home *** CD-ROM | disk | FTP | other *** search
- type new
-
-
- <<< MEMO EMS - 77 lines, 1 append(s) >>>
-
-
- Append on 02/10/87 at 22:38 by Rob Larson:
-
- There are at least 6 EMS simulator products available. Since I wrote
- one of them, I can give you as much detail as you can take on the topic.
-
- EMS provides a 64k window below the 1 megabyte address. Up to 4 16k pages
- may be active at any one time within that window. While paged memory
- hardware typically places the 64k window in controller memory (between 640k
- and 1 meg), software simulators allocate the buffer in conventional memory.
- Since the EMS 3.2 specification describes the software interface only, the
- actual hardware used may vary considerably.
-
- The performance of any EMS implementation is a function of the real hardware
- used, the software driver performance, and the application using it. The
- amount of page turning, as opposed to data access within a page, is the
- dominant factor in describing the performance difference between paged
- hardware, extended memory, and disk based implementations of EMS. Given
- all that "your mileage may vary" mumble, I'll identify a few points in the
- spectrum. The following are measurements of performance on my 10 Mhz,
- Cheetah 80386 equipped, 40ms access hard disk, early IBM AT BIOS machine.
- The hardware EMS board is a recent version AST Rampage, and the simulator
- is LIMSIM using extended memory.
-
- Program Function Paged Hardware Simulation Ratio
- Ready Load program 12.4 seconds 12.5 seconds 1:1
- 123 Load 394k file 36.6 seconds 38.4 seconds 1:1.2
- Bench EMS Benchmark 4.2 seconds 252.1 seconds 1:60
-
- I recall a test in which a character string was duplicated across about
- a megabyte of cells in Symphony. The simulation took about twice as long
- with LIMSIM as opposed to with an AST Rampage.
-
- Obviously a page thrashing program such as the PC Magazine EMS benchmark is
- not a viable application to run with a simulator. Ready is clearly a very
- good fit, and 123/Symphony are degraded somewhat but are still in the area
- of viability.
-
- The advantages of a simulator are that it:
-
- 1. Supports existing extended memory.
- 2. Allows use of (generally) less expensive memory.
- 3. Works with high clock rate machines.
- 4. Tends to be more cost effective.
-
- The disadvantages of a simulator are that it:
-
- 1. Runs slower than a paged memory board.
- 2. Requires 64k of conventional memory for the page frame.
- 3. Doesn't work with all EMS programs.
-
- Early versions of Desqview (1.20 and before) go direct to the Intel/AST
- hardware. The current version of Javelin maps the same physical page
- into two different locations simultaneously (and requires storage into
- either to be reflected into the other immediately). Late versions of
- Desqview work with some simulators and not with others. On a "well
- behaved" scale of 0 to 10 I would give Desqview a minus 10. It is,
- however, an important program that can be supported by a simulator.
- Since access to extended memory includes enabling interrupts, situations
- that require interrupts to be disabled cannot be handled by a simulator.
- Disk based simulators (such as Above Disk) additionally have restrictions
- on whether or not a disk i/o call can be issued at the point where a page
- mapping request comes through. Most single task programs run successfully
- with a simulator, but multi-taskers tend to create all sorts of "you can't
- get there from here" situations. Or as James Burke said at a talk I went
- to last night, "In order to get there, you wouldn't start from here."
-
-
-
- Now a question. Can anybody tell me how to get the hardware description
- (port addresses, etc.) of the XMA memory board for 3270 PC/ATs? I'd like
- to add XMA memory board support to LIMSIM. The idea is to provide an EMS
- style interface to the XMA memory.
-
- /Rob Larson
-
- *** APPENDED 02/10/87 22:38:06 BY TYMSHARE/ROB ***
-
-
- <<< MEMO NETWORKS - 8 lines, 0 append(s) >>>
-
- Network access requirements.
- Could someone please take the time to document access requirements to any
- and all networks referred to on this conference such as BITNET, USENET,
- DELPHI, INFO-MAC ETC. Information such as access restrictions, membership
- fees, telecommunications charges etc. would be very useful. Thanx in
- advance to all who take the time.
-
- *** CREATED 02/10/87 22:49:11 BY RCA/KRAIG ***
-
-
- <<< MEMO TOSHIBA - 19 lines, 1 append(s) >>>
-
-
- Append on 02/10/87 at 22:36 by Kevin Strong - Interactive Data Corp - IDC:
-
- I mean that after all these years dealing with a 5 1/4 inch world that it
- feels strange to use a 3 1/2 drive, the medium from which I can safely
- stuff in my shirt pocket (good), but which I can`t use anywhere else (bad).
- It don't bend, but it isn't transportable. The PC-Link board is one of the
- major items in making this box usable, My opinion is that this unit is
- useless without the Link board - I used it to load **all** of the software
- that I'm using. (I may have cheated slightly on software licenses, but I
- DID order a new copy of everything I installed.) Other than DOS, I don't
- think any of them are available on 3.5 anyway - and I'm using MS-DOS, not
- IBM. At this point, I'm with Arny - I plan to use the 3.5 for backup
- only. At least it holds twice what a 5 1/4 does.
-
- I forgot to mention above: The unit came with SIDEKICK and SUPERKEY
- included on 3.5 disks.
-
- *** APPENDED 02/10/87 22:36:32 BY IDC ***
-
-
- <<< MEMO 386 - 68 lines, 1 append(s) >>>
-
- 80386 Product Discussion
-
- This is intended to be a home for 386 product rumor, discussion, experience,
- and whatever. To start it off, I'm aware of the following products that are
- either currently available or are supposed to be Real Soon Now.
-
- 1. Compaq 386. $7000 or so and available now. Solid product that
- performs like a 16 Mhz 286 now, plus has much higher potential
- when software becomes available in 32 bit mode.
-
- 2. Intel 386 accelerator board. About $2000, "limited availability
- late January". RSN.
-
- 3. Intel 386 PC/AT replacement mother board. About $3000, available
- now from various board vendors. Probably equivalent performance
- to the Compaq mother board.
-
- 4. Orchid Technology 386 accelerator board. About $1200, available
- RSN. Still waiting for my "come get it" call that was scheduled
- for the last week of January.
-
- 5. Cheetah 386 Adaptor. $495 RSN. I've got one (!!!). My unit is
- what I would call "late prototype", and I had to push hard to get
- it (without documentation or 80287 compatability). The model
- shipped to customers obviously will have both documentation and
- 80287 compatability.
-
- 6. American Computer 386 translator. $795 Now. "Now" not tested
- since Cheetah came thru with the same functionality at several
- hundred dollars less.
-
- There were many more 386 products displayed at COMDEX, but these are the
- ones that I know about as being either available or with a credible RSN.
-
- /Rob Larson
-
- *** CREATED 02/10/87 23:15:59 BY TYMSHARE/ROB ***
-
- Append on 02/10/87 at 23:16 by Rob Larson:
-
- I've had a Cheetah 80386 board now for about 2 weeks. The version I
- have is what I would describe as "late prototype". I convinced them to
- send me a board even though the documentation is not ready and the 80287
- hardware is not functional. My documentation consists of a phone number
- to call for assistance. The latest story I have from Cheetah is that
- the adapter board will not be sold separately from the 386 chip, since a
- special press is required to install the chip. Guess you don't just
- push the 132 pin chip into it's mounting by hand.
-
- Performance of the 386 adaptor board is roughly equal to a 6 Mhz AT,
- provided that you install the 8+ Mhz crystal that came with my kit.
- After using the adapter board for about a week, my AT died. To make a
- long story shorter, I now have a 6/10 Mhz clone mother board running at
- 10 Mhz with the remaining parts from the AT. I'm running with the case
- off, the two slots next to the 386 empty, and all the floppy/hard disk
- cables as much out of the air circulation as possible. The AT mother
- board seems to have healed itself, but I think there was a heat problem
- with the adapter board installed. At 10 Mhz, the PC Magazine BENCH
- program describes the speed as roughly equal to an 8 Mhz PC/AT. Sounds
- to me like about 20% degradation from a native 286 board. I'm using 150
- ns memory chips (120 ns. chips on the AST Rampage), and the jumper set
- to 1 wait state. The memory rating is too slow for 10 Mhz usage, so I
- expect problems, but I'll leave the 6/10 switch at 10 as long as it
- works. The 386 is rated at 16 Mhz so that's clearly no problem.
-
- /Rob Larson
-
- *** APPENDED 02/10/87 23:16:09 BY TYMSHARE/ROB ***
- R;
-
- 7.