stever@videovax.Tek.COM (Steven E. Rice, P.E.) (11/11/86)
In article <1150@cit-vax.Caltech.Edu>, Lyle N. Scheer (lyles@tybalt.caltech.edu.UUCP) writes: > A few questions. First, is it truely possible to use a 68020 replacing > the 68000(ie, are there any problems, bugs, etc(I have a program that is > said to make a 68010 work with the amiga, but I don't know if it will do > the same thing with the 68020). Second, what is the relation of the > 68881 to the 68020, and, if you can replace the 68000 with the 68020, is > there a spot for the 68881? 1. There are a couple of problems in replacing a 68000 with a 68020, one of which is also present when replacing a 68000 with a 68010. In the 68000, the MOVE SR,<ea> (Move from Status Register) instruction is a non-privileged instruction. In the 68010 and 68020, this instruction was made privileged (because visibility of the "S" bit, indicating supervisor state, would prevent a virtual memory system from being transparently layered over an existing operating system). The 68010 and 68020 added the MOVE CCR,<ea> (Move from Condition Code Register) instruction to allow user-mode access to the condition codes. The program you have is designed to paper over this problem in software that uses the MOVE SR,<ea> instruction by substituting MOVE CCR,<ea> instructions. There is a call in the Amiga operating system that gets the condition codes the correct way, no matter which processor is used. The second problem is physical. The 68000 (and the 68010) come in a 64-pin DIP ("aircraft carrier"). The 68020 is packaged in a 114-pin Pin Grid Array (PGA). There needs to be a socket adapter added. 2. The 68020 and the 68881 share a common interface -- the Coprocessor Interface. This is a protocol, rather than extra pins, and it can be emulated by the 68000 or 68010 (or, for that matter, the 68008 or a Z80!). However, it is *much* faster with the 68020 because the protocol is built into the processor's microcode. Something else that helps speed up the 68020-68881 combination is that they can communicate over a 32-bit-wide bus, while the 68000 and 68010 must perform almost twice as many data transfers to accomplish the same results. The 68881 is packaged in a 68-pin PGA. It can be connected more-or- less directly to the bus, so adding it to a 68020 adapter card, along with a few control chips, would be the most effective solution. Steve Rice ---------------------------------------------------------------------------- {decvax | hplabs | ihnp4 | uw-beaver}!tektronix!videovax!stever
aburto@marlin.UUCP (Alfred A. Aburto) (11/12/86)
In article <1150@cit-vax.Caltech.Edu> lyles@tybalt.caltech.edu.UUCP (Lyle N. Scheer) writes: >Organization : Calfornia Institute of Technology >Keywords: 68020, 68881 >From: lyles@tybalt.caltech.edu (Lyle N. Scheer) >Path: tybalt.caltech.edu!lyles > > > > > > >A few questions. First, is it truely possible to use a 68020 replacing the >68000(ie, are there any problems, bugs, etc(I have a program that is said to >make a 68010 work with the amiga, but I don't know if it will do the same thing >with the 68020). Second, what is the relation of the 68881 to the 68020, and, >if you can replace the 68000 with the 68020, is there a spot for the 68881? > > Wonko the Sane Yes it is possilble to replace the Amiga's 68000 with the 68020 and the 68881. My Amiga is now running with a 68020/68881 in place of the 68000. It works very well with KickWorkbench V1.2. V1.1 does not support the 68881 (multi-task switching problems). The best feature of the 68020 is its internal cache memory which helps to increase the Amiga's performance by (roughly) a factor of two. This is much more of an improvement than you would get from the 68010. Also for dramatic improvements in floating-point processing speed the 68881 is a must. The 68881 is designed as a co-processor for the 68020---they work together and the 881 floating-point instructions are treated as an extension of the 020 instruction set. That is with these co-processors an 80-bit floating-point add (FADD.X FP0,FP1 for example) is treated as any other 020 instruction would be (such as ADD.L D0,D1). With the 68000 or 68010 the 68881 must be treated as a peripheral---you need to trap and decode 68881 instructions and read/write to a port. The 68000 and the 68020 are not pin-for-pin compatible. An adaptor board is needed to replace the 68020/68881 directly with the Amiga's 68000. Or the 68020/68881 can be interfaced to the Amiga via the expansion bus(this has advantages because you can run the 020/881 at a higher clock speed--- CHIP memory must however still run at the standard 7.16 MHz rate). As far as software goes there is only one compiler right now that generates code for the 68881 with the Amiga and thats Absoft's Fortran 77 68020/68881 Amiga compiler ($495). Also Quelo is beta testing an Amiga 68020/68881 assembler. I have a Quelo 68020 Amiga assembler at this time. Hope soon to get the 68881 update. There may be others, but I'm aware of only one company that makes 68020/68881 adaptor boards and expansion boxes for the Amiga and thats Computer System Associates , San Diego, Calif, (619) 566-3911.... If there is any interest I will upload benchmark comparison tests I've conducted with these Amiga 68020/68881 processors and other systems. Al Aburto
tsub@pnet02.cts.com (Tom Wang) (01/22/89)
How does the Commodore A2620 (68020, M68881, and 68851 MMU) compare with the
Ronin Hurricane Board? It seems to me that they are both basicly the same
thing, except the Ronin is upgradable to 68030 and M68882 (am I correct?).
--
Tom C. Wang
UUCP: {ames!elroy, <backbone>}!gryphon!pnet02!tsub
INET: tsub@pnet02.cts.comme128-aw@kepler.Berkeley.EDU (me128 student) (01/23/89)
In article <11061@gryphon.COM> tsub@pnet02.cts.com (Tom Wang) writes: >How does the Commodore A2620 (68020, M68881, and 68851 MMU) compare with the >Ronin Hurricane Board? It seems to me that they are both basicly the same >thing, except the Ronin is upgradable to 68030 and M68882 (am I correct?). From what I've heard, the Commodore board is almost exactly 1/2 the speed of the Ronin board, due to the fact that it has 2 wait states and the Ronin has none. The Ronin '020 board, however, does not have an MMU, and therefore will not be able to run UNIX. The Ronin upgrade path is simply an 030 board which plugs into the 020 socket of their 020 board, and should work with any 020 board, not just their own. If you need to run UNIX, I suggest you call Ronin, as I don't want to say anything about anything they might be waiting to unveil. If you don't need UNIX, I'd strongly recommend the Ronin board. It's fast, extremely-well engineered, and a fine board all around. Don't even consider a CSA. Just my opinion, not necessarily that of my school, friends, professors, associates, or relatives. -Vince Lee
bader+@andrew.cmu.edu (Miles Bader) (01/23/89)
me128-aw@kepler.Berkeley.EDU (me128 student) writes: > In article <11061@gryphon.COM> tsub@pnet02.cts.com (Tom Wang) writes: > >How does the Commodore A2620 (68020, M68881, and 68851 MMU) compare with the > >Ronin Hurricane Board? It seems to me that they are both basicly the same > >thing, except the Ronin is upgradable to 68030 and M68882 (am I correct?). > > From what I've heard, the Commodore board is almost exactly 1/2 the speed of > the Ronin board, due to the fact that it has 2 wait states and the Ronin >... > If you don't need UNIX, I'd strongly recommend the Ronin board. It's fast, > extremely-well engineered, and a fine board all around. Don't even consider > a CSA. Didn't it come to light a while ago on this newsgroup that the Ronin board didn't support DMA to its on-board memory? Or have they fixed that by now? -Miles
lphillips@lpami.wimsey.bc.ca (Larry Phillips) (01/25/89)
In <27669@ucbvax.BERKELEY.EDU>, me128-aw@kepler.Berkeley.EDU (me128 student) writes: >From what I've heard, the Commodore board is almost exactly 1/2 the speed of >the Ronin board, due to the fact that it has 2 wait states and the Ronin >has none. Well, I've heard difefrent things than you about the Ronin vs. the 2620. The figure I hear most often is that the Ronin is 8% faster than the 2620. The MMU adds one wait state. >If you don't need UNIX, I'd strongly recommend the Ronin board. It's fast, >extremely-well engineered, and a fine board all around. Don't even consider >a CSA. There's one more reason to buy a 2620 in favour of a Ronin. The 2620's memory has DMA capability. -larry -- Frisbeetarianism: The belief that when you die, your soul goes up on the roof and gets stuck. +----------------------------------------------------------------------+ | // Larry Phillips | | \X/ lphillips@lpami.wimsey.bc.ca or uunet!van-bc!lpami!lphillips | | COMPUSERVE: 76703,4322 | +----------------------------------------------------------------------+
rokicki@polya.Stanford.EDU (Tomas G. Rokicki) (01/25/89)
> From what I've heard, the Commodore board is almost exactly 1/2 the speed of > the Ronin board, due to the fact that it has 2 wait states and the Ronin > has none. Nope. My Ronin board has one wait state. Ronin claims I can change an R/C time constant to run it at 0 wait states, but not only do I need 80ns RAMs to do it, they can't guarantee that it will work. Not very reassuring. I was under the impression that the Commodore card was one wait state as well, so I'd expect equivalent speeds. I'll post benchmarks if someone with a 2620 will. A not to be forgotten fact, though, is you can't DMA into Ronin memory, which slows my lightening-fast CDC Wren III/SCSI/DMA drive dramatically. > Just my opinion, not necessarily that of my school, friends, professors, > associates, or relatives. Just my experience. I'm getting a 2620 to replace my Ronin and am taking bids . . . -tom
daveh@cbmvax.UUCP (Dave Haynie) (01/26/89)
in article <11061@gryphon.COM>, tsub@pnet02.cts.com (Tom Wang) says: > > How does the Commodore A2620 (68020, M68881, and 68851 MMU) compare with the > Ronin Hurricane Board? It seems to me that they are both basicly the same > thing, except the Ronin is upgradable to 68030 and M68882 (am I correct?). Any properly designed 68881 interface will support the 68882, since they're 100% pin compatible. The A2620 comes with 2 megs of 32 bit DRAM, upgradable to a total of 4 megs. The A2620 memory supports DMA from expansion devices, which may result in hard disk performance 150%-200% that of a plain 68000 Amiga. Non-DMA performance will be increased somewhat too. The A2620 memory runs in 2 wait states (the extra one being because of the 68851 MMU). The Ronin lets you add on a board that'll support up to 4 megs. The Ronin memory add on is completely non-standard, and requires a special configuration program to add it in. With fast filesystem, a DMA driven hard disk, and the suggested Ronin configuration, your hard disk performance will be 50%-75% that of a plain 68000 Amiga. A non-DMA driven hard disk interface will go faster than on a plain 68000 with Ronin. The Ronin memory board runs in 1 wait state. Ronin also sells a 68020 to 68030 converter, which allows the use of a 68030 in a 68020 socket. I would expect it to work in just about any 68020 board, including the A2620, CSA, and LUCAS boards as well as the Ronin, though naturally it's mechanically designed to fit with the Ronin board. Such an adaptor does not let you use any of the faster features of the 68030, so, like the Mac IIx and most other 68030-dropped-into-a 68020 systems, you may see something around 15% increase in system speed with the 68030 adaptor. The main advantage of adding a 68030 adaptor to a Ronin board will be the 68030's MMU (already present on the A2620). > Tom C. Wang > UUCP: {ames!elroy, <backbone>}!gryphon!pnet02!tsub > INET: tsub@pnet02.cts.com -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy Amiga -- It's not just a job, it's an obsession
daveh@cbmvax.UUCP (Dave Haynie) (01/27/89)
in article <27669@ucbvax.BERKELEY.EDU>, me128-aw@kepler.Berkeley.EDU (me128 student) says: > In article <11061@gryphon.COM> tsub@pnet02.cts.com (Tom Wang) writes: >>How does the Commodore A2620 (68020, M68881, and 68851 MMU) compare with the >>Ronin Hurricane Board? It seems to me that they are both basicly the same >>thing, except the Ronin is upgradable to 68030 and M68882 (am I correct?). > From what I've heard, the Commodore board is almost exactly 1/2 the speed of > the Ronin board, due to the fact that it has 2 wait states and the Ronin > has none. Where did you hear this? It's complete bunk! The Ronin memory board runs with 1 wait state, the A2620 with 2. They're both using the same speed DRAM, but the A2620's MMU requires an extra wait state for reliable operation. On the other hand, the Ronin memory doesn't accept DMA, so if you're using a DMA driven hard disk, your hard disk performance can drop to 50% or worse of what you get with the plain 68000, depending on system load and display resolution. The A2620 allows 16 bit DMA devices to access it's memory, so most DMA driven devices run significantly faster on an A2620. If you're using non-DMA devices like the GVP hard disk controller, the Ronin, A2620, CSA, etc. boards with 32 bit memory will all give you a performance increase, though not to the same degree as DMA into 32 bit RAM. > If you need to run UNIX, I suggest you call Ronin, as I don't want to > say anything about anything they might be waiting to unveil. I really doubt a 3rd party the size of Ronin will try to market their own UNIX. It's possible that they could find a way to make the Commodore UNIX run on their board. Actually, the "run" part should be no problem, but early on in the startup process, Commodore UNIX is looking for some things that are A2620 specific and would have to be emulated on a 3rd party board. > If you don't need UNIX, I'd strongly recommend the Ronin board. It's fast, > extremely-well engineered, and a fine board all around. The construction struck me as a bit cheap, but the thing worked just fine in several A2000's I tried, and all three boards looked well laid out. TINAR, but if you want one, I wrote a review of the Ronin board, with both '020 and '030, lots of benchmarks, etc, last fall for Amiga Sentry. If you're considering a Ronin board, you should check this out. > Just my opinion, not necessarily that of my school, friends, professors, > associates, or relatives. > -Vince Lee -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy Amiga -- It's not just a job, it's an obsession
sjm@well.UUCP (Stephen Moehle) (01/27/89)
In article <5795@cbmvax.UUCP> daveh@cbmvax.UUCP (Dave Haynie) writes: |The A2620 comes with 2 megs of 32 bit DRAM, upgradable to a total of 4 |megs. The A2620 memory supports DMA from expansion devices, which may |result in hard disk performance 150%-200% that of a plain 68000 Amiga. |Non-DMA performance will be increased somewhat too. The A2620 memory |runs in 2 wait states (the extra one being because of the 68851 MMU). |-- |Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" | {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy | Amiga -- It's not just a job, it's an obsession Does the number of wait states depend on whether the 68851 is enabled or not? Are there alway two wait states? Stephe {ucbvax,pacbell,hplabs}!well!sjm or well!sjm@lll-winken.arpa "You heard the weirdo man. What is truth?"
tope@enea.se (Tommy Petersson) (01/28/89)
In article <27669@ucbvax.BERKELEY.EDU> me128-aw@kepler.Berkeley.EDU (me128 student) writes: :In article <11061@gryphon.COM> tsub@pnet02.cts.com (Tom Wang) writes: ::How does the Commodore A2620 (68020, M68881, and 68851 MMU) compare with the ::Ronin Hurricane Board? It seems to me that they are both basicly the same ::thing, except the Ronin is upgradable to 68030 and M68882 (am I correct?). : :From what I've heard, the Commodore board is almost exactly 1/2 the speed of :the Ronin board, due to the fact that it has 2 wait states and the Ronin :has none. : :The Ronin '020 board, however, does not have an MMU, and therefore will not :be able to run UNIX. The Ronin upgrade path is simply an 030 board which :plugs into the 020 socket of their 020 board, and should work with any :020 board, not just their own. : :If you need to run UNIX, I suggest you call Ronin, as I don't want to :say anything about anything they might be waiting to unveil. : :If you don't need UNIX, I'd strongly recommend the Ronin board. It's fast, :extremely-well engineered, and a fine board all around. Don't even consider :a CSA. : :Just my opinion, not necessarily that of my school, friends, professors, :associates, or relatives. : :-Vince Lee I'm interested in buying a '20 or '30 card for my B2000 (PAL) machine. I would appreciate info about Ronin and other boards, including speed, RAM, coprocessors, quality, price, availability, compatibility problems aso. I would also like to have phone numbers and/or addresses to reliable sources to buy the stuff from. Thanks in advance, Tommy P.
mp1u+@andrew.cmu.edu (Michael Portuesi) (01/28/89)
rokicki@polya.Stanford.EDU (Tomas G. Rokicki) writes: > > Does the Ronin board fit in the processor slot on an A2000? Is the Ronin > > board the same one that used to Finally Tech. or some such? > > Yes, and yes. And I get 2670 drystones (dry from fish disk 1) . . . > > -tom which means absolutely nothing unless you also tell us what compiler you used. -- Michael Portuesi / Information Technology Center / Carnegie Mellon University INET: mp1u+@andrew.cmu.edu / BITNET: mp1u+@andrew UUCP: ...harvard!andrew.cmu.edu!mp1u+ "I'm very sorry, Master, but that WAS the backup system" -- Slave
rokicki@polya.Stanford.EDU (Tomas G. Rokicki) (01/28/89)
> > Yes, and yes. And I get 2670 drystones (dry from fish disk 1) . . . > > which means absolutely nothing unless you also tell us what compiler > you used. Read the message again. I used the file `dry' from Fish Disk 1. I *did not* recompile. (Had I, using Lattice 5.0, I probably would get around 4600 dhrystones. When I get home I'll try this trick.) -tom
mcr@julie.UUCP (Michael Richardson) (01/31/89)
>:The Ronin '020 board, however, does not have an MMU, and therefore will not >:be able to run UNIX. The Ronin upgrade path is simply an 030 board which >:plugs into the 020 socket of their 020 board, and should work with any >:020 board, not just their own. >: >:If you need to run UNIX, I suggest you call Ronin, as I don't want to >:say anything about anything they might be waiting to unveil. >: >:If you don't need UNIX, I'd strongly recommend the Ronin board. It's fast, >:extremely-well engineered, and a fine board all around. Don't even consider >:a CSA. >: >>:-Vince Lee I quite agree that without a MMU, Unix just isn't realistic - but what bugs me about all the '20 and '30 boards is that I can't `detach' them from the Zorro II bus and let the 68000 do I/O, interupt processing, maybe even the low level buffer caches... AND run an X server! (As an AmigaDOS task of course...) One of the CA guys said something to effect that they wished that all this work on Unix would go into 1.4: It seems to me that AmigaDOS with an MMU just isn't realistic (and Unix without isn't) It would be nice if AmigaDOS moved closer and closer to Unix, with vnodes or GFS (Generic File System - a la DEC) and real processes. Keep most of Exec and Intuition but junk AmigaDOS itself. Also - as long as we are doing this, don't make it '020 or Motorola specific, (beyond msb first) and replace Janus too... Who'd want to run MS-DOS on your '386 board anyway? I WANT the features of Unix. I would like to keep a faster interrupt sensitive OS. `People are Real-Time events too' -- :!mcr!: Michael Richardson Amiga v--------+ UUCP: uunet!attcan!lsuc!nrcaer!julie!mcr | INTERNET mcr@doe.carleton.ca Fido: Michael Richardson @ 1:163/109.10<--+ Alter @ 7:483/109.10
c60a-1fy@web-2a.berkeley.edu (Anon) (02/04/89)
I must apologize for some figures I gave concerning the ronin board, as I found they were not for the standard ronin board, but for one clocked at a much higher clock rate. The new ronin mem board which has been shipping for 3 or 4 months does indeed run at zero wait states, so if you have one wait state, it is an earlier board. I don't know what upgrade policy they may have. I was told by them that you can expect an overall 20% increase using their new mem board vs. the commodore board. I was also told by a friend working for them that their present 030 board will plug into their 020 board, a CSA board, and even a MACII, but NOT a commodore board due to a design flaw in commodore's accelerator. As far as future products, UNIX (AMIX), etc, they are of course working on stuff. I have been told things, but promised to secrecy, so you should call them 'cause I don't know what is and is not ok for me to divulge. remember. you didn't hear it from me. -vince lee
daveh@cbmvax.UUCP (Dave Haynie) (02/07/89)
in article <19939@agate.BERKELEY.EDU>, c60a-1fy@web-2a.berkeley.edu (Anon) says: > The new ronin mem board which has been shipping for 3 or 4 months does indeed > run at zero wait states, so if you have one wait state, it is an earlier > board. Perhaps so, but if that's the case, they really screwed themselves. The board I reviewed for Amiga Sentry was in fact a 1 wait-state design. That review was only last fall, and you'd certainly think, if someone's going to send you a thing for review, they'd send you the latest. Oh, well, it really doesn't matter anyway. > I was told by them that you can expect an overall 20% increase using their > new mem board vs. the commodore board. That would be a fair statement for memory bound things, all else being equal. But all things aren't equal. For instance, if you run my latest SetCPU program on an A2620, you can get things like layers and FFP bound stuff running about 200%-400% faster than on any Ronin '020 board. But it doesn't really matter. > I was also told by a friend working for them that their present 030 board > will plug into their 020 board, a CSA board, and even a MACII, but NOT a > commodore board due to a design flaw in commodore's accelerator. Ronin has an unfortunate habit of equating things that don't understand or design decisions they don't like with "Commodore design flaw". They're wrong. For more examples of this, see my Sentry article. > As far as future products, UNIX (AMIX), etc, they are of course working on > stuff. I have been told things, but promised to secrecy, so you should > call them 'cause I don't know what is and is not ok for me to divulge. It would be possible to run AMIX on a third party '030 board. But it won't be easy. In addition, this was a conscious decision, there's no reason AMIX couldn't have been made specific to the A2620 if we wanted it to be so. Ronin doesn't have the resources to do their own UNIX port, obviously, but if they get a hardware guy that'll stop trying to do all the dirty work in software (see the '030 tower for an example of this -- they can't use the data cache on non-Ronin memory), the could get AMIX up on their board. Of course, they don't yet have AMIX. > -vince lee -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy Amiga -- It's not just a job, it's an obsession