Well, check Adobe Minimum System Requirements for CS 5.5.2, make sure you have upgrade ?, Check approved video cards in Adobe usually Nvidia Quadra , mine is Fx 3800 does work on PC but only Fx4000 on Mac i think, as some rendering will show as red on time line, normally green 0or even yellow. I use H.264 1920 x 1080p PAL mov from a Canon 5D Mark II, works well . Only Fx4000 for MAC, others may be (will be more helpfull).
EXEON i7 Windows Professional 64 bit only Quad Core 2.27 GHz 12GB RAM 1 TB 15,000rpm Hard drives Raid 0 is fast enough , but, hyper - threading may be handy for complex editing , run you computers performance monitor , it will tell you, I hope. is ok, all the best., incidentally Sony Vegas Pro 11 runs on only 2GHz, amazing .
Editing is a breeze, your unit looks big enough, but, Video cards FX 1700 looks old ?????????????????????????????
Message was edited by: Bob Dix Photographer
I recently installed a GTX 580 from EVGA (3 GB Ram), costed my $ 500,, and I wanted to see how much performance gain I got.
For now I was just testing playback with mutliple layers,, I know some effects are not CUDA-enable, thats why I did the test withouth any effect at all,, like I say,, just three layers (the top two scaled down on top of the main one in the background).
When I have only one layer the playback is great, fast,, but when I put three layers of video the playback becomes chunky after forward 2x of more,, and backwards is more noticeable.
I know Im not using the correct terminology, Im not a profressional editor, this is just family stuff and I since I already had a workstation for other reasons I though I could use it for Premiere.
It looks to me that I have a bottleneck somewhere on my system (Im concerned about my XEONs E5420) because they show a lot of utilization even when Mercury Playback is enabled.
I have no reference so I dont know if my system should perform well or not.
I did the PPBM5 test (little confusing to me because it seems to be configured for people that have media, projects and scratch disk all on the same HD and I have them in three RAIDs so I was running in circles making it work)
I sent the results and my system scored 377.
377 2012-02-09 11:58:35 Base2008PT1 Desktop Dell T7400 Intel Xeon E5620 2.5 2 8 32 425 3.41 198 94 123 156 10 15.6 5.5 GTX 580 SATA 128 SSD 2 Raid 0 SATA 128 SSD 2 Raid 0 4 DEL SAS i6
Thanks guys for the advise!!
Thanks Harm, I posted my results for the PPBM5 test on my last post,,,my system name is "Base2008PT1",, maybe you could spot something there.
As you can see from the rendering test, hardware MPE takes 10 seconds and software MPE takes 156 seconds, so that is clear proof that your CUDA card is used very effectively. Your MPEG score is in line with similar systems, of course helped by the 580, which accelerates the scaling in that test. The H.264 test is somewhat disappointing, but that is mainly caused by the relatively low clock speed of your CPU's. However, your Disk test is disappointing and I guess that is caused either by background processes, such as indexing or compression not being turned off, anti-virus software running or similar, or the Dell SAS i6 not supporting the trim function on your SSD's and then writing a 13 GB file can show the weakness of SSD's on write speeds.
Regarding your remark on the submission form, my preference would be an overclocked i7-3930K or even an i7-3960X with 64 GB memory and a large parity array with an Areca ARC-1882iX/16/24 with 4 GB cache, rather than a SR-2 with dual Xeon X5690's. Even if money is no object, I rate the benefits of a very good disk setup higher than the benefits of a dual CPU system, at least at this moment in time, because the architecture of the current crop of Xeons is rather dated.
Thanks for letting us know where to find your entry. I remember processing it and my question to you and your prompt reply, but had not made the link with your screen name here.
I did order an ARC-1882iX , expecting to receive it tomorrow,, I will definetely run the Benchmark again and compare.
Regarding the i7 vs Xeon, I have a T7400 right now (Harpertown, LGA 771 socket) so the highest I could upgrade my CPU would be to a Xeon X5492 , and neither of them has Hyperthreading, etc etc.
My other options would be replacing the Mobo on the T7400 for the new T7500 and upgrade to the latest Xeon Gulftown (LGA 1366 socket) , it will fit in my chassis and my current PSU can handle,, but then no overclocking for me,, plus the comment you made about i7s being more suitable for Premiere than Xeons.
The last choice would be a complete new rig, (case, Mobo, etc etc),, what Mobo would you recommend? (Ive been reading about the SR-2 for days, looks great for Xeons, but I didnt read much about LGA 2011 - based boards)
I would like to take a look at the LGA 2011 Mobos,
Were you using an Anamorphic preset for the project you spoke of in the first post?
Im using Square Pixels 1.0
What sequence preset did you use when you created the sequence?
DSLR 1080p30 (that matches my EOS 7D Video I use for test)
Editing mode: DSLR
Frame size: 1920h 1080v (1.0000)
Frame rate: 29.97 frames/second
Pixel Aspect Ratio: Square Pixels (1.0)
Fields: No Fields (Progressive Scan)
The funny thing is that playback backwards hits the CPU harder.
If I playback forward and backwards about 5 seconds continuosly, the CPU goes down and it is smooth playback,, but if I let it run forward or backwards a few seconds more then it becomes chunky again.
Im looking forward to test the ARC-1882iX this week.
I dont know if it helps, but I have my projects sitting where the OS lives (on SSD disks Raid 0), I have the media on a individual SATA drive, and I have the scratch on two SATA Raid 0 HDs.
What video driver do you have installed for both cards?
220.127.116.1162 for both
(thats the latest one I guess,, the 285.62 on Nvidia website)
sorry for adding more trouble,, but I just realized that I have the same problem even with one layer of video.
At 1x playback is ok,, at 2x is ok but after 15 seconds it becomes chunky, and at 3x or more it is really bad.
I dont event know if Im supposed to be able to see smooth playback above 2x,,, am I?
Well CS5.5 has a bug with shuttling back and forth through the timeline at faster speed. The Caching model changed with 5.5 and it really effected that so that is normal right now. Change the MPE mode to software and delete the Preview files. Change it back to hardware and delete the preview files again. See if the average load on the CPU changes faster when you start playback. It will always start out higher as the frames are caching but it should drop to normal load fairly quickly.
I see, Im running Premiere 5.5.2 (latest update),, I tried already what you suggested, a few times, comparing MPE vs Non-MPE.
When I have Software Mode, I create the Previews and the playback flies at all speeds,, the problem is when I have MPE turned on.
The true is, I dont need to Shuttle so fast, it is just that I was using it as test for my CUDA card,, because thats what Adobe advertised on their videos about MPE "hay, look how fast we can Shuttle, even with effects and multilayer".
So, when I saw that performing real bad on my machine I started thinking that my Premiere was not using properly the CUDA,, I saw the COPU compsution real hight,, I saw the GPU utlization real low , thats why I though I had issues.
At this point I dont really know if I have a good system or not,, my PPBM5 rank is 377 (above are the details),, looks like I coudl benefit from better CPU and better Disk Controller,,,, I will ge the controller this week,(eventhough it seems to be ok since with the Preview files on Software Mode I get very good shuttling performance).
Im scared of upgrading the whoel thing just to find out that I wont gain that much from it. (on the CPUs).
The most difficult part of the MPE engine for editors to understand is that it's a pipeline now. The CPU has to process the decode and data before the data transits to ram and then down to the GPU ram for GPU processing. Then it transits back to system ram for CPU encoding. This means the CPU GHz or speed essentially decides the overall performance of the MPE engine once you have enough threads/cores to handle the decode efficiently. Your Xeons have the threads but are very slow on the GHz which is why your GPU load is so low. The GPU you have can handle such a high load that it's essentially waiting on the CPU to get it's work done before it essentially has anything to process. Upgrading your drive controller or drive speed will only have a marginal impact on this. The Software mode is keeping all of the processing up at your CPU which threading will compensate for lack of speed. That is why the Software MPE is working better to some extent. The shuttling performance issue is just an issue with CS5.5 and hopefully will be addressed soon. Adobe went backwards on their caching model performance and I am not sure why or how it made it past testing.
Eric,, in your opinion, my Xeons are falling behind, right?
If you had the chance of changing system, what would you recomment?
- Dell T7500 with two Xeon X5690 (no overclocking capabilities)
- Stay with my Dell T7400 and upgrade to two Xeon X5492 (no overclocking capabilities)
- New System with EVGA SR-2 and two Xeon X5690 (or X5680) (with overclocking capabilities
- New System with single CPU i7-3960X
If I may ask, please rank the options above.
All of above including 64GB of RAM, ARECA RAID card, Nvidia GTX 580 , etc.
Not Eric, but let me rank them as I see it:
1. Dell T7500, 3-rd place
2. Dell T7400, last
3. SR-2, 2-nd place
4. 3960X, 1-st place
Let me add that I don't like Dell, or HP for that matter, for editing machines. Too limited, crippled BIOS, no overclocking capability, limited PSU options and way too expensive.
I agree with Harm's rankings pretty much. Your particular Xeons are both pre-Nehalem - in other words, based on the obsolete Core 2 Quad architecture. As such, they fall short of even i7-920-class performance, and even an i3-2100 can beat a single Xeon E54xx-series CPU in CS5.5 performance.
Im almost making my mind here now :-)
So you all think a single i7 3960X will beat a SR-2 with Two Xeon X5690 (dual CPU) on Premiere CS5.5?
If so, then I have to be careful with the Motherboard I pick, because the LGA 2011 Motherboard landscape is super populated with brands and options, and looks is very confusing to me.
The ASUS P9X79 WS looks good to me.
Thanks a lot guys,
I agree with RJ and Harm and I would add any other until the Xeons are updated to the new platform is a waste of money. The MPE engine will work best with the SB-E platform right now and especially either the 3930K or the 3960X at over 4GHz. Essentially the biggest advantage to the SB architecture right now is the clock speed ceiling which is allot higher on average than the previous platforms. Once you exceed 8 threads with the MPE with most codecs, the GHz decides the performance. Now once the Dual Xeon SB's release, we will have to see what that brings to the table.
Well, now I will be in the quest for the right Motherboard with space for:
- Two GPUs (I know MPE uses only one but I have three monitors)
- One ARC-1882iX Controller PCI-E 2.0 x8
- Plenty of RAM
I hope if I get the investment right then it will last me three years as my current T7400 did.
Up to 64 GB, GTX 580 as the primary card in a PCIe-16x slot, GTX 550 as the secondary card in a PCIe-8x slot and the Areca in a PCIe-8x slot.
Mr Millaard and Bowen, I received and installed the Areca 1882-ix-24 , the only slot I was able to use was the PCI-E 16 where my second GPU was installed, so, no three monitor setup for me anymore.
I left the GTX 580.
I got some more HDs and set the following:
- Two SSD Force GT120 GB each on RAID 0 for the OS
- Four SAS 10K RPM (not SATA) 300GB each on RAID 0 for Scratch Disk
- Two SAS 15K RPM 175GB each on RAID 0 for After Effects Cache (and also Lightroom Preview 1:1 renders)
- One SATA 7.5K RPM (no RAID) 3 TB for Media (Videos, Audio).
My media disk is in live one way backup with a external disk.
I took an image of the OS after full installing it using the free Macrium Reflect (I used it before, it is awesome).
I dont care about backing up Lightroom previes or Scrath files.
Premiere Pro projecst are still on the MyDocuments on the OS SSD Drives.
So far, nine HDs, I have room for (believe it or not) six more HDs on my Dell T7400 but Ill stop here since thats pretty much all the drives I had sitting around.
The areca cards is awesome (runs a little hot to my taste, but seems fast to me, and I can barely hear the fan),, but Im not sure if I configured the stuff right,, Im using HD Tune Pro to measure the HD performance,, Im getting 800Mbits on the OS, 800Mbits on the scratch disk and 200-600 on the rest. (all using the first tab "Benchmark",,) if I use the other tab "File Benchmark" the numbers for transfer on OS array and Scratch goes all the way to 2000 MB/s (not quite sure what the difference is).
I wonder if those numbers looks right.
Also,, I will run the PBM5 next week and compare my results before and after the Areca card,,, but my renders, previews, transfer and scrubbing seems a lot faster than before.
But Im afraid you guys are right, my processor is falling behind, when I render complex effects it goes to 100% , seems that it will be time for a new set up after all.
I read about the new Xeon E5 processors, also the upcoming new Dual CPU boards (from EVGA and ASUS), so I guess I will sit tight and wait what will come around (and to which price :-) )
If you have any suggestion for me on my current HD configuration or where to place the files so they are on more optimun location for Premiere, I will be very thankful.
Thanks a lot
7D video is encoded on the camera using h.264 codec. While you scrub the timeline or play a preview in real time, your CPU has to decode h.264 (which is processor intensive by itself) before any of the good stuff gets to use the current frame of video. No effects, CUDA cores, sending to monitors, color corrections etc etc until your CPU has dug into the h264 file, decoded that frame (a lot of work) and then sent it out to the have all that other stuff done.
Because of this, your bottleneck becomes your CPU like others have said and upgrading your CPU will help of course. One other thing you can do (a big thing) is transcode your video files before using them in your applications (like premier etc). This basically means converting your videos from the "compressed" format of h264 into a "inflated" format made for editing. The best one is Apple ProRes if you're on a mac and have Final Cut installed (thanks Apple, for making your awesome editing format only available by buying your crappy software). If you're on PC, download a program called MPEGStreamclip and use it to convert your files into whatever intermediate formats (a.k.a "inflated" formats) it gives you (i dont remember which is best to use when on a PC, but you can google the formats yourself to find out). Then work with the resulting files. They will be huge files, and for good reason.
This will result in a pretty substantial performance boost. The reason is these inflated formats aren't compressed. So Premiere is like "hey this guy wants to add 40 effects to frame number 1034, go get the frame mr CPU" and that frame is already decoded and waiting to be picked up. Its easier on your CPU and your Hard Drive and therefor results in better performance in all aspects of editing (at the cost of disk space).
After a few days trying I confirmed exactly what you just said.
h.264 is heavy on the CPU, and Transcoding to an intermediate format is one way of improving the scrubbing performance wihtouth having to upgrade the CPU.
I tried NeoScene from Cineform and it yields much bigger files, jus tlike you said, I read that it basically pre-creates many of the frames in between than h.264 presents to the CPU as a mathematical calculation,, effectibly offloading the CPU from the task since after the transcoding the frames are "already there" so the CPU does not have to 'work' to create all the frames in realtime.
It definetely improves a lot the scrubbing.
HDD space is not a concern for me,, what is a concern is that h.264 seems to be already a 'less than ideal quality' file,, and Im thinking that just like when you save and re-save JPG images,, everytime I transcode im basically losing detail on the footage.
I feel concerned about transcoding because I dont know how much 'visual data' Im losing everytime I do so.
Do you have experience on that or any advise to offer me?
If Im not losing data, then it is fine to me, I could transcode as standard practice right after my PC consumes the material from the memory cards.
This is for all of you who are thinking SR-2 as a cheaper way to gain CPU speed...
EVGA told me that my SR-2 was for gaming, not video editing. I could not get any audio I/O to work with the PPro timeline. Everthing else, and I mean everything else works. Not PPro. the timeline hangs and stutters while trying to deliver either Firewire or even ASIO to a X-Fi Titanium Pro. Thank God, really, thank you Lord, for B and H having a liberal RMA process. I tried every combo I could afford and no success. I switched to an Intel Dual (non-OC) mainboard to get PCI so I could go back to an M-Audio Delta I/O.
Here's a list of what didn't work: Saffire Pro 14, EMU 1616, M-Audio Firewire Pro, 2 different TI chipset Firewire cards, the Soundblaster I mentioned, Black Magic Intensity Pro...None of these worked on the SR-2.
I also called DIgital Storm to ascertain if they could get their audio I/O working, and they informed me that they too had little success.
Save your money...unless you spring for 5690's, you will not be able to use the SR-2 as a video editing solution. Now, I have an email in to ADOBE (Hi ADOBE GUYS!) to try to get a trial of CS6, just to see if it will work on an SR-2 timeline with a firewire installation. If it does, all of us are in business!
If you guys in ADOBELAND can try this, please post it quickly! We could really use this test information.
Please pull the Quadro out, including its drivers. Reboot, check your cudasupported card .txt file for the 580 and go back to your timeline.
Adobe does not do well with 2 cards, especially if one is a Quadro and the other a regular card like the 580.
I have a couple posts regarding the SR-2. Do not get one. It does not fully support PPro.
As soon as you install some sort of audio I/O (Firewire) etc, your PPro timeline will either hang or stutter, hopelessly.
Also, When it comes to After Effects, a dual Xeon with tons of RAM can't be beat, even with the extremly fat 3960. AE takes full advantage of all the threads and will process very quickly.
BUT!!! DOn't buy anyrhing until we have a chance to determine if CS6 will run better on the SR-2. If it's really fixed (PPRO that is) an SR-2 will be choice number 1. with a couple 5645's OCd to 3400MHz, it would be a smokin workstation.
Also, CS6 will need a newer video card to use the new colorizer....
MPE is used for DVD encodeing mostly. Render something out to DVD and use GPU-Z to monitor useage. Let us know.
MPE is used for DVD encodeing mostly. Render something out to DVD and use GPU-Z to monitor useage. Let us know.
Actually that statement is incorrect without additional information. The fact is that for simple MPEG2-DVD encoding only the CPU is used, now if you are scaling down from HD to DVD or are using some other of the MPE assisted video effects then you will see heavy GPU usage.