fsdal1@acad3.fai.alaska.edu (LEVINSON DONALD A) (09/20/90)
occasionally I see ads for EISA motherboards that sya they can use either EISA or ISA boards. how can this be? is this part of the EISA spec? fsdal1@acad3.fai.alaska.edu don levinson
grege@gold.GVG.TEK.COM (Greg Ebert) (09/20/90)
In article <1990Sep19.233544.16757@hayes.fai.alaska.edu> fsdal1@acad3.fai.alaska.edu writes: >occasionally I see ads for EISA motherboards that sya they can use >either EISA or ISA boards. how can this be? >is this part of the EISA spec? > EISA is an upward-compatible backplane architecture which allows both ISA (standard PC and AT-bus cards) AND EISA cards to be used. It is impossible to put an EISA card into an ISA bus. The maximum tranfer rate is 33 MB/sec for 32 bit-wide transfers. [Damn, and I left my EISA spec with my previous employer..] There are 2 tiers of connector 'fingers' on an EISA card. _______________________________________________ / \ | | | Circuit board | | | | | | | \_______________________ ____/ |____________________| <---- ISA fingers |________________| <---- EISA fingers When you put an ISA card into an EISA slot, the physical width of the ISA bus connector area doesn't allow the card to reach the EISA signals. When you put an EISA card into an EISA slot, the card seats all they way down with ISA signals going to the ISA fingers, and EISA signals going to the EISA fingers. When you put an EISA card into an ISA connector, DUCK!!!! :-O Personally, I feel EISA is a crock of horse-puckey contrived to throw a monkey wrench into the PS/2 marketplace. So what if it's 32 bits. What on earth do you need it for ? For a variety of reasons, it's optimal to put RAM on the CPU board (if you have plug-in CPU's), or on the motherboard. What about disk controllers ? Sheesh, an AT bus-master can transfer about 4 MB/sec, which is 2-3 times faster than the fastest hard-disk transfer rate. Video, anyone ? Try caching video BIOS; now it's too fast - No joke, I'm totally serious. ETHERNET fizzles-out at 1.25 MB/sec. I don't think MODEMs will run that fast for awhile (sarcasm). I even heard it from the mouth of someone who personally attended the 'Gang of Nine' EISA meetings that it was definitely not the wave of the future. My advice is to put the money you would otherwise spend on a EISA system towards a 33Mhz 486, and a 'bus-master' type of hard disk controller (such as the Adaptec). ------------------------------------------------------------------------- ##### {uunet!tektronix!gold!grege} Register to vote, then ## | ## grege@gold.gvg.tek.com vote responsibly # | # # /|\ # Support the First Amendment, not the party that attacks it #/ | \# "I was, BANNED in the USA" - 2 Live Crew #######
marshall@wind55.seri.gov (Marshall L. Buhl) (09/21/90)
fsdal1@acad3.fai.alaska.edu (LEVINSON DONALD A) writes: >occasionally I see ads for EISA motherboards that sya they can use >either EISA or ISA boards. how can this be? >is this part of the EISA spec? Yes. EISA just stands for extended ISA. Just as the 16-bit AT bus (ISA) can handle XT cards (8-bit), the 32-bit EISA bus can handle ISA cards. Or even XT cards for that matter. They won't have the same functionality, but they should work. -- Marshall L. Buhl, Jr. EMAIL: marshall@seri.gov Senior Computer Missionary VOICE: (303)231-1014 Wind Research Branch 1617 Cole Blvd., Golden, CO 80401-3393 Solar Energy Research Institute Solar - safe energy for a healthy future
KDM101@psuvm.psu.edu (Kevin Maher) (09/21/90)
In article <1990Sep19.233544.16757@hayes.fai.alaska.edu>, fsdal1@acad3.fai.alaska.edu (LEVINSON DONALD A) says: > >occasionally I see ads for EISA motherboards that sya they can use >either EISA or ISA boards. how can this be? >is this part of the EISA spec? > >fsdal1@acad3.fai.alaska.edu >don levinson As far as I know (from what I've read in the trade mags, it is part of EISA's spec to accept the standard ISA boards. There is a special "key" on the EISA boards that allow it to slip the whole way into the bus slot and connect with all of the slot contacts. Standard ISA boards will slip down into the slot 100%. PC magazine just reviewed the 3 major bus architectures about a month ago. If you know somebody that gets this magazine, you may want to check it out. (as far as they were concerned, ISA will be around for a while yet. Neither EISA or MCA are that much better that they'll wipe out ISA) --------------------------------------------------------------------------- | OOO | | | O O SSS | Why do today what's due tomorrow? | | O O PPP U U S | | | O O P P U U SSS | Hey...sounds good to me! =-) | | O O PPP U U S | | | OOO P UUU SSSS | KDM101@PSUVM O04@PSUVM | | P | KXM@PSUARCH kmaher@psusun01 | --------------------------------------------------------------------------- Why should I want to disclaim anything??? It only makes me look guilty!
malloy@nprdc.arpa (Sean Malloy) (09/21/90)
In article <1471@gold.GVG.TEK.COM> grege@gold.GVG.TEK.COM (Greg Ebert) writes: >There are 2 tiers of connector 'fingers' on an EISA card. > > \_______________________ ____/ > |____________________| <---- ISA fingers > |________________| <---- EISA fingers > > >When you put an ISA card into an EISA slot, the physical width of the ISA bus >connector area doesn't allow the card to reach the EISA signals. > >When you put an EISA card into an EISA slot, the card seats all they way down >with ISA signals going to the ISA fingers, and EISA signals going to the >EISA fingers. Close, but no cigar; you've got the basic principle right -- the EISA cards do go in deeper to hit a second row of contacts -- but you've got the mechanism wrong. The EISA cards have the same length of edge connectors (measuring the long way on the card), but the connector is twice as deep. Where a 16-bit ISA card has one notch (between the 8-bit and 16-bit extension connectors), the EISA card has three; the additional two only go halfway up the connector flange: ISA: |_________________________ __ ___| |________||___________________| EISA: |_________________________ __ ___| | __ || __ | |___||___||__________||_______| (The positions of the additional EISA cutouts may or may not be as depicted here; I'm working from memory) EISA sockets have flanges across the width of the socket halfway down; when an EISA card is inserted, the flanges match the cutouts, and the card can seat all the way down the socket. When an ISA card is inserted, the flanges prevent the card from going any farther in than is necessary for the card to connect to the first row of contacts, which match the ISA bus contact layout. | "The three most dangerous Sean Malloy | things in the world are a Navy Personnel Research & Development Center | programmer with a soldering San Diego, CA 92152-6800 | iron, a hardware type with a malloy@nprdc.navy.mil | program patch, and a user | with an idea."
marshall@wind55.seri.gov (Marshall L. Buhl) (09/21/90)
grege@gold.GVG.TEK.COM (Greg Ebert) writes: >What about disk controllers ? Sheesh, an AT bus-master can transfer about >4 MB/sec, which is 2-3 times faster than the fastest hard-disk transfer >rate. I'm certainly no expert at this, but what if you're data striping to a whole bunch of disks? Compaq, Dell and Northgate (and maybe others) use data striping in their new tower systems. -- Marshall L. Buhl, Jr. EMAIL: marshall@seri.gov Senior Computer Missionary VOICE: (303)231-1014 Wind Research Branch 1617 Cole Blvd., Golden, CO 80401-3393 Solar Energy Research Institute Solar - safe energy for a healthy future
RREED@ucf1vm.cc.ucf.edu (09/21/90)
In article <1990Sep19.233544.16757@hayes.fai.alaska.edu>, fsdal1@acad3.fai.alaska.edu (LEVINSON DONALD A) says: > >occasionally I see ads for EISA motherboards that sya they can use >either EISA or ISA boards. how can this be? >is this part of the EISA spec? > Exactly! That is one of the biggest advantages the EISA bus sports, high performance and backwards compatibility, something that the Microchannel bus machines can not claim. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Reginald Reed -- Struggling Computer Science Student -- Does Bo Know Comp Sci?? BITNET=+> rreed@ucf1vm INTERNET=+> rreed@{ucflan,ucf1vm}.cc.ucf.edu
phil@brahms.amd.com (Phil Ngai) (09/21/90)
In article <1471@gold.GVG.TEK.COM> grege@gold.GVG.TEK.COM (Greg Ebert) writes: |There are 2 tiers of connector 'fingers' on an EISA card. | | | | | \_______________________ ____/ | |____________________| <---- ISA fingers | |________________| <---- EISA fingers | That was the original plan but then someone came up with the scheme described by Sean and it was adopted. Clever idea. As for the comments on the need for EISA, I would expect them to be accurate for at least 5 years, if not forever, with the possible exception of video. I don't know if there will be a plug-in XGA card (1024x768, 16 bit color) but if so it can probably use all the speed you can get. -- Phil Ngai, phil@amd.com {uunet,decwrl,ucbvax}!amdcad!phil Freedom is dead, long live privacy!
grege@gold.GVG.TEK.COM (Greg Ebert) (09/22/90)
> I write: > >>What about disk controllers ? Sheesh, an AT bus-master can transfer about >>4 MB/sec, which is 2-3 times faster than the fastest hard-disk transfer >>rate. > marshall@wind55.seri.gov (Marshall L. Buhl) writes: >I'm certainly no expert at this, but what if you're data striping to a >whole bunch of disks? Compaq, Dell and Northgate (and maybe others) use >data striping in their new tower systems. I guess that might be a valid question, but most of the time involved in disk operations involves seeking. UNIX has a neato scheme where accesses are queued and sorted so that head movement is generally 'sweeping' instead of erratic as in FIFO requests. Adding yet another level of complexity, transfers to/from another drive(s) is/(are) done while other units are involved in seek operations. Watch the activity lamps on a multi-drive UNIX system; it's kinda trippy. I presume network servers would do the same. Maybe this is termed 'data striping' ? A 512 byte sector can be eaten or barfed in about 130-140 microseconds by an AT. Assuming an average access time of 15 milliseconds for a rather speedy drive, the transfer overhead is ~1%. Even a track-track access will take >5msec, so your overhead jumps to 3%. For network-server applications, my hunch is that an ISA controller will still be acceptable because the _maximum_ data rate for Ethernet is 10Mbits per second, or 1.25Mbytes/second (neglecting overhead) which is another bottleneck. Contrast this to ~4Mbytes/sec for an ISA (AT) bus-master. The trick is how the files are laid-out physically on the disk, and how multiple requests are handled. People do PhD dissertations on this stuff (no joke, it's a massively complex problem). [lawyer fodder] UNIX is a trademark (or other silly possession) of AT&T. ------------------------------------------------------------------------- ##### {uunet!tektronix!gold!grege} Register to vote, then ## | ## grege@gold.gvg.tek.com vote responsibly # | # # /|\ # Support the First Amendment, not the party that attacks it #/ | \# "I was, BANNED in the USA" - 2 Live Crew #######
cgordon@vpnet.chi.il.us (Gordon Hlavenka) (09/22/90)
>occasionally I see ads for EISA motherboards that sya they can use >either EISA or ISA boards. how can this be? The EISA spec was designed to be backwards-compatible with the ISA (AT) bus, which in turn is backwards-compatible (mostly :-) with the XT bus. So... >is this part of the EISA spec? Yes. ----------------------------------------------------- Gordon S. Hlavenka cgordon@vpnet.chi.il.us Disclaimer: Yeah, I said it. So what?
carroll@sunc7.cs.uiuc.edu (Alan M. Carroll) (09/22/90)
What about more than 16Meg of memory? I've got an ISA box here, with 16Meg, and it's starting to get a little tight (I'm stucking doing LISP development with Emacs, X-windows, and Unix. Memory -> poofta).
barton@holston.UUCP (Barton A. Fisk) (09/27/90)
In article <1477@gold.GVG.TEK.COM> grege@gold.GVG.TEK.COM (Greg Ebert) writes: >marshall@wind55.seri.gov (Marshall L. Buhl) writes: >>I'm certainly no expert at this, but what if you're data striping to a >>whole bunch of disks? Compaq, Dell and Northgate (and maybe others) use >>data striping in their new tower systems. > >erratic as in FIFO requests. Adding yet another level of complexity, transfers >to/from another drive(s) is/(are) done while other units are involved in seek >operations. Watch the activity lamps on a multi-drive UNIX system; it's >kinda trippy. I presume network servers would do the same. Maybe this is >termed 'data striping' ? > Don't know about Dell and Northgate, but Compaq's machine syncronizes the hard disks (same sector is presented to the same head on all drives) for reads and writes, so in theory a four drive system will read/write 4 times faster than a one drive system. The catch is drive failure causes loss of pieces of all files. Compaq offers several solutions to guard against this including mirroring and controller duplexing and data guarding. The later being an interesting approach where the data in a 4 drive array is encoded and placed on one of the drives (transparent to the user). If a failure occurs, the defective drive is fixed and the system restores itself. I like this. Needless to say, I am waiting for my Systempro to arrive. -- uucp: holston!barton pseudo: barton@holston.UUCP
davidsen@sixhub.UUCP (Wm E. Davidsen Jr) (09/30/90)
In article <5777@holston.UUCP> barton@holston.UUCP (Barton A. Fisk) writes: | The catch is drive failure causes loss of pieces of all files. Compaq | offers several solutions to guard against this including mirroring | and controller duplexing and data guarding. The later being an interesting | approach where the data in a 4 drive array is encoded and placed on | one of the drives (transparent to the user). If a failure occurs, the | defective drive is fixed and the system restores itself. I like this. | Needless to say, I am waiting for my Systempro to arrive. I can see how that would work using any of several schemes, but all of them seem to require not using the other drives until the failed unit is replaced. The simpler solutions also would require that for every bad sector the corresponding sector be marked bad on all srives, although you can get by that if you are willing to put up with the delay caused by mapping every sector on every drive on every access. Use of sector sparing would make the first method easier to live with. -- bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen) sysop *IX BBS and Public Access UNIX moderator of comp.binaries.ibm.pc and 80386 mailing list "Stupidity, like virtue, is its own reward" -me