• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Need a video card for encoding

Valicious

n00b
Joined
Oct 6, 2009
Messages
15
I rip and encode a lot of VHS, a LOT. I don't have that big of budget, but I finally need to replace my failing card I've had for years.

Few questions for you guys.
1. AMD or nvidia? The benchmarks I've been reading say AMD is top of the pops when it comes to OpenCL. Can anyone confirm or deny this?
2. The card I've been looking at is the R9 280x, but I'd like to spend a little less than that if I can get away with it.
http://developer.amd.com/community/blog/2014/02/19/introducing-video-coding-engine-vce/
This sounds really good, and like something I would benefit from. The 280x can only use v1.0 though. Are any of the cards compatible with v2.0 worth it?
3. Any suggestions for cards? Like I mentioned, the primary use would be ripping, editing, and encoding VHS concert tapes. Quality is paramount as I only have the opportunity to do this once/tape, but so is speed as I'm trying to regularly get through hundreds of tapes.
 
You're better off inquiring about this in a Audio/Video oriented group instead as you need to first establish the workload, or have you done that already (eg. what software)?

Hardware accelerated encoding (VCE, NVENC, Quicksync), OpenCL, CUDA, and x86 (CPU) all have differing characteristics going for them.

For example VCE would not have a relation to OpenGL performance or an ecoder which uses OpenGL.
 
I'd advise getting a cheap Maxwell, likely the 950/960 and use NVENC to encode. The speed is great, and from what I've seen with my own eyes and read so far, has higher quality than CPU encoding in some cases.

Maxwell 2.0 supporting HVEC fully doesn't hurt, either.
 
AFAIK all major platforms (AMD, NVIDIA, Intel) all have hardware-accelerated encoding, if you already have a modern Intel CPU give QuickSync a try before jumping to a dGPU just for encoding.
 
What CPU are you using? Both AMD and intel can do encoding with integrated graphics and its quite fast.
 
I'm using H264 and an i5-2500k. The suggestion of Haswell intrigues me, but I just bought the i5. I'm encoding VHS of concerts, and a big reason is archival though so quality is paramount.
 
I'm assuming you're using 'lossless' MPEG-4 as opposed to actual lossless (like Lagarith or HuffYUV)?
 
For maximum quality you want x264 and a good cpu.

This.. i've used various Cuda and QuickSync encoders and the quality sucks compared to the purely cpu based ones like x264. (and by extension my favorite encoder, handbrake given it uses ffmpeg which uses x264)
 
also one thing I will mention since you said this is for maximum quality archival purposes from VHS I'm guessing home videos of the sorts, you should use the switches; --qcomp-1 and --mbtree-0. This will significantly minimize reductions in quality and blockiness in motion and dark areas. This will however increase file size.

Or if using handbrake; mbtree=0, qcomp=1

Are you are fluent on the rest of the settings?
 
Which is why the OP needs to use lossless codecs for archival purposes and re-encode for playback as needed. I get that it's massively inconvenient (it's my daily job, so I get the inconvenience) but if you want to preserve the quality you're doing yourself a disservice to encode to MP4 (even at 'lossless' settings) as you can't recover what you're throwing away later down the chain.

Lossless is in single quotes because even though it's called lossless you can see a color-shift doing a side-by-side comparison between encode and source,
 
Which is why the OP needs to use lossless codecs for archival purposes and re-encode for playback as needed. I get that it's massively inconvenient (it's my daily job, so I get the inconvenience) but if you want to preserve the quality you're doing yourself a disservice to encode to MP4 (even at 'lossless' settings) as you can't recover what you're throwing away later down the chain.

Lossless is in single quotes because even though it's called lossless you can see a color-shift doing a side-by-side comparison between encode and source,

A couple of years ago I captured a bunch of Hi-8 cassetes and encoded with h264 using quicksync. Even with lossy compression the quality was great.

I don't recall the settings I used but since you only need 480p at most then a bit rate of about 1500kbps should be plenty.
 
The videos I produce for web distribution are lower than 1500 Kbps and they visually look great and they sound great (HandBrake is a great program) but you are throwing away a lot of information which limits how much you can do with them down the line without further visual degradation. (One of the files I'm looking at right now for web distribution is 184 Kbps for video stream and 111 Kbps for audio; this is great for web distribution but terrible for archival purposes, especially if you intend on doing anything to the files later on).
 
Sure if you can name a good lossless format that doesn't take up a ridiculously ludicrous amount of space. If a 290x is currently out of OPs price range, then multiple TB of disks are as well unless he already has them.
 
also one thing I will mention since you said this is for maximum quality archival purposes from VHS I'm guessing home videos of the sorts, you should use the switches; --qcomp-1 and --mbtree-0. This will significantly minimize reductions in quality and blockiness in motion and dark areas. This will however increase file size.

Or if using handbrake; mbtree=0, qcomp=1

Are you are fluent on the rest of the settings?

They are actually VHS releases of Japanese visual kei bands. As most of the releases are from the 90s/maybe early 2000s, getting the best quality and cleaning it up via editing is a task in itself. Some of the releases were limited to like 50 copies and sold at a single concert in 1992 or something ridiculous like that, so I'm not kidding when I say I'll never see some of this stuff again.
I was looking through some of my lists from earlier this year, and the number of cassettes and VHS I rip and encode is in the thousands not hundreds. (What am I doing with my life?)
I need to get SOMETHING as my current card is causing daily black screen crashes with I fear is degrading my hard drives. Since encoding is mostly cpu-based (I do use a few plugins that use the gpu), do I really have to agonize over what card to get TOO much?

Offers on the table:
Powercolor (ick) 280x for $150 (warranty hasn't been registered yet)
EVGA Geforce GTX 680 SC Signature 2 $160
MSI gtx 760 2gb $110
Asus Radeon HD 7850 DirectCU 2 $110
or should I wait and keep looking? I'd rather not buy a used card from a company that I'm not sure if they'd RMA it (that's why I love XFX, double-lifetime warranty)

AMD beats the pants off nvidia for this stuff right?
 
If you're doing mostly x264/x265 encoding then there's really not much point for a powerful graphics card, pretty much anything will do.

If you're looking into maybe switching to hardware-based encoding, then you should try out NVENC. I've seen some claims of it being equal to or better than x264 medium preset, while being much, much faster. Don't ever use CUDA encoding. QuickSync is also pretty meh quality-wise, IMO.

What kind of plugins do you use?
 
If you're doing mostly x264/x265 encoding then there's really not much point for a powerful graphics card, pretty much anything will do.

If you're looking into maybe switching to hardware-based encoding, then you should try out NVENC. I've seen some claims of it being equal to or better than x264 medium preset, while being much, much faster. Don't ever use CUDA encoding. QuickSync is also pretty meh quality-wise, IMO.

What kind of plugins do you use?

Generally a deinterlacer, noise/chromanoise removal, and cropping/resizing.
 
I'd advise getting a cheap Maxwell, likely the 950/960 and use NVENC to encode. The speed is great, and from what I've seen with my own eyes and read so far, has higher quality than CPU encoding in some cases.

Maxwell 2.0 supporting HVEC fully doesn't hurt, either.

This isn't entirely correct..For example, Maxwell based cards cannot (hardware) decode HVEC Main 10, which is the format NetFlix (and rumors point to others) settled on to stream their 4K content..

As it stands, IIRC only some Intel Skylake SKUs will have full HVEC hardware decoding, including Main 10.
 
This isn't entirely correct..For example, Maxwell based cards cannot (hardware) decode HVEC Main 10, which is the format NetFlix (and rumors point to others) settled on to stream their 4K content..

As it stands, IIRC only some Intel Skylake SKUs will have full HVEC hardware decoding, including Main 10.

Oops, brain fart.

I meant full encoding support since that's the topic at hand, but yeah they only support hybrid decoding.
 
Skylake only has full hardware support for HVEC Main. Main 10 support is hybrid (specifically listed as GPU assisted on Intel's slide deck - http://images.anandtech.com/doci/9562/med-cap-07.png).

Not all Maxwell 2.0 GPUs use the same VP block. GM200/GM204 use the same VP block with only hybrid decode support.

GM206 uses a new VP block (shared with Tegra X1). It is listed as having full HVEC Main and Main 10 hardware decode. The only test I know of off hand that documents this though is Anandtech's test of the Sheild TV with Netflix (HDCP 2.2, HDMI 2.0, and HVEC Main 10) - http://www.anandtech.com/show/9289/the-nvidia-shield-android-tv-review/7
 
Back
Top