[Dave Haynie]

Native DVD is 720×480, so yeah, you need to “stretch” your 640×480 a bit. I did SD resolution editing for years on several PCs much less powerful than yours. One render (high quality conversion of a 2 hour video from NTSC to PAL) too 8.5 days on my 1GHz AMD machine.

Higher quality MPEG-2 will take longer than basic quality… sometimes. However, if you’re making a relatively short video, you can render at constant bit rate (CBR), 8Mb/s or so, and get top quality without some of the computational overhead. If you’re filling the DVD, you’ll want to spend the time on VBR rendering to optimize quality based on content. The “custom” button on the “Render As…” dialog is where you change these, if you haven’t found it yet.

-Dave

[Dave Haynie]

I agree with John… even on my six core AMD, NeatVideo takes a really long time. If I need it, I render to an intermediate format (Cineform, back when that worked with Vegas, now probably MXF/MPEG-2), then move on from there. It’s not the kind of thing you’d like to have to do twice in a project.

-Dave

[Dave Haynie]

Yeah, they claim to be supporting OpenCL. The current Vegas supports CUDA, which is a nVidia-only thing.

ATi/AMD have their own branded interface, Stream. Stream actually covers a bunch of interfaces. It used to include a low-level GPGPU interface called “close to metal”, but that’s been deprecated. OpenCL is now AMD’s go-to interface for GPGPU; Stream also included other GPU-related bits, like Windows-standard interfaces, things like DXVA and DirectCompute support (also managed by nVidia, but not necessarily listed as part of CUDA). DirectCompute is Microsoft’s “native” answer to OpenCL. No one uses it 🙂

OpenCL is, oddly enough given the name, an open spec, and is now supported by nVidia as well. OpenCL programs should be able to use nVidia, AMD, eventually even Intel GPUs as general purpose computing engines (if Intel ever gets around to building that kind of GPU). Since OpenCL was released after CUDA (and maybe because it originated at Apple, but is now independently managed by the Khronos Group), support has been lagging.

It may also reflect the fact that AMD’s been on a roll. They’ve increased GPU sales 15.4% (24.8% total), while nVidia as dropped 28.4% (20% total). And Intel’s hanging out at just over 50%… if they do deliver a GPGPU capable GPU at some point, every cheap PC will have it, and it won’t run CUDA either.

It may also help that nVidia is falling behind AMD in GPU sales these days

-Dave

[Dave Haynie]

Far as I know, Vegas will only output decoded audio. S/PDIF (IEC60958) is not fast enough for 5.1 channel uncompressed audio. It was originally engineered for stereo only, then the IEC61937 standard defined how compressed multichannel audio, AC-3 or DTS usually, could be sent via the same protocol. As far as I know, you can’t get Vegas to compress to AC-3 on the fly.

If I look at the setup in my system, the S/PDIF output (which I’m not using…) indicates compressed types it’ll pass on: AC-3 and DTS. Clearly, DVD/Blu-ray software is capable of sending pre-encoded bitstreams to a decoding amplifier this way, so there’s a Windows standard of some kind for this interface. But it’s not going to do the AC-3 encoding for me.

It’s technically possible someone could make a “fake” audio device, which shows up as a 6 channel uncompressed output, but does the conversion to AC-3 and pipes to S/PDIF on the fly. Modern PCs are plenty fast enough to do this, but I’ve not heard of it actually being done.

-Dave

[Dave Haynie]

Ha! I used to work for Scala! And I did actually use it to author a 2 hour film… back in the days of tape. Scala plus a Genlock did the titles, controlled the camcorder and the recording deck, controlled a device called a TBCplus, which did the usual TBC stuff plus color corrections, video effects, etc.

For pure video authoring, yeah, I would definitely use Vegas over Scala. They ship Scala with many tools these days, and I’m sure you could do video authoring with it too. But the point of Scala is aggregating assets for digital signage presentations. You can build all the video you like in Vegas, then use it in Scala to create complete presentations.

You have to first understand how you want to use your digital signs. Scala basically treats it like television: you may want to build different segments/shows, present them at specific times, etc. And Scala lets you do this for one display or 10,000.

Same idea as DVD Architect: I wouldn’t use DVDA for editing or rendering video (technically you could, a little bit). But I also wouldn’t use Vegas to assemble a finished DVD (you could, as long as its just one video). The right tool for the right job.

-Dave

[Dave Haynie]

Sorry, that’s not proof. Sure, they may have multiple reasons for implementing download caps. Over half of Comcast’s customers, for example, use less than 100GB per month, all told… streaming a regular evening’s worth of TV at Blu-ray quality make it 100GB+ per day. They could not possible supply that to every customer, right now.

And in particular, streaming across the internet is far more limited than from the head end. And of course, it’s only IPTV using head end to client data bandwidth… cable and FiOS systems use digital broadcast for their TV channel offerings.

Any internet user knows there’s a hard limit… just log on after dinner, and things are much slower than you’ll find at 3AM, on practically any system. There is a limit on bandwidth for a cable system… using 256QAM (state of the art modulation), you’re going to see 30Mb/s-60Mb/s per analog channel slot (6MHz slots, as low as 1GHz of total analog bandwidth, probably a bit under 10Gb/s of total digital bandwidth). Or about 250 people watching different full bandwidth Blu-ray downloads… IF you shut down all of the analog TV and just ran internet on every channel. Each analog slot normally delivers 3-6 HDTV channels (if they’re using AVC encoding), or a while mess of SD slots (most cable and satellite systems use MPEG-2 TS, so they’re probably limited to 32 subchannels on any given analog slot).

Now, of course, you don’t have all 14.7 million Comcast customers sharing the same 166-or-so physical channels. In fact, in a modern cable system, the head end will run optical links to the neighbood distribution node, which typically serves up to around 500-2000 people, depending on the setup. But of course, most of the cable channels are dedicated to television. And the 6MHz slots dedicated to internet are unidirectional… you’re downstream or upstream. So the actual internet bandwidth to your neighborhood is way less than most people would figure. You may only have a few channels devoted to downstream (and many systems are on the lower side of my 30-60Mb/s estimate… these things are not being upgraded to the latest tech every year)… you need many more for upstream to manage the same number of simultaneous internet users, even though the upstream links are much, much slower.

Of course, you’ll have much more bandwidth on FiOS. In fact, they’re already frequency-division multiplexing most fiber: one laser color for TV, one for internet, and one for voice, usually. But it won’t matter once you leave the head end and go out to the internet.

And that’s just to the home. Most ISPs have a tiny, tiny fraction of their aggregate bandwidth out to the internet. The infrastructure: you to the head end, they pay for that once in a generation, you pay every month. Their internet backbone is something they have to pay for every money. I’m not claiming they’re being good guys by limiting this, but the truth is, they don’t have nearly enough for everyone to be streaming full HD from local servers, much less across the internet. No one does. The primary reason they’re capping people probably is to protect this imbalance, which saves them the most expensive kind of infrastructure.

It was ten years ago, but one of my old companies, Metabox AG, ran an ISP out of our main office. We had thousands of customers, and two OC3 lines to serve them.. that’s a total of 155Mb/s. This was multipied, like all of the big ISPs do, by local caching of popular web content at the head end (and, in our case, on the customers’ STBs themselves… we could push 900MB or more data to each STB per day over analog TV).

-Dave

[Dave Haynie]

[Jim Greene] “Yes, agreed. I was one of the first people to program on Windows NT and Windows 95, which introduced multi-threading. A lot of race conditions to worry about.”

The big problem I found with early(ish) Windows NT (I was working on some software for Windows NT 4.0) was that the Windows API itself had a nasty habit of serializing things that would normally be done in parallel on other OSs. Signals between independent threads turned into message though the single Windows message port, stuff like that (don’t recall any specific examples, this was back in like ’94-’95).

Building a proper state machine, locking resources with semaphores, etc. isn’t all that difficult if you’ve had the proper introduction to multithreaded programming. Bugs certainly can be squirrelly to track down, though.. particularly when you’re on a multiprocessor system, thus real concurrency. Maybe it helped that I started out primarily in hardware.. I’m used to everything happening all at once.

[Jim Greene] “I certainly feel that their focus on 3D has allowed way too many bugs to appear in a very solidly built code base before version 10. I have no interest in 3D and personally think it’s a passing trend, like the Laserdisc and Blu-ray (I think we are all going to digital cloud storage).”

I agree on 3D, at least at present. There’s too much poor quality material right now to justify the format. I disagree on Blu-ray. Yes, online is going to happen eventually, but there’s not close the per-user bandwidth to support Blu-ray quality viewing on a regular basis. Most of the online video is of dramatically lower quality… maybe ok for the average consumer with a cheap 720p television. Apple’s HD is about 1.5Mb/s 720p, Netflix goes all the up to about 3.6Mb/s last I checked (though they’re streaming in VC-1, mainly, they need a bit more bandwidth than you would for equivalent H.264).

Sure, the “right now” has largely triumphed over quality; people buy music online at reduced quality, even for the same price as CD (ok, a few of us buy online at places like HDTracks, but its very limited material). More people listening though cheesy earbuds, fewer though good speakers, has allowed this. But I think TVs are getting better.. the price of LCD panels has plummeted, as production facilities crank them out in garage-door-sized panels. Of course, half the people with 1080p displays are probably sitting too far away to tell the difference between 1080p and 720p… but there has to be some demand for a quality image.

Blu-ray is already pretty much at critical mass. It’s expected to eclipse DVD in player sales this year — it’s basically moving from the high-end feature in DVD players to the midrange-on-up feature of DVD players… same price point that upscaling and 480p once were.

I have no interest in cloud storage of my own personal stuff… I don’t support I’d mind being able to access my own network via “the cloud” (my SlingBox allows that, but satellite internet, not so much). But cloud rentals, the cloud as a replacement for cable/satellite, that’s something you can do today, if you’re not too concerned about quality. It will get better, too.

The problem today, too, is that, were everyone to do this overnight, the ISPs would be saturated. Some are already implementing monthly download caps, and those are targeted directly at feature video streaming. If you could get Blu-ray quality streaming, that’s roughly 50GB per 3 hours or so… with the download caps coming in aruond 250-340GB/month, you won’t get a great deal of viewing done in full quality HD.

Me, I’m waiting for the 4K upgrade to Blu-ray… it’ll probably come with 3D anyway. I have the room for a larger TV (mine is only 71″), then I’d need a new disc player (I think you’d need about 200GB… BD-XL is halfway there today), a couple new cameras, much faster computers for editing… ok, maybe I’m happy with plain old Blu-ray for the next 5-10 years 🙂

-Dave

[Dave Haynie]

[Rodney Ballard] “Have any idea why there is no audio when put into DVD Arch 3? I dragged the MPG file from the explorer tab, into the main working area. When played after rendering in Studio6 it is with audio. The template is NTSC DVD. “

It’s kind of traditional for DVD applications to work on low-level assets. I’m not exactly sure why, only, that things like AC-3 or DTS in an MPEG-2 wrapper aren’t standard (DVD extended this, but it wasn’t part of the MPEG-2 spec). So that would normally be a separate audio and video file, with the video in MPEG-2 elemental stream format. And for many DVD authoring systems, it is. Vegas produces a more common program stream, but not to add audio, but instead to allow the inclusion of markers and other metadata.

So normally, at least in Vegas Pro, you’d render to the MPEG-2 “DVD NTSC” or “DVD NTSC Widescreen” template. These produce the MPEG-2 files without audio. Then you find the AC-3 (or WAV, if you like) option, and render the audio. If they’re in the same directory, DVD Architect should pick up the audio when you select the video.

-Dave

[Dave Haynie]

In theory, it is now a MOOT point. You should, of course, be sure that your particular ATi/AMD is supported with OpenCL in the latest versions of Catalyst. But this is a good sign that Sony intendeds to be GPU-agnostic going forward; OpenCL can support nVidia, AMD, maybe other GPUs.

-Dave

[Dave Haynie]

[Stephen Mann] “I remember when we (programmers) made the evolutionary steps from 8-bit to 16, then 32. Somewhere along the way we picked up multiprocessing in the O/S. In every step we were introduced to more instruction sets and system calls. But, it was evolutionary. Our programs required little rewrite to adopt the new technology and our underlying function remained largely unchanged. “

Actually, learning to do multithreading and multiprocessing correctly was revolutionary… and the revolution isn’t over yet. On the Amiga, back in the 80s, we considered “Task” just another programming construct… but a decade later this still hadn’t made it into either UNIX or Windows, as a rule. It’s not just the OS, it’s how you use it.

[Stephen Mann] “The CUDA API is not evolutionary as it requires a significant rewrite of how the video is processed – an underlying function of the program. And, as you point out, the competing CUDA platforms have different, proprietary API’s.”

Yes, it’s a whole new API. It’s not a whole different way of doing video… it’s a different way of writing some of the blocks, some of the “black boxes” in the overall video workflow. As was learning SSE, or 3D graphics, or any major new thing that came along.

[Stephen Mann] “Given that non-use of the GPU is probably the number one slam that Vegas-haters and some adherents use to denigrate Vegas, if it were as easy as you suggest, don’t you think that Sony would have done it by now?”

I’m not claiming it’s easy.. I am claiming it’s very doable. And yeah, I think if Sony weren’t so preoccupied with 3D, they would have done this by now. The very fact they’re moving to OpenCL in 10d says they’re thinking about it. And the very fact that most if not every other major NLE is doing it or already has, without any fundamental visible shift to the workflow (and they all, today, work far more like Vegas does than they did back in the day — it’s all non-destructive, it’s native or largely native video formats, etc).

-Dave