evil@bbn.com (DavE Nye) (05/03/90)
Hello World... I got to play for an hr with the Amiga 3000 today... WOW...that is one slick machine!!! Three Cheers for the Group at CBM... While I was playing, I decided to do a non-scientific benchmark.. I loaded MandelVroom 2.0 in my stock 2000 and generated a screen size picture of the 'Aug_85Cover' (I think that is the name) and it took 9.47 min to complete. I took the MandelVroom disk down to the Bit Bucket and tested the same thing on the A3000 using the FFP mode. It took 1.21 min to complete!!!! I was very impressed as this is what I do alot of. Now that I have an Amiga 3000 within walking distance I'd love to try some other benchmarks but I don't have any. Can someone tell me where I can find some? (ie. Drystone 2.0 etc..) The Bit Bucket also has a 2500/30 that I'd like to compare the 3000 with. Thanks, -=TheEvilAmigaHacker ps. I know most people say that most benchmarks are a waste of time but I'd be interested and I would let the net know of my findings... ------------------------------------------------------------------------------- --- --------- ----------------------- DavE Nye, a.k.a. -=TheEvilOne - --- --------- ----------------------- evil@bbn.com THE Evil Empire - --- --- --- ----- evil%bbn.com@relay.cs.net - --- --- --- --- --- --- ----- ...!harvard!bbn!dnye - --- --- --- --- --- --- ----- BBN, Inc., Cambridge, Mass. - --- --- --- --- -Inc- Employers computer, MY opinion.. - -------------------------------------------------------------------------------
bryan@cs.utexas.edu (Bryan Bayerdorffer @ Wit's End) (05/03/90)
In article <55662@bbn.COM> evil@mikey.bbn.com (DavE Nye) writes:
=-
=- While I was playing, I decided to do a non-scientific benchmark..
=-I loaded MandelVroom 2.0 in my stock 2000 and generated a screen size
=-picture of the 'Aug_85Cover' (I think that is the name) and it took
=-9.47 min to complete. I took the MandelVroom disk down to the Bit Bucket
=-and tested the same thing on the A3000 using the FFP mode. It took
=-1.21 min to complete!!!! I was very impressed as this is what I do alot of.
Yeah, and that's WITHOUT using the 882. A new version of MV (2.1) has
been promised that uses the 882. If you want it soon, send irritating mail to
kevin@amdahl.uts.amdahl.com :-)
kevin@uts.amdahl.com (Kevin Clague) (05/07/90)
In article <498@mohawk.cs.utexas.edu> bryan@cs.utexas.edu writes: >In article <55662@bbn.COM> evil@mikey.bbn.com (DavE Nye) writes: >=- >=- While I was playing, I decided to do a non-scientific benchmark.. >=-I loaded MandelVroom 2.0 in my stock 2000 and generated a screen size >=-picture of the 'Aug_85Cover' (I think that is the name) and it took >=-9.47 min to complete. I took the MandelVroom disk down to the Bit Bucket >=-and tested the same thing on the A3000 using the FFP mode. It took >=-1.21 min to complete!!!! I was very impressed as this is what I do alot of. > > Yeah, and that's WITHOUT using the 882. A new version of MV (2.1) has >been promised that uses the 882. If you want it soon, send irritating mail to >kevin@amdahl.uts.amdahl.com :-) FFP mode is the worst case. It does floating point all in software. IEEE mode at least uses the 881 (or 882). MandelVroom 2.0 has a bug in it's 020/881 menu item. You can use newzap to patch the binary. Patch sector 221 byte 0a1 to a value 0f (it was 1e). This should allow 020/881 to run on 020/881 and 030/882 machines. Rick Sterling from Commodore called me about this one. He provided the patch and had it running on a 3000 before they were released. He said it was extremely fast. This and other problems are being fixed in MV2.1. Kev -- UUCP: kevin@uts.amdahl.com or: {sun,decwrl,hplabs,pyramid,seismo,oliveb}!amdahl!kevin DDD: 408-737-5481 USPS: Amdahl Corp. M/S 249, 1250 E. Arques Av, Sunnyvale, CA 94086 [ Any thoughts or opinions which may or may not have been expressed ] [ herein are my own. They are not necessarily those of my employer. ]
evil@bbn.com (DavE Nye) (05/10/90)
This is the Amiga 3000 25 MHZ. running WB 1.3 Amiga System Information 1.0 Gerald Brandt 21-Jul-88 EXEC reports a 68020 CPU and a 68881 Math Co-Processor Available memory is 1147 KBytes ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ since this is an old program I don't think it knows about an '030, and it has the memory wrong as well. CPU Timings Copy Blocks in CHIP memory is 206 ticks vs 621 301 % Copy Blocks in FAST memory is 104 ticks vs 621 597 % Move from register to CHIP memory is 171 ticks vs 855 500 % Move from register to FAST memory is 99 ticks vs 855 863 % Move from data reg to data reg is 69 ticks vs 476 689 % Unsigned divide in data register is 184 ticks vs 1812 984 % Unsigned multiply in data register is 130 ticks vs 856 658 % Unsigned multiply in CHIP memory is 208 ticks vs 911 437 % Unsigned multiply in FAST memory is 152 ticks vs 911 599 % Stack manipulation ( MOVEM.L ) is 105 ticks vs 598 569 % Total percentage speed: 619 % This is now running on A3000 25 MHZ running WB 2.0 beta. Amiga System Information 1.0 Gerald Brandt 21-Jul-88 EXEC reports a 68020 CPU and a 68881 Math Co-Processor Available memory is 1103 KBytes ^^^ again wrong.. I don't have the code to this to find out..why. CPU Timings Copy Blocks in CHIP memory is 211 ticks vs 621 294 % Copy Blocks in FAST memory is 105 ticks vs 621 591 % Move from register to CHIP memory is 173 ticks vs 855 494 % Move from register to FAST memory is 99 ticks vs 855 863 % Move from data reg to data reg is 70 ticks vs 476 680 % Unsigned divide in data register is 187 ticks vs 1812 968 % Unsigned multiply in data register is 132 ticks vs 856 648 % Unsigned multiply in CHIP memory is 212 ticks vs 911 429 % Unsigned multiply in FAST memory is 154 ticks vs 911 591 % Stack manipulation ( MOVEM.L ) is 3990 ticks vs 598 676 % Total percentage speed: 623.4 % When I tested the A2500 with the '020 I got a 350% increse. I could not test it with the '030 in it as they had it being fixed. Thanks to John Sparks for sending this program to me. If anyone else has any other benchmarks that they would like me to do send them to me and I'll do my best to get them done and post. -=TheEvilAmigaHacker ------------------------------------------------------------------------------- --- --------- ----------------------- DavE Nye, a.k.a. -=TheEvilOne - --- --------- ----------------------- evil@bbn.com THE Evil Empire - --- --- --- ----- evil%bbn.com@relay.cs.net - --- --- --- --- --- --- ----- ...!harvard!bbn!dnye - --- --- --- --- --- --- ----- BBN, Inc., Cambridge, Mass. - --- --- --- --- -Inc- Employers computer, MY opinion.. - -------------------------------------------------------------------------------