smith@NRL-AIC.ARPA (05/11/84)
From: Russ Smith <smith@NRL-AIC.ARPA> A previous note of mine had a wish that someone would develop an AI machine with a HUGE physical (as opposed to virtual) address space. Well, no sooner said then done. Motorola has decided to put their 68012 processor in an 84-lead pin-grid-array-style IC (*). Thirty one (31) of the pins are for address...2 Gigabytes directly non-virtually addressable memory...time to start thinking about building another home machine...Okay memory manufacturers, now where are those 500 Mbyte RAM chips I've been waiting for? Where goes Motorola (et al), where goes the world? Russ <Smith@nrl-aic> NCARAI (*) According to the May 3, 1984 edition of ELECTRONIC DESIGN, page 31.
yamauchi@fortune.UUCP (05/15/84)
#R:sri-arpa:-67000:fortune:28000037:000:58
fortune!yamauchi May 14 14:10:00 1984
68012?????????? Don't you mean 68020?!
tim@nmtvax.UUCP (05/15/84)
<> He didn`t get it wrong. There is a (or will be) a 68012. It is not a 32bit wide data cpu. It is the same as the 68000 with the difference that all the address lines are available. The 68000 throws away 8 of the address lines normally. His reference article can it explain it better than I. Look it up... Tim
abc@brl-tgr.ARPA (Brint Cooper ) (05/19/84)
See the May, 1984, issue of IEEE Transactions on Computers for a fascinating paper by 3 chaps from Princeton. The paper deals with the general usefulness and value of machines with massive amounts of main memory and modestly fast processors.
wsmith@umn-cs.UUCP (Warren R. Smith) (05/24/84)
#R:sri-arpa:-67000:umn-cs:6900028:000:150 umn-cs!wsmith May 23 14:16:00 1984 There is an article in the most recent issue of IEEE Transactions on Computers on the subject of such machines that you might want to take a look at.
mwm@ea.UUCP (05/25/84)
#R:sri-arpa:-67000:ea:7100012:000:570 ea!mwm May 24 16:36:00 1984 /***** ea:net.micro / nmtvax!tim / 4:44 am May 17, 1984 */ <> He didn`t get it wrong. There is a (or will be) a 68012. It is not a 32bit wide data cpu. It is the same as the 68000 with the difference that all the address lines are available. The 68000 throws away 8 of the address lines normally. His reference article can it explain it better than I. Look it up... Tim /* ---------- */ Minor correction - the 68012 is the 68010, not the 68000. This buys you virtual machine/memory capabilities, as well as memory-speed block moves/compares/etc. <mike
max@bunker.UUCP (Max Hyre) (05/25/84)
[ Can we store all these lines in a WOM somewhere? ]
I'm surprised no one has mentioned an idea I read somewhere a couple
of years ago. Remember, we're talking about write-once memory here.
What you do is put 400M of whatever on your 500M CD and leave the
rest blank. You have cunningly built write capability into the drive,
and when a bug comes up, you've got 100M of space to fix it in. Unless
you write obscenely buggy software, 25% should be sufficient to put
corrections in. This also allows the user to put his home movies,
favorite TV shows ( fair use only :-) ), whatever, on it. Just send
an update floppy to the user and the system will take care of the
rest.
The article I read (source forgotten) also pointed out how useful
this short of thing is for audit trails and such. Every time you update
a record in the database, you just rewrite it, and mark the previous
version of that record as expired.
All this leads up to: Just what sort of equipment is needed to
*write* a laser disk, and what are the chances it can be packaged
suitably for home consumption?
"Give me ROM, lots of ROM, ... don't fence me in!"
Max Hyre
(Somewhere in the vicinity of decvax!ittvax!bunker!max)
P. S.: Make that "this *sort* of thing" in paragraph 3.
P. P. S.: Anyone else remember trying to figure EPROM patches which
differed from the existing code only in having certain 1s flipped
to 0s so you didn't have to erase & burn the whole thing over?Kyle.wbst@XEROX.ARPA (05/26/84)
What was the bottom line in that article?
crp@stcvax.UUCP (Charlie Price) (05/28/84)
> What you do is put 400M of whatever on your 500M CD and leave the > rest blank. You have cunningly built write capability into the drive, > and when a bug comes up, you've got 100M of space to fix it in. A nice idea, but I'm not sure you can do it. Another division of Storage Technology is building a 4 Gigabyte write-once-read-many optical disk. You record information by burning spots onto a disk-shaped media and you read it back by looking at the spots. It isn't possible to mass-produce this disk with pre-recorded information on it -- you would have to write each disk. This is quite different from a video disk (analog) or CD where the disks can be reproduced by a mastering process quite cheaply (on the order of $10 for reasonable volumes). The whole point of using a video disk or CD is that it is cheap to reproduce once you have done the master. I don't think that you COULD record anything with a laser on a video disk or CD once it had been reproduced; it simply isn't that sort of media. A video disk reader for a computer is really just a heavy duty Analog to Digital device with a *LOT* of error correction. CDs are a better story in that regard I think, but still don't have good enough error characteristics for storing correct digital data. Reference Technology, a small startup in Boulder, is building a video disk reader that works like a disk drive (i.e. a fairly fast servo positioner, a "block" structure on the device, and a heavy-duty ECC scheme). The video disk is just a standard old video disk and you could play it on a TV though the result might be disappointing (a Michael Jackson video will go into their reader with equally disappointing results when considered as a data-storage disk). They expect to have at or just under a gigabyte on a video disk. They are just at the point of first customer ship of an evaluation unit. The interface is something standard (SCSI?) and it should be able to hook up to both larger systems (like VAXen) and PCs that have an interface of that sort. They are definitely planning that it should be usable with a PC, but it is sufficiently expensive that it isn't a hobbyist sort of device (I either don't know or can't tell you a price). Their market isn't what people have talked about in this group -- it is much more the static database market -- and there is a larger one than I ever thought there was. For instance, somewhere in the US there is a legal database that contains something like every brief ever submitted to a federal court. Lawyers query this by significant words and phrases. Generating the index information for all significant words and phrases takes WEEKS of large mainframe time, but then the retrieval is fairly fast. There is a large mainframe whose sole purpose is to retrieve items from this database in response to textual queries. This service is apparently fairly expensive and for various reasons of equipment or convenience not all lawyers are able to use it. They want to blast the database onto video disk(s) and ship the database to lawyers (or firms more likely) who have a PC or a larger machine. Retrieval from the fully indexed database is something that even a PC or a small UNIX machine can do quickly enough. I couldn't speculate on any future plans they have for CDs other than to note that people in the company think they are interesting. Incidentally, the ECC required for their reader is very heavy-duty. The error rate they have to expect is 1 bit in 100. They have an ECC that should allow correction of a 1 inch error burst on the disk. If the physical damage didn't matter, you could take a video disk, shoot a hole in it with a 45, and their reader could still read it without error. -- Charlie Price - STC (disk division) uucp: { hao, ihnp4, decvax}!stcvax!crp { allegra, amd70, ucbvax }!nbires!stcvax!crp USnail: Storage Technology Corp - MD 3T / Louisville, CO / 80028 DDD: (303) 673-5698