mwm@opal.berkeley.edu (Mike Meyer) (09/17/86)
This was run on an Amiga 1000 w/ 68010 cpu and the RS Data Systems 2 Meg Pow-R-Card on a MicroForge single-slot expansion. The only things in memory were the Workbench and the CLI the memory test program was run from. This version of the memory test was compiled under Lattice, with the while loop containing the asm code replaced by a call to memcpy(). This should produce slightly slower code than the Manx version, but shouldn't invalidate the test, as the question of interest is the time to access memory. The test was run several times under slightly different conditions. The value below is as described above, and seems typical. CHIP: 156451473 FAST: 150751446 Difference: -5700027 Or a net gain of 3.6 percent in the fast memory. Not bad - the fast memory is fast! Now, someone who understands the hardware better can tell me why the test is invalid, and what I need to do to make it valid :-). The Pow-R-Card is in strange shape, mostly because the rest of the world didn't behave as RS Data Systems expected to. Rather than try and explain what's going on, I'll tell you how things have happened, as reconstructed from my talks with RS Data Systems. In January or February, RS Data decides the Amiga is going to be a hot machine (pretty sharp people, right?) and starts asking who is going to be building expansion boxes for it based on the then-current expansion specs. They find five companies who claim to be working on such, and so start on their memory card. They get their card ready in June/July, only to discover that four of the five people they talk to don't have expansion boxes. The only things available are from MicroForge. Therefore, they start on their own expansion box. In early August, I call them to ask about the board. See the end of the article for a description of it. They tell me the above, state that they are shipping the board and the MicroForge single-slot expansion board. The 2 Meg Pow-R-Card is $850 or so, the MicroForge card is $84. Their expansion box will be available in 2 months for $84. I decide that $15 or so a week for 2 Meg is reasonable, and order a Pow-R-Card and MicroForge expansion, planning on getting their expansion box later. Shipping time is roughly 2 weeks (have to get the card from MicroForge). At this time, fate intervenes, and I wind up out of town for 2 weeks starting about when the card should arrive, followed by a one-week vacation, so I don't get to look at the cards for a while. When they arrive, I find that it doesn't autoconfig with 1.2. They say "yes, that's they way the old specs are." I'm told that everybody who had the Pow-R-Card and MicroForge expansion would be getting one of their expansion boxes gratis, sometime around the end of September. In early October, when 1.2 is officially released by CBM, I'll be able to buy a board for $15 or so that goes in the second slot in their expansion box and makes their memory card autoconfig. After a weeks use, I decide that the board seems somewhat flakey. Call RS Data up, and complain. They send me a floppy with a ram test program for their card. The floppy arrives DOA, so I call them again, and download a copy of the software. At the same time, they give me a cute little program that opens a one-gadget window, and will upon request allocate or free all available fast memory (not sure what it does if more fast memory becomes free after it thinks it's got it all. Haven't tried it). Sure enough, the memory test program finds a RAM chip that fails intermittently. It even draws a picture showing where all the chips are, with the bad one in red. Pull the chip, check that the pins aren't corroded, re-insert the chip. Everything works like a charm, and the board has been rock-solid ever since. Call them back to tell them this, and they let me know that the expansion boxes are running right on schedule (meaning about a week earlier than they had told me. No dummies, these guys - they give themselves time to fix things!) so I should be seeing mine in a week or so. I expect it any day now. Now, a description of the board: It's a one-slot 100-pin board. You get either 2, 4, 6 or 8 meg. The 2 and 4 meg versions will run off of Amiga internal power. The 6 and 8 meg versions need to be in an expansion box with its own power. The memory is zero wait-state, "unless you access it while it's refreshing that bank. But if you're actually using it, that should never happen." I'm not a hardware hacker, but the board appears solid enough to me. All chips are socketed so that you RS Data can do chip-swapping with you should some go bad. The 6 and 8 meg boards use a daughter board. I have not seen this. Upgrading requires new PALs from RS Data. They are willing to talk about selling those without memory. I'm going to investigate that later. As stated above, it does NOT autoconfig. Available later, for a slight extra fee. This also eats up a slot in your expansion box. The expansion box has been described to me. It passes the bus, and does the same for the mouse port. Whether it passes the second mouse port is unknown. Just so long as I can stack the second disk drive on it, this doesn't bother me [Listen up CBM - the only ergonomically bad thing I've found about the Amiga is that there IS NOT a good place to put the second drive. It HAS to be on the right side of the box, where it interferes with the mouse cable]. I'll post commentary on that when I get the box. I'm happy with it. I wish it autoconfiged, but can live without it. The RamCheck is great, and lets anyone do chip-swapping for bad memory chips. MemGrab is something that's nearly trivial, and something like it should come with every memory expansion, at least until everyone starts writing code that works with fast memory. I haven't decided whether to buy the autoconfig card when it shows up, or put the hack in startup-sequence and use the slot for something else (SCSI controller for an IOmega box, maybe?). RS Data can be reached at: RS Data Systems 7322 Southwest Freeway, Suite 660 Houston, Texas 77074 713/988-5441 Naturally, I have no affiliation from them except as a satisfied customer. <mike