CompuServe Thread

#Old CBM 68020 boards?

16 messages in this thread
#44963From: Jim DuttonJan 1, 1992 5:44 AM
Does anybody know if the old CBM 68020 accelerator boards are still available (who sells them/how much) ?? – via Whap! (from Southern Illinois)
#44976From: Kelly JordanJan 1, 1992 10:40 AM
Jim, The 68020 accelerator board (A2620) is no longer made. Can only buy them used. The 68030 accelerator board (A2630) is the current accelerator available from CBM. Should be able to get one from any CBM dealer. Don't know the price. 73, Kelly
#45319From: Jim DuttonJan 4, 1992 11:23 PM
Thanks for the info – hopefully, I can snag an A2620 somewhere. The '30s and '40s acclerators are 'too rich' for my bank account! – via Whap! (from Southern Illinois)
#45512From: daniel wolfJan 7, 1992 4:28 AM
i think you'll find the 020 is just as fast as an 030 at equal clock speeds, anyway. except for the MMU (and slightly larger cache) I never understood why folks paid such high prices for 030s. Now the 040 is another story entirely -FAST math right on board.
#45523From: Gregory B. TibbsJan 7, 1992 7:21 AM
reply If the 020 is paired with a 68851 PMMU, as with the A2620, you incur a one wait state penalty for the address to translate through the 851. The 030 does not incur this delay. The 030 also has a data cache which when properly enabled, allows a typicaly speed increase of over 50% in most benchmarks. At the same clock speed, the 030 will blow any 020 away. The only time your claim even comes close to being true is when the both have only 16 bit memory to talk to. -Greg Tibbs The 040 has only hardware multiply, add, subtract, and divide. All other FPU instructions have to be implemented. Many of the transendental functions are 30% slower than an equivalently clocked 030/68882 combination.
#45524From: Gregory B. TibbsJan 7, 1992 7:23 AM
Change All other FPU instructions (of the 040) have to be emulated, not implemented, or implemented in software. 😉 -Greg
#45801From: robert fisherJan 10, 1992 12:21 AM
Excuse me for jumping in like this… but you guys seem to be talking about just the kind of stuff I need to know about! What would you think of upgrading an A2620 board with a 25MHz 68882? Would this be close to an 030/68882 board for a number cruncher like Imagine? I can't afford another accelerator and was wondering if it would be worth spending a couple hundred bucks on a faster FPU instead. Robert Fisher
#45810From: Gregory B. TibbsJan 10, 1992 5:38 AM
If the software uses the FPU, it will see a performance increase. If not it won't. I would think you might get as fast as 75% of the 030/68882 in some situations, but if you get where the 68882 is always waiting on the 020 to complete its work, then the performance will me much less. As they say: you're mileage will vary… ! -Greg – via Whap!
#45858From: robert fisherJan 11, 1992 1:01 AM
Well, the only program I'm really interested in speeding up is my ray tracing program. Everything else runs fast enough for me with the 020. It's all those floating point calculations that keep my machine crunching through the night! Robert
#46558From: daniel wolfJan 18, 1992 4:14 AM
you'll be surprised how close the 2620 board is to a similar speed 030/882! the 882 isn't that much faster in most real cases and the 030 isn't any faster at all. it just has an mmu and a bit more cache. the 040 is the only real power development since the 020. if you can get a good buy, go for it!
#46576From: robert fisherJan 18, 1992 11:55 AM
Well, I went for it. Wyle Electronics has 25 MHz 68882 chips for about $95. If I can manage to solder the 25 MHz oscillator onto the board without screwing anything up, then I'll have firsthand knowledge on the performance of an upgraded 2620 board by midweek!
#46676From: Bill ThompsonJan 19, 1992 11:39 AM
Could you please post the address/telephone for Wyle Electronics… That seems like a pretty good price. Course I guess I should wait til you post your results! ..Bill 😉
#47207From: robert fisherJan 25, 1992 12:47 AM
Bill, I'll try to remember to look it up next week. It's an 800 number in Santa Clara, CA. It was kind of a pain setting up an account with them…something about Motorola's paranoia and exporting their proprietary technology to Iraq or some such thing! Whatever…! In any case, I should have the 68882 in my anxious hands shortly after SuperBowl Sunday. I'll let you know if I've just blown $100 or if it's all worth it! Robert
#45579From: John PendergrassJan 7, 1992 8:12 PM
Daniel, I don't know the in's and out's of it, but my 25mhz 030 (with 32bit ram and math chip) is about four-eight times faster than the 14mhz 020 (also with 32bit ram and math chip) CSA board I had earlier. I remember running timing tests when I originally upgraded. On comm'l type software (tracers and such… shoot, I probably timed a FractPro render as well 🙂 Isn't the 040 on-board math non-transcendental? How does it perform on your FractalPro stuff that uses those functions? (or maybe my info is off) –jp–seattle–
#45743From: Jim DuttonJan 9, 1992 6:48 AM
I agree that the 040 boards are more worth their money than the 030 boards, but they are both too expensive for me. An "old" 020 board, right now, will fit the bill quite nicely. If I ever "strike it rich", I will definately latch onto a 68040 board with LOTS of memory, but I haven't hit the "pot 'o gold" yet. – via Whap! (from Southern Illinois)
#45022From: Betty Clay/SYSOPJan 1, 1992 8:13 PM
Not to my knowledge, Jim. I have not seen them offered for quite a long time. Perhaps someone else will know. Betty