xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) (12/15/90)
walrus@wam.umd.edu (Udo K Schuermann) writes: [about Imagine performance:] > Benchmarking will be difficult because lighting, positioning, and the > properties of objects can affect the rendering time greatly. Here are > a few bits, though, which may be useful to you: > My system: 68030/68882 @ 32MHz > I have a scene with a very large diamond (12 faces) has a disturb > texture on it (to produce obvious flaws in the mostly transparent > material), three spheres, a multi-color pyramid, and all this on a > black ground with a thick grid mapped onto it. A torus with an explode > effect on it goes into several hundred facets that fly outward. > The scene requires about 15 minutes in Scanline mode at 352x470 > interlaced HAM. > Another scene with a crystal sphere, multiple reflective objects, > others with roughness and some with brushes wrapped onto them, as well > as three light sources, takes (again at 352x470) about 40 minutes in > full trace mode. I'm still a bit confused. Are those the times to render a single frame of the animation, or to render an entire animation? In either case, how many frames of animation did you create, and what was the missing (total or frame) time to do the animation? > Summa summarum: Full traces with refraction, object reflection, and > cast shadows require more time, and for these 20 minutes is only > possible if you have relatively simple scenes. An unaccellerated 1000, > 500, or 2000 will not be too impressive. Probably so since the unaccelerate machine should take at least six to twelve times longer, and possibly much more if you're using an FPU as well, but one of the joys of multitasking is to put something like this in the background and let it grind while you put your system to other uses. The limiting quantity here seems to be chip ram, so the obvious question: is it possible to render to fast ram, or better, directly to a file, to prevent a squeeze in chip ram over the long period of an animation? Is it possible to checkpoint and restart an animation to avoid a restart from scratch due to power problems or system guru's during software development? These are some of the features that would make such a system "fully professional". On a separate subject, one of the real roadblocks to doing animations these days is the high cost/low availability/marginal quality of single frame videotape recording. Frankly, recording a single frame to a videotape is an abuse of the technology. Is anyone doing work onto other media, such as floppy optical storage, that would allow an animation to be created and then dumped to videotape at normal recording speeds? This is a tough problem, since, as the CD versus CD-ROM converstations finally convinced me, it is laser videodisks that use a recording technology similar to FM radio in how it encodes the signals, so a digital floppy optical recording couldn't be just blindly dumped to tape, but would require a digital to analog step in real time. Input? Kent, the man from xanth. <xanthian@Zorch.SF-Bay.ORG> <xanthian@well.sf.ca.us>
walrus@wam.umd.edu (Udo K Schuermann) (12/17/90)
In article <1990Dec15.042510.14514@zorch.SF-Bay.ORG> xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) writes: >walrus@wam.umd.edu (Udo K Schuermann) writes: > >> [times to render stuff] > >I'm still a bit confused. Are those the times to render a single frame >of the animation, or to render an entire animation? Single frames. Keep in mind that full raytracing is _much_ slower than Scanline rendering. My machine traces nearly anything in 30 minutes or less. >In either case, how >many frames of animation did you create, and what was the missing (total >or frame) time to do the animation? The Scanline rendered scene came out well, so I went for a "final" with full trace. 60 frames at 352x470 interlaced HAM, each frame roughly 25 to 35 minutes, takes it a good 30 hours to trace. >but one of the joys of multitasking is to put something like this >in the background and let it grind while you put your system to other >uses. The limiting quantity here seems to be chip ram, so the obvious >question: is it possible to render to fast ram, or better, directly to a >file, to prevent a squeeze in chip ram over the long period of an >animation? Is it possible to checkpoint and restart an animation to >avoid a restart from scratch due to power problems or system guru's >during software development? I always have Imagine running at priority -1 so that the system responds well when *I* want it to respond. Imagine just gets all the left-over cycles. No (or very little) chip RAM is used during rendering, and the frames are written directly to disk. Chip RAM is used when the animation is created. Thus, I render an animation's frames, store them all on disk (this requires disk space, of course) and when the whole thing is ready, I'll create the animation without killing the rendered frames (in case of trouble). Once that's done, I can wipe the frames. This approach means that when the machine does go down I lose only a minimal amount of work: Never more than one (incompletely) rendered frame, and Imagine updates the project status immediately to reflect the presence of a new rendered frame. It goes to say here that I don't create an animation unless all the animation's frames have already been rendered. This way I kill as little Chip RAM as possible. The size of individual frames can range anywhere from 20K to 140K -- depends on the output format as well as the size and complexity of the scene. >These are some of the features that would make such a system "fully >professional". I agree. Imagine _is_ professional. ._. Udo Schuermann "How is American beer similar to making love in ( ) walrus@wam.umd.edu a canoe?" -- "Both are f***ing close to water."
mark@calvin..westford.ccur.com (Mark Thompson) (12/18/90)
In article <1990Dec15.042510.14514@zorch.SF-Bay.ORG> xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) writes: >Is anyone doing work onto other media, such >as floppy optical storage, that would allow an animation to be created and >then dumped to videotape at normal recording speeds? I would like to hear someone tell me I am wrong but I am fairly certain that floptical/CD/CD-ROM and the like are not capable of the data bandwidths required for high quality video storage. The recordable laser video disks are currently the only things capable of this. Of course you could always stack a BUNCH of them in parallel to achieve it but then whats the point. Then again, MPEG will probably change the situation. >Kent, the man from xanth. +--------------------------------------------------------------------------+ | Mark Thompson | | mark@westford.ccur.com | | ...!{decvax,uunet}!masscomp!mark Designing high performance graphics | | (508)392-2480 engines today for a better tomorrow. | +------------------------------------------------------------------------- +