the buffer has to be filled in advance of the cursor correct? Well in 4.6 the VSTs didn't work well, the video layout tool wasn't as robust and other new features have been added as well. Well as you are filling the buffer the computer has to read ahead and process stuff to the buffer quicker than the buffer is unloading it to your output. With added features new things have to be accounted for. Even if you aren't using features the system still has to read ahead and decide if a feature is turned on.

If a computer could talk it would be saying something like this, "Here I am reading ahead to fill the buffer. I can read up to 169 frames ahead in my buffer. Right now I am on frame 82. Oh look, there is frame 83. Is there anything I have prerender. Hey you user, here is my list of stuff to check for. Dang everytime they make me better this list gets longer. I have all this new stuff to check for. Even if none of it is implemented I still have to evaluate whether each feature is active at frame 83. Quit bugging me about being slower. I am still working as fast as I can but now I have to check for more items. Give me more coffee and I will run faster and your RT will be the same as it used to be".

Of course the above seems kinda silly but with more features that we all love and long for more things have to happen for RT to be available. This things take processing cycles to work. More cycles mean less RT. I challenge anyone to still find an NLE with better RT abilities. Even at V5 we smoke all the others.

Sure, if you are using an older computer things will start to bog down a little. I use a laptop for SDE editing and my performance is not near my quad desktop. I will probably have a little lost performance on the laptop as well. When you upgrade your computer in the next year or so you will see a huge increase in performance and RT will be good again.

Like I said earlier, I don't have any proof that this is what is causing reduced RT but being an ex-programmer I understand completely why RT may be a little reduced, especially on slower systems. There are just more things to "compute" with V5.

Hope that all makes sense and if I am incorrect in my assumptions someone from Grassvalley can feel free to correct me. I think they have created a pretty incredible piece of software for us.

Just an attempt to explain why you may see slight reductions in RT with newer versions. I may be wrong but it sounds logical to me.

[gerbow]

I have seen performance improvements. Was running edius 4 on WinXP home. Now running Edius 5 on Vista Ultimate. Program loads faster, realtime playback is smoother (I have up to 5 tracks of video layers). Export too seems quicker although I have not done comparative test.

[vienna1944er]

about the 3D-PIP issue ......That's nothing new .......... from 09-16-2008, 08:40 PM

There has been an overhead introduced with 3DPiP in v5,
however it appears to be a bit too severe than was intended.
It is being investigated as per beta test reports.
.

friendly greetings from the "inhouse-librarian" ....... old (but not morbid) Hans ;-)

[wags]

Philip, nice explanation there ...made me laugh. I have bee building a MAC/PC computer program for the last 4 months here, you kind of have to end up thinking how code thinks.

Anyhow it's not just the 3D pip filter, just on normal multiply tracks is down as well.

Anyone else find they get less tracks in realtime now?

[Philip]

Philip, nice explanation there ...made me laugh. I have bee building a MAC/PC computer program for the last 4 months here, you kind of have to end up thinking how code thinks.

Anyhow it's not just the 3D pip filter, just on normal multiply tracks is down as well.

Anyone else find they get less tracks in realtime now?

I learned to try and think and talk like a computer when I was troubleshooting code in my programming days. It was a good trick a few bosses taught me. No matter what function you may or may not be using it just makes sense that the program may have to evaluate more variables with each release cause more is happening. I'm sure with the new releases additional resources need to be factored in somewhere to accomodate the new features thus RT could be down but remember, I am no longer a programmer (there may be a reason why I am not :-) ). I could be wrong in all my assumptions so I hope no one takes them as fact. I was just laying out some possible ideas explaining the reduction in RT by some.

I have not noticed any significant RT reduction with V5 but then I am running a quad core and normal event like stuff. Multicam still works great with 3 cams of HQ right off a single eSATA drive. That is about as intensive as I push it. I seldom use multiple PnPs at once so haven't noticed those issues. I am happy with the new release. By the time we hit V6 someday most users will have bigger computers and be happily editing along in V5 with plenty of RT. That's how technology goes. :-)

[wags]

I'm running a QUAD core as well here.

Are you on VISTA? I'm on XP still, maybe it runs better on VISTA.

[Philip]

I'm running a QUAD core as well here.

Are you on VISTA? I'm on XP still, maybe it runs better on VISTA.

I'm still on XP. From what they said when V5 was announced it is still running in 32bit mode but is written work run on Vista. Of course most computers outfitted for Vista will have more than the 2gb of RAM that my XP machine has. That could easily factor into it....especially if outfitted with 8gb or more.

So what kind of application are you writing for PC/Mac. Is it video related? I may be game to be a tester if you need one since I used to be a programmer I have been known to try an kill applications before. :-)

[wags]

Hey Philip
Can you put a Canopus HQ 1440/1080 clip on the timeline and then add the 3D Picture in Picture to it to see if it plays back realtime there. If it does can you see how many you can get playing realtime.

I'm only using FLASH action script to build the program with.
It will be going into schools, sorry ND. FLASH AS code has come along way. We had better keep this topic to EDIUS though or we will get told off.

[Philip]

Hey Philip
Can you put a Canopus HQ 1440/1080 clip on the timeline and then add the 3D Picture in Picture to it to see if it plays back realtime there. If it does can you see how many you can get playing realtime.

I'm only using FLASH action script to build the program with.
It will be going into schools, sorry ND. FLASH AS code has come along way. We had better keep this topic to EDIUS though or we will get told off.

I tried a HQ clip with 3DPnP and got no RT from it. I prefilled the buffer and it chewed through the buffer in about 10 seconds. I have a Quad Q6600 with 2GB RAM on XP. I think my ram is 800mhz...not sure on that. Does look that could be a problem though cause on a Quad I would think you could get at least one layer of 3DPnP.

On your other thing...no big deal...just thought if it was something special you may need some beta testers and you could always send a PM if you needed them. Sounds like a fun project though.

[wags]

That's the same this end when I try it as well. If I open up 4.60 I can have 3 playing realtime so I would say it needs to be looked at or it could be a bug.
I don't know how something like that could be overlooked though. Lets see what others say.

I have a Q6600 here as well, it's overclocked to run at 3.2Ghz which is does very well with 4 GIGS of ram but XP will not use that much.

I'll send you a PM if I need some beta testing....Thanks for your offer.