Archived from groups: rec.video.desktop (
More info?)
Thanks for your reply Frank.
I can imagine that there are indeed opportunities for parallel processing in
the rendering process, perhaps something as simple as one thread working B
and P frame generation while the other is doing audio processing of some
sort. My comment was really meant to question why Ulead gets an apparent
speed benefit of (maybe) 5X over others who also exploit multi-threaded
techniques like Vegas.
The SmartRender regions appear to be operating at (essentially) disk I/O
speed, which is to say that the renderer spits out finished output at a rate
which is way above real time. This "lack of action" is certainly key to the
overall speed, since the areas which do not need to be rerendered are output
extremely quickly. Such is not the case for Vegas or FCP. More impressive to
me is the speed the rerendered areas seem to be processed at. Even in the
transitions, titles, etc., the renderer moves very swiftly. I had to check
my settings and file output to be sure I wasn't rendering normal DV footage.
Sure enough the output file was of the proper resolution for HDV, has a
bitrate of about 25 MBits/sec, and looks fabulous.
My hunch is that the renderer is written with a combination of assembler,
very strategic use of the P4 SSE / SSE2 instruction set, an optimizing
compiler (if they did use C++ or some other higher order language
whatsoever), and hand optimized code for the tightest loops where the real
work is done. The best aspect of all of this is the very likely prospect
that HDV NLE software will support pretty robust workflow without the need
for hardware acceleration. Such was not my impression when working with the
other (Vegas and FCP) products, whose previews and rendering are far too
slow for efficient production.
I too was a KoolAid junkie, and would actually eat the powder before it even
went into the pitcher. My mom would therefore mix it up before I could get
my hands on the envelope........and she also discouraged us from drinking
the stuff. I'm sure some of the amalgam in my mouth owes it origin to all
this junk.
The price/performance of this UV9 product is really quite remarkable. I have
only done relatively short clips to make comparisons between the 3 products,
but I don't see why they shouldn't scale about linearly for longer jobs. I
will do more experimenting and see what I find, and I will certainly follow
other reports here and in other forums to see if my Dell has some magic
affinity for UV9.
I feel like the guy who (believes he) is getting 60 miles per gallon in his
Hummer H2................
Smarty
"Frank" <frank@nojunkmail.humanvalues.net> wrote in message
news:1s3pi1l9s4rkct5agra1sei63i3ov1vhpd@4ax.com...
> On Sat, 17 Sep 2005 11:59:45 -0400, in 'rec.video.desktop',
> in article <Re: HDV - Too good to be true? (FCPHD versus Vegas 6
> versus Ulead VS9)>,
> "Smarty" <nobody@nobody.com> wrote:
>
>>Thanks for your reply Frank.
>
> You're quite welcome.
>
>>I have visited your website often and consider
>>it a superb and truly authoritative / comprehensive place to get HDV
>>information.
>
> I'm pleased to know that it helps people.
>
>>Yes, I did mean HT (not GB),
>
> You meant GHz (giga Hertz, not giga bytes), I think.
>
>>but am surprised that
>>hyperthreading gains much (since the processing doesn't inherently seem
>>likely to benefit by parallelism) and also since Cineform / Vegas should
>>have presumably exploited the same HT benefits if they do exist.
>
> Any given piece of software would have to be purposely designed and
> written in a manner such that it could take advantage of multiple
> execution units whether in the form of HyperThreading, dual cores,
> multiple processors, etc. The trick to achieving parallelism is to
> create multiple worker threads which can be executed in parallel but
> without conflict in the form of logical out-of-order execution,
> deadlock conditions, race conditions, etc. Some types of applications
> tend to lend themselves to being written in a multi-threaded manner
> more than others. Applications which can only be executed in a
> top-to-bottom fashion can't really benefit from running on a system
> capable of parallelism, although it some cases - using a dual
> processor system as an example - one processor could be assigned to
> running the application and the other processor could be assigned to
> executing only operating system functions such as memory management,
> supervising I/O operations, timer operations, screen updates, etc.
>
> I understand that Vegas itself contains code to take advantage of
> parallelism, but I don't know if this is also true of the Cineform
> codec.
>
>>Your number 2 point is especially interesting to me. I was unaware that no
>>proxy or intermediate was being used in the Ulead codec,
>
> Yes, it's merely (losslessly) converting the Transport Stream into a
> Program Stream. The raw data is unchanged, hence the lossless nature
> of the conversion from a quality standpoint.
>
>> and this would
>>indeed account for faster renders, particularly for material which is
>>unaltered in the edit.
>
> That's the SmartRender feature in action, or lack of action, actually,
> if you know what I mean.
>
>>The amazing performance, however, seems to take place
>>even heavily edited material where the long GOP structure ***MUST BE***
>>broken such as titling and transitions and blasts right through it.
>
> I don't know what to say. Maybe someone swapped the innards of your
> single processor Pentium 4 system with a dual processor Xeon system
> when you weren't looking.
(Wouldn't *that* be nice?!)
>
>>The
>>edited frames are, no doubt, "lumpy" (to borrow the phrase somebody else
>>coined to describe MPEG streams which have their P and B frames altered).
>>I
>>guess it is conceivable that Ulead has chosen to do I frame only edits,
>>effectively taking out hunks as long as one half second, at the edit
>>points,
>>but they must redo the GOPs for the smooth and continuous effects they
>>achieve with transitions and smoothly scrolling titles.
>
> They may also be creating open GOPs. This I don't know, but it could
> be checked by running the output file through an MPEG datastream
> analysis program. In the early days of MPEG, I used to have one or two
> such programs, but no longer. They were commonly used to validate MPEG
> datastreams for compliance to the standard.
>
>>In addition, since
>>they do not apparently use a codec which creates a proxy or intermediate,
>>you would think their editing and previewing would go very very slowly. To
>>the contrary, their timeline seems way way faster to me than Vegas 6. It
>>just seems contra-intuitive that they can avoid the conversion to a proxy
>>/
>>intermediate, achieve fast timeline / scrubbing / preview effects, AND
>>render so damn fast without trading speeds off the way the others do.
>
> Are you working with short clips, or doing long-form work? This
> difference could have a major impact upon performance/responsiness.
>
>>Certainly they must be making some extremely good design decisions in the
>>algorithms and even better choices in their code / programming since the
>>performance is so sweet.
>
> It would certainly seem that way.
>
>>On an absolute basis, I will readily admit that
>>Ulead's product is very limited in functionality compared to the other two
>>products (FCP and Vegas) and does not warrant the high prices of the other
>>two, but it sure does make a nice, simple, competent, and ***FAST*** HDV
>>solution.
>
> Based upon your report, it would seem to provide good value for money,
> or to put it another way, a high price/performance ratio, which I
> think makes it a good fit for the low-end consumer market to which
> it's designed to appeal, where most users are complete novices when it
> comes to the technical aspects of video and video editing.
>
> In the world of Windows HDV editing products, the only other low-cost
> contender that I know of is Vegas Movie Studio+DVD Platinum Edition
> for $129.95.
>
>>Their Media Studio professional product is probably fast also, or
>>so I would imagine.
>
> Don't quote me on this, but I understand that MediaStudio Pro version
> 8 will contain some pleasant surprises.
>
>>To answer your question, all of my work is with HDV from a Sony FX-1 at
>>1080i. And thus my comparisons are all using the same footage.
>
> Good, that's what I figured.
>
>> Did I mention that you can arbitrarily mix HDV and normal 720 by 480
>> MPEG2
>>as will as high res stills in the same Video Studio 9 time line and Ulead
>>blasts right through them?
>
> You didn't mention that, but I did know that.
>
>>Try doing that it FCP............it doesn't know
>>how to handle it at all as far as I can tell. I have been using Video
>>Studio
>>since it came out as well as most of Ulead's other consumer programs, and
>>this version is in a whole different league IMHO. The early versions were
>>buggy beyond belief, and Ulead's support was awful. Both have changed
>>dramatically.
>
> All good news.
>
>>I would really like to see if anybody else has played with HDV on VS9 to
>>see
>>if I am maybe drinking the wrong KoolAid or smokin' something here.....
>
> Must be the KoolAid. I know I just *loved* the stuff when I was a kid,
> despite being constantly told that it would rot my teeth.
>
>>Smarty
>
> --
> Frank, Independent Consultant, New York, NY
> [Please remove 'nojunkmail.' from address to reply via e-mail.]
> Read Frank's thoughts on HDV at http://www.humanvalues.net/hdv/