(Kind of) New Vegas Pro Benchmark

TheRhino wrote on 11/11/2013, 7:07 PM

Bruce---

I am not saying that V12 uses less cores - all 6 cores are used - but they are not running at 100% during a cpu-only render...

I have the SAME project loaded in V10 & V12 (gpu off). V10 finishes the SAME VEG render 40% faster than V12 set to cpu-only mode. During renders, CPU usage for V10 is 100% from start to finish whereas V12 shows only 60%-65%. Because the CPU is not maxed-out the render takes longer in V12 - about 40% longer, the same as the difference between V12's 60% & V10's 100% cpu usage. Make sense? Again - same project, same hardware...

Now if I turn-ON the GPU, V12 renders faster than V10 even though the CPU is still not pushed to 100%. This is because the GPU is handling much of the processing now.

Until I upgraded my GPU to the GTX 570, V12 renders were 40% slower than V10e renders on my system so I did not use V12 until I added the GTX 570.

OldSmoke wrote on 11/11/2013, 7:24 PM

-TheRhino

I apologize for the downgrade it was not intentional. So we would have to compare it with the latest i7-4960X?
I am seriously thinking of a dual Xeon setup for my next upgrade.

I do multitasking but never when it comes to rendering. I don't trust Vegas so far not to screw it up and my projects are usually 30-60min; not something I would like to redo or check for missing frames. I use Vegasaur to batch render to multiple formats but still only one at a time. Maybe a smaller 5min project I wouldn't care so much.
My two 512GB RAID 0 (total 1TB) SSDs deliver 1000+ MBps according to Crystal Mark drive test. Very useful for multi-camera editing; I render to a dedicated SSD too but I use spinning disks RAID 10 for archive. Footage is usually HDV but now I have all AVCHD 1080-60p.

Edit:
Since you have both V10&V12 on the system it would be a good test to see if a project without any FX, just a clip on the timeline, output format same as timeline but different codec, would show equal render performance differences. If you get same result in both versions, then the now GPU accelerated FX and others are no longer CPU only optimized. If there is still the same difference then the whole application is, as you mentioned, no longer CPU only optimized but still can do the job.

skeeter123 wrote on 11/11/2013, 9:19 PM

Ordered a R9 290 AMD space heater/ear-splitter...should be here Wednesday....hopefully some results this weekend...

..

BruceUSA wrote on 11/11/2013, 9:26 PM

Nice..... I want one :) 280X is pretty d a m n good to.

skeeter123 wrote on 11/11/2013, 10:15 PM

Nice, yeah. But if it's TOO loud/hot, it goes back. Then I'll wait for some more sane cooling....Or is it time to augment the the CPU liquid cooling with GPU liquid??

Hulk wrote on 11/11/2013, 10:18 PM

skeeter,

I have a feeling that since Vegas won't load that GPU like a game does that you might be able to dial the fan back a bit to a more comfortable noise level. Should be interesting and blazing fast. Can't wait to see the stats.

skeeter123 wrote on 11/11/2013, 10:21 PM

Great minds think alike, Hulk....we'll see.....or hear as the case may be...

..

OldSmoke wrote on 11/12/2013, 2:18 PM

Well… that should get your system finally in the right spot, on top of the list.

skeeter123 wrote on 11/12/2013, 2:24 PM

OldSmoke:

They were done in 8 bit mode....methinks....I will rerun them tomorrow to be sure.

...

OldSmoke wrote on 11/12/2013, 4:29 PM

Hmmm… Have you tried changing your driver to 296.10? It is a good 25% faster; at least on my system.

skeeter123 wrote on 11/13/2013, 10:04 PM

@OldSmoke.

I'm Win 8.1...I'd love to try 296.10 if it would work!

I installed the R9 290 today. Some mixed results and I need to put some science behind the results before posting. But it will be forthcoming.

The "(Kind of) New Benchmark" rendered to XDCAM EX in 30 secs vs. 55 secs for the GTX580.

However, some of my personal benchmarks are a bit scattered.....And these are the ones that matter to me....mostly high bitrate (225Mbps+) GoPro via Mercalli Prodrenalin sources...

The "18 minute test" took 4x longer to render to Main Concept 1920 x 1080 60i. AVC/AAC (OpenCL) than the 580 (CUDA). This can partially be explained by the not-so-swift implementation of OpenCL...

I plan to re-install the 580 and carefully run the tests this weekend...

EDIT: On the plus side for sure, it isn't nearly as loud as I expected it to be. Just a tad louder than the OC'd 580. It does run hot....95C at 100% load...but it is designed for the heat...I hope...

Stay Tuna'd..

ddm wrote on 11/14/2013, 12:33 AM

I have a dual boot Win 7 and Win 8.1 machine and I just ran all these benchmarks with some interesting results. I am using the latest nvidia driver 331.65 on my Windows 8.1 boot, and 296.10 on my W7 boot with my 570 graphics card. I was quite surprised to find my render times and preview times almost exactly the same on both setups, and that's a first, as the last time I benchmarked the times favored the old driver on W7 by a good margin. Overall, I was quite pleased with the preview times as they are dramatically different than CPU only. On both setups I can get perfect playback (starting at 25 seconds in the timeline). Not sure if it's because of the type of file that the benchmark test uses or what but it is nice to finally get some payoff from that video card. The file rendering is equally impressive 1:10 vs 4:10 or more. And no crashes either. I did find that lowering preview ram from my normal 4096 to 16 did improve render and preview times a wee bit.

skeeter123 wrote on 11/14/2013, 1:23 AM

ddm,

Excellent!

Are you still running the i5 rig?

I've been sooooo tempted to go back to Win 7 on my new build...or dual boot it...

Any other comparisons would be great. Especially rendering Sony AVC-mp4 to 60i or 60p....and at 200MB preview....that seems to be a sweet spot for me...

thanks!

..

skeeter123 wrote on 11/14/2013, 3:52 AM

Ok. Some encouraging news...

On the first part of my "long-benchmark", the R9 290 bested the 580 by 77%. However, preview is not as good/fast/smooth through the NB titles.... Need more testing here..

My full "long-benchmark" is running now and I'll let it run overnight... The time to beat is 01:30:00 (HMS) for the 41 minute, 32 bit project ... (200Mbps+, 60p 1920x1080 source) rendered Sony AVC/mp4, 26Mbps, 1920x1080 60p.

Nerd Heaven here...

ddm wrote on 11/14/2013, 6:14 PM

>>>>>Especially rendering Sony AVC-mp4 to 60i or 60

rendering to Sony AVC from any camera format in particular?

TheRhino wrote on 11/16/2013, 12:47 PM

Bump... skeeter123 - any new results to share regarding the new AMD GPU?

skeeter123 wrote on 11/18/2013, 6:48 PM

...sorry for the delay...been learning lots about overclocking the new AMD boards...

..stay tuned..

skeeter123 wrote on 11/18/2013, 8:56 PM

My camera's are Sony RX100 MKII and GoPro HeroBlack HD, both in 1920 X 1080 60p modes...

I then stabilize them in Mercalli Prodrenalin 1.0 which is where the uber high bit rates source come in...

The R9 290 is a keeper for me, since I render to Sony AVC/mp4 and it knocks about 75+% off my render times....As Hulk and others have pointed out the OpenCL implementation is iffy in the Main Concept CODECs. (Hulk, correct me here if needed). My "old" GTX580 beats the R9 290 with MC CODECs....but I rarely use them..

The 290 plays really nice with New Blue titles as well. Really nice...so far...

If you'd like to run my "Short Benchmark" PM me and I'll send you a link to my Google Drive. It's 2.34 GB...

BruceUSA wrote on 11/19/2013, 12:29 AM

Skeeter123,

The R9 290 gpu accelerate everything else? except it won't accelerate MC avc , is that correct?

Stringer wrote on 11/19/2013, 9:27 AM

At the referenced WEB site it says:

" Running the Test

Forgive me for being a bit dense, but I can't seem to locate the" instructions " other than the description of how the results were obtained in the Sony Demo package..

Can someone please help ?

Also where do you see " fps " during playback ? ( Vegas ver 12 )

skeeter123 wrote on 11/19/2013, 11:17 AM

@Bruce,

It does to a degree. My "Press Release" Benchmark was 89 seconds with GPU on and 129 off for MC. The XDCAM EX numbers were 30 on, 133 off.....

The numbers for my GTX580 were 36 on, 125 off for the MC CODEC.

So, it seems the MC OpenCL implementation is not as robust....YMMV...

@Stringer: fps pops up under the preview window after "Display" when previewing...
..

Stringer wrote on 11/19/2013, 4:23 PM

Thanks skeeter.. I see it now.. It wasn't apparent to me because my FPS was locked solid at 29.97..

I'm still looking for the instructions Bruce was referring to..

Hulk wrote on 11/19/2013, 4:40 PM

Stringer,

My bad on that. Originally I had test instructions. Then I decided it would be easier just to include an excel download asking for the testing information. If you look closely at the header for each column you'll understand. Just use the default templates as indicated and everybody will be on a level playing field.
If you have questions just e-mail me. My e-mail is on the benchmark site.

- Mark

Stringer wrote on 11/19/2013, 4:48 PM

That's what I need to know.. Thanks..

I'm just a hobbyist, but your benchmark tool is useful for letting me see how various hardware and settings let me make the most of what I have..

I had never given much thought to the ' fps ' during preview, thinking " it is what it is ",
but I just learned with video acceleration enabled, I'm getting a solid 29.97 ( The Sony demo ), while with it turned off it stutters considerably, and that's with a 3930k ..

I'll try to add to your spreadsheet..