new computer build for virb edit / virb 360

for my experience, the better GPU you have the faster it will render. There are several utilities out there that allow us to see GPU % usage just like a CPU and when exporting videos the Video Engine is always at 100% on the gpu.
( I use this one: https://www.techpowerup.com/gpuz/ )

Since version 5 of VirbEdit, if you have Windows 10 and all the proper drivers, it will use your GPU and little CPU, throwing in a massive CPU is wasting money, better put it on the GPU.
RAM as well, it will not use lots of RAM, mine never uses more than 1Gb of RAM at maximum.

disk wise, as long as your HDD can feed the GPU crunching the videos, an SSD may be overkill, but that's your wallet.

SLi I do not know, never tried. maybe someone can elaborate on this.

And yes, I do *alot* of videos in VirbEdit:
https://www.youtube.com/channel/UCkXHx2sVdEv8YBGIGqscHnQ

RAM as well, it will not use lots of RAM, mine never uses more than 1Gb of RAM at maximum.

Using HD or FullHD, you are right. 4k or even 360 (5,7k) will need much more RAM!!!
As you can see here 16GB are recommended.

All, if you have a 1080ti, please share your experience with Redwolf. How fast does a 4k video export?

@Redwolf,
VirbEdit uses a variety of technologies to decode and encode videos. Since the 5.1 release, VirbEdit takes full advantage of Nvidia's video encoding and decoding APIs, including leveraging Nvidia cards' video encoding ASIC (specialized circuitry on the card that is designed to do video encoding).

I don't have access to a 1080ti so I'm afraid I can't give you real world performance numbers, but I can throw a big pile of information at you:
There are a couple operations that VirbEdit does that benefit from a more powerful GPU: encoding and color space conversion (YUV->RGB and back). Color space conversion speed depends largely on the number of cores the GPU has. A 1080ti has about 3x as many cores as a 1060, so I would expect a fairly significant speed improvement there. Video encoding is done by specialized hardware on the card. A quick search of google found this performance comparison of the various 10 series Nvidia cards: https://devtalk.nvidia.com/default/t...ed-comparison/

The basic finding was that the 1080 (not ti as far as I can tell) can encode about twice as fast as the 1060.
With a card that powerful, you may find that decoding speed becomes the performance bottleneck, in which case you would get only marginal performance gains over less expensive 10 series cards.
I have a 1070 in my home PC, and it is marginally faster than the 1060.
From past experience, real world performance is less impressive than the numbers make it seem. So I wouldn't expect to get twice as much performance out of a 1080ti as a 1060, but I would expect it to be better.


In terms of RAM, VirbEdit is (for now) a 32 bit application, so it is not capable of using more than 4GB of RAM. I've gone to quite a bit of work to make sure that the application works within that limitation. Working with 5.7k videos is pushing up against that limit, so we may eventually take the 64 bit route, but even if we do, the app will not take up anywhere near 16GB. 32GB is probably overkill.

EDIT: VirbEdit doesn't currently make any attempt to leverage SLI, so that will not currently help at all.

Thank you for the feedback. It will be Windows 10. I also use the computer for photo processing which is all CPU with the software I use. It will also copy a bunch of files often, the m.2 ssd drives are amazing and highly recommended, I have one in another machine.

I ordered both the 1060 and 1080ti cards for testing. I'll let you guys know what I find out.

Final hardware:
AMD ryzen 1800x
AX370 "Gaming" K3 mobo
16gb 3000mhz ram
512GB M.2 SSD (2GB/sec read speed)
WD Black 2TB platter HD
GTX 1060 SC ($289)
Aorus 1080ti xtreme ($750 *cry*)

Purdington a question for you: Is 5.7k video stitching planned for VirbEdit any time soon? how about batch exporting? thanks

@RuiTerrinha I believe the capability to do 5.7k stitching was part of our marketing of the Virb 360 camera, and we are eager to fulfill that promise. As always, I cannot provide exact release dates, sorry :( We don't currently have any plans to add batch exporting as a feature in VirbEdit, though I do think it is a good idea.

Here are my results...

Test rig:
Windows 10 64bit Fresh Install
AMD ryzen 1800x
AX370 "Gaming" K3 mobo
16gb 3000mhz ram
512GB M.2 SSD (2GB/sec read speed)
WD Black 2TB platter HD
Latest drivers at the time (Early Sep 2017). I tried with regular Nvidia G-Force drivers and also with and without CUDA toolkit. Same results.

GTX 1060 SC ($289)
GTX 1070 SC ($465)
Aorus 1080ti xtreme ($750)


Software and Video:
Virb Edit 5.1.3
360 source video was 21:12 (21 minutes, 12 seconds) long
Some gmetrix gauges like speed, track map, g-forces, laptimes
Encoding engine: FFMPEG
Decoding engine: DXVA
Exported as 2.7k at Medium Quality


Results:
8:55 - GTX 1080Ti 11GB Aorus Xtreme
9:17 - GTX 1070 8GB SC
9:33 - GTX 1060 6GB SC

I was expecting a much bigger difference in render times between a 1080Ti and 1060. It seems like the CUDA cores/Video Engine are not being fully utilized, or I did not have all the right supporting software installed/configured correctly. If anyone has any tips on that front, please share.

Also, I wasn't consistent about monitoring the max CPU, GPU and Video engine, but here's what I have using CPUID HWMonitor app

Max Percentages
1080Ti: CPU: 30, GPU: 47, Video Engine: 69
1070: no data
1060: Max CPU: ??, GPU: 26, Video Engine: 55



Other tests:
1060 @ 1080p/med: 8:16
1060 @ 1080p/med (Enc: Microsoft Foundation, Dec: DXVA): 8:22
1060 @ 2.7k/med (Enc and Dec: FFMPEG): 12:14