Article 236 of comp.sys.amiga.reviews: Path: menudo.uh.edu!usenet From: dgilbert@jaywon.pci.on.ca (David Gilbert) Newsgroups: comp.sys.amiga.reviews Subject: REVIEW: VXL*32 32-bit RAM expansion for Amiga 500 Followup-To: comp.sys.amiga.hardware Date: 28 Mar 1993 18:31:03 GMT Organization: The Amiga Online Review Column - ed. Daniel Barrett Lines: 254 Sender: amiga-reviews@math.uh.edu (comp.sys.amiga.reviews moderator) Approved: barrett@math.uh.edu Distribution: world Message-ID: <1p4qt7$t4@menudo.uh.edu> Reply-To: dgilbert@jaywon.pci.on.ca (David Gilbert) NNTP-Posting-Host: karazm.math.uh.edu Keywords: hardware, memory, 32-bit RAM, A500, commercial PRODUCT NAME VXL*32 32-bit RAM expansion for Amiga 500 BRIEF DESCRIPTION As I joyfully announced in c.s.a.hardware and c.s.a.misc, I got my VXL*32 8 meg RAM expansion. I have the 25Mhz VXL30 68030 accelerator, and have been desiring this upgrade for a long time. This review is based on approximately six months experience with the product. TEST SYSTEM For the reader's reference, here is the current review system: - Amiga 500 - VXL30 68030 accelerator with 68882 - VXL*32 8 meg RAM (80ns) expansion - GVP Series II (4 megs deactivated RAM) - 105 meg Quantum LPS - 155 meg Wren III (ESDI w/ Emulex board) - 44 meg Bernoulli - 2 Commodore floppy drives - 1 meg Chip RAM via Supra board - 2400 Zoom modem - SupraFAX v.32bis v.42bis modem - 2 Original XT 200-watt power supplies (paid $30 Cdn) powers all units, including the Amiga, and the three hard drives - Deskjet printer PRODUCT INFO Company: Microbotics Address: 1251 American Parkway Richardson, TX 75081 USA (214) 437-5330 Product: VXL*30 RAM expansion Configuration: 8 megs of 80ns RAM installed Price: $800 (Cdn) approximately Availability: Got it from my local dealer IMMEDIATE OBSERVATIONS When I originally got the VXL30 board, there was some speedup in some programs. I was rather disappointed, actually. When I bought the 68882, I found some applications (such as Post) much faster, but overall, again, I was unimpressed. When I installed the 8 meg RAM board, the machine even booted about four times as fast! Every single application (especially Emacs) showed a marked improvement in speed. I was impressed. THE SAME BOX??? When I got to my dealer, I almost thought he'd got the wrong item for me. The box for the VXL*32 is exactly the same as the box for the VXL30! In fact, the only difference is a stuck-on label proclaiming the new product. The installation disk is also re-used -- underneath the VXL*32 label, you can see the VXL30 label. No problem, I guess, they probably just had too many disks printed. It looks like the VXL*32 disk will also be shipped with the new VXL30's. They also shipped the same 1-page, cheaply-printed instruction sheets. Not that the instructions are hard to follow or too short, just that they might think of making a little booklet --- it would look more professional. Actually, the VXL*32 came with 4 sheets of paper, with lots of information on them. They only lacked one tidbit, but that's for the next section. INSTALLATION I cannot recommend that any of you install this for yourself: I'd get sued a million times if I did. But: it isn't that hard, and the layman should be able to do it. Well... not the layman who has never pulled chips (he should have a friend to help him); but otherwise, the installation is simple, and should go off without a hitch. The first thing you all should know is that all of the V1 and V2 type serial numbers will require a number of chips to be replaced on the VXL30 itself. These are provided free of charge to all purchasers (currently). This might have been the reason that it took so long to develop the RAM expansion. In my unit, all but 4 of the DIP packaged, socketed chips had to be replaced. I own a V1 unit. This operation is not hard, and all chips are numbered for easy identification. The toughest of thing that must be done to these older boards is a line that runs from one chip to the underside of the RAM board. This line is apparently for DMA access. They provide chip "8" with a soldered wire, and there is a socket on the underside of the RAM board for the wire. No problem. After I had done all this, I installed my 2.04 ROM on the RAM board. I will caution buyers of the VXL*32 NOT to buy a Kwickstart board. You can place one ROM on the RAM board, and one ROM on your motherboard. The ROM on the RAM expansion cannot be accessed in 68000 mode; but if you think about this, it is not a problem. Wherever the ROM is, it can be loaded into 32-bit RAM after bootup. This operation does not seem to require a re-boot (as some others do). Benchmark programs, however, seem to verify that the ROMs are in 32-bit RAM. This does NOT require an MMU. CAUTIONS, WARNINGS, PROBLEMS, AND GRIPES First of all, it's about time! I bought one of the very early VXL30 units, and at that time, the RAM was promised "soon." Well, they said "soon" for a long time. But, it's here, so I'll stop griping about that. I almost considered getting the GVP530 instead. One thing that they don't mention is what to do about the jumper on the 2.04 ROM. On my motherboard, it's required that it remain. On the VXL*32 RAM board, it must be cut. There is no mention of this in any of the VXL*32 RAM documents, and I was very reluctant to cut it (it would be difficult to put back together). The symptom of this was that it wouldn't boot, and the screen was purple. I have no way of testing the following, so I'll just pass on the information that I got. The RAM expansion supposedly will work perfectly with accelerators up to 40Mhz, but with 50Mhz they require ONE of the following: - 60ns RAM (256x4 are available, 1MBx4 available Fall92). - Higher speed FPGA part. - Defeat burst mode. I have not done any of the above as I have a 25Mhz model. In addition, they recommend that RAM be mapped out of DMA address space for the 50Mhz mode OR burst turned off. This is another item I suspect gave them enough trouble to delay it. Among the other warnings that came with the product are that a new power supply should be considered. (I have a 200 watt supply.) This is only sensible due to the nature of having 8 meg of high speed (and power-demanding) RAM in the system. BENCHMARKS What review of processors and RAM would be complete without a benchmark? First off, the unit scored as a whole similar to an Amiga 3000. In integer and floating point performance, it will identical, if not a few fractions of a percent higher. This is to be expected as the RAM and processor that I have are identical to the 3000. The unit is slower, however, in the Chip RAM access department. It, of course, is still dealing with the 16-bit bus of the A500. Although I would like to have a 3000, I think I'm going to wait for the dust to settle and the new machines to come out. Despite this, AIBB's "Writepixel" test declared that this board was slightly faster than the 3000. The 8 meg of RAM, however, seems to be slightly faster than the 3000. This could be due to design, or it could be due to the fact that the moon is in the wrong phase. I've never really trusted benchmarks. The reported difference was in the range of 1-5%. I will upload the AIBB module that I created to wuarchive.wustl.edu (128.252.135.4). You can all take a look at the specs. Suffice it to say that I am pleased! FROM BENCHMARKS TO THE REAL WORLD You can look at the AIBB module, and compare to your heart's content. This section, however, is dedicated to the observed speedup in programs that I use every day over their performance before the RAM expansion. The difference between the stock system and adding the VXL*30 was not terribly large. Probably similar in magnitude to when I added the 68010. There were some things ran perceptibly faster, but not too much. Similarly, when I added a 68882 to the setup, several applications more than doubled in speed, but others were not affected. With the addition of the RAM, however, there was an overall increase in speed. Post, for instance, which was vastly sped up by the addition of the '882, went from 1min/page to 30sec/page to 3sec/page for TeX generated Postscript going from 68010 to 68030+882 to 32-bit RAM. Another application that received a major jump is IBeM. My personal theory is that since the author used instructions that access odd addresses, a worst case scenario happened: access byte 3 or 4 of a longword, and the 68030 puts the address of the longword on the address bus to find that it's 16-bit (remember that memory signals the 68030 as to what width it is), then it has to cycle again to get the byte. I'm just guessing that this happens... I don't know much about the processor, but if it doesn't know that the memory's not 32-bit, wouldn't it ask for the longword first? Anyways, IBeM gets 100x speedup --- from waiting for characters to almost as fast as my V30 laptop. In fact, some games play faster --- to the point of unplayability. Other applications such as Desktop Publishing, Emacs, gcc (GNU C compiler), and even MED are all perceptibly more responsive. OF SCSI DRIVES AND AUTOCONFIG One of the first things I found out is that my older Series II hard drive does not like memory it can not DMA to. I may be able to adjust parameters in the menu of FaaastPrep for it, but at this point, I don't know. When I configured my 4 meg 16-bit RAM to be Autoconfig, and the 8 meg of 32-bit RAM to be outside Autoconfig, I got SCSI errors randomly. I decided to leave it for now. I'm pretty sure the problem is the GVP, and not the VXL. One very obvious potential culprit is the GVP ROM --- it's quite old at this point. 3.something, I believe. As far as DMA ability, the RAM passes with flying colors. I have tried all the standard DiskSpeed tests, and my drives are almost exactly the same as they were before --- if not a tiny bit faster. The CPU Availability Index has gone up significantly, too. With only the 8 meg RAM installed, I have had no complaints from my drive. The RAM Autoconfigs wonderfully. It comes up automagically when the card is set to map the RAM into Autoconfig space. It even appears as a full Autoconfig board. FYI, it has a product number of 68 (the accelerator shows up as product 69). The keeper of `Sysinfo' might want this information. One interesting thing is that you can change the position of the RAM using software (and I don't have an MMU). You can also map the ROM into RAM --- again without the MMU. I thought that you had to reboot when you move the ROM, but this does not seem to be the case with the VXL. Maybe the software is smarter? THE MOST SURPRISING THING HAPPENED... The most surprising thing happened when I switched to 68000 mode, however. The full 8 meg of 32-bit RAM *becomes* 16-bit RAM. I would assume the circuitry for this is similar to that which makes it so DMAable. However... I *never* expected it. The ROM on the RAMboard is not available in 68000 mode, but I use the 2.04 ROM on the RAMboard, and the 1.3 ROM on the motherboard... this makes sense... when you *must* drop back to the 68000, chances are you don't want the 2.04 ROM either. This can be turned off... I think there might be games that don't like 8 meg of RAM, but I haven't found them. It's still nice to know. Dave. ---------------------------------------------------------------------------- |David Gilbert, University Of Guelph, Ontario | Two things can only be | |Preferred : dgilbert@jaywon.pci.on.ca | equal if and only if they | |Possible : dgilbert@snowhite.cis.uoguelph.ca| are precisely opposite. | ---------------------------------------------------------GLO---------------- --- 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