csimmons@hqpyr1.oracle.UUCP (Charles Simmons) (10/26/88)
Okay... random thought for the month. As we all know, Steve Jobs help popularize the concept of the 3M machine; a machine with 1 Megabyte of memory, 1 "Mips" of processing power, and a 1 Megapixel display. (The machine was also supposed to cost minimal numbers of dollars.) In some sense, these types of machines are fairly common (even though the 1 Megapixel display still costs a lot). So, I'm interested in thinking about the 3G machine. Obviously, the 3G machine (selling for $1995 retail :-) contains a Gigabyte of memory and a 1 "Gips" processor (1000 times as fast as a Vax 780). This isn't all that hard to imagine. At present trends, that kind of memory and processing power should be common in low-end workstations somewhere around 2001 (depending on how optimistic your crystal ball is). But what I'm really interested in is, what's that third G? A gigapixel display? That's something like 30,000 by 30,000 pixels. How big does the display need to be in order that all of the pixels produce information that can be perceived by a human? Or maybe it would be a 3-D display, 1Kx1Kx1K pixels on each side? The most awesome display that I can imagine having (guess what I want HDTV to standardize on...) is a 2Kx2K color display having 24 bits per color. That basically requires 12 Megabytes, which seems a far cry from the type of display that a 3G machine should have. Retrenching for a moment... it seems clear that the third M in the 3M machine is a rough measure of the communications bandwidth between the computer and the human user. So, rather than concentrate on the number of pixels that fit on the display, maybe other factors should be taken into account, such as the keyboard, mouse, microphones, sound-synthesizers, etc. [Bizarre example: we might say that a Mac+ has a .25Megapixel display, and that the mouse is equivalent to .1 Megapixels of user interface, and that the keyboard is equivalent to .01 Megapixels of user interface, and that the 4-voice not-overly-great- quality sound-synthesizer is worth an additional .2 Megapixels of user interface; giving us a total of .56M.] The real purpose of this posting, however, is to see if I can get people to fantasize about what kind of user interface the 3G machine will have. Any takers? -- Cheers, Chuck
grenley@nsc.nsc.com (George Grenley) (10/26/88)
In article <470@oracle.UUCP> csimmons@oracle.UUCP (Charles Simmons) writes: >But what I'm really interested in is, what's that third G? >A gigapixel display? That's something like 30,000 by 30,000 >pixels. How big does the display need to be in order that all >of the pixels produce information that can be perceived by a >human? Or maybe it would be a 3-D display, 1Kx1Kx1K pixels >on each side? >The most awesome display that I can imagine having (guess what I >want HDTV to standardize on...) is a 2Kx2K color display having >24 bits per color. That basically requires 12 Megabytes, which >seems a far cry from the type of display that a 3G machine should have. Your first guess seems more reasonable - 3D. TI has recently demo'd what seems to be workable 3D technology - I suppose one will want bits on the order of 1.5K by 1.5K by 24 bits deep, then 1K or so of Z axis. But I suspect that the next (NeXT?) major growth area of man/machine interface is sound & speech. George Grenley
peter@ficc.uu.net (Peter da Silva) (10/27/88)
In article <470@oracle.UUCP>, csimmons@hqpyr1.oracle.UUCP (Charles Simmons) writes: > [Bizarre example: we might say that a Mac+ has a .25Megapixel > display, and that the mouse is equivalent to .1 Megapixels of > user interface, and that the keyboard is equivalent to .01 > Megapixels of user interface, and that the 4-voice not-overly-great- > quality sound-synthesizer is worth an additional .2 Megapixels > of user interface; giving us a total of .56M.] Well, let me pick one little nit. The mouse is nice, but I think you would have trouble entering text without the keyboard. Windowing interfaces that work without the mouse... even that work well... are possible. Look at Microsoft Windows, for example. Not only can you do anything you want to with the keyboard, but it's actually convenient! Give the keyboard at least as good an I/O rating as the mouse. > The real purpose of this posting, however, is to see if I can get > people to fantasize about what kind of user interface the 3G machine > will have. Any takers? NASA's VIVED system looks like a good start. 3-d high-res display, using the dataglove as feedback. Two gloves, at least, with IR links to the headset. Replace the CRTs with see-through LCDs so you can concentrate on the real world without taking the headset off. At least 16 channels (8 for each ear, via the headset) of 16-bit stereo sound. Leave the keyboard in place, but attach a touch tablet for precision input work (I suspect drawing with the dataglove is like drawing with a mouse). Note that there is no seperate screen. If you don't care for feedback, or if you can get feedback from the dataglove, you don't need a keyboard either. An office full of people gesturing at the air or typing on their desks would be weird. If your headsets are hooked together, the computer could show you some indication of the cloud of data around each person... you'd never want to take it off at work. Walk down the hallway reading a listing. Nothing to tie you to a desk. I'd think that you'd need well over your 2K by 2K resolution to simulate virtual screens floating in the air. 8K by 8K with a couple of bits of antialiasing would be more reasonable. For fine work you could drop shades over the LCDs and shut out reality. Simulate reality in wireframe so you can still move around without being distracted. Antisocial behaviour (nerdishness) could end up being strongly enforced. Oh yeh, voice input. Whether it's used for deata input or just used to talk to people, it's useful. "Computer" "Yes, Captain" "To Charles Simmons" "You mean csimmons@hqpyr1.oracle.UUCP" "Yes" "OK" "Check this out. End." "Ready" "Attach file" (point to /usr/spool/news/comp/arch/176) "Ready" "Send" "Yes, Captain" -- Peter da Silva `-_-' Ferranti International Controls Corporation "Have you hugged U your wolf today?" uunet.uu.net!ficc!peter Disclaimer: My typos are my own damn business. peter@ficc.uu.net
muir@postgres.uucp (David Muir Sharnoff) (10/27/88)
In article <470@oracle.UUCP> csimmons@oracle.UUCP (Charles Simmons) writes: >The most awesome display that I can imagine having (guess what I >want HDTV to standardize on...) is a 2Kx2K color display having >24 bits per color. That basically requires 12 Megabytes, which >seems a far cry from the type of display that a 3G machine should have. You can do much better than 2Kx2K... I've always wanted a 3'x4' display at >300dpi and >= 24 bits deep.. The display would take >445MB and that should be enough to burden even a 3G machine. On the other hand, I don't think anyone is going to think it worth building such a display until machines that are 1G by other criteria are OLD news. Perhaps it will be a possible display for the 3T machines. -Dave
koopman@a.gp.cs.cmu.edu (Philip Koopman) (10/27/88)
In article <470@oracle.UUCP>, csimmons@hqpyr1.oracle.UUCP (Charles Simmons) writes: > As we all know, Steve Jobs help popularize the concept of the 3M > machine; a machine with 1 Megabyte of memory, 1 "Mips" of processing > power, and a 1 Megapixel display. (The machine was also supposed to > cost minimal numbers of dollars.) > In some sense, these types of machines are fairly common (even > though the 1 Megapixel display still costs a lot). So, I'm interested > in thinking about the 3G machine. > But what I'm really interested in is, what's that third G? > A gigapixel display? That's something like 30,000 by 30,000 > pixels. How big does the display need to be in order that all > of the pixels produce information that can be perceived by a > human? Or maybe it would be a 3-D display, 1Kx1Kx1K pixels > on each side? How about gigapixels/second. Animation is the key to interaction, and interaction is the key to understanding. A 2Kx2K display at 1 gigapixels/second would update at 250 frames/second, which is probably overkill. How about multiple screens updating at, say 25 frames/second? Phil Koopman koopman@maxwell.ece.cmu.edu Arpanet 5551 Beacon St. Pittsburgh, PA 15217 PhD student at CMU and sometime consultant to Harris Semiconductor.
ralphw@ius3.ius.cs.cmu.edu (Ralph Hyre) (10/29/88)
In article <3405@pt.cs.cmu.edu> koopman@a.gp.cs.cmu.edu (Philip Koopman) writes: >In article <470@oracle.UUCP>, csimmons@hqpyr1.oracle.UUCP (Charles Simmons) writes: >> thinking about the 3G machine. > >> But what I'm really interested in is, what's that third G? > [rhetorical part, maybe I'll crosspost to computers and society digest later] I'm amazed that an order of magnitude increase every 4 years is happening in this industry. When will/can it stop? What's the next great challenge, after computers? A shame the political issues (like the DRAM crunch delaying new products, and RISC/CISC wars [quit reading netnews and get back to your real work and designing these 3G machines], and OS/2 vs Unix) are shaping the future more than technical ones. [now, on to the technical content.] >How about gigapixels/second. Animation is the key to interaction, >and interaction is the key to understanding. >A 2Kx2K display at 1 gigapixels/second would update at 250 frames/second, >which is probably overkill. How about multiple screens updating >at, say 25 frames/second? I believe that only 2-3 screens/person are needed, but remember you want 24-bits/pixel, so its really 2kx2kx24 at 10 frames/second is the limit. Unless timesharing makes a comeback, the silicon graphics 'Super-Personal' IRIS GTX can come out a year or two before the 21st century. I'm more intrigued by the 3d comments, but am impressed by THOSE bandwidth requirements (with 'stupid' encoding, 1kx1kx1kx32bits/pixelx66 frame/second) is 2.27 TERAflops/second, another three orders of magnitude and (presumably) 20 years away. [I suspect no-one will use stupid encoding, so we'll see vector-based 3-D graphics systems may make a comeback.] Snooping the transoceannic phone cables will consume a gig/second of INPUT capacity, so the NSA is in for some high-quality competition. -- - Ralph W. Hyre, Jr. Internet: ralphw@ius3.cs.cmu.edu Phone:(412) CMU-BUGS Amateur Packet Radio: N3FGW@W2XO, or c/o W3VC, CMU Radio Club, Pittsburgh, PA "You can do what you want with my computer, but leave me alone!8-)" -- - Ralph W. Hyre, Jr. Internet: ralphw@ius3.cs.cmu.edu Phone:(412) CMU-BUGS Amateur Packet Radio: N3FGW@W2XO, or c/o W3VC, CMU Radio Club, Pittsburgh, PA "You can do what you want with my computer, but leave me alone!8-)"
loren@pixar.UUCP (Loren Carpenter) (10/31/88)
In article <471@oracle.UUCP> csimmons@oracle.UUCP (Charles Simmons) writes: >In article <3440@pt.cs.cmu.edu> ralphw@ius3.ius.cs.cmu.edu (Ralph Hyre) writes: >>I believe that only 2-3 screens/person are needed, but remember you want >>24-bits/pixel, so its really 2kx2kx24 at 10 frames/second is the limit. > >Um... I was kind of thinking that the 2Kx2Kx24 color screen would update >at something like 40 to 60 frames per second. Presumably that would >require a bandwidth of something like 480 to 720 Megabytes per second. > >(The intent is to have >a screen that has no artifacts. The screen has to have resolution better >than you can get in a theater. I remember watching a movie once, and >being bummed because updating a screen at only 24 or 30 frames per second >makes candle flames jump around all over the place.) > >-- Chuck You don't want artifacts?... Well, first you have to quit thinking 24 bits (8 each RGB) is sufficient. Of course, with perfectly matched and compensated dac's, monitors, ambient lighting, etc. you might get away with 8 bits. Also it's important to distinguish the number of bit planes per color in the memory from the number of bits going into the dac. It's not hard to find examples where 10 in or 10 out are barely enough, especially as the dynamic range of displays improves. The monitor frame rate had better be more than 60. 60-72 frames per second is demanded by folks who look at imagery all day. 35mm projection resolution in a typical theater is roughly equivalent to 1200-1500 horizontal lines. 70mm is not quite twice as good. However, a projected film image jumps around with projector frame-to-frame misalignment, has grain noise that averages out somewhat, is run through a 2 (or 3) bladed shutter which flashes each frame twice (or 3 times), and so on... The multi-bladed shutter is responsible for the strobing of the candle flame (as well as the strobing of non-motion-blurred animation). Yeah, I'd like a 2kx2k (2kx3k would be better) 64-bit (16 each RGBA) frame buffer with a 66 hz frame rate. It's possible to build such an animal today, but you couldn't sell it for less than $200k. Loren Carpenter ...{ucbvax,sun}!pixar!loren
snoopy@sopwith.UUCP (Snoopy T. Beagle) (11/01/88)
In article <3405@pt.cs.cmu.edu> koopman@a.gp.cs.cmu.edu (Philip Koopman) writes: |How about gigapixels/second. Animation is the key to interaction, |and interaction is the key to understanding. A 2Kx2K display |at 1 gigapixels/second would update at 250 frames/second, |which is probably overkill. How about multiple screens updating |at, say 25 frames/second? 250 frames/second is overkill. 25 is a little slow. 50-60 should be about right. Standard motion pictures are 24 fps and are sometimes jerky. Also note that normal film projectors open and close the shutter 2 or 3 times per frame to reduce flicker. _____ /_____\ Snoopy /_______\ |___| tektronix!tekecs!sopwith!snoopy |___| sun!nosun!illian!sopwith!snoopy