Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!ames!oliveb!amiga!cbmvax!daveh From: daveh@cbmvax.UUCP (Dave Haynie) Newsgroups: comp.sys.amiga Subject: Re: new chips questions Message-ID: <4627@cbmvax.UUCP> Date: 1 Sep 88 15:45:49 GMT References: <2838@pt.cs.cmu.edu> Organization: Commodore Technology, West Chester, PA Lines: 57 in article <2838@pt.cs.cmu.edu>, bader@spice.cs.cmu.edu (Miles Bader) says: > > In article <4601@cbmvax.UUCP> daveh@cbmvax.UUCP (Dave Haynie) writes: >>The same problem now occurs again in the new Denise, only more so. We're >>now dealing with 35ns pixels instead of the previous 70ns pixels, since 400 >>line non-interlaced must double the horizontal scan rate... > My understanding of how interlaced output works must be wrong. I would > have thought that the horizontal rates would be the same, and it would just > stall every other scan line (so other people could use the bus). Nope. With a 60Hz refresh rate, and the 15.6-something kHz rate of the NTSC standard horizontal line, you only have enough time to display about 200 lines (or whatever you get with appropropriate overscan). Interlacing is a way to get twice the number of lines by interleaving two separate 60Hz frames, you get twice the resolution, but it takes twice as long to refresh the full frame, so you drop to 30Hz per frame and get flicker. When you go to the new non-interlaced mode, you're displaying at about 31kHz per horizontal line, so obviously, within a 60Hz frame, twice the normal information can be displayed. > From what you said, I gather that both modes output continuously, the slower > rate of the interlaced mode letting the beam drift farther vertically (and so > leaving the gap between lines). You only get gap between lines in standard non-interlaced modes. That's because the display device is designed to allow interlacing, but you're only supplying 1/2 of the possible frame, at twice the rate. There are really two limiting factors here; display rates and memory bandwidth. Simply put, the reason that the new Denise allows only 64 colors in noninterlaced hires is a display rate limitation, the reason that the new Denise/Agnus only allow 2 bitplanes in noninterlaced hires is a memory bandwidth limitation. In the former case, you're switching from 70ns pixels to 35ns pixels. The Denise color registers only run at 140ns, so they must be quadplexed. Normal 640 pixel modes mux the registers before the CLUT; in order to be fast enough, noninterlaced hires must do it's second MUX after the CLUT. So instead of turning 16 registers into 8, it turns 12 color bits into 6. In the latter case, nothing about the Agnus memory bus has changed. In normal 640 line modes is was possible to fetch 2 bitplanes without any CPU contention or 4 bitplanes with heavy CPU contention, at the nominal rate of 200 lines in 1/60th of a second. Now in non-interlaced 640 line hires it's possible to fetch 1 bitplane without any CPU contention or 2 bitplanes with heavy CPU contention, at the nominal rate of 400 lines in 1/60th of a second. Add 'em up, and you see it's exactly the same about of memory transfer in each case, it's just that Denise is doing different visual things with the raw data Agnus provides. > -Miles -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {ihnp4|uunet|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy "I can't relax, 'cause I'm a Boinger!"