blf@wookumz.gnu.ai.mit.edu (Bill Frandsen) (05/11/91)
ICD AdSpeed Test Results: (4 Megs fast Ram, 1 Meg Chip Ram, A2000) Other hardware: Supra WordSync Hard Card (ST-157-N1 Drive), Supra 4/8 mem. All tests were done with fast memory on, speed increase w/o fast ram is minimal (between 5 and 10 percent only). AIBB3 Test: 7.32MHz - 14.80MHz (% incr) ----------- ------- - -------- -------- WritePixel: 45.74 - 32.38 ( 41.3%) DhryStone: 1009d/s - 1736d/s ( 72.2%) Matrix: 115.20 - 63.68 ( 80.9%) Fibonacci: 98.44 - 58.76 ( 67.6%) Sieve: 34.60 - 18.36 ( 88.5%) Sort: 20.34 - 10.58 ( 92.2%) LLines: 22.02 - 21.30 ( 3.4%) IMath: 78.86 - 42.22 ( 86.8%) FMath: 90.72 - 52.08 ( 74.2%) Savage: 428.72 - 230.92 ( 85.7%) FMatrix: 98.10 - 55.52 ( 76.7%) BeachBall: 1075.02 - 586.40 ( 83.3%) Kramden SI: 7 Mhz - 14 Mhz (% incr) ----------- ------- - -------- -------- Integer: 1.0 - 2.0 (100.0%) Floating Pt: 1.0 - 2.0 (100.0%) (IBM INT:) 3.3 - 6.6 (100.0%) (IBM FP:) 7.7 - 14.6 ( 89.6%) SysInfo2.0: 7.25Mhz - 14.5Mhz (% incr) Chip Only: --------------- ------- - ------- -------- ---------- A500 STANDARD 1.33 - 2.55 ( 55.0%) 1.00 1.07 B2000 EXTRA RAM 1.00 - 1.94 ( 94.0%) .77 .81 B2000 GVP A3001 .12 - .24 (100.0%) .09 .10 A2500 A2620 .38 - .74 ( 94.7%) .29 .31 A3000 25MHz .17 - .32 ( 88.2%) .13 .13 IBM PC/XT 4.34 - 8.32 ( 91.7%) 3.31 3.49 LZ 1.01: 7 MHz - 14 MHz (% incr) --------- ----- - ------ -------- Create: 3:04 - 5:10 ( 68.5%) Dissolve: 1:28 - 2:31 ( 71.6%) (Tests were done from Fast Ram to Fast Ram) (using all three Test Drive II course scenery directories) (601299 orig, 350476 arcd, 41% compression, 86 files) The A64 Package: C64 - 7 MHz - 14 MHz (% incr) ---------------- --- - ----- - ------ ------- 1000 iterations 60 - 139 - 77 ( 80.5%) 10000 iterations 600 - 1455 - 806 ( 80.5%) Program was: ^^^ A real C-64, x/60'th of a sec. 1 TI$="000000":FORA=0TO 999:NEXT:PRINT TI 2 TI$="000000":FORA=0TO9999:NEXT:PRINT TI This is a little better list of the real results of the increase in speed as compared to plain "count" tests. My package included instructions about using with the Xetec HD Interfaces, and the GVP series controllers. It will NOT work with the bridgeboard, in fact if you have a hard drive WITH a Bridgeboard, and boot MS-DOS from a hard drive partition, you must have the jumper set at 7MHz otherwise it will lock up the system when booting. To use the Bridgeboard, you must turn off the acclerator the whole time you are using the Janus software by using the software switch program (9k, loaded all the time as a process). For $215 + $6 shipping, the best place to buy currently (They have them in stock, unlike many of their competitors) is Manta, in New Jersey. Arrived to North Dakota via UPS Ground in 8 days. Their number is: 1-(800)-477-7706 - See ads in Amiga World.
bmccnnll@vax1.tcd.ie (05/15/91)
Just how many Dhrystones can the Amiga do? I have an A500 with 1Mb Chip RAM and 2Mb Fast RAM, and AIBB2 tells me 661, AmigaBench tells me 1430, and now apparently AIBB3 gives 1009 on this person's A2000. What the hell is going on??? If anyone has any _reliable_ figures, please let me know (e-mail is preferable). I'm also interested in comparing an Amiga 3000 to an Archimedes speed-wise... Barry.
lkoop@pnet01.cts.com (Lamonte Koop) (05/18/91)
bmccnnll@vax1.tcd.ie writes: >Just how many Dhrystones can the Amiga do? I have an A500 with 1Mb Chip RAM and >2Mb Fast RAM, and AIBB2 tells me 661, AmigaBench tells me 1430, and now >apparently AIBB3 gives 1009 on this person's A2000. What the hell is going >on??? You must remember that the dhrystone benchmark is dependent upon several factors...none the least of which is how the code was compiled, the language used, and the optimization level. AmigaBench was written entirely in assembly, and optimized as such, therefore the higher number. When I wrote AIBB v2.0, I was in some kind of funky Modula-2 kick, so you see a lower number, as I have yet to see an M2 compiler for the Amiga which optimizes as well as the popular C compilers. AIBB v3.0 saw my return to C, (SAS/C 5.10a to be exact), and thus the higher number over AIBB 2.0. As for the code, and optimization, AIBB v3.0 is uses a version of the dhrystone program VERY close to the popular dhrystone 2.1 floating around. Optimization wise, the code uses register variables where possible, and of course register function arguments. It was not run through SAS/C's global optimizer as AIBB was meant to compare system performance...having GO eliminate chunks of code as 'dead' eliminates this purpose. >If anyone has any _reliable_ figures, please let me know (e-mail is >preferable). I'm also interested in comparing an Amiga 3000 to an Archimedes >speed-wise... > I would have e-mailed this, but I've been having mail bounce all over the net lately, an besides, perhaps someone else is interested in this as well. >Barry. LaMonte Koop Internet: lkoop@pnet01.cts.com ARPA: crash!pnet01!lkoop@nosc.mil UUCP: {hplabs!hp-sdd ucsd nosc}!crash!pnet01!lkoop "It's a dog-eat-dog world...and I'm wearing Milk Bone underwear"--Norm