SCRATCH and its non-realtime in 4K.

[laguun]

So,

first of all, scratch is a good system.
And as any system, it has limits.
In performance and in its business model.

As there are some different definitions (between post-veterans/studio owners and employees and customers of the manufacturer assimilate, the company behind scratch) of 4K requirements, market outlooks and postproduction workflows, this thread should easily give orientation for customers of red what scratch, being the only system which can operate at 4K allowed by red so far,
what the scratch software exactly -can- and -cant- do.

- 4K display. Some people (i am one of them) who mastered fullfeatures @4K say that it is crucial to monitor 4K at 4K, as noise, aliasing, and grain might be invisible at 2K monitoring. The scratch system so far has no 4K display. It is possible to view 25% of the images content, pixel by pixel, on the maximum display of the system which is 2K resolution.
- 4K realtime. This is a main source of confusion. Scratch cant decode 4K redcode raw in realtime in its full quality. By switching to lower qualities decode (as raw wavelet datareduction as used by the red camera easily allows), the station can read 4K @ 2K in RT. This is a great advantage if working in 2K - but the redcode raw source footage being processed at 4K requires a minimum 1:3 overhead in time.
- 4K FX. As any nvidia graphics card basing computer today, scratch, equipped with decent graphic cards (Quadro 5600 etc) is able to process 4K media in realtime. However, this does not apply to redcode raw, as the high-quality-debayer of this format overloads such as computer.

These are the basic limitations of the so far -only- allowed 4K redcode raw system. No 4K display (in general), no 4K RT (with redcode, no problem with decoded data), no RT 4K FX (once more, available with -dpx- sequences, but not with redcode raw).

I have been pointing out that these things seem to be a restriction and limitation to users who are used to a 4K RT workflow - in many posts.

In the recent week, a employee of Assimilate has been erasing these posts - backed up by the argument, that customers might be misunderstanding these facts. Many of the posters and readers of this discussion board were anrgy about this.

Therefore, i -condensed- these topics in this thread.

Beside these technical restrictions, people have been complaining about the business model of assimilate. While i agreee that this is understandable, this is a business decision of assimilate, the 48-month old company which has only one product: Scratch. They have to protect their product, as they have no other one. And especially if they are aiming at a exit-strategy (pure speculation).

Even as i pointed out months before the red camera was released, that not having a software for the customers of RED would cause the stir we are seeing now, it is important to understand that assimilates business method right now is to give customers an -individual- pricing, and therefore listprices might be causing problems. This has to be respected, even if some people (as me) say that this business model is the -exact- opposite of what the company red brought to the markets.

Closing this post, it is important to know, that there are many 4K realtime display, FX and processing alternatives in the market - which have more firepower than scratch regarding 4K RT i/o. And also more companies having more expertise, tradition and sheer number of supported raw-workflow.

Not only from the country i am living in (Germany, that would be DVS clipster and Iridas Venue among others), but also from the UK (filmlight baselight, quantel) , japan etc.

Scratch is a good software, and we might want to add it even to our -heavy- DI gear, but it is important to discuss these limitations and the market outlook.

If this sub-forum of reduser.net will be continously censored when discussing, besides the good things scratch offers, its weaknesses, then it is time to take this discussion elsewhere - and have this subforum become a marketing device instead of a discussion forum. I hope that this thread will prove the opposite - as anything else would be the opposite of the red approach.

[Lucas Wilson]

Good post, Laguun... nice summary of what is and is not currently possible.

Only small correction I have is that we *do* have list prices, I just choose not to disclose them on this forum. I have sent about 100 emails to various members of reduser with these list prices. And they are all the same.

You should also know that the deletion of posts was not unilateral. It was discussed among the moderators and was given a go-ahead. I do not make those kinds of changes (and neither do the other moderators) without discussing it first.

And honestly, I only realized yesterday that just the Moderators see which posts have been deleted and why. To everyone else, it just appears that they disappeared. Based on that discovery - I agree that the authors of removed posts should be notified, and I will do so in the future. I can't speak for the other mods, though. : )

Best,

Lucas

[Tai Wah Lim]

Not only from the country i am living in (Germany, that would be DVS clipster and Iridas Venue among others), but also from the UK (filmlight baselight, quantel) , japan etc.

Scratch is a good software, and we might want to add it even to our -heavy- DI gear, but it is important to discuss these limitations and the market outlook.

If this sub-forum of reduser.net will be continously censored when discussing, besides the good things scratch offers, its weaknesses, then it is time to take this discussion elsewhere - and have this subforum become a marketing device instead of a discussion forum. I hope that this thread will prove the opposite - as anything else would be the opposite of the red approach.

Laguun, thanks for your concerns to the RED community and pointing out limitations of available software for the RED workflow, not only SCRATCH. Like any buyer, we weigh what are available, their benefits and the limitations, ROI and decide on what we should buy at the present time. We did that and had decided to buy SCRATCH. We will certainly consider other more competitive products in the future if Assimilate does not deliver.

Lim

[Andrew M.]

Lagun, I don’t know of any 4K RT post system out there at full 12 bit linear quality.
If there is I would like to hear about it.
It is easy for 10 bit 4K with 8 bit to the video card to be real time.
But try it with full 12 bit or 16 bit linear.

Actually even if the Mysterium is 12 bit the RED RAW once imported in to the system have to be processed in 16 bit for grading.
Playing it to the screen could be done in 8 bit but I agree with you that it should be at full 4K resolution in order to estimate artifacts and FX keying.

So far quick shortcut to 1:1 zoom is doing great on 2.8K screens, sure full 4K would be more convenient but I don’t feel like spending 50K + for just the video for the computer.


What I am amazed about SCATCH is how tide the code is! 3MB that is all!
And as far as I remember the biggest active module on SCRATCH is 1.8MB
As compared with 150MB AE or Premiere it is a killer.
I don’t know how they did it but it looks like they went down to the ASSEMBLER language to program some crucial operations, and programming in ASSEMBLER is very expensive. It is like BIOS work.

I agree though that we have studios/post houses and Indies.
And I thing Assimilate can serve both markets.

Probably the Indie market in terms of profit will be larger very soon, so you have to bet on the future. I guess they have to take the crystal ball and predict the future and gamble a bit. For me it is no brainer gamble, maybe I have better crystal ball.

[laguun]

Lagun, I don’t know of any 4K RT post system out there at full 12 bit linear quality.

There are many.
DVS Clipster operates at 16bit. We are starting to roll out another -8K- 12bit Iridas Venue. Disneyland Europe was, iirc, the customer before us.

If there is I would like to hear about it.
It is easy for 10 bit 4K with 8 bit to the video card to be real time.
But try it with full 12 bit or 16 bit linear.

Systems are clipster are not limited by video cards.
they have ther propietary designed card. dvs has been using 16bits since many years in their 20 year comapny history.
this also has disadvantages, but performance and qualiyt-wise, it gives them the dge over the competition.

Other companies, as iridas or filmlight, use clusters.
The planned 8K system for our customer right now is a 36-core, 10 cluster system with 8* Nvidia Quadro FX4600 graphics combined - for one display. (http://spotlight.iridas.com/200604/index.html)

So far quick shortcut to 1:1 zoom is doing great on 2.8K screens, sure full 4K would be more convenient but I don’t feel like spending 50K + for just the video for the computer.

My project right now is an 8K multiscreen venue.
100K+ is not covering the entry costs for the display.

I don’t know how they did it but it looks like they went down to the ASSEMBLER language to program some crucial operations, and programming in ASSEMBLER is very expensive. It is like BIOS work.

I highly doubt that scratch is assembler coded. a) its to slow for that, b) all important i/o is basing on 3hrd party drivers (you dont want to interface with directX10 with assembler), c) debugging assembler on windows is pretty complicated, d) multithreaded as scratch is, assembler is not your preferred language e-z).
Several systems are impressively tight (as scratch), but i would credit that to their extensive and clever use of external libraries.

Probably the Indie market in terms of profit will be larger very soon, so you have to bet on the future. I guess they have to take the crystal ball and predict the future and gamble a bit. For me it is no brainer gamble, maybe I have better crystal ball.

As long the redcode monopoly of scratch is in place and the new 4K realtime raw solutions arent sold to end user (2-4 months) it is a good business to offer scratch, as it is the only redcode online system. Especially for non network-centric shortform 2K productions, i think scratch is the fastest way to go. 4K network-centric, workgroups and full-feature VFX extravaganzas however, are certainly better served with decoded redcode, as the postproduction industry support for redocde in full quality is basicly non.existing. We are operating more powerful systems as scratch since many years. scratch is a good product. in comparision to filmlight baselight or iridas venuehowever, its becomes obvious why these products have much higher pricepoint. scratch is a mid-level DI software - not in the most expensive cluster / dedicated hardware class, not in the non-gpu accelerated software class. The company behind can decide where to position it. We have our Di systems, we have our desktop-NLEs, we have our mid-level softwares.

let me make my point -precisely- clear.
3D animation software (as today cinema 4D, lightwave, maya, max, xsi or open sourced - blender) was >100.000$.
Audio DAW software was.
The first Avid media composer i had was ~80.000. today its 2.500

This is valid for -all- software applictaion classes, and "DI" being nothing else than editing with colorcorrection, no matter if matrix, nucoda, discreet, quantel, apple color, iridas, autodesk/discreet, dvs, baselight, davinci etc is (apple color) and will be no different.

Exit strategies are the most important business consideration for many of the DI software companies right now, not marketshare or customer relations. Keep in mind, we are talking mostly about 10-20 people companies here.

[laguun]

Laguun, thanks for your concerns to the RED community and pointing out limitations of available software for the RED workflow, not only SCRATCH. Like any buyer, we weigh what are available, their benefits and the limitations, ROI and decide on what we should buy at the present time. We did that and had decided to buy SCRATCH. We will certainly consider other more competitive products in the future if Assimilate does not deliver.

Lim

i think scratch is an very good ideal system for many needs. if we would do commercials and not offer the services as posthouse to agencies (who prefer conservative avid/flame) it would be a preferred choice. However, being 4K longform, we have different needs.

[Andrew M.]

Lagun, thanks again for very informative comments.
I see your pain, network centric houses have to have some common platform for al these diverse horse power they use.
I agree the exit strategy or if successful market adapting skills are the key to success here.

This is why Bill Gates got lucky with his CPM operating system and moved to DOS and got out from OS2/IBM stronghold.
Well he flipped his strategy few times in right time and right place, so did Adobe.

I hope SCRATCH will be successful here since I like the package and the GUI interface, the right size for my business. You couldn’t be more right, not network-centric, short form 4K>2K final.
This is my profile, you guessed it!

BTW I checked on the http://www.dvs.de/
Nice products, but bit out of my league.

Andrew

[I Bloom]

I don’t know how they did it but it looks like they went down to the ASSEMBLER language to program some crucial operations, and programming in ASSEMBLER is very expensive. It is like BIOS work.

It's fairly common for the "inner loop" of realtime systems and heavy lifters such as rendering engines and game engines, to be written in ASSEMBLER. This is in fact a very good explanation of why Red's utilities are all Intel only, because they have been optimized with code specific to one family of processors. Programming in assembler is essentially a black art, practiced only by those who can do better than a compiler at optimizing code and those willing to risk some very deep and difficult bugs.

In my mind scratch is kind of like finishing a movie on the Quake engine, a low footprint heavily optimized program made for pushing out pixels as fast as possible. Processors got very fast, and programmers got lazy, it's nice to see when someone has to really push the limits a system with deep optimization. If you can get your hands on a copy of this book: http://www.amazon.com/Zen-Code-Optim.../dp/1883577039 much will be revealed.

IBlooom

[Lucas Wilson]

It's fairly common for the "inner loop" of realtime systems and heavy lifters such as rendering engines and game engines, to be written in ASSEMBLER. This is in fact a very good explanation of why Red's utilities are all Intel only, because they have been optimized with code specific to one family of processors. Programming in assembler is essentially a black art, practiced only by those who can do better than a compiler at optimizing code and those willing to risk some very deep and difficult bugs.

In my mind scratch is kind of like finishing a movie on the Quake engine, a low footprint heavily optimized program made for pushing out pixels as fast as possible. Processors got very fast, and programmers got lazy, it's nice to see when someone has to really push the limits a system with deep optimization. If you can get your hands on a copy of this book: http://www.amazon.com/Zen-Code-Optim.../dp/1883577039 much will be revealed.

IBlooom

Ian,

Really cool that you would say this. You and our Chief Architect must share some brain cells. When you put a few beers into him and start talking about his theory behind SCRATCH, he always says he is building a game engine...

...and he does a lot of stuff in Assembly Language.

Lucas

[I Bloom]

Ian,

Really cool that you would say this. You and our Chief Architect must share some brain cells. When you put a few beers into him and start talking about his theory behind SCRATCH, he always says he is building a game engine...

...and he does a lot of stuff in Assembly Language.

Lucas

If you ship him out to NAB this year, I will put some beers in him ;)

Ian