Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!iuvax!watmath!watcgl!wsflinn From: wsflinn@watcgl.waterloo.edu (Scott Flinn) Newsgroups: comp.sys.ti Subject: Re: Myarc Geneve Keywords: Graphics capabilities Message-ID: <10677@watcgl.waterloo.edu> Date: 14 Jul 89 19:32:37 GMT References: <4752@psuvax1.cs.psu.edu> <1643@csm9a.UUCP> <176@zip.eecs.umich.edu> <10655@watcgl.waterloo.edu> <5507@pt.cs.cmu.edu> Reply-To: wsflinn@watcgl.waterloo.edu (Scott Flinn) Organization: U. of Waterloo, Ontario Lines: 138 In article <5507@pt.cs.cmu.edu> bjm@RIO.MT.CS.CMU.EDU (Bret Musser) writes: >To which Atari ST are you making these comparisons to? The 1040ST that I have owned for ~2 years. >If it is to the earlier models, all this is outdated. >The newer ST's have dedicated graphics chips, like >the Amigas do. Most people who own Atari ST's own 1040's or 520's, neither of which can use the new graphics chips (I assume you mean Blitter) without a) fancy hardware mods (like an accelerator board with a blitter socket) and b) an upgraded version of TOS (ie. 1.4, which isn't officially available yet). But this is probably not the place to be discussing Atari stuff. >This is all good and fine about graphics, but what about >processor speed? The 9995 runs at the SAME MHz as the >9900 in the old /4A. Just that the 9995 already had >the dedicated 8 bit bus, and it took fewer cycles to >execute the same instructions and it pre-fetches >instructions. I'm not sure what you mean here, but the 9995 in the Geneve and the 9900 in the 4A are definitely not running at the same speed. I forget the exact stats I produced (it has been a while), but I ran comparisons of counting loops which used the processor's on-board RAM, on both chips. This avoids most of the problems of data bus width, memory speed and wait state insertion. I always forget the clock speed of the 4A (I used to look it up in my EA manual, which I no longer own), but I do recall the ratio of the counting loop speeds *precisely* equaling the ratio (12 MHz) / (MHz of 9900 in 4A). > Also, it is not really a 12MHz machine. >When you calculate instructions-per-second you use >something like 4 MHz. The internal frequency is not the >same as the external crystal. I find this interesting (I'm not pretending to be a Real Expert ... I appreciate your comments). I thought that the custom chip provided hardware support for 4A mode which, among other things (like changing the memory map) introduced extra wait states into memory cycles. And since there was almost no Geneve specific software available when I sold my machine, I only assumed that running in "Geneve mode" would alleviate the problem. I didn't have info telling me how to enter this mode myself, but I thought that Myarc Advanced Basic used it, and I certainly know that this Basic really flew ... was that simply due to replacing the old GPL interpreter? >Also, when you buy a machine like the Atari or Amiga with >their 68000's, you get much more memory addressable. >The 9995 is 16 bit. It can address 64K. As I recall, >the 68000 has a 20 or 24 bit bus so can access X megabytes >(sorry, I don't have a calculator here). It can access 2^24 = 16M (although the Atari ST MMU restricts it to 4M on that machine). But there is no way you can compare the Geneve *in a general way* to the 68000 machines (Mac, St, Amiga) ... the software just isn't there. If a decent compiler (read: very tricky to do right) was available which handled Geneve's memory pages the same way PC compilers cope with the Intel memory segments, then things might be different (I realize the since Geneve's pages must be on 8K boundaries whereas 8086 segments are only restricted to 16 byte boundaries, the compiler could not be as flexible). >I don't wish to burst anyone's bubble here... but like >when I see Myarc saying that "Well, if we up the crystal to >18MHz [or whatever] we'll be knocking on the door of >the 80386 machines." (That's almost a direct quote from >a recent Micropendium.) First, when that was written, they >had never reached 18MHz. Also, the max rated crystal >frequency is 12.1MHz, although things are always rated >conseratively. But look at it this way, 18/12 = 1.5. >If the Geneve is about the speed of a PC, an 18MHz Geneve >won't even be near an AT (which is like 2-5 times a PC). > >I used to own a Geneve for 2 years. I really liked it. >However, I got really sick of Myarc not pulling through. >It wasn't their fault entirely. I'm glad to see it >start to pull itself together...BUT, remember that when >you buy a Geneve, you are buying 10 year old technology. >And the software support seems to be like 20 years old. Was Geneve ever intended to be competition for PC's, Mac's, Amiga's and the like? There isn't a stand alone version yet is there? If not, then its market is "doomed" from the very beginning ... it is only useful for those very few TI enthusiasts who have a PE box and can afford a new toy. Unless a "modern" software environment is provided (like GEM's desktop, the Amiga workbench, Mac Finder, etc.), nobody will want to use it for "real" things (ie. suitable for a business environment). But ... there are lots of interesting things that happy Geneve owners can make themselves happier doing. Before you die of boredom, two brief examples: 1. I still contend that the machine has greater graphics processing potential than a stock Atari ST (or Mega ST). Without a floating point unit, quick 3D graphics will still be difficult, but the Geneve environment should allow new things (suited to its specific hardware) to be concocted. 2. I have just finished writing (for a PC type machine) a real-time multi-tasking kernel. I am now kicking myself for having sold my Geneve ... such a kernel would be perfect for the Geneve, for two reasons: a) the architecture of the TMS99xx processor provides complete context switches for free (well, for the cost of executing a single BLWP). This would give incredible multi-task performance (comments anyone?). b) having a system based on many concurrently running light-weight tasks would make many of the memory paging problems go away, having the kernel take care of all communication between processes running in different memory pages. Anyway, the point is that the Geneve is not competition for the 68000 machines, but need not be relegated to the status of a machine outdated in every way by 20 years. It still provides a platform for some useful things that other machines do not, and above all, gives a more able computer to TI people who can't bear to part with their beloved beast. >[If you disagree, post it... this is a good discussion we >have here... this boards picking up a little!] I have taken your suggestion and run with it. My apologies for the length. >=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= >Bret J. Musser --- Carnegie-Mellon University | "If you can count your > | money, you don't have >Internet: bjm@f.gp.cs.cmu.edu | a billion dollars." >Bitnet: bjm%f.gp.cs.cmu.edu@cmccvb | >UUCP: | -- J. Paul Getty >=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= -- Me: Scott Flinn / "If it doesn't fit, force it. Domain: wsflinn@watcgl.waterloo.edu / If it breaks, then it didn't UUCP: watmath!watcgl!wsflinn / fit anyway."