50 ms·
Inside the stacked RAM modules used in the Apple III
- kens 6y agoAuthor here for your vintage DRAM questions :-)
- zimpenfish 6y agoDon't have any vintage DRAM questions, alas, but I thank you for one of my favourite blogs!
- crest 6y agoHow did we get from two chip modules to multi chip modules? Was the upper address decoding logic its own chip, just part of one of the chips or replicated in each chip?
- kens 6y agoIt depends on the module, but for example a 1 gigabyte DRAM module can be built from sixteen 64 meg x 8 chips. Each chip is 8 bits wide, so 8 chips give you 64 bits. This is duplicated for two banks. The row and column addressing is repeated in each chip, while the circuitry to select the right module and bank is external. I'm sure someone else can provide more details on modern modules. One reference is http://www.softnology.biz/pdf/lecture05-dram.pdf http://www.softnology.biz/pdf/lecture05-dram.pdf
- morcheeba 6y agoYes, what Ken said. There was another way to do it if the memory chips were already wide enough. Some static RAMs had multiple enables of different polarities. For example, the 6264 had an E1* (active low) and an E2 (active high)... wire an address bit to E1* on one chip and E2 on the other and you'll get the decoding done without adding a third chip. http://users.ece.utexas.edu/~valvano/Datasheets/MCM6264.pdf http://users.ece.utexas.edu/~valvano/Datasheets/MCM6264.pdf
- johndoe0815 6y agoAn interesting related fact is that the 48k Sinclair Spectrum also used 8 32kBit DRAMs (in addition to 8 16kBit chips). However, these were not the special modules used in the Apple 3, but 64kBit DRAMs which had a fault in one half of the matrix, such as TI TMS4532-20 or OKI M3732. The defective half of the chip was simply disabled.
- klelatti 6y agoIf I've understood correctly the one half was disabled on the Spectrum (i.e. the Spectrum would never try to access the defective part and the half that wasn't working had to be identified by Sinclair) - so they really were defective chips, in contrast to chips where the manufacturer deals with the defect.
- Zardoz84 6y agosome people had lucky (not defective) or replaced it for good 64 kBit drama and add a switch to change between the lower and upper bank. (my father did this)
- amelius 6y ago> Unfortunately, the Apple III was a business failure due to reliability issues and competition from the IBM PC introduced a year later. Curious about these reliability issues, this prompted me to read more about the Apple III on Wikipedia and it's quite interesting: > There was way too short a time frame in manufacturing and development. When the decision was made to announce, there were only three Apple IIIs in existence, and they were all wire-wrapped boards. > The case of the Apple III had long since been set in concrete, so they had a certain size logic board to fit the circuits on ... They went to three different outside houses and nobody could get a layout that would fit on the board. > They used the smallest line circuit boards that could be used. They ran about 1,000 of these boards as preproduction units to give to the dealers as demonstration units. They really didn't work ... Apple swapped out the boards. The problem was, at this point there were other problems, things like chips that didn't fit. There were a million problems that you would normally take care of when you do your preproduction and pilot run. Basically, customers were shipped the pilot run. https://en.wikipedia.org/wiki/Apple_III https://en.wikipedia.org/wiki/Apple_III
- kens 6y agoAnother problem with the Apple III was that Steve Jobs didn't like cooling fans or vents so the Apple III (like the Apple II and Mac) didn't have a fan. Reportedly, the Apple III got hot enough to warp or melt floppy disks. Another problem with heating/cooling cycles was that integrated circuits would migrate out of their sockets. Apple's recommended "fix" was for customers to lift the computer a few inches and drop it, hopefully reseating the ICs. https://www.techjunkie.com/apple-iii-drop/ https://www.techjunkie.com/apple-iii-drop/
- formerly_proven 6y agoIt's curious to see how far back the form over function at Apple goes. Truly part of the company spirit.
- ksec 6y agoAnd then 20 years later Steve Job repeated the same mistake again with Power Mac G4 Cube. ( No Fan ) Though ultimately the tech got good enough. Another 20 years passed the G4 Cube we now have 5nm A14 on a Smartphone or iPad Air that are powerful enough for 80-90% of consumer use cases. All without a Fan. We also have a roadmap of 3nm and 2nm ~2025. Unfortunately Steve did not live long enough to see it.
- bogomipz 6y agoAnother great post! I had a couple of questions: >"The second layer of polysilicon (poly 2) is arranged in diagonal regions to implement the selection transistors." Since the diagonal polysilicon covers two different capacitors how is each one selected independently? Does each one have separate metal wire connected to it? Is the sense line really a bunch of different sense lines? Does each bit have its own sense? Is the sense line part of the metal wiring then?
- kens 6y agoThe bits are in a matrix, with sense lines in silicon vertically and row select lines in metal horizontally. When a row select line is activated, 128 transistors are turned on and each sense line is connected to the single capacitor in that row. Thus, 128 bits are read in parallel. Thus, each horizontal row select line is connected to 128 different transistor gates, in other words connected to 64 of the diagonal poly 2 regions. The confusing part is that the rows zig-zag a bit to improve the density. In other words, the physical rows in the photo don't quite match the electrical/logical rows. Look at the photo and imagine pulling it apart like a zipper. Then shift the middle part up half a step so the capacitors line up. Then the logical rows are all in a straight line. Hopefully that explanation isn't entirely confusing. In brief, the two capacitors with the same poly 2 diagonal are in the same electrical row, so they are selected together (along with 126 other capacitors). I'll probably write more about this, so let me know if there is a clearer way to explain this :-)
- bogomipz 6y agoThanks, this explanation was really helpful. I think your zipper analogy would actually be a nice addition to your post. It also helped me appreciate just how ingenuous this design is. Cheers.
- exabrial 6y agoDumb question from a clueless noob... Since the stacking required a select line, why not skip the process and just add more ram slots? Seems like that'd be cheaper than custom hardware
- kens 6y agoUsing these modules gives you twice the density that you'd get otherwise. If you have unlimited space, you can add more RAM slots. These make sense if you have a fixed area for memory chips.
- xattt 6y agoTangential: in the heyday of PocketPC PDAs, you could send away an iPaq for a storage upgrade that used stacked DRAM chips. They came with a special driver for WinCE that permitted such addressing.