Hi, I've a concern about performance and don't really understand what is the problem. I've this system
System architecture: Intel Xeon E5-2687W
Physical processor count: 8
Logical processor count: 16
Processor speed: 3092 MHz
Corsair Dominator Platinum
Built-in memory: 65489 MB
Free memory: 53725 MB
Video Card: NVIDIA Quadro 6000
OpenGL Drawing: Enabled.
OpenGL Drawing Mode: Advanced
OpenGL Allow Normal Mode: True.
OpenGL Allow Advanced Mode: True.
OpenGL Allow Old GPUs: Not Detected.
Video Card Vendor: NVIDIA Corporation
Driver Version: 184.108.40.2065
Driver Date: 20130131000000.000000-000
Video Card Driver: nvd3dumx.dll,nvwgf2umx.dll,nvwgf2umx.dll,nvd3dum,nvwgf2um,nvwgf2um
Video Mode: 2560 x 1440 x 4294967296 colors
LSI MEGAraid 9260-8i with 5 Corsair Force 3 GT SSD plugged in RAID 0 with 512 Mo Cache.
I get amazing transfert rate of (2.0Gb/s in Read and 1.5Go/s in Write)
I also use Ramdrive for project and get terrific (3.6Go/s in Read and 3.7Go/s in Write with 1311Mb/s (4k) random read and 829.8Mb/s(4k) write
Windows 7 64 bits & Windows 8 64 bits as been tried for test...
Motu MK3 ASIO
Now, as you can see, I run a monster (I Think...)! I work with HD video (1080P 29.97 fps)
When I preview render video (using enter key), It's faster to use software mode vs using mercury engine GPU (10s less for a 16 seconds video). It take to me minutes and minutes to render 30 seconds video. I don't use any special effect. If I just revserse speed of video (100% - no effect), I'm not able (even if I render video using enter key and wait many minutes) to get a none "laggy" video. I always get lag in my video when playback the video. I'm not able to get clean preview with correct reel frame rate. I can believe that. I've the last Quadro 6000 drivers installed on my computer, I've a clean installed OS. If I check my ressource monitor, my disks are used at 2 or 3% only. I've more memory than what I need for 16s video. I've a clean install of Adobe Master Collection CS6. All update as been done (Windows/Adobe).
Now, I'm speachless ! What's wrong with my setup ? It's really frustrating to not be able to play correctly a 50% speed down HD video without lag with a more then 10K $ computer... Any help will be REALLY appreciated. I will send some good cookies of my wife to anybody get me a solution :O) !
My main language is french, so sorry for the mistake ! I think you get the general idea...
Thank you !
This is the sequences settings...
Editing mode: AVCHD 1080p square pixel
Frame size: 1920h 1080v (1.0000)
Frame rate: 29.97 frames/second
Pixel Aspect Ratio: Square Pixels (1.0)
16:9 Fields: No Fields (Progressive Scan)
Sample rate: 48000 samples/second
I use AVCHD, H.264 and .MGEP video (All format gives me same trouble).
I only use 1 video track at time (for my tests....)
I benched my computer to be sure everything is correct with it
I used passmark 64 bits to do it.
I got 5197 scores
CPU Mark - 14715 (Terrific speed)
2D Graphics Mark - 756 (Very good)
3D Graphics Mark - 3724 (Very good)
Memory Mark - 2641 (Terrific speed)
Disk Mark - 11316 (> 1Go/s Terrific speed)
The computer seems to be in perfect health !
How long are the clips? Does the following apply?
Look at the video I did. 17s length. You will see the problem. At the end of playback, there is 1 sec where the problem disapear (The video became smooth) (Like it's supposed to be). It always like that !
I will try some things on your link (I effectively work with AVCHD and effectively see very high CPU requirement (But the XEON E5-2687W seem to deliver that power). I was thinking that exporting the media with codec like H.264 will change the video format. Do I really need to use Prelude to transcode de video ?
It seems many people here with very good results are using a regular GTX card - like a GTX 580 with 3 GB. Did you tried with a regular card like this ? Since you are willing anyway to do so many tests (windows 7 and windows 8).
I'm willing, but Adobe will have to answer the inconsistance between that kind of "marketing" http://tv.adobe.com/watch/digital-video-cs6/how-to-optimize-after-effe cts-cs6-for-high-performance/?go=12437 and the real world. Adobe recommands and never stop to says that Nvidia Quadro is the product to buy to get the best performance. I agree with you that If I've to pay 3500$ less for a better card (Obviously, I'm not against that), why Adobe continue to promote that product. I can believe it's so simple. If I had the certitude that GTX card like 680 can give me better performance, I will sell right now my Quadro 6000, get 3000$ back and will buy 2 GTX Titan (next generation) in SLI to get better performance ! And I will have 1000$ in my pocket after the transaction. Thanks for your suggestion, but I still continu (for instance) to believe there is a solution for that problem !
First, as far as I know - the GTX 580 used to be better for programs not 680. The 680 is very good with games, and probably will be good for programs with certain newer drivers. Perhaps they are already good, I don't know. Just the last time I checked the 580 was recommended.
And second, as far as I know you won't get an advantage with two card in SLI for most of the programs. SLI is good mostly for games.
There is no sense in getting two Titan cards in SLI, simply because SLI configurations are not supported and give serious problems. Not to mention it would likely require a new PSU.
With regard to your Passmark results, they are certainly good, but not spectacular. Look at (click on the image)
and notice that my system has a GTX 680 / 4G card that easily runs way faster in this benchmark test than your Quadro 6000 and is only a fraction of the cost. Your Passmark results may indicate that some tuning on your system might be needed. How many processes are running on your system when you use PR?
Your scores are significantly lower than mine by respectively 33% - 9% - 36% - 68% - 33% and 63% which tells me you can optimize your system a bit.
For more testing results of my system, look at Final Results - Reflections
Also take a look at Practical System Requirements which tells a different story than Adobe, for the simple reason it was not made by Marketing. Quadro cards have always been pushed by Adobe Marketing, but are in fact overpriced and underperforming.
Thanks for posting. I understand your point, I also understand I can get better performance (for gaming) with a GTX card. Making 3D and floating point computation is not the same business than gaming. (Because, I work with 3D also). But for video making, it's really not clear what is best or not.
My passmark score is a fair score. I understand I can push the computer way above this limit (I got near 7000 with a 4 years old computer). But the fact, I don't want to create ramdrive and bench virtual setup just to get better score (The computer are not stable and I don't really want to run this kind of system). If you're able to get 18K with disk can you say exactly to me what is the transfert rate, what is the setup (The card/Disk/Controller/RAID/SAS configuration, ...) you have. This kind of information will help me to understand what you do and will help me ! We have to be careful with passmark score, so many people get score putting their computer into freezer just to get the maximum of speed. Second, you get 16K with your CPU (By overclocking it over 4/GHz). The only reason why I used XEON it's because I can install a second (none overclocked) CPU to get near 30K. The passmark video testing is only gaming testing score. The thing I wan't to know is why Adobe say that Quadro is better then GeForce. I can see the spec (Geforce has more cuda, faster processor, faster clock bus, faster rendering capability, cheaper price), but I can't understand logic behind the scene ! So now, If you can confirm to me you have GeForce 680 and can play AVCHD video (The most difficult video to render) easily, I will consider to do some test soon. But actually, I'm not sure the video is the problem. Thanks for suggestions.
There is a whole site behind this link: Final Results - Reflections that may contain answers to your questions. Take some time to look around the other pages as well. It is required that you register and login to the site to access for instance the Detailed Results page.
But the fact, I don't want to create ramdrive and bench virtual setup just to get better score
Neither have I, as documented on the page I linked to.
Why Adobe is pushing Quadro so much, there has to be a commercial reason for that I think. Marketing people have a natural tendency to leave out facts and only proclaim 'the ultimate dream' that is often far from the truth. There is ONE, and only ONE situation where you want a Quadro card over a GTX card and that is when you NEED 10 bit output to expensive 10-bit monitor(s).
BTW, the Quadro 6000 has 448 cores, the GTX 680 has 1536 cores, the 6000 has a memory bus of 384 bits versus 256 for the 680, the memory bandwidth is 144 versus 192.2 GB/s and most importantly the 6000 is a much older architecture.
Thanks for link. I will read everything there to get better idea of what's up with video editing vs performance. You didn't answer to the question (What si your disk setup to get a 18K result ?)
It is also documented on the site, but essentially it is 3 x 7 disk raid3, striped to raid30 plus 3 global hot-spares of Seagate Constellation ES 1 TB disks, making a total of 24 disks and a net capacity of 18 TB.
HDTunePro, CrystalDisk and AJA benchmarks are included on the page I linked to. Just scroll down a bit.
Bill and I are still struggling with PPBM6, so it is not yet available. I have received your PPBM5 results and added them. Rank #22 among the CS6 results, solid Q3 results with a very good H.264 score, due to the 8 core CPU. The MPE results are comparable to a GTX 570, which makes sense since the specs are very close to the Quadro 6000, apart from the amount of VRAM of course, but the amount does not impact performance in this test.