stevef@well.UUCP (Steven Robert Fordyce) (07/12/89)
Since there has been so much interest lately in the barrage of accelerator choices that have suddenly come to the ST market, we thought that a comparison of them all, side by side, was in order. Here are the results! SOFTWARE: The software that was used is QuickIndex 1.4. The Mandelbrot test of the FPU was one distributed on Atari's FPU004 development disk. HARDWARE: Atari-8: Atari ST Mega2, unmodified, hi-res monochrome, no blitter JATO-16: Atari ST 1040, JATO-16, med-res color, no blitter FAST-T16: Atari ST Mega4, FAST-T16, hi-res monochrome, no blitter CMI: Atari ST 520, CMI Processor Accelerator: blitter; 16 MHz Math FPU; FastROM, hi-res monochrome RESULTS: The table below contains the tests of the different accelerators. The tests of the JATO and FAST boards were as submitted to GEnie by two different posters. All values are represented as percentages. The tables between the results -------------| are the difference between the CMI Accelerator and the system | results listed in the column to the left. The tests are | listed in the approximate order they would be used in a normal | application. The number in parentheses next to the test name is | an importance value that is used to obtain the weighted average | of each test. | CMI Proc Accel Atari-8 V JATO-16 FAST-T16 Price($U.S.) $299 $339 $99 $399/299? (Blitter) CPU memory (3) 101 101 100 100 130 GEM draw (3) 220 279 158 151 187 BIOS text (3) 201 210 143 162 177 GEMDOS I/O (2) 112 112 112 112 75 DMA read (2) 161 161 161 161 161 CPU divide (1) 183 183 100 183 202 CPU shifts (1) 180 180 100 179 205 Mandelbrot (FPU,1) 153 976 100 111 ** --- --- --- --- --- Weighted Average: 164 228 128 141 157 * The numbers for the stock ST are not all 100% because of the difference in speed between 1.1 and 1.2 of the operating system. ** Speed for the FAST board was not available, but it's speed is probably about the same as our board without the FPU or less, and wouldn't appreciably affect it's weighted average. COMPARISON: (CMI vs Atari) (CMI vs JATO) (CMI vs FAST) Weighted Avg: 28% 14% 4% Weighted Average (blitter): 38% 26% 13% Weighted Avg (FPU & blitter): 78% 62% 45% ^ |___ CMI is this much faster than the competition. FEATURES: CMI Proc Accel Atari JATO-16 FAST-T16 Blitter support !! YES !! Mega only NO NO Math FPU support !! YES !! NO NO NO FastROM mode !! YES !! NO NO NO Toggle 8/16 !! YES !! NO YES YES Total Compatibility !! YES !! YES NO ??? (Cache On/Off) Expansion connector !! YES !! Mega only NO NO Cache RAM NO NO NO YES Since we were not able to test all products at the same time with the same equipment these results will have to be labeled "unofficial". Some people may argue with the manner or purpose of the weighted average. We feel very strongly that this was chosen in a fair manner. If the results are not weighted, a person will not be able to get a realistic picture of OVERALL system performance. Some people may question the "obvious bias" that this test has, leaning towards the CMI Processor Accelerator. Quite frankly, we don't give a damn. We think that without question the CMI Processor Accelerator is by far the most superior product that you can purchase to enhance your ST. Richard N. Rodgers, President Creative Microsystems Inc. 11952 SW 90th Court Tualatin, OR 97062 (503) 691-2552 If you don't like my test, give me a call and tell me why. Or better yet, do your own test and post the results. If you want to place an order, or get product information: Lilliane Carter, (Customer Service & Support / ST Dealer & End User sales)
gl8f@astsun.astro.Virginia.EDU (Greg Lindahl) (07/13/89)
In article <12662@well.UUCP> stevef@well.UUCP (Steven Robert Fordyce) writes: > Since there has been so much interest lately in the barrage of accelerator > choices that have suddenly come to the ST market, we thought that a > comparison of them all, side by side, was in order. If you run Minix, your mileage will vary... In particular, the CMI board has its "Fast ROM" feature, which won't help Minix, while the FAST-T16 board's cache should help Minix same as before, as long as you don't use self-modifying code (presumably). It's nice to see benchmarks; the only thing I could have wished for was (1) a benchmark of a REAL program with loops that does a fair amount of computing in registers with low cycle-count instructions (not shifts or floating point), and (2) less marketing hype... ------ Greg Lindahl gl8f@virginia.edu I'm not the NRA.
lharris@gpu.utcs.utoronto.ca (Leonard Harris) (07/13/89)
As requested, here are the quick index performance values obtained using the Datafree 16Mhz Accelerator board. My system is a 520ST upgraded to 4 megabytes with the original 1985 tos 1.0 ROMS. Quick Index 1.5 was used with drive C being a 30 Meg Supra harddisk. Settings were Mono, tos 1.0, drive C. No upgrade Datafree board Datafree + Turbost v1.5 ------------ -------------- ----------------------- CPU memory 100 103 103 CPU register 100 102 102 CPU divide 100 185 185 CPU shifts 100 197 197 DMA read 5038 5038 5038 Gemdos I/O 110 118 118 Disk RPM 3604 3604 3604 Bios text 100 126 330 Bios string 100 122 763 Bios scroll 100 115 181 Gem draw 100 120 149 Using applications like ARC, I get about a 35% increase in speed. Spectre (actually Mac programs) for some reason gets a 45-50% increase in speed. The board is only $99 US, is extremely small ( the exact same size as a 68000) uses a custom chip, 100% compatible AND AVAILABLE NOW!! (actually 2 weeks ago). Seems it is faster and cheaper than the CMI and Jato boards as well. Hope this helps /leonard
lharris@gpu.utcs.utoronto.ca (Leonard Harris) (07/13/89)
Sorry, I forgot - Datafree Industries is at: 391 Ormont Drive Weston Ontario M9L 1N8 (416) 741-9825 / /leonard
bds@lzaz.ATT.COM (B.SZABLAK) (07/13/89)
In article <1732@hudson.acc.virginia.edu>, gl8f@astsun.astro.Virginia.EDU (Greg Lindahl) writes: > It's nice to see benchmarks; the only thing I could have wished for was > (1) a benchmark of a REAL program with loops that does a fair amount > of computing in registers with low cycle-count instructions (not shifts > or floating point), and (2) less marketing hype... I will repeat my earlier suggestion: Why not run the MSDOS benchmarks on the ST using PC Ditto I and the accelerator. A comparison of the SI values would be interesting, as would the BYTE benchmarks...
jlemon@cory.Berkeley.EDU (Jonathan Lemon) (07/13/89)
In article <1989Jul12.233551.3813@gpu.utcs.utoronto.ca> lharris@gpu.utcs.UUCP (Leonard Harris) writes:
%As requested, here are the quick index performance values obtained using
%the Datafree 16Mhz Accelerator board.
%
%My system is a 520ST upgraded to 4 megabytes with the original 1985 tos 1.0
%ROMS. Quick Index 1.5 was used with drive C being a 30 Meg Supra harddisk.
%Settings were Mono, tos 1.0, drive C.
%
% No upgrade Datafree board Datafree + Turbost v1.5
% ------------ -------------- -----------------------
[ values deleted ]
What about the performance of TurbotST w/o the Datafree board?
--
Jonathan ...ucbvax!cory!jlemon or jlemon@cory.Berkeley.EDU