cjl@ecsvax.UUCP (05/27/84)
First of all, the 32R16 DOES have some applications. Granted, not as many as the 16R8, but for the same reason as a 27256 is not needed as much as the 2716- bigger is not always better, but it can sometimes come in handy. The 32R16, for all its faults, can do things that 4-5 16R8s can't due to the larger OR plane groupings. And not everyone needs 30 ns prop delay. AND- if one chip will take the place of 4-5, I'll pay $50 each in most cases. Now for CMOS- I used to use TTL and even ECL in some applications. CMOS will not replace either one in very high speed applications. CMOS DOES- -have a higher noise immunity -reduce power supply needs greatly -reduce heat problems in circuitry -allow digital/linear circuitry on the same chip (Now if only someone would make a CMOS 32R16...) -Charles Lord cjl@ecsvax
phil@amd70.UUCP (Phil Ngai) (05/28/84)
> First of all, the 32R16 DOES have some applications. Granted, not as many as > the 16R8, but for the same reason as a 27256 is not needed as much as the > 2716- bigger is not always better, but it can sometimes come in handy. I think there are some problems with your analogy. 2716s are obsolete and it's not wise to design them in. In the EPROM market, things move so fast that at AMD, 2764s are officially obsolete (but 2764As are still good). The price of 2764s is down in the dirt and 27256s soon will be. The pinout and functionality is obvious up to the 27512 level so there are many second sources, the yields are high, and therefore it makes sense to use a 2764 when a 2716 would have been enough. The PAL market is not as neat as the EPROM market. Everyone seems to be inventing their own architecture and no one wants to second source anyone else's. The yields don't seem to be as good on the new complex parts, hopefully this will change but EPROMs are inherently simple devices while PALs aren't as simple. So I claim the big PALs coming out will see a less steep price decline than 27256s. > The 32R16, for all its faults, can do things that 4-5 16R8s can't due to the > larger OR plane groupings. And not everyone needs 30 ns prop delay. AND- > if one chip will take the place of 4-5, I'll pay $50 each in most cases. Well, I wouldn't pay $50 for $12 worth of functionality in almost the same amount of board space that is single sourced. Do they really let you design in single sourced products? I wouldn't even if I were allowed to. -- Phil Ngai (408) 749-5286 {ucbvax,decwrl,ihnp4,allegra,intelca}!amd70!phil