Mark, thank you for taking the time to set up the webpage and create a process.
I added another wrinkle to my tests, running renders and previews (Vegas 12 build 670, Win 7 64bit) with Preview RAM set to zero and then to 16. I found a considerable difference on my i7-3770k/GTX 660ti machine, shown in the following table:
Do you think it is worth doubling the number of renders to check on this long-maligned variable?
ok here is what I got and Eff # 0f cores x Mhz avg. I am confused. If you can show screen shot of how you get that in calculation.
Vegas Pro 12 build 367 64 bit
Windows 7 Pro 64 bit
3930K @4.6Ghz
GPU Acceleration ON (radeon XFX HD6970 2gb card
AMD 13.8 driver.
preview performance fps 25 sec into timeline, Set to Best/full 29.970( real time)
XDCAM EX render 1920 x1080 60i > 41 sec
CPU load 24% GPU load 82% (not sure how you get average fps in render (Need a screen shot of this)
MainConcept AVC
CPU load 13%
GPU load 90% (OpenCL)
Render times 43 secs
If you can help figure out how the add the rest. I will do that.
Thanks for diving in here. I have rewritten the testing instructions so that they are clear. Basically we need results with GPU assembly off, GPU render off, and then GPU assembly on with GPU render off and then on. If you follow my directions I'll get all the info needed and update the spreadsheet tonight while I'm watching the MMA fights;)
dxdy - I'm including dynamic RAM preview settings in the table. Good idea.
Thanks for bearing with me guys. I've run a site like this years ago for MediaStudio Pro and it take a bit of massaging to get things going correctly. Already we have a ton of data to analyze.
Don't worry about making "eff" or other computations. I do that in the spreadsheet.
To finish up can you answer the following?
What is your dynamic RAM preview setting?
What is your preview fps at "Best/Full" setting with GPU off at 25 seconds on timeline? Remember to restart.
With GPU off, what preview setting will provide full frame rate preview at 25 seconds section?
Finally can you do the Mainconcept render test with GPU on in preferences and CPU only in Mainconcept dialog?
Results sent to you. These test will be good indicators of what hardware we should be using. I do want to mention that I ran these tests on Saturday. On Sunday I ran XDCAM render over again, same settings, and it took 30% longer with same apparent CPU load. Maybe stuff happening in the background, maybe the effect of moving the veg out of the C:downloads directory to a data drive. So many variables...
980X 6-core @4.0Ghz (3+ years-old now!)
GTX 570, 296.10 Driver ($70 AR about a year ago)
Windows 7 Pro 64 bit, All Updates/No Junk
Source on SSD & Target on fast RAID
Preview RAM = 200 for all tests
Vegas Pro 12.726, GPU ON:
Preview Performance @ 1920x1080 Best(Full) = 29.970 (real time)
XDCAM EX render 1920 x1080 60i = 0:53
CPU load = 55-65%
GPU load = 75-85% (CUDA)
MainConcept AVC render 1920x1080 = 0:45
CPU load = 45-55%
GPU load = 75-85% (CUDA)
Vegas Pro 12.726, GPU OFF:
XDCAM EX render 1920 x1080 60i = 2:41
CPU load = 50-60%
GPU load = 0% (OFF)
Vegas Pro 10e 64, CPU only
Preview Performance @ 1920x1080 Best(Full) = 29.970 (real time)
XDCAM EX render 1920 x1080 60i = 1:17
CPU load = 98%
GPU load = 0% (N/A)
MainConcept AVC render 1920x1080p = 1:21
CPU load = 90%
GPU load = 0% (N/A)
My Typical Workflow = (2) Renders at Once:
(2) XDCAM EX renders 1920 x1080 60i = 1:15
CPU load = 65-70%
GPU load = 75-85% (CUDA)
(2) MainConcept AVC renders 1920x1080 = 1:27
CPU load = 65-75%
GPU load = 85-95% (CUDA)
NOTES:
Vegas 12 does not utilize 100% of a 6-core CPU during GPU renders, so there is little difference between my 3 year-old 980X & friends' newer 6-core 3930K & 4930K. The new AMD 290 appears to have amazing OpenCL performance so I am interested in how that $400 card compares to my $70 GTX 570... However, I will not buy new hardware until I see how it performs in the next release of Vegas...
I submitted mine by email toady. I disagree with you. My 3930K is utilized on all 6-cores as much as your 4-core but the question is, can other components keep up with it. If they could, all cores would be running at 100%. You also have to compare it with a 3960X or the the latest extreme processors as yours is one of that kind.
But yes, the 980X is a great performer. I must take some time and do some test with my old Q6600 and GTX470 system since I still have it in the corner.
Benchmark site updated with new scores. Will enter Rhino's scores now.
Results spreadsheet is also updated so you can download and work with it if you like.
If you submitted scores and have blank spaces in your results I would greatly appreciate it if you could sent me the rest of the data when you get a chance to do it. With that being said, I realize this is a labor or love (and science) of sorts and we all have "real lives and work" to get to so I appreciate any time given to this endeavor. I have a 1 year old and 5 year old myself so I know how hard it is to squeeze in time unnecessary extras.
Also I'm thinking of adding a "Stability" column to the results. I need some help coming up with various stability rating levels. Here's a starting point for us to work it out if it seems like a good idea.
Rock solid (never or very rarely crashes),
Very stable (crashes occasionally on 1+ hour project with third party plugs), Stable (crashes on 1+ hour projects without third party plugs),
Moderately stable (usable, but crashes a few times a day),
Unstable (crashes all the time for no apparent reason)
The 980X is an I7 6-core, not a 4-core. At 3+ years-old it still out performs ALL newer 4-cores & with the same GPU can hold its own against the newer 6-cores in Vegas 12.
IMO Vegas 10e was optimized for CPU renders and V12 gave-up some CPU efficiency to offload processes to the GPU. Even with GPU turned-off, faster CPUs are not fully utilized in V12 like they were in V10...
The real test is how systems multi-task. With (2) internal 400+ MBps hardware RAIDs, we are able to render (2) VEGs in the background while burning (2) Blu-ray ISOs and editing another VEG in the foreground - without editing pauses or bad burns. Most of our source video is DNxHD but we also work with uncompressed HD. When we finish a VEG, most clients today want more than one file type so rendering multiple VEGs & multi-tasking is a must for us.
(Quote) Even with GPU turned-off, faster CPUs are not fully utilized in V12 like they were in V10...
That is not true.. V12 do use all cores rendering to CPU only. How much its use is depends on what plug ins you are using. Here was my test, render cpu only with newblu FX color correction, cpu usage all 6 cores 98% solid. While using the same footage but using BCC7 color correction. CPU usage ranging 42-53% and used all 6 cores. This test is a 4.7Ghz 3930K. With that being said. Everything else is equal. The 3930K is still 15-20 % faster than the over price 980X when it was first relased to the public.