|
|||||||||
|
Thread Tools | Search this Thread |
January 14th, 2010, 03:49 PM | #16 |
Regular Crew
Join Date: May 2009
Location: Monument, CO
Posts: 109
|
That could very well be Rob. Do you by chance have something that isn't AVCHD to work with to test that theory? I have an EX-1 so I don't routinely work with that format. By chance do you know how to run Performance Monitor and catch a screen cap of that for us?
Specifically, set up counters to capture % Processor Time, Processor Queue Length, Current Disk Queue Length & % Disk Time(for each physical drive) and Memory Pages/sec. This will give us any idea on what your system looks like while rendering. Want we want to look for is tasks waiting to be executed or queuing up somewhere. If the disk counters are low then you're theory will be proven correct and it will be a matter of digging into individual processes to see how they're performing. What you want to do is show how hard the hardware is working and if it isn't then you can bet that it's an application problem. |
January 14th, 2010, 08:09 PM | #17 |
Trustee
Join Date: May 2004
Location: Knoxville, Tennessee
Posts: 1,669
|
I have an I7 920 with non-RAID 7200rpm disks. It's about average for an I7-920 on the Premiere Perfomance benchmark scoresheet.
When I encode AVCHD clips to 1080i H264 using CS4 AME I get all cores active with CPU utilization fluctuating between 74 and 81%. You mentioned your problem is with "edited" AVCHD - do you have color correction or whatever applied? Some filters may not be multicore aware ... might explain why Avisynth works OK and (your) AME doesn't seem to. |
January 14th, 2010, 08:56 PM | #18 |
Regular Crew
Join Date: Jul 2007
Location: Ft Myers, Florida
Posts: 65
|
Steve, I'll get on it. Please give me a list of the most relevant counters to set up in PM and I'll grab a screen shot. I'll also try encoding to/from some other formats, perhaps ruling out the AVCHD and/or QT codecs playing badly with AME and I7.
|
January 14th, 2010, 09:09 PM | #19 |
Regular Crew
Join Date: Jul 2007
Location: Ft Myers, Florida
Posts: 65
|
Graham, hmmm, now that IS interesting. I've encoded AVCHD both straight and using a few simple filters, such as a HSL or levels adjust. Same rendering problem on both accounts.
Only difference I can see is you're encoding to H.264 1080i whereas I'm encoding to MPEG4 (Part 2) 1080p. As I said to Steve above, it will be worth a shot to try encoding to these different formats. Based on what you've just said I'll start with that. |
January 14th, 2010, 09:45 PM | #20 | |
Regular Crew
Join Date: Jul 2007
Location: Ft Myers, Florida
Posts: 65
|
Quote:
Harm, Thank you for sharing your information and posting the link to PPBM4. I just downloaded it and I'm sure it will prove to be helpful. |
|
January 16th, 2010, 07:23 AM | #21 | |
Regular Crew
Join Date: Jul 2007
Location: Ft Myers, Florida
Posts: 65
|
Quote:
Regarding the AME video export tab, from top to bottom, would you be so good as tell me exactly what your export setting are? I feel I'm making progress and would like to take this further. |
|
January 16th, 2010, 08:46 AM | #22 |
Trustee
Join Date: Aug 2006
Location: Rotterdam, Netherlands
Posts: 1,832
|
Although directed at Graham, allow me to give you my results, exporting to H.264-BR VBR-1 pass, target 25 and max 30, 1920 x 1080 25i with AC3 5.1 448 Kb sound, results in these rates:
|
January 16th, 2010, 05:19 PM | #23 |
Trustee
Join Date: May 2004
Location: Knoxville, Tennessee
Posts: 1,669
|
Rob - these are the export settings for the encoding I mentioned earlier. The original clip was 1080i AVCHD from a Canon HF10.
|
January 16th, 2010, 06:59 PM | #24 |
Regular Crew
Join Date: Jul 2007
Location: Ft Myers, Florida
Posts: 65
|
Graham and Harm, thanks much for sharing your settings. Graham, I'll try yours tomorrow (Harm is using XDCAM/HDV/DV) ... see what's what with CPU usage and post the findings. If I get similar utilization as you then at least I know it's the encoding codec and not AME.
By the way, Harm. Here are some of the results of your benchmark on my current configuration. What's the verdict? 77.7, secs Total Benchmark Time 10.6, secs AVI Encoding Time 35.1, secs MPEG Elapsed Time 32, secs Rendering Time |
January 17th, 2010, 08:20 AM | #25 |
Trustee
Join Date: Aug 2006
Location: Rotterdam, Netherlands
Posts: 1,832
|
The 10.6 AVI result is pretty disappointing. Compare it to Jim Simons results, he is in the same league and like his system, that score is in the D1 range, belonging to the lowest 25% of all scores. This means that either your disk setup can be improved or you have too many disk related background processes running, like indexing, compression or disk optimizers.
Your MPEG encoding is quite average for the i7. Nothing wrong with that result. It shows that you have a good CPU and in comparison to others you are in the top 25-50% on this test. However, due to your disk intensive results, your rendering test is only mediocre. Your system can be optimized to give better results. Bill will include your data shortly and he probably will give you similar feedback. On Total time your results are MED, but in RPI only Q1. What that means can be found in the legends on the results page or in the methodology on the conclusions page. |
January 19th, 2010, 02:32 AM | #26 |
Major Player
Join Date: Mar 2003
Location: Sweden, Stockholm
Posts: 469
|
I rendered alot of material yesterday(SD material, 720x576) - I render from a striped HDD array to a single SDD disk. I have a Quad 9450 @ 3.2ghz with 8gb mem and Win 7 x64.
I noticed that renders where I only had cuts(no color processing or so) - AME used about 25-30% cpu - disk usage was about 200mb/sec(resource monitor). When AME rendered projects where I used blur/color processing/etc the cpu was rock solid at 100% - disk down to about 25mb/sec. My conclusion: When "joining" files(ie cuts) - only one thread is running When processing media, all available cores are used. For me, the latest version of AME works very well(compared to a year ago) // Lazze
__________________
Sony HDR-CX130,Canon 5DMKII,i7 930@4.0, 12gb memory, 3x SSD + 2x1Tb(stripe), Geforce 480, Dual 24", Win 7 x64 |
January 19th, 2010, 06:11 AM | #27 | |||
Regular Crew
Join Date: Jul 2007
Location: Ft Myers, Florida
Posts: 65
|
Quote:
Quote:
First result set: 435MT, Computer Model 93.1, secs Total Benchmark Time 10, secs AVI Encoding Time 37.1, secs MPEG Elapsed Time 46, secs Rendering Time Intel, CPU Manufacturer I7-920, CPU Model 2.66, GHz CPU speed Total render queue time was 9:02 Second: 435MT, Computer Model 80.5, secs Total Benchmark Time 10.5, secs AVI Encoding Time 37, secs MPEG Elapsed Time 33, secs Rendering Time INTEL, CPU Manufacturer I7-920, CPU Model 2.66, GHz CPU speed Total render queue time was 8:42 With some minor exceptions, it looks like everything is relatively the same. Quote:
As far as some 3rd party filters, Magic Bullet Looks, for instance, seems to slow down CPU utilization to about 25% no matter what. I haven't used it in a few months since a few upgrades but assume it might be the same. These days I edit AVCHD almost exclusively. So far I've found the biggest determiner of CPU utilization and render time to be the codec/format I'm exporting to. For example, exporting to QT MPEG4 gives me only 25%. Yet exporting to QT H.264 gives me 75% (all 1920x1080p). I've tried a number of other export formats per this thread. I've had around 75% with most. The only time I've seen 100% is during Harm's benchmark test. When rendering out the AVI file, (there are 10 in the test), during about the last 1/5th of the render, all 4 cores will peg to 100%. Looking at the timeline, it is during the multiple color-bar portion of the test. Last edited by Rob Johnson; January 19th, 2010 at 07:01 AM. |
|||
| ||||||
|
|