grenley@nsc.nsc.com (George Grenley) (11/09/88)
I sure wish we'd get more input from people who are neither current nor former NSC employees. I enjoy GBS's participation, but this is supposed to be a user's group... Is there anyone out there with reasonably fitm plans to actually BUILD something? In article <1051@stratus.UUCP> gbs@stratus.UUCP (George B. Smith) writes: >In article <7594@nsc.nsc.com> grenley@nsc.nsc.com.UUCP (George Grenley) writes: >>I said a lot of work would be required. However, for a hobbyist on >>a budget, it would not be too difficult to add a DRAM controller and a couple >>of meg of DRAM; and also add a SCSI port. >Amen to a lot of work! My viewpoint here is that it would be so much work >that a blank PC board may be a better starting point. In other words, what >advantage does the 532DK give other than the 532 chip? It is in a >non-standard physical format, and it has only 128k of static RAM and a serial >port. The great advantage to starting with a DK is that you have a basic working board to begin with (more or less). For instance, if you build some DRAM onto it and it doesn't work, you can write little wiggle-the-lines programs to help figure out why, even if the only test equipment you have is a 'scope. Starting from scratch is a lot more work. >>Here at the ol' factory we have >>a list of app notes we're working on; these subjects are to be covered, as >>quick as we can get to them. (By the way, while we cannot commit to anything, >>we do sometimes work with outside engineers to write such app notes.) >I have to say that I am skeptical about these app notes. The staffing level >in the 32k group is the issue here. Now the cg16 group has done some *very* >good things in the way of app notes as is readily apparent in the new 32k >data book. For netlanders, please note that the 32k app group and the cg16 >app group are two totally different groups with very different management. GBS makes some good points here. There are ALWAYS more app note titles to be writ than there are people to do it. This is one of the reasons I mentioned (in an earlier posting) that NSC is very interested in hearing from anyone who is planning on doing any DK based work - if it is any good, we'd like to publish it, and you will get money for it. (much discussion of the relative merits of cg16 versus 532 systems deleted) GBS points out that minix, et al, do not require an MMU. I stand corrected; with only the defense of being merely a hardware type... He also refers to a soon to be announced (I hope) CG16 based computer. I can't talk much about it, but if it looks a lot like a PC-AT, well... As a CG16 development platform it will be nice. But it will cost $$$. Let's compare: A 532 system based on the DK board, even with a 3 waitstate DRAM design, will run about the same CPU performance as a fast 386 design. There's just no way a CG16 can provide as much CPU power. Frankly, I think the interesting thing to build would be a multiple CG16 based hi res display controller. Say 16 cpus or so.... you could draw some fast mandelbrots... >Since we have been discussing low-cost, "hobbyiest" systems, I don't think >many of them will be porting actual UNIX to this type of hardware. Besides, >I don't think you can beat the Ziaz deal right now anyway. I agree. I recommend the 532 DK ONLY to the student/hobbyist who is more concerned with learning than with having a machine tomorrow... (comments about the virtues of NSC's business decisions deleted) >I am afraid that I *don't* know. That is why I asked the question. Let's >look at your half-empty glass a different way. Suppose we looked at it as >a half-full glass, i.e. National, with the resources of a $2 billion company, >can it do something that small companies can do with far, far fewer resources >can do? I think it is a matter of *desire*. If you wanted to produce a >neat low cost, useful, fun system, then you probably would find a way. If >you *don't* want to do this sort of thing, then it doesn't matter what you >have in the way of resources or overhead or what, you probably won't do it. George, since you used to work here, you know the situation as well as I do. You also know that it won't do me any good to comment about it on the net. Charlie Spork may well be interested in my opinions, but there are a lot of other people who aren't. Let's go have a beer & talk about it. >George B. Smith Disclaimer: personal opinion only. >Stratus Computer, Inc. >gbs@stratus George R. Grenley Disclaimer: my opinions are unchanging law... National Semi grenley@nsc
dtynan@sultra.UUCP (Der Tynan) (11/10/88)
In article <7642@nsc.nsc.com>, grenley@nsc.nsc.com (George Grenley) writes: > > (much discussion of the relative merits of cg16 versus 532 systems deleted) > > GBS points out that minix, et al, do not require an MMU. I stand corrected; > with only the defense of being merely a hardware type... > > George R. Grenley Disclaimer: my opinions are unchanging law. As I pointed out in an earlier posting, if an MMU is available, it should be used. The *only* reason MINIX doesn't require an MMU, is because the brain- dead IBM-PC doesn't have one. George, you were correct in the first place. Just because MINIX doesn't demand that you put one in, is no reason to leave it out. It is far and away easier to develop code when the system is protected by some sort of MMU. Furthermore, I don't think anyone is interested in designing/building a system which is comparable to '82 designs (ie, the PC). Not having an MMU means that if you want to run multi-tasking, then the OS is forced to relocate the code each time it executes. For those of you who argue that they don't want to do multi-tasking, don't forget that print- spooling can technically be considered as multi-tasking. In sum, the MMU is an outstanding piece of work. It alone, is what dragged me away from the grip of Motorola. So, PLEASE, don't leave it out. - Der -- dtynan@sultra.UUCP (Dermot Tynan @ Tynan Computers) {mips,pyramid}!sultra!dtynan --- God invented alcohol to keep the Irish from taking over the planet ---
gbs@stratus.UUCP (George B. Smith) (11/11/88)
In article <2632@sultra.UUCP> dtynan@sultra.UUCP (Der Tynan) writes: >As I pointed out in an earlier posting, if an MMU is available, it should be >used. The *only* reason MINIX doesn't require an MMU, is because the brain- >dead IBM-PC doesn't have one. George, you were correct in the first place. >Just because MINIX doesn't demand that you put one in, is no reason to leave >it out. It is far and away easier to develop code when the system is protected >by some sort of MMU. > > dtynan@sultra.UUCP (Dermot Tynan @ Tynan Computers) *** flame on *** First off, lets not throw stones at other companies. Until we are sure that the 32k is going to be even a modest success in the marketplace, there is little room to insult others. Lets just work on promoting the 32k. *** flame off *** Now, I agree with you 100% on the MMU issue. I just wanted to point out to George Grenley that an MMU was *not required* for MINIX or XINU. I agree that it *is desirable* however. George B. Smith disclaimer: I just had to get that off Stratus Computer, Inc my chest. gbs@stratus.stratus.com