svermeulen%Janus.MRC.AdhocNet.CA@UNCAEDU.BITNET (Steve Vermeulen) (08/07/88)
Regarding the 68030...
+From: cthulhu@athena.mit.edu (Jim Reich)
+Mac compatibility is pretty unlikely, considering that Jobs isn't with Apple
+anymore... and as for the 68030, I forgot to say to buy a CSA 68030 board...
+(BTW, has anyone out there tried one? Does it use all the 68030's features?
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+Does it support Memory Management compatibly with the A2620?)
+ -- Jim
While at AmiExpo in January I pestered the CSA sales rep until he allowed
me to do a crude compiling benchmark on his 020 and 030 board. I had
brought with me a copy of Manx C 3.4 and a LARGE piece of source code.
On both the 030 and 020 machine I put all the files into RAM: and ran
the same short compile and link script. Note the only fast ram on both
machines was 32 bit wide static ram. What I wanted to know was if the
030 card was faster than the 020 card for the work I do a lot of.
Anyway, when we ran it (timing by SEIKO) both the 020 and 030 did the
compile and link in exactly (to less than 1%) the same time. I did
observe the MandleBrot program being run on both systems, in this case
a 68030+68882 ran about 1.7 times as fast as a 68020+68881.
Note that the 030 board that I tried was a pre-production model and
---MOST IMPORTANTLY--- did NOT run with the 030's data cache enabled,
I don't know if the production version of CSA's 030 board does allow
the data cache to be used.
BTW, the 020 and 030 were about 3 times as fast as a stock Ami 2000 with
no true fast ram.
Stephen Vermeulen
Author: Express Paint
Chairman: AMUC.daveh@cbmvax.UUCP (Dave Haynie) (08/09/88)
in article <880806111207.003@Janus.MRC.AdhocNet.CA>, svermeulen%Janus.MRC.AdhocNet.CA@UNCAEDU.BITNET (Steve Vermeulen) says: > Regarding the 68030... > +From: cthulhu@athena.mit.edu (Jim Reich) > +Mac compatibility is pretty unlikely, considering that Jobs isn't with Apple > +anymore... Do Marketing and Head Honcho types at Apple determine hardware and software compatibility all that much? Given that the 68020 runs a decent portion of Apple software in a Mac II, the 68030 would be expected to do pretty similarly. Turning on the data cache would cause the most trouble with self-modifying code in the Apple Mac world, since they really aren't concerned with DMA or true context switching. > +Does it support Memory Management compatibly with the A2620?) Yup. All 68030s have an on-chip MMU that's a subset of the 68851 MMU used on the A2620, and Commodore-Amiga only supports the 68030 compatible subset. > While at AmiExpo in January I pestered the CSA sales rep until he allowed > me to do a crude compiling benchmark on his 020 and 030 board.... > Anyway, when we ran it (timing by SEIKO) both the 020 and 030 did the > compile and link in exactly (to less than 1%) the same time. That's pretty much what you'd expect in a plug-replacement setup with the 68030 data cache disabled. In the ideal situation, a 68030 will go just over twice as fast as the 68020+68851 combination at a given clock speed, with data cache off. Even faster, of course, with it on. However, that extra speed is fully dependent on memory architecture -- with the same memory system, and the data cache off, you won't be able to tell the difference. Even the ability to run the data cache depends on a few simple but important elements in the memory system, and any adaptor board that doesn't take these elements into account won't allow data caching. The Amiga adds an extra twist to this, in that CHIP memory and registers must never be cached, and that DMA devices on the expansion bus must really account for data caching in their device drivers. > Stephen Vermeulen > Author: Express Paint ^^^^^^^^^^^^^ Saw this at C Ltd's Suite at Comdex. Nice program. Who says DPaint has no competition? > Chairman: AMUC. -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {ihnp4|uunet|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy "I can't relax, 'cause I'm a Boinger!"