Announcement

**Marc44** · 09-15-2007, 08:36 AM

Originally posted by Angelo View Post

In any case anyone looking at a E6600 to a Q6600, the Q6600 is the BEST deal going...Angelo

Yes, but at the same price than Q6600 , the CPU E6850 will go faster.

Marc

**GrassValley_KH** · 09-15-2007, 03:06 PM

For the record, there is some form of multi-threading optimisation at a codec level, within EDIUS. I don't know the extent of it, but it's what makes it possible to get realtime playback of native AVC-Intra content on the timeline (1-2 streams) whereas only about two months ago, this was not possible.

The same multi-threading optimisation is currently being worked on for native J2K content, to improve its realtime performance.

**alias** · 09-15-2007, 03:39 PM

I am hoping that Edius 5.0 will be a total rewrite of the code. I would like to think Edius 5 will make full use of all four cores and support 10 bit.

**Angelo** · 09-15-2007, 06:48 PM

Originally posted by GrassValley_KH View Post

For the record, there is some form of multi-threading optimisation at a codec level, within EDIUS. I don't know the extent of it, but it's what makes it possible to get realtime playback of native AVC-Intra content on the timeline (1-2 streams) whereas only about two months ago, this was not possible.

The same multi-threading optimisation is currently being worked on for native J2K content, to improve its realtime performance.

So improving NLE systems on multi-threading optimization is limited to the codec? Is GV going to improve the HQ codec for multi-threading beyond 2-core systems? Any hope for an update to the v4.51c on this?

I was a bit disappointed with the Quad-Core upgrade when I already had a Dual-Core.

I was doing some math and to get the Core2Duo in RT with the 1 3DPIP would only require the CPU clock to be at least 2.7Ghz. Looking at the two 3D-PIP or the 3 HDV-HQ test times, for the Core2Duo to match the Core2Quad 2.4Ghz performance would require the Core2Duo to be clocked at least 3.0Ghz to get the same performance. Therefore, with the same clock speed Quad and Duo the Quad is gives a 25% gain over the Duo, not quite the 100% I was hoping for.

...Angelo

**Jerry** · 09-15-2007, 10:56 PM

Therefore, with the same clock speed Quad and Duo the Quad is gives a 25% gain over the Duo, not quite the 100% I was hoping for.

If you use AE CS3 you will notice a big jump. That and Photoshop will put your quad to good use.

**Rusty** · 09-16-2007, 05:26 AM

Angelo,

You must love these comparisons, and thanks for doing them!

I was actually disappointed when I moved from E6300 to Q6700 a few months back, but I may not be disappointed long. Nice to see Edius making better use of resources.

I need to dig into AE more now and see what they will do.

Rusty

**lalx** · 09-16-2007, 08:06 AM

Originally posted by Angelo View Post

In any case anyone looking at a E6600 to a Q6600, the Q6600 is the BEST deal going.

...Angelo

Hi Angelo,

Have you tried the older 4.1~4.3 versions? I did not upgrade to 4.51c yet as I was waiting to see how far the version numbers climbed before stabilising. I did read on some of the posts that the latest version was slightly slower in performance than the older ones - any truth in that?

An update on my Q6600 experience:

I got the latest version of OnLocation which gives you raw m2t files. I did the same test as what I had done earlier - but this time on a normal clocked (2.4)machine as I am lapping my HSF etc. I expected to see a drop in RT - as these files are not HQ. I was shocked that the whole thing still played in RT. (ie 2 layers of colour corrected chromakey HQ and a base layer of level corrected, lumakeyed m2t). 3x3D pip over the same level corrected lumakeyed m2t played in real time too.

One strange glitch I have noticed and not really seen on the forum is playback surprises. Sometimes if you place the cursor on the begining or middle of a problem clip (such as a PIP) and it does not play. But if you drag back the cursor a couple of seconds and play back - it plays back in RT! Strange behaviour. Anyone seen this before or has it been posted about?

- Lex

**Angelo** · 09-16-2007, 09:28 AM

Originally posted by lalx View Post

Hi Angelo,

Have you tried the older 4.1~4.3 versions? I did not upgrade to 4.51c yet as I was waiting to see how far the version numbers climbed before stabilising. I did read on some of the posts that the latest version was slightly slower in performance than the older ones - any truth in that?

An update on my Q6600 experience:

I got the latest version of OnLocation which gives you raw m2t files. I did the same test as what I had done earlier - but this time on a normal clocked (2.4)machine as I am lapping my HSF etc. I expected to see a drop in RT - as these files are not HQ. I was shocked that the whole thing still played in RT. (ie 2 layers of colour corrected chromakey HQ and a base layer of level corrected, lumakeyed m2t). 3x3D pip over the same level corrected lumakeyed m2t played in real time too.

One strange glitch I have noticed and not really seen on the forum is playback surprises. Sometimes if you place the cursor on the begining or middle of a problem clip (such as a PIP) and it does not play. But if you drag back the cursor a couple of seconds and play back - it plays back in RT! Strange behaviour. Anyone seen this before or has it been posted about?

- Lex

No, I didn't try older versions of 4.xx, but do recall people posting about a performance drop with v4.51 and HD. I've only just started using HD footage. Edius 4.0x to 4.51c I was editing with DV and the Core2Duo was "major" overkill and didn't see any performance loss with DV when upgrading to v4.51c.

Was your 3x3D-PIP using M2Tfiles? If so there must be a code issue with Edius v4.51c to have such a BIG drop in performance from earlier version of 4.xx.

Originally posted by Jerry

If you use AE CS3 you will notice a big jump. That and Photoshop will put your quad to good use.

I have older versions of AF & PS. I currently don't use them that often for me to justify the upgrade. If I do have a real need for them I can justify the upgrade. Since newer upgrades are always around the corner, I wait till I need them.

Originally posted by Rusty

I was actually disappointed when I moved from E6300 to Q6700 a few months back...

Have you tested earlier version of 4.xx since Lex is reporting better performance?

...Angelo

**Gordon B** · 09-16-2007, 11:45 AM

I have posted about Quad core performance in other postings and since I built my Q6600/Asus PK5 Premium/2GB DDR2 PC I have been checking in Task Manager when I do any HDV work and I see that all cores "appear" to be responding and the CPU activity is around 20%.

Am I just imagining that all four cores are working with Edius or am I reading Task Manager all wrong?

I get full RT playback of HD, even with four 3D Pips.

Gordon.

**Angelo** · 09-16-2007, 05:40 PM

Originally posted by Gordon B View Post

I have posted about Quad core performance in other postings and since I built my Q6600/Asus PK5 Premium/2GB DDR2 PC I have been checking in Task Manager when I do any HDV work and I see that all cores "appear" to be responding and the CPU activity is around 20%.

Am I just imagining that all four cores are working with Edius or am I reading Task Manager all wrong?

I get full RT playback of HD, even with four 3D Pips.

Gordon.

All my cores are active too. Do you use HDV in your 3D Pips? Are the HDV 1440x1080 files in HQ or M2T format? Does your buffer say full or increase during four 3D-PIP playback?

...Angelo

**Gordon B** · 09-16-2007, 10:21 PM

Hi Angelo,

I have done test captures in both Canopus HQ and native M2T, both play OK with the buffer staying full and only dropping during the four PIP's but still managing to play. I could of course reneder the PIP's and then the buffer would stay healthy.

My HDV is in 1440x 1080i.

I am about to do a serious edit in HDV and I will report my findings then.

Gordon.

**Angelo** · 09-16-2007, 10:27 PM

Originally posted by Gordon B View Post

I get full RT playback of HD, even with four 3D Pips.

Gordon.

I did some more testing with Four 3D-PIP under different Presettings. This time the 3D-PIP had no borders or effects enabled, no background and no CC to the 1440x1080 59.94i footage. Core2Quad Q6600 at stock speed 2.4Ghz.

I tested this test project with:
OHCI HD 1440x1080 59.94i buffer=512MB (5.57sec HD buffer)
OHCI SD 720x480 29.97p buffer=512MB (23.4sec SD buffer)
DVStorm 720x480 29.97p buffer=192 (6.3sec SD buffer)

All run times were tested with pre-filled buffers. Also, Calculated CPU FPS (frames per second) which is the max fps the CPU can process during playback.

Preset.............Time..........CPU FPS (or fields/s)
OHCI HD ......... 34sec ...... 25.06 (or 50.12i)
OHCI SD ......... 54sec ...... 17.0
DVStorm ......... 14sec ...... 16.5

Looking at the OHCI HD results I am getting the same results as Gordon. Here in Canada the frame rate is 29.97fps (or 59.94i) but in Europe (and the UK) the frame rate is 25fps (or 50i). As you can see my 2.4Ghz Quad is processing at a frame rate of 25.06fps, this is RT in PAL but not in NTSC.

Looking at the two SD projects there is a drop in CPU calculated FPS that must be from the HD to SD scaling overhead in RT.

Using the same HD footage on a HD timeline yields the best performance. Also the CPU utilization on the Quad Core was around 50%-60%.

I 'm going doing some more SD picture quality (PQ) tests to see if converting the original HQ 1440x1080 to HQ 720x480 will show any PQ through the DVStorm's component-out when comparing the two on the same timeline.

UPDATE: I converted (with PC2) the original HQ 1440x1080 59.94i to HQ 1280x720p and HQ 720x480p and viewed the results via DVStorm's component-out card (SD 480i) to 23" Panasonic 720p HDTV LCD. There is a noticeable PQ difference between HQ 1280x720p and HQ 720x480p, HQ 1280x720p has no jaggies just smooth straight lines (unlike the HD 720x480p with jaggy lines and stair-stepping) on a SD timeline. Plus 720p HD offers better RT performance than 1080i.

...Angelo

**manbart** · 09-17-2007, 06:08 AM

Re: Quad-Core

At the risk of elaborating on the obvious I'll just say that generally just a handful of applications are currently optimized such that they can take anything near full advantage of a quad-core processor. The good news is that we have increasingly affordable processors available that are capable of very impressive real time editing of HD. . . just the software is lagging behind a bit and not taking full advantage. I think this is usually how it is. As the quad-core's move in to the mainstream of the desktop market, this will likely change pretty quickly. I wonder how the fast the quad-core can render 1080i HD to a file.

Right now I have a bit older Dual processor 3.2 Xeon system (Irwindale) with the 800mhz front side bus, and rather slow memory. I wonder how much better performance I would get from a 2.9 quad-core on new Asus Intel P35 board with 2 gigs of Corsair Dominator 8500 1066. I am a former DV Storm Premiere user, then Avid Liquid Edition 7 Pro. Now thinking about Edius NX Express. I was using Sony VX 2000, and now also have a Canon HV20 with Rode Mic. I use video to document my historical and genealogical research in Eastern Europe. Delivery medium would still be DVD, but I would like to start shooting with Canon HV20 and editing in HD. Put edited HD project out to tape for archiving, and put timeline out to DVD in 16:9:

**Angelo** · 09-17-2007, 10:16 PM

In all my Edius v4.51c tests I have outlined the specs for each test with the DVStorm and/or OHCI HD & SD. Just use my playback times as a reference to your current system to my 2.4Ghz Quad-Core to give you some idea if the upgrade is worth the gains.

I would guess your current system is equivalent to a Core2Duo 2.4Ghz to 2.6Ghz with Edius, but don't quote me on it :)

...Angelo

Announcement

Edius v4.51 & Quad-Core Performance.

Edius v4.51 & Quad-Core Performance.

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment