There are many terms to get your head around when dealing with computer memory. One of the least well-documented is the term “Composite Memory Module.” There are two big reasons for the lack of documentation on the subject. The first is that the terminology has been irrelevant since the mid-1990s. The second reason is that Apple almost exclusively used the term.
Memory modules at the time were fairly different from those of today. Instead of being DIMMs, memory modules were SIMMs, though the DIMM was in the process of being introduced.
Tip: DIMM stands for Dual Inline Memory Module, whereas SIMM stands for Single Inline Memory Module. The electrical contacts on a SIMM are identical on either side. On a DIMM, they are distinct. This difference doubles the number of electrical connections on a DIMM in a given space.
Memory modules also had a much lower capacity than they do today. 32MB SIMMs were cutting edge and expensive, in the ballpark of $2000 in 1993. Compare that with getting a 32GB DDR4 DIMM for less than $100. That’s 1024 times the capacity for 5% of the price, and that’s not accounting for inflation which would cut the cost to 2.5%.
Structure of a Memory Module
One of the things about any memory module, be it SIMM or DIMM, is that many individual DRAM chips are mounted on the module. One of the things you may notice about the vast majority of modern memory is that they have 8 DRAM chips on each DIMM. A 16GB DIMM will have eight 16 Gb DRAM chips mounted on it. An 8GB DIMM will use 8 Gb DRAM chips, and a 32GB DIMM will use eight 32Gb DRAM chips.
Tip: This isn’t always the case with modern DIMMs, but we’ll come back to those cases later and try to explain the differences.
This wasn’t always the case, though. The technology to create DRAM chips larger than 4Mb took longer to commercialize than the pressure to create SIMMs larger than 4MB. Therefore, some enterprising memory manufacturers mounted larger numbers of 4Mb DRAM chips. 16 chips could be used to make an 8MB SIMM, while 32 chips could be used to create a 16MB SIMM. Using more than 8 DRAM chips was termed composite memory, while SIMMs and DIMMs with 8 DRAM chips were termed non-composite memory.
At the time, making and using composite memory was financially sound. A 16MB composite SIMM could often be half the price of a non-composite SIMM. This was because producing the lower number of 16Mb DRAM chips was significantly more expensive than making more of the smaller chips.
The Effect
For many computer systems, it didn’t make much difference whether composite or non-composite memory was used. Unfortunately, that was not the case for every computer. The problem was primarily felt in some lines of Apple computers of the time. Specifically, the Quadra 800 was notorious for being incompatible with composite memory. Because this issue mainly affected Apple computers, the terms are mostly associated with the company.
The problem arose from the extra circuitry and management chips needed, on the SIMM, to handle the high number of DRAM chips. This large number of chips increased the power draw while the extra circuitry increased the latency. The computers that suffered compatibility issues with composite memory tended to be specifically tailored to high performance, especially regarding memory. This meant that the small deviations from the expected performance characteristics resulted in unintended and unexpected behavior, often system crashes.
Modern Differences
Some modern DIMMs don’t come with 8 DRAM chips but 9. These are ECC DIMMs; ECC stands for Error Correcting Code. ECC DIMMs are primarily meant for servers where memory errors could have a more serious impact than on a home computer. ECC DIMMs can detect and correct single-bit memory or transit errors to ensure accuracy.
Some modern DIMMS have 8 DRAM chips on each side. At first glance, this may seem the same as a composite SIMM, but it isn’t. This type of DIMM is “dual rank.” Each rank of 8 DRAM chips offers the full data width of the RAM data bus. A composite SIMM needed 16 or 32 chips to be able to fill the width of the RAM data bus. This means that a dual-rank DIMM operates closer to two DIMMs in one. It also doesn’t come with any performance or stability issues, as the process is entirely standardized.
Tip: Some dual-rank DIMMs are also ECC DIMMs; these would have 18 chips and are a combination of the two with the implications of both.
Conclusion
Composite memory is an obsolete term from the early- to mid-1990s. It dates from a time when the requirement for memory capacity in a memory module exceeded the ability to commercialize DRAM chips to fill them as standard. To get around this issue, memory manufacturers added more of the smaller DRAM chips to SIMMs. This achieved the desired capacities but required extra management overheads.
These overheads had the knock-on effect of causing performance and stability issues in certain computers. The term composite memory expressly referred to SIMMs and possibly DIMMs that featured more than 8 DRAM chips per module. Modules that featured the standard 8 DRAM chips were referred to as non-composite. All modern RAM is non-composite.
puzzle says
Can you explain why a 8GB DIMM has 8 times 8 GB DRAMs? What is the relationship between the 8Gb DIMMS and 8GB DRAM?
From what I understand a 64-bit processor will read and write 64 bit at a time. So here does 8GB DIMM refer to being able to read 8 billion different address to fetc 8bytes?
Technically with 8 times 8GB DRAM we have enough memory cells to store 64GB. I would really appreciated your help.