“New” graphics card

[Rich Kutnick]

Earlier today I ran a battery of tests using the 7248_rendertestjr.veg file. As do you John, I also have 16GB memory, ATI Radeon HD 5870, SSD boot (C: Drive) and 2TB HDD (D: Drive). My PC is two years old this summer and has an i-7 3820 processor running at 3.60 Ghz. You noted that it has 4 cores; is that correct, John? I don’t know the number of threads, though.

However, before I offer my results I want to note my observations. All playback was viewed on FULL/BEST. When I first played back the file on the timeline with GPU ON it started around 22fps and leveled off at 25fps, but on successive playbacks immediately following the initial one it was 29.97fps rock solid! Monitoring GPU usage using GPU-Z was fruitless, for it seemed to be delayed in its reporting values and most of the time bounced all over the place when it did report usage, so I relied more heavily on the fps and how the video looked to the naked eye. Also observed, as did you, John, was that with GPU OFF, playback on the timeline was around .5fps, but again, if I played it again successive playback yielded a wee bit better results (up to .7fps).

Rendering: Sony AVC renders were done using the AVCHD 1920X1080 60i template (unaltered); MainConcept renders used the Blu-ray 1920X1080 60i 25Mbps (unaltered)—I believe that this template defaults to two-pass 40Mbps max.

Here are my results:

Timeline GPU Acceleration OFF (Playback 0.5fps)
————————————————
Sony AVC–(Auto). . . . . . . 2:24
Sony AVC–(Auto)–Rendered it a second time for kicks. . . . . . . 2:21
Sony AVC (CPU Only). . . . . . . :22
Sony AVC (GPU). . . . . . . . :21
MainConcept AVC (CPU). . . . . . 1:05
MainConcept AVC (OpenCL) . . . . :09

Timeline GPU Acceleration ON (Playback 29.97fps)
————————————————
Sony AVC–(Auto). . . . . . . :32
Sony AVC (CPU Only). . . . . . . :23
Sony AVC (GPU). . . . . . . . :22
MainConcept AVC (CPU). . . . . . 1:05
MainConcept AVC (OpenCL) . . . . :08

OK, John, please interpret…BUT FIRST, bear in mind that my Dynamic RAM Preview is set to 8GB—might this affect these numbers PLUS might it also account for the initial GPU ON 22-25fps going up to a full 29.97 upon successive playbacks? Might this explain why with the AVCHD footage that I have shot I do not get full 29.97 when playing back anything with movement the first time (bounces around between 22-27fps) but upon successive playback of the same footage I do get 29.97fps (well, 29.97fps jitters a bit during playback)? Maybe this set-aside RAM has no bearing, but in any case these numbers are NOT typos, but real-world figures from my testing. In addition, then, with such stellar numbers why does some of this same footage refuse to play back with the GPU engaged and therefore freeze SVP 12? Just a minute ago I tested this observation again—and when I moved ahead in the timeline the video froze and the audio kept going strong!!

I do not know how these numbers can be true, though, for explain to me how the MainConcept OpenCL numbers using the GPU acceleration option are all but identical with or without the GPU being active! I also corroborate that the Sony AVC template numbers are the same, using the GPU acceleration option, with or without GPU support. So it appears, that, for my results at least, when a template that is supposed to use GPU acceleration is chosen, no matter if the GPU actually is turned on or off the results are the same!! Something smells very fishy to me, but that is why I am deferring interpretation to you and the experts here. Far be it from me to tell you what is going on.

Thanks in advance, John!

Rich Kutnick
VIDEO IMPRESSIONS

[John Rofrano]

[Rich Kutnick] “OK, John, please interpret…”

OK, we are not comparing apples to apples here (no pun intended) because I made a serious flaw in not telling you what render template I used. I’m really sorry about that. The render templates do make a difference.

Here are the templates that I used:

Sony AVC | Internet 1920×1080-30p
MainConcept AVC | Internet HD 1080p

Luckily you did post the templates that you used so I re-ran my tests using your templates (i.e., Sony AVC | AVCHD 1920×1080-60i and MainConcept AVC | Blu-ray 1920×1080-60i). Here are my new results:------------------------------------------------
Timeline GPU Acceleration OFF (Playback 0.5fps)
------------------------------------------------
Sony AVC - AVCHD (CPU Only) . . . . . . . 2:57
Sony AVC - AVCHD (OpenCL) . . . . . . . . 2:57 (0x)
MainConcept AVC - Blu-ray (CPU) . . . . . 3:40
MainConcept AVC - Blu-ray (OpenCL). . . . 2:47 (1.3x)

------------------------------------------------
Timeline GPU Acceleration ON (Playback 29.97fps)
------------------------------------------------
Sony AVC - AVCHD (CPU Only) . . . . . . . 0:33
Sony AVC - AVCHD (OpenCL) . . . . . . . . 0:39 (-0x)
MainConcept AVC - Blu-ray (CPU) . . . . . 1:27
MainConcept AVC - Blu-ray (OpenCL). . . . 0:25 (3.5x)As you can see, Sony doesn’t seem to use my GPU for rendering even though I have “Use OpenCL if available” and the “Check GPU” shows that it’s available (in fact it was slower with GPU enabled!). Also MainConcept is consistently 3.5x better with GPU enabled for rendering and 10x better overall with both timeline GPU and rendering GPU enabled.

[Rich Kutnick] “BUT FIRST, bear in mind that my Dynamic RAM Preview is set to 8GB—might this affect these numbers PLUS might it also account for the initial GPU ON 22-25fps going up to a full 29.97 upon successive playbacks?”

Yes, I should have said that my Dynamic RAM Preview is set to 200MB which is the default. Setting it higher will affect the results and is why you saw successive passes get smoother because they were getting cached. The more you ran the tests the more primed the cache got. Your render results are being skewed by cached frames. You need to re-run the tests with Dynamic RAM set back to the default. It looks like you may only get 25fps playback even with GPU on. See if you still get 29.97 fps with Dynamic RAM set to 200MB.

[Rich Kutnick] “I do not know how these numbers can be true, though, for explain to me how the MainConcept OpenCL numbers using the GPU acceleration option are all but identical with or without the GPU being active!”

Did you exit Vegas Pro and restart it after turning the GPU off? When you change it, a dialog box pops up to tell you that it won’t take affect until after a restart. You must exit Vegas and restart it every time you change this GPU Acceleration parameter. It sounds like you either did not do this or the 8GB Dynamic RAM Preview cache is getting in the way of the true frame rates.

The way I tested was to turn GPU Acceleration OFF. Press APPLY then OK, Exit Vegas Pro, start Vegas Pro and render Sony AVC and MainConcept AVC. Then turn GPU Acceleration ON. Press APPLY then OK, Exit Vegas Pro, Start Vegas Pro and render Sony AVC and MainConcept AVC. Also don’t forget to reset the Dynamic RAM Preview to Default (200MB).

Overall, your results were a little better than mine which can be attributed to my 6 year old 8-core/8-thread 2.8Ghz CPU vs your 2 year old 4-core/8-thread 3.6Ghz CPU which is both newer technology and 25% faster clock speed.

What this test does seem to suggest is that the type of media you are using in your projects may be causing your problems because with generated media you seem to be getting a pretty good boost from the GPU.

BTW, Have you run Sony’s “Red Car” test project? My 8-core 2008 Mac Pro + Radeon HD 5870 beat the render times of all the servers and cards in the test (even the Radeon HD 6870!!!). I’m quite happy with my recent eBay purchase. 😉

~jr

http://www.johnrofrano.com
http://www.vasst.com

[Rich Kutnick]

I just ran the 7248_rendertestjr.veg file tests again, exactly as you specified I run them. My results are not too different from yours, so there is consistency. Here are my numbers:

Timeline GPU Acceleration OFF (Playback 0.5fps)
————————————————
Sony AVC–(Auto). . . . . . . 2:26 (No GPU Load)
Sony AVC (CPU Only). . . . . . . 2:25 (No GPU Load)
Sony AVC (GPU). . . . . . . . 2:25 (No GPU Load)
MainConcept AVC (CPU). . . . . . 3:09 (1-4% GPU Load)
MainConcept AVC (OpenCL) . . . . 2:20 (78% GPU Load spike at start, then went down to 0%)

Timeline GPU Acceleration ON (Playback 29.97fps)
————————————————
Sony AVC–(Auto). . . . . . . :30 (60% GPU Load)
Sony AVC (CPU Only). . . . . . . :24 (59% GPU Load)
Sony AVC (GPU). . . . . . . . :29 (62% GPU Load)
MainConcept AVC (CPU). . . . . . 1;06 (No GPU Load)
MainConcept AVC (OpenCL) . . . . :22 (87% GPU Load)

These tests seem to show that the best render times require an active GPU state. Sony AVC rendering works best with the CPU- Only template (GPU rendering for me, too, John, is slower) and the MainConcept AVC render requires the OpenCL mode. The former makes no sense while the latter makes absolute sense! No wonder everyone seems confused about which GPU states and render templates to use!!!

Now this begs two questions:

1) Why is the AVCHD footage from my Sony NX-5U freezing up SVP 12 when the GPU is active for the timeline?

2) Why NOT set the Dynamic RAM Preview for 8GB? Isn’t this added caching helping, or in addition to its advantages will it cause other headaches elsewhere in the editing/rendering processes?

Since you asked, here are my results using the “Red Car” Test (Full PressReleaseProject.veg file) with all of the same parameters as the 7248_rendertestjr.veg (I bumped up the Project Properties Full-resolution rendering quality from Good to Best and made it an interlaced project):

Timeline GPU Acceleration OFF (Playback–Region 1: 10 fps; Region 2: from 15fps to 29.97fps at end of region; Region 3: 29.97 fps except 15-22 fps during clock wipe; Region 4: 7-9 fps; Region 5: 2.2 fps; Region 6: 29.97 fps; Region 7: 7 fps average)
————————————————
Sony AVC–(Auto). . . . . . . 4:17 (No GPU Load)
Sony AVC (CPU Only). . . . . . . 4:19 (No GPU Load)
Sony AVC (GPU). . . . . . . . 4:10 (No GPU Load)
MainConcept AVC (CPU). . . . . . 6:50 (1-3% GPU Load)
MainConcept AVC (OpenCL) . . . . 3:39 (45% spike near the beginning, then 0% GPU Load balance of test)

Timeline GPU Acceleration ON (Playback–Region 1: 29.97 fps; Region 2: 29.97 fps; Region 3: 29.97 fps; Region 4: 29.97 fps; Region 5: 29.97 fps; Region 6: 29.97 fps; Region 29.97 fps)
————————————————
Sony AVC–(Auto). . . . . . . 1:38 (32-46% GPU Load)
Sony AVC (CPU Only). . . . . . .1:45 (33% GPU Load)
Sony AVC (GPU). . . . . . . . 1:38 (32-46% GPU Load)
MainConcept AVC (CPU). . . . . . 4:04 (No GPU Load)
MainConcept AVC (OpenCL) . . . . :56 (71-90% GPU Load)

The red car test is very interesting, too– MainConcept AVC (OpenCL) takes the most advantage of my graphics card and also renders hands down in the least amount of time!!

So with 10 tracks of constant motion/effects and my GPU enabled for the timeline, I AM getting rock solid playback—29.97fps without so much as a hiccup. Yet my Sony NX-5U footage causes grief. Any ideas why? Anyway, we all would appreciate your current interpretation of these four test scenarios, John!

Rich Kutnick
VIDEO IMPRESSIONS

[John Rofrano]

[Rich Kutnick] “1) Why is the AVCHD footage from my Sony NX-5U freezing up SVP 12 when the GPU is active for the timeline?”

Could it be your hard drive? The Sony NX-5U shoots AVCHD FX (24Mbps) 1920×1080/60i at the top end. My Sony AVCHD camera only shoots 16Mbps. Could the 24Mbps files (which must be larger and contain more data per frame) be slowing things down? This is just a guess but something to consider. Maybe try a test with shooting at 24Mbps and 16Mbps and see if there is a difference?

Also the amount of FX that you add to the footage will affect playback. You only need to add one FX that is not GPU accelerated or is single-threaded to bring playback and rendering to a crawl. Try disabling FX’s and see if performance improves. If it does, then try and narrow it down to the FX that is causing the problem.

[Rich Kutnick] “2) Why NOT set the Dynamic RAM Preview for 8GB? Isn’t this added caching helping, or in addition to its advantages will it cause other headaches elsewhere in the editing/rendering processes?”

It depends on how much RAM you have. That 8GB is dedicate to RAM preview. It is not used for anything else. It is only used for caching frames. If you have 16GB of memory then I guess giving 1/2 or it for preview caching is OK, but try different settings and see if it affects your render times. I was not saying that using 8GB was bad, only that it skewed the results of our tests because it cached frames and didn’t allow you to measure raw performance correctly. I usually have my Dynamic RAM set to 1GB as that’s all I usually need for previewing.

[Rich Kutnick] “So with 10 tracks of constant motion/effects and my GPU enabled for the timeline, I AM getting rock solid playback—29.97fps without so much as a hiccup. Yet my Sony NX-5U footage causes grief. Any ideas why? “

So other than my wild guess of the higher bit-rate and single-threaded/non-GPU enabled FX having something to do with it, I really don’t know why. I have no problems with my AVCHD footage.

The key thing is that your final render time for MainConcept AVC on the Red Car project was 0:56 which beats all of the cards that Sony tested including the $1700 Quadro 5000 with a Radeon HD 5870. That is quite impressive! (I’m really happy I got the Radeon HD 5870) You also saw it make good use (90%) of your GPU. So Sony GPU acceleration is definitely working and definitely worth it.

You should ask Sony why the footage from your NX-5U is not working well in Vegas Pro 12.0. This may be a bug. We recently found a bug where GoPro Hero footage that plays back at a solid 29.97 in Vegas Pro 11.0, stutters terribly at a few frames per second in Vegas Pro 12.0. I would definitely report your problem to Sony and see what they say.

~jr

http://www.johnrofrano.com
http://www.vasst.com

[Dave Haynie]

[John Rofrano] “BTW, Have you run Sony’s “Red Car” test project? My 8-core 2008 Mac Pro + Radeon HD 5870 beat the render times of all the servers and cards in the test (even the Radeon HD 6870!!!). I’m quite happy with my recent eBay purchase. 😉
“

Not a surprise, that result. AMD changed around their naming a little between the 5xxx and 6xxx series. Under the 5xxx conventions, the 59xx are all dual-GPU cards. Vegas will only use one GPU at a time, so these don’t buy you anything. And even today, they’re pretty expensive.

When the 6xxx series came out, they released the HD 6970 as the fastest single-chip card, and the HD 6990 as the only dual-chip card. It’s also the case that nearly all of the 6xxx series is based on the same VLIW5 architecture used in the 5xxx series, but die shrunk (smaller, cooler chip) and in some cases cost reduced (slower clocks, narrower buses). The HD69xx series use a newer VLIW4 architecture (yeah, that’s correct… the number isn’t the edition, but how many operations happen per core in one cycle), it’s not directly based on the HD5xxx chips. Still, the HD6970 and the HD5870 are directly comparable, in each case the fastest single-chip GPU of their generation (well, from AMD anyway).

-Dave

[Dave Haynie]

[Mike Kujbida] “A few folks on the Sony Vegas forum are really liking the new AMD R9 290 card. It’s not the cheapest (nor the most expensive) but the reviews so far have been very positive.
https://www.hyperactivemusic.com/vegaspro/vegaspro.html is a chart comparing the performance of this card to others from AMD and Nvidia. CPU specs are included making this a good chart.”

That’s a pretty interesting chart. It’s clear from the results than the R9 290 is not getting OpenCL support from Main Concept (as expected), but it’s outperforming my HD6970 on the part of the rendering pipeline internal to Vegas, where the GPU is going to help. This also proves my theory that the nVidia GTX580 might sometimes be faster than the HD6970…. I benchmarked against the GTX570 when I bought mine, both being in the lower $300 range back when Vegas 11 came out.

-Dave

[Dave Haynie]

[John Rofrano] “There is the Options | Preferences | Video | GPU Acceleration switch for the timeline. Vegas Pro uses CUDA for NVDIA cards and OpenCL for ATI cards so one switch toggles whatever you have. “

That’s actually a point of some confusion out there. Vegas itself is only OpenCL, not CUDA per se. nVidia these days says CUDA includes OpenGL, so they might technically be correct saying that a nVidia card is using CUDA under Vegas. But it’s only the OpenCL part of CUDA.

For some CODECs, like the often mentioned MainConcept AVC CODEC, you are selecting between CUDA and OpenCL acceleration, and that only on very specific chips. It should be possible, for example, to benchmark a nVidia card NOCPU, CUDA, and OpenCL. But MainConcept will just ignore the acceleration if your card’s chip is not specifically on its list of approved devices. So no OpenCL for nVidia in the CODECs. But only OpenCL for nVidia in Vegas proper.

-Dave

[Rich Kutnick]

Not wanting to have a knee-jerk reaction, I viewed an edited veg timeline for a Bat Mitzvah I shot two weeks prior to the one in question. Except for the expected stutters during transitions, my AVCHD footage was a rock-solid 29.97fps! I don’t think that my Nvidia GTX 560 ever gave me that! Anyway, the AVCHD files for both events are on Drive W: (External USB 3.0) and the veg files for both events are on Drive Z: (External Esata). My guess is that there is some obscure reason the second Bat Mitzvah Ceremony veg file is experiencing these freeze-ups. So I believe that everything is working properly now, save for this anomaly which I will try to track down. Thanks for all of your help and insight, John. It’s also good to know that my choice of a different graphics card worked out in my favor (due to you and several other colleagues here who recommended the ATI 5870).

Rich Kutnick
VIDEO IMPRESSIONS

[Rich Kutnick]

I found the problem–thank goodness it’s not my AVCHD footage, either!! I have placed a good number of NewBlue VE Flash Remover-OpenFX effects into this timeline, and evidently when my graphics card is active this causes the timeline to keep freezing up. Well, I replaced all of these effects with the standard NewBlue VE Flash Remover plug-in (non-Open-FX version) and this seemed to do the trick–when I activate my GPU on the timeline it does not freeze anymore! So if anyone from New Blue is reading this…

Rich Kutnick
VIDEO IMPRESSIONS

[Edward Troxel]

Rich, please send full details to support at newblueinc.com

Edward Troxel