Nick

The cards appear as two identical cards, not one. I always have HWinfo64 running in the background and both GPU load indicators in my taskbar.

The load depends on where the playback video is displayed. I have 4 monitors, two on each card and that could be the reason why I am seeing both loaded. However, during render it doesn't matter which windows layout I use, both cards are almost equally loaded up to 80% depending on what media is rendering.

Bruce.

I do have a ASUS R9 290 lying around and I am waiting for a second one at a good price. I am reluctant to try the one cards first as my two 580s are watercooled and ripping them out for a test isn't a small task.

However, if I believe what you are saying it might well be worth even trying one R9 290?





I enabled crossfire on mine and always left it at that. In After Burner I see gpu loads mostly on the first card and the 2nd card is still load but not loading equally with the first card. Since you already got 1 290 and you definitely heading toward AMD Land. Why not tear down the GTX580 and pop in the 290. It will definitely perform better than the GTX580 for sure. You got to think about this. This benchmark is a 10 tracks and heavily FX applied and 4K footage is some serious stuff. . In the general, let say you got 5 tracks and some FX, this will definitely perform much more better than the benchmark. Don't you agree?

Old Smoke,

I finally have my rig (5960x, 32 GB RAM, R9 290) up and running enough to run your benchmark. I am not quite able to get the overclocking performance that Nick got. I can run at around 4.4GHz. It’s tough to make comparisons, but I think mine is a tad smoother than what he showed. I ran at Best Full and left all the other tweaks and pre-sets alone (at least that is where I seemed to get the best results). The transitions are still “blocky”. They could do for editing in a pinch, but I don’t know if I would really be satisfied with them.

I tried to focus on the fps count and noted at least 7 instances of the frame rate dropping below 29.97. The lowest count was approximately 24 fps.

I also ran some tests at stock settings (no overclocking) with preview at Best Auto. I had roughly the same number of instances of below 29.97 fps but the length of those drops was much longer in duration with the lowest observed fps at around 21 fps.

Hope this helps.

Jeff

@Jeff

Thanks so much for running the test on your system. How is the playback at Preview/Full? Are transitions smooth at Preview/Full?

I think in general we can say if you want to do serious 4K editing, not even a 5960X system will get you 29.97p at Best/Full with no proxies.

Old Smoke,

You are correct. Transitions are not smooth.

My performance looked very similar to Nick Hopes, maybe a tad better, but not enough to be definitive.

Not sure where the bottleneck is (CPU or GPU). Given that the transitions is where the problem occurs, I'd guess it is the GPU. Perhaps two R9's could pull it off.

Jeff

@Jeff

I am getting another R9 290 on Tuesday and I will test my current 3930K system with 2x R9 290. I currently have 1x R9 290 and 1x GTX580 in my system with R9 290 set for GPU acceleration. The SCS HD Benchmark project is flying! I can still render to MC AVC with CUDA and I am down from 28sec. to 25sec; that's a good 10% improvement. However, MC AVC OpenCL rendering seems broken, there is no GPU load and render times are about as good as GPU only.
The SCS 4K Benchmark project is also running a lot smoother at Best/Full but not acceptable yet.

Old Smoke,

Once again you are correct. Main Concept CODEC’s are broken. Here is a quote from Dave Haynie over on Creative Cow.net; “Main Concept, on the other hand, delivers a dramatic speedup if you have a compatible card. Unfortunately, they've rather subverted the whole point of APIs like OpenCL or CUDA (they actually support both) by hard-wiring their CODEC to use OpenCL ONLY for a small set of AMD GPUs -- those available circa 2010. And same with CUDA -- just those nVidia GPUs available around 2010... GPUs that exited prior to Vegas 11 shipping.”

There has been a TON of stuff discussed on this issue over there.

I would love to see a listing of each CODEC and the “optimal” settings for playback and rendering with that CODEC. I understand that some actually run better if you limit the number of cores.

Jeff

@OldSmoke

I'm breaking in a 5960X system @4.5ghz
GPU is HD6970

Without GPU acceleration, on Best/Full, the 4k bench preview dips into single digits quite a bit.
Have to go to Preview/Quarter for almost solid 29.97..

With GPU on, it dips into the teens sporadically at Best/Full..
Have to go to Preview/Half to get almost solid 29.97, with a couple of dips into high 20's..

I'll edit with some render times later on..

@OldSmoke,

I finally have my 5960 cooking along (well, I hope it is not really cooking) and would like to try your 4K test. The Google link in your original post is not working for me, is it gone?

@dxdy

Sorry, I had to remove it, it took up a lot of space. If you have the original SCS Benchmark project, you can turn it into a 4K project very easily. All I did was rendering all project media clips to XAVC-S, 29.97p and set all generated media to 4K 3840x2160, 29.97p.

@OldSmoke:

I think I have duplicated your 4K test, here is what I did with the video:

Mercedes 1 took 11:41 to render to XAVC 4K 3950x2160 Intra 29.97p. GPU (R9 290) got used in occasional spikes, up to 90%. All 16 threads looked like they were about equally used, overall usage was reported in the mid-60% range. CPU i7 5960 temp hit 40C (water cooled), GPU hit 53C (air cooled). Mercedes 2 and 3 were similar. Project Render Quality set to best.

The files created have the following MediaInfo listing:

General

Complete name                            : F:\RenderJKL\Mercedes1.MXF

Format                                   : MXF

Format profile                           : OP-1a

Format settings                          : Closed / Complete

File size                                : 7.35 GiB

Duration                                 : 4mn 30s

Overall bit rate                         : 234 Mbps

Encoded date                             : 2014-12-29 15:24:54.000

Writing application                      : SONY Vegas 13.0.0.428

Writing library                          : SONY Vegas 13.0.0.428



Video

ID                                       : 2

Format                                   : AVC

Format/Info                              : Advanced Video Codec

Format profile                           : High 4:2:2 Intra@L5.1

Format settings, CABAC                   : No

Format settings, GOP                     : N=1

Format settings, wrapping mode           : Frame

Codec ID                                 : 0D01030102106001-0401020201323001

Duration                                 : 4mn 30s

Bit rate                                 : 232 Mbps

Width                                    : 3 840 pixels

Height                                   : 2 160 pixels

Display aspect ratio                     : 16:9

Frame rate                               : 29.970 fps

Color space                              : YUV

Chroma subsampling                       : 4:2:2

Bit depth                                : 10 bits

Scan type                                : Progressive

Bits/(Pixel*Frame)                       : 0.934

Stream size                              : 7.30 GiB (99%)

Transfer characteristics                 : BT.709

Then I loaded a copy of the test veg, changed it to a 4K project, renamed the videos via Windows Explorer so Vegas asked where the new ones are, and changed the 2 PNGs up to 4K via Photoshop Elements 13. EDIT: Bad assumption. See below. I ASSUME the generated media got converted to 4K when I resized the project (maybe not?).

Best Full playback performance is fine. The 5960 is overclocked via an ASUS utility (which overclocks based on demand), I see the CPU running at up to 3 GHz (not the 4 maximum I see in renders, or the 1.2GHz at idle) and frame rate is pretty solid at 29.97. There are half a dozen stumbles, never below 25 fps, and never lasting more than a second. GPU-Z definitely show the GPU being used, but not constantly and nowhere near capacity.

Edit: With the generated media properly resized with the Vegas script, playback is not as good as it was. However, at Best Full it never goes below 22 FPS, and really does not go below 29.97 for very long when it does - the exception being 2 or 3 seconds around the 25 second mark. Apologies to OldSmoke for not reading the original post carefully.

I also have to say based on my experience with Canon MXF 2k 50 Mbit/sec files that I think sheer file size and bitrate of the source is a major player in all this. When I get a chance, I am going to spread the three mxf files across three different 7200 RPM SATA drives and see if there is a difference.

@dxdy

It seems you are getting similar preview on your system as I get on my 3930K@4.3GHz and 2x R9 290 which is good and bad. Good because upgrading to a 5960 isn't realy getting me what I expect from a 4K editing system hence I won't do that. Bad because I may now really have to look into getting a dual Xeon setup but only if I have prove it will deliver.

Is this what is known as the Red Car benchmark? I was looking for that, but in less than 4K maybe. I keep finding mention of the SCS benchmark but all my searches have been for naught. Some 404 errors and just chasing my tail it seems.

Actually, even the old benchmark test from 3-4 years ago would work. I used that and submitted the results back to the forum (I think it was here) right after I built this workstation. I don't know if I need the full "original SCS Benchmark project". I do have 60 Mbps download capability.

Any help appreciated. . .

Wayne

I still have the project "PressReleaseProject.Veg" that came out in 2011 but all the assets come to 2.47 GB and I only have 1.1 GB of storage left in my free dropbox. If you have an ftp site or some other storage, I'll upload it for you. Right now, I'm going to finish mowing my yard.

http://www.sonycreativesoftware.com/vegaspro12benchmark ?
Same as the one below?
http://download.sonymediasoftware.com/whitepapers/vp11_benchmark.zip

Thanks John but I'll save you the trouble since vkmast had the direct link. Not sure why no searches showed it up.

Mow your lawn, that hurts. Mine is under 10 inches of ice and snow :-)

Wayne

Got it downloaded and up and running on VP13. Also finally got to review the PDF document with reports for various cards on VP11. Love that the music was created with CineScore. I wish it would get resurrected, but at least I have a good selection of themes to tide me over.

Should have the replacement card later this week, an ASUS R9 270-DC2OC-2GD5. Certainly not my top choice but all that will physically fit in current workstation case.

I filed all the link info in case I run across someone with my question.

Thanks again for your help.

Wayne

Link to download that benchmark?

You mean this? https://www.vegascreativesoftware.info/us/forum/vegas-pro-faqs--104787/ point 29. (Where can I download the Sony Vegas Pro 11 "red car" benchmark project?) leads you to https://web.archive.org/web/20131113061953/http://download.sonycreativesoftware.com/whitepapers/vp11_benchmark.zip

May be a bit slow but worked for me as I tried. See also.