08/1997
Path: news.uh.edu!barrett From: per-espen.hagen@ffi.no (Per Espen Hagen) Newsgroups: comp.sys.amiga.reviews Subject: REVIEW: Blizzard 1230-II accelerator/RAM expansion for A1200 Followup-To: comp.sys.amiga.hardware Date: 31 May 1994 13:44:44 GMT Organization: The Amiga Online Review Column - ed. Daniel Barrett Lines: 365 Sender: amiga-reviews@math.uh.edu (comp.sys.amiga.reviews moderator) Distribution: world Message-ID: <2sff0c$g48@masala.cc.uh.edu> Reply-To: per-espen.hagen@ffi.no (Per Espen Hagen) NNTP-Posting-Host: karazm.math.uh.edu Keywords: hardware, accelerator, 68030, RAM, A1200, commercial Originator: barrett@karazm.math.uh.edu PRODUCT NAME Blizzard 1230-II BRIEF DESCRIPTION Highly expandable 68030/FPU/RAM/SCSI-II/etc. board for the Amiga 1200. AUTHOR/COMPANY INFORMATION Name: phase 5 digital products Address: Homburger Landstrasse 412 D-60433 Frankfurt/Main Germany Telephone: +49 69 5481844 LIST PRICE Starts at around GBP 250 for a version with a 68EC030 CPU at 40 MHz, no FPU, no RAM, and no SCSI. SPECIAL HARDWARE AND SOFTWARE REQUIREMENTS HARDWARE Amiga 1200 with unused trapdoor expansion slot. SOFTWARE None. COPY PROTECTION None. Well... the board is multi-layered (four or five layers?), so I wouldn't want to copy it anyway :-) . And copying the chips would probably take a few evenings too... The board is not hard drive installable, but you could probably install it into an 8" floppy disk jacket. :-) MACHINE USED FOR TESTING Amiga 1200. Toshiba 250 MB 2.5" IDE hard drive. Commodore 1940 multisync monitor. AmigaOS 3.0 (Kickstart 39.106, Workbench 39.29). Blizzard 1230-II board with 40 MHz 68EC030 CPU 40 MHz 68882 FPU 12 MB RAM (4 + 8 MB 70 ns SIMMs) INSTALLATION After any SIMMs, FPU, and add-on cards are fitted, the card plugs into the 150-pin expansion connector in the bottom of the A1200. It is a rather tight fit. Connecting it isn't really a problem though, unless you use two SIMMs -- and if the SIMM in Bank 2 is two-sided, it's REALLY tight. (See next section about attaching memory, etc., to the board.) PHYSICAL DESCRIPTION The card has the same basic shape as any other vendor's "1230" card, but the placement of the various chips and connectors is a bit different. There are components on both sides. On the side that faces the trapdoor lid are the 68030 CPU, FPU sockets (both PLCC and PGA), a connector for an optional FAST SCSI-II controller card, oscillator(s), a number of jumpers, plus some other chips. On the other side are two standard 72 pin SIMM sockets, a battery for the real-time clock, and even more chips. On the edge opposite the 150-pin connector is a second (custom) expansion connector. The 68030 is socketed (PGA), so a 40 MHz 68EC030 can be upgraded to a 40 or 50 MHz full 030 later. The reason for the two FPU sockets is that the PLCC versions are cheaper, but they're not available at frequencies higher than 40 MHz, in which case you need the PGA one. For some strange reason, the board requires a separate FPU oscillator even if the FPU and the CPU are clocked at the same speed (as is the case with my board). The oscillator costs about $6, so this isn't much of a problem. The board accepts industry standard 32- or 36-bit, 72 pin SIMMs, the same kind that are used in A4000s, Microbotics/Paravision 12x0, and lots of PCs. Any size from 1 MB to 32 MB is accepted. A very rare and extremely useful feature is that the two SIMMs can actually be of different size! Any combination is possible, from 0+1 via 4+8, 2+16, etc, up to 32+32 MB; I use 4+8. The memory is configured with the jumpers. But the Blizzard's capabilities don't stop there. As I mentioned, the board has two connectors for add-on cards. One is used for an optional (currently available) FAST SCSI-II controller sub-board. This is a cute little thing that rests on the Blizzard "motherboard". It connects via a cable to a SCSI port, to be attached instead of the blanking plate on the back of the 1200. The other expansion connector is more of a mystery, although according to some magazines, Phase 5 have announced two possible add-ons, one of which is an MPEG decoder card (like the one for the CD32). How Phase 5 will solve this problem, which the Commodore engineers considered "unsolvable", remains to be seen... Finally, the board has a jumper to enable the mapping of Kickstart ROM into RAM (also on boards with the MMU-less 68EC030 CPU). More about this later. IN USE The board is COMPLETELY auto-configuring; just "plug and go". There are no software patches, "Expansion" directory files, or anything. It doesn't even need a "setclock load" command in the Startup-sequence to set the internal Amiga clock! With my configuration, the 4+8 MB RAM was mapped to one contiguous chunk at address 0DC0 0000 - 0E7F FFFF. Strange placement, but it probably always places the border between the two SIMMs at 0E00 0000. A full 64 MB system would thus occupy the space 0C00 0000 - 0FFF FFFF. I have used the card extensively since I first got it some weeks ago, and it is simply a dream to use. I upgraded from an A1200 with 4 MB fast RAM and a 68881/14, but it's still a very noticeable step up. I use the Amiga for various types of applications; like ImageFX, Imagine, Scala, Final Writer, OctaMED, WinGnuPlot, and Amiga Oberon. Software that's sluggish on the 68020/14 system -- and that can include several of the above applications, with large projects -- now runs MUCH smoother and faster. Of course, the extra RAM also makes the machine a lot more stable. Upgraders from stock A1200s will probably notice much faster hard disk access, too (SysInfo reports 2.1 MBytes/second from my internal IDE; I got the same results with only fast RAM, but only about half that with the unexpanded 1200). At first I was a bit worried about heat problems -- the plastic trapdoor lid became rather hot, not to mention the 68030 itself. But I have used it for longish (18+ hours) sessions, and haven't encountered any problems. The temperature seems to stabilize -- at a rather high level, though. (Hint: I wouldn't like to sit with the 1200-030 in my lap for any period of time!) Some A1200 accelerator manufacturers recommend that the machine be used without the lid present, but this does not seem necessary with the Blizzard. As previously mentioned, the card has a MapROM option that copies the Kickstart ROM into RAM. Note that MapROM, as opposed to the standard AmigaDOS "CPU FASTROM" command, also works on MMU-less (68EC030) systems. While this makes a lot of sense on an A500 or A2000 (which have 16-bit ROMs), it's not very critical with the 32-bit (but rather slow) A1200 ROMs. AIBB tests show a speedup ranging from zero (most of the integer tests) to more than 50% (WritePixel) with MapROM enabled, but the typical speedup is less than 10%. Obviously, using MapROM will cost you one half meg of RAM. BENCHMARKS Any accelerator test has to give a few benchmark results, so here goes. I have listed the results for all the AIBB 6.1 tests. All figures give the performance relative to an unexpanded A600 (7 MHz 68000, ECS, no Fast RAM); that is, the A600 score is 1.00 on all the tests. In addition to the internal AIBB modules (stock A600, stock A1200, A3000/25, A4000-040), I have included the results achieved with my previous setup: A1200 with Microbotics MBX1200z, 14 MHz 68EC020/68881, and 4 MB Fast RAM. The Blizzard was tested with both CPU caches ON, instruction burst ON, data burst OFF, and MapROM enabled. 68020 and 68881/2 code was used wherever possible. I have not listed the A4000-040 FP results with 68040 specific FP code, mainly because it is my experience that very few programs are optimized for this (and you can find the 040 FP results in AIBB anyway). For quick reference, the fastest machine in each test is marked with an asterisk. Computer A1200 A1200 A3000 A4000 A1200+Blizzard 1230 CPU 020-14 020-14 030-25 040-25 030-40 FPU No FPU 881-14 882-25 040-25 882-40 Code RAM Chip Fast Fast Fast Fast Integer math ---------------------------------------------------------- EmuTest 1.86 3.63 5.05 16.09 * 7.87 Sieve 4.34 5.18 9.28 11.72 14.85 * Dhrystone 2.04 3.65 5.59 19.03 * 9.09 Sort 2.69 3.83 7.13 19.66 * 11.41 Matrix 3.65 5.80 10.07 16.08 16.15 * IMath 8.50 10.40 17.94 41.28 * 28.93 MemTest 2.74 5.91 6.54 3.48 10.51 * InstTest 1.75 3.52 5.47 10.05 * 8.95 ---------------------------------------------------------- Average 3.45 5.24 8.38 17.17 * 13.47 Graphics ---------------------------------------------------------- Writepixel 2.97 4.36 3.54 15.33 * 7.85 EllipseTest 2.39 3.26 2.34 5.46 * 4.39 TGTest 2.14 2.67 1.94 3.82 * 3.41 LineTest 1.73 1.86 1.08 1.87 1.92 * ---------------------------------------------------------- Average 2.31 3.04 2.23 6.62 * 4.39 Floating point ---------------------------------------------------------- Savage 2.10 115.81 207.72 165.74 332.49 * FMath 1.73 10.45 23.96 214.40 * 38.68 FMatrix 2.32 3.68 7.15 19.21 * 11.57 BeachBall 2.57 25.61 50.24 155.97 * 86.56 Flops 2.09 29.86 70.12 405.22 * 111.32 TranTest 2.12 61.77 101.35 106.43 158.94 * FTrace 2.13 60.07 123.25 125.75 202.24 * CplxTest 1.93 4.21 7.07 27.88 * 11.43 ---------------------------------------------------------- Average 2.12 38.93 73.86 152.58 * 119.14 I also ran AIBB's "All Tests|Make Module" option, which (after some time...) reported that my machine was 1.05 times as fast as the A4000-040 for integer operations, 0.67 times for graphics operations and 0.47 times for floating point operations. The AIBB results are about as expected and very similar to those achieved with other similar boards. Of course, the floating point results are the most impressive ones -- up to 160 times as fast as an unexpanded 1200 -- but the integer tests are the most important ones. Thankfully, the results there are also rather healthy -- typically 4-5 times the speed of the 1200. The A4000-040 is heavily outperformed by the 1230 in a few tests: MemTest, which fully reveals the slow memory system of the 4000-040, and Savage, which is slower because the 68040 must emulate the transcendental functions of the 6888x in software. On some other tests however, the 4000-040 is several times faster. A Blizzard-equipped 1200 is somewhere between the 4000-030 and the 4000-040 in terms of performance, probably closer to the 040. I have also tested some programs with the Blizzard board. "mp 1.03" (Michael van Elst's Berkeley-based MPEG decoder) runs at up to 8.4 fps in 8-bit grayscale with "DebbieHarry.mpeg" (a 160x128 IPB MPEG stream which is available on aminet, and which, incidentally, I made myself on the very same machine), and 5.5 fps in HAM6. This is about 2.5 times the performance I got from my Microbotics MBX1200z setup. "aMiPEG 0.4" plays the same MPEG at 4.6 fps in full-screen (scaled) DblPAL HAM8 resolution. As for floating-point performance, CineMorph runs a LOT faster. I never timed it with my old setup, though, so I don't know the exact speedup. And "rot3d" now flows quite well in full-screen mode. Just for fun, I also ran a benchmark that we use at work to compare our various machines. It came in 30 (thirty) times faster than an IBM PC/AT, five times faster than a Compaq 386-387/20 PC, well above the VAXstation 3100, at about the same speed as the Apollo DN4500 (which is also 68030 based) and VAX 8600, slightly slower than the HP-Apollo 400t, but was outperformed by a Cray X-MP by a factor of 12, and a HP 9000/730 by a factor of 25. Ah well... Finally, recall that my Blizzard is the 40 MHz version. The 50 MHz 1230 should be up to 20-25% faster. DOCUMENTATION The board comes with a 16-page Users Manual in English. The language is clear and concise, although the manual is a bit vague about the necessity for two oscillators even with a CPU-synchronous FPU. But otherwise, the manual is very good. It includes pictures, figures and tables explaining FPU and SIMM installation, jumper configuring, etc. LIKES I like EVERYTHING about this board: the speed, the unmatched flexibility and expandability, the full AutoConfig, and the price. I really can't think of anything negative to say about it. One of the things I REALLY like is the flexible memory system which allows different-sized SIMMs to be used simultaneously, a feature I haven't seen in any other board. This system means that you never have to scrap more than one SIMM when you upgrade. For instance, when 16 MB SIMMs become more affordable in the (near?) future, I might sell my 4 MB one to some PC user, keep the 8 MB one, for a total of 24 MB Fast RAM. DISLIKES AND SUGGESTIONS None! COMPARISON TO OTHER SIMILAR PRODUCTS I waited this long with buying a "real" accelerator board for my A1200 because I saw major flaws in all the ones previously available. Here are my complaints: Microbotics 1230XA: Only one SIMM slot. Memory doesn't autoconfigure under AmigaDOS 3.0. No expansion capability. GVP A1230 Series I and II: Uses non-standard GVP SIMMs, which are more expensive and more difficult to sell when upgrading. If two SIMMs are used, they must be the same size. Series I didn't have much expansion capability, there's no 50 MHz 68030 option for it, and it doesn't have a real-time clock. The new Series II board fixed many of these problems, but it's significantly more expensive than the Blizzard and offers no advantages over it. Cheap accelerator boards like the Blizzard 1220 (28 MHz 68EC020) can in my opinion not be considered to be "similar products". I guess another similar product to compare a Blizzard-equipped A1200 to would be the A4000-030. The 4000 is quite a bit slower, more expensive, and, in contrast to the 50 MHz 1230-II, doesn't have an MMU. On the other hand, it's more expandable -- 68040 accelerators, 24-bit graphics cards, 16-bit audio cards, etc, can easily be added in a 4000. Then again, the Blizzard has its two expansion sockets, and the 1200 also has a PCMCIA connector. One reason why I've chosen to stick with the A1200 is it is physically very portable.. I can just put it in a medium-sized shoulder bag and take it along to work, friends' houses, hotel rooms, the summer house, or wherever. If there's not a multisync or VGA monitor around (there usually is), I can always connect it to a TV -- the 1200 has both SCART, composite, and RF output. BUGS The need for two oscillators might be a (hardware) bug; other than that, I have not found any bugs or problems. SUPPORT phase 5 have listed a phone number, fax number, and mail address for owners in Germany. All other users are asked to contact their distributor or dealer concerning guarantee claims or technical inquiries. WARRANTY One whole page of the manual describes the terms of the guarantee. I have no intention of quoting all of it. Basically, the board comes with a 12-month parts and labour guarantee against material and manufacturing defects. CONCLUSIONS It seems to me that the Blizzard 1230-II combines the best from all the other 1230 boards. Two standard SIMM slots. Different size SIMMs can be used simultaneously. Full autoconfig. As expandable as the new GVP boards. 40 or 50 MHz options. Both PGA and PLCC FPU sockets. Very reasonable price. On a scale from 1 to 10, I can't help but give it a score of 10. I really can't imagine how this board could be improved -- at least not without renaming it "1240" (probably infeasible due to heat and/or power problems) or "1260".... COPYRIGHT NOTICE Copyright 1994 Per Espen Hagen. All rights reserved. The author can be reached by Internet e-mail as Per-Espen.Hagen@ffi.no. --- Daniel Barrett, Moderator, comp.sys.amiga.reviews Send reviews to: amiga-reviews-submissions@math.uh.edu Request information: amiga-reviews-requests@math.uh.edu Moderator mail: amiga-reviews@math.uh.edu Anonymous ftp site: math.uh.edu, in /pub/Amiga/comp.sys.amiga.reviews