Path: utzoo!attcan!uunet!husc6!think!linus!eachus From: eachus@linus.mitre.org (Robert I. Eachus) Newsgroups: comp.sys.amiga Subject: Re: ..Most Excellent Message-ID: Date: 1 Jun 90 22:21:11 GMT References: <18017@snow-white.udel.EDU> <1990May7.184749.20324@digibd> <18283@well.sf.ca.us> Sender: usenet@linus.mitre.org Organization: The Mitre Corporation, Bedford, MA Lines: 57 In-reply-to: farren@well.sf.ca.us's message of 1 Jun 90 09:19:27 GMT In article <18283@well.sf.ca.us> farren@well.sf.ca.us (Mike Farren) writes: eachus@linus.mitre.org (Robert I. Eachus) writes: > I have seen a 3000 and spent a couple of hours hands on. > A very nice machine, but the price is the real killer for now. >That's the only way in which it currently obsoletes the A2500/30. >The 16/2 (32-bit FAST/32-bit CHIP) Meg of on-board space is probably >the biggest difference from the 2500/30 (4/1 Meg 32-bit FAST/16-bit >CHIP) from a current performance point of view. As will undoubtedly be pointed out several times over, that is A biggest difference, but not THE biggest difference. The Zorro III bus provides another biggest difference, and the 32-bit DMA provides Yet Another biggest difference. The net result is significantly better speed than a 2500/30, especially on disk-intensive applications. Two points. First there is a difference between "is better than" and "makes obsolete." The 3000 is better than the 2500/30 in a lot of ways, but if you have a 2500/30 or a GVP Impact board, there isn't enough difference TODAY to justify replacing it with a 3000. When 4x1 Meg SCRAMs are available in quantity, or when someone has a 68040 board for the 3000, or even when Zorro III boards with significant performance improvements show up then it may be a different story. But for now my impression is: don't buy a 2500/30, they cost too much, unless you need more (yeech!) IBM slots than in a 3000. I will probably put a GVP Impact or 2630 in my 2000 at work, however, if I find a good price. Second, I managed to get my hands on a 3000 (actually two) with 80ns 4x256K SCRAMs installed. It do make a difference. In fact the 3000 at 25 MHz slightly outperformed a GVP Impact at 25 MHz (your milage may vary...) The numbers are not statistically significant (7.73 to 7.65 MIPS) but this puts the 3000 way past any Sun 3 I have measured, and close to the SPARCstation. (I get 8.6 of MY MIPS on a SPARCstation...other people whose mix is less character oriented get other numbers.) Incidentally, my measurements come very close to those in both Amigo Times 1.10 and Amiga Software & Information (Volume 3 Issue 2 May 1990) on machines we both tested...I get a slighty lower number for the 2000 (0.79 MIPS) but all in all a very close match. This seems to be normal for the 680x0 family, while the RISCs and the Intel 80x86 families are both a lot more sensitive to instruction mix. For the record, AmigoTimes has the Impact A4000 at 33MHz at 10.34 MIPS, and ASI has the Impact A3001 at 8.32 MIPS. Not shabby at all. (For those who perfer Dhrystones, the AmigoTimes figures are A2000: 769 (which seems to indicate Dhrystone 2.1 done right), A2620: 2380, Impact A3001 (25MHz): 6250, Impact A4000: 7142. -- Robert I. Eachus Amiga 3000 - The hardware makes it great, the software makes it awesome, and the price will make it ubiquitous.