[-k]

patdawg wrote:

hyrlvlrec wrote:

so when the flame console says “700mb allocated to framebuffer” thats what its doing?

Exactly…it is allocating 700MB of your SGI’s system memory to be used for the framebuffer. Note that you can change that number up or down by modifying the memory keyword in your init.cfg file. Once we get 64-bit memory addressing in the next year we’ll be able to add gobs of memory, and crank that memory keyword up to ridiculous values.

I’m not an expert but this hype sounds a bit too easy to me.

1st) Someone posted a test on flame-news:

—
I made some benchmarks in the last week for texture fill rate only. And I want to post my results to here. This tells both texture fill + accumulation buffer performance of the gfx subsystem. The cpu speed is almost unimportant.

The test is like this, just get a 2K_8bit image into action, add an image, and scale it down to 50, and render 100 frames of PAL. But set the antialiasing to 16 samples, and motion blur to 50 samples.

1. inferno_onyx350_ir4(single RM11-1024): ~4min.
2. inferno_onyx2_ir2(dual RM9-64): ~4min.
3. flint_fx300g(dual 3Ghz xeon): ~18min.
4. tezro_V12(dual 700Mhz): ~32min.

—-

Tezro does not look that good (in this case)….!

2nd) I thought the whole point of the AGP bus is that you could use system memory on PCs. Nobody did it of course but should it not be possible?

3rd) Just because you can use system memory does not tell you how fast this solution is.

4th) Even if so. It might be fun to have 100 layers in Action, DVS but depending on what you do you might only have a few layers but hundreds of blurs , keys, edge detects, color corrects, logic ops, gmasks etc. (that’s at least the case for me far more often than your scenario). Which probably benefits much more from cpu speed

5) Honestly I think it is a JOKE that FFI has not been a 64 Bit application since years. Now they try to sell you this for lots of $€$€. A joke really…

-k

[-k]

Once again of course you can compare… what’s the problem with that? I did’nt say they are the same, cause a comparison would be rather boring in that case.
They are really different that’s why you have to compare!

I totally agree with demand on realtime playback but you can do the same with combustion. After rendering to memory you put the contents of the cache to disc and that’s it… you know stones are not the only highspeed disc arrays in the world… we just got ab Xraid on a g5 which does 2k playback. I don’t like them particulary but they work.. granted I would not want to work on lots of footage with this in a supervised session. But for short sequences it does the job…

What does bandwith help if your CPUs are too slow to deliver… The only discreet product which does alot in realtime is lustre which runs on a windows PC and apparently not even with StoneFS on the raid (anybody knows this? from what I understood it’s actually NTFS???)… so much for the bandwith…

And now with the increasing power of GPUs already or soon to be used for real realtime image processing…

I think discreet has to look at these trends and better make toxic a real good thing…

-k

[-k]

Quote:

Even if you work on a system faster than an inferno or a flame or a flint, combustion is not the better tool (then any other desktop app) for working with a client or on a complex shot when you don’t have so much time.

In most cases you are right. But I can only repeat myself. There are quite a few exceptions to this rule. I’m working on project in HD 12 bit on a [email protected] Quite some heavy comp stuff. Some setUps render ~45 min with proxy mode still ~25 min.
What’s left of the FFI advantage now…? If it wasn’t for burn…

As for the sparks… yes you can live without almost all of them (50% are nonsense actually). I prefer to build my glow by hand for example. But it’s a joke FFI still don’t have a real “defocus” or blurmask (greyscale determines the blur radius), cause it is really tricky to fake something like this…
Other people might have different needs though…

-k

[-k]

Quote:

‘m a french compositor working on flint on octane 2, and have a powerbook with combustion on it.

YOU CAN NOT SERIOUSLY COMPARE THOSE TWO SOFT

Of course you can compare. Rather easy to do so…
And seriously, lets not talk about combustion on powerbooks, not even G5s (the mac ports are a joke).
As I said, as long as you don’t have a burn farm in the background, for a few short sequences with lots of CPU intensive stuff there is good chance you are faster on a decent workstation with Combustion than on a [email protected]…
Plus the paint module is much better in combustion (ok, its different anyway but I like it better).
Granted I’d still prefer our flame or inferno for most of my work and yet there are quite a few occasions, where I decide I’m better and faster in combustion or shake. And as the others pointed out, having to pay some xx.xxx€ for inferno sparks to get some functionality which most other apps offer for a fraction of the cost or are directly integrated and network render for free…
That’s probably my biggest problem with FFI at the moment cause burn solves the performance issues on CPU heavy stuff really well…

wondering what toxic will be like in the end…

-k

[-k]

I would not deny that there is a difference. And yet you do have to wait for stuff to render, and as soon as you have lots of CPU intensive stuff in your setUp (couple of sparks, blurs, keys, ccs, etc.) the difference starts to become smaller.
Granted we have a 8 burn servers which help alot.
But over the past few years the performace advantage of FFI systems has decreased quite a bit depending on what you do, plus the toolset of the desktop apps have become rather sophisticated and “built in”. Seriously the whole sparks thing on FFI is a bad joke… burn licenses for sapphire anyone…?

-k

[-k]

You can buy just the software. It’s done quite often.
We had the gear before for other reasons, so then we just bought stones and software from discreet. Even videoboards, etc we got somewhere else…

Of course you can get FFI on a CR-Rom. That’s at least what I saw the admin putting in our onyx… 😉
I mean, there’s no mystery about IFF. It’s just some software which needs certain hardware, so what? With some knowledge you could set it up all by yourself… that’s in parts what we did. Whether that’s a good idea to do is a different question.

Considering the fxguide china report, there are apparently quite a few FFI “hacks” running out there…

As for the “instataneous” results…
Hmm… lets stick to: you get realtime playback. Which is great, but I seriously doubt that a scenario as you described would be solved “instantly”, though probably faster than on desktop apps…

-k

[-k]

Quote:

for the lord of the rings story. Rumour has it that weta had scripts linking shake to their render manager. so artists could press render when they leave their work station and their shots came out already comped.
i’ve personally done that, it its a lovely way to work.

You can network render with flame too. Works very smoothly
Check out “burn”.
It’s just terribly expensive… 😉

-k

[-k]

Well, if you want to stay 8 Bit you don’t need to use DPX which is 10 Bit. Go with TGA or TIF.
Plus you might want to consider XStoner cause it makes transfers more easy…

-k

[-k]

Hm…the prism looked promising but as you said, the Itanium does not seem very safe…
Bad luck for SGI I think.
And maybe for Discreet too.

but maybe this whole toxic thing is gonna shake up things anyway, even though discreet says it won’t

-k

[-k]

I think there are some tips on this to be found on fxguide.
In short: think like an animator! Look for extreme “poses” and set the keyframes for your masks there first. This way you usually get better results than with setting keyframes every 10 frames or so.
Use seperate masks for different body parts. This way you become much more flexible in terms of overlapping and also concerning the interpolation.

-k

[-k]

You might also want to check out ARRI’s development:

http://www.arri.de/prod/cam/d_20/d_20.htm

-k

[-k]

Quote:

-k wrote:

I got different numbers / experience :
HDCAM ~7 fstops, Viper ~8-10, Film ~14

The difference still exist.

yes, I would’nt deny, it’s just not as extrem as your numbers might suggest.

Quote:

there are HDR sensors out there which can capture 26 f stops.

There are 4K digital projectors and OLED large-screen displays, 500G rewriteble Blue-ray discs and many other experimental laboratory devices which not yet implemented as end-user products icon_smile.gif We are talking about commercial cameras, am I right ?

Oh, yes. You can buy the camera I mentioned. It’s just stillphoto panorama cam.
http://www.spheron.com/products/SpheroCamHDR/spherocam_hdr.html
So, I was just wonderin…

Quote:

-k wrote:

Yip. But there are adapters to give you 35mm DOF and you can use those lenses. Whether that’s a practical solutions is a different question.

It is only adapter to use existing lenses. The differense in sizes still exist. You can see it when checking image on the screen. You can see it if print onto the film original negative and digital source. I saw it on real content.

I think this is not true. The whole point about it, is the rotating mirror inside it on which the image is projected thus giving you a bigger area of projection -> 35mm DOF. From what I heard however this technique softens the overall image a bit. Haven’t really seen it though.
Check Pro35 under
http://www.35digital.com

-k

[-k]

Quote:

In-first, you must not confuse with 12 bits. Bits in digital cameras and in negative is differs. Both measure the accuracy of dynamic range’s representation. But dynamic range of digital cameras smaller than film’s range ([email protected] and [email protected] vs 13-17D for different Kodak emulsions). Especially, CCD sensors lose details in hilites in compare with negative.

I got different numbers / experience :
HDCAM ~7 fstops, Viper ~8-10, Film ~14

btw.: there are HDR sensors out there which can capture 26 f stops (had the pleasure to use this for onset capturing environments for CGI). Although it produced quite some noise at the ends of the range, I wonder when we do see higher range CCDs in digital movie cameras. That would however add to the bandwith problem 😉

Quote:

2. Don’t forget about smaller size of this sensors vs 35mm framesize which produce the worser depth of field.

Yip. But there are adapters to give you 35mm DOF and you can use those lenses. Whether that’s a practical solutions is a different question 😉

-k[/quote]

[-k]

Quote:

When applying the patch the inferno on Onyx2 gave the message
“driver not found for ide device”
before the usual barrage of messages.

Got the same message on our onyx2 though that does’nt seem to be a problem(?). Smoke(octane) and flame(octane2) patched without problems….

-k

[-k]

Mindmachine wrote:

🙄 Hi Guys, i tried to track a locked off shot with a character moving.I tested all the mayor 3D trackers,like boujou, 3d equalizer, matchmover and also syntheyes, but all of them had problems in tracking a strong motionblur. I mean are there no intelligent algorithms for interpolating between tracked and not tracked frames? (I mean a tracker that guess in which direction the point went ?) Especially boujou, which is not a object tracker, but it found many trackable features on my shot, but when it comes to solve the 3D camera, boujou totally failed. Even with supervised trackers,locaters and gold flagged points, it was not able to reconstruct the scene. I use boujou a lot and mostly i get almost perfect tracks, but these
tricky motionblur shots are still a problem, for all of them !! (The shot is done now by hand, but any suggestions or hints a still welcome)

Thanks MArk[/b]

I just had a similar problem. Head of actress coming into frame, looking into camera, going out of frame again. Lots of motion blur at the beginning and the end. I went through all trackers available (3DE, matchmover, boujou, inferno. In the end I had the best results in Maya Live.
I think this is really quite a good tracker if you don’t need automated tracking. The 2d tracking really works well ( better than the combustion/flame tracker, I think)
Use a higher number of root frames than the default and smaller filtering level.

worked for me.

-k

[Author]

Posts