While I don't have a reliable source at hand, it should be obvious that semiconductor memory is not intrinsically random-access. See charge-coupled device and bubble memory. The issue is that there are many sources that state otherwise. This is however not because semiconductor memory is intrinsically random-access, but because of the nature of those sources. These sources tend to be modern textbooks, and currently, there aren't any sequential-access semiconductor memories available, so there is a tendency to ignore historical sequential-access semiconductor memories. This does not mean that a statement claiming an intrinsic property of semiconductor memory is random-access is true. AZ1199 (talk) 01:11, 4 January 2016 (UTC)
It is a good point @AZ1199: You are correct. I have rewritten the sentence to reflect that it's a property of most modern memory implementations, as there are examples of modern devices like the FIFO memory that do not allow a random access. Regards. --Crystallizedcarbon (talk) 07:53, 4 January 2016 (UTC)
Regarding the edit made in response to my OP: I'm sorry if I give the impression that I am being overly critical, but the provision of a FIFO as a counter-example of a semiconductor memory that isn't a RAM isn't entirely correct. While a FIFO memory doesn't present a number of addressable locations as its interface (it isn't possible to read or write to entry #10 in a 64-entry FIFO), that's only because its interface abstracts away the fact that it's commonly a RAM. In a RAM, the user provides the address for an operation; but in a FIFO, it's a pair of wrap-around counters that do (and these are incremented upon the assertion of a read or write enable signal). Thus, from the perspective of logic external to the FIFO, the FIFO appears as a black box that simply accepts values and reads out the oldest written value. Internally that's not quite true. That being said, there are FIFOs that are actually sequential (if my quick search of Google Books is accurate), so there is some truth to the current lead section.
Having had some time to think about this article, I'm actually questioning its necessity. The term "semiconductor memory" doesn't really mean much more than that the memory is constructed from semiconductor materials. Contrast this to what early computers used for their memories: devices such as electromechanical relays, vacuum tubes, and Williams tubes. There can't be too much to say about semiconductor memories, other than perhaps why it has displaced other forms of non-electronic or electronic memory. AZ1199 (talk) 23:58, 4 January 2016 (UTC)
Hello @AZ1199: How the FIFO queue is implemented internally is not relevant, by definition a queue is not random access. As far as implementation, most are designed using pointers, but a LIFO queue could be easy to implement sequentially by cascading flipflops with a bit of additional logic and without any need for pointers. I think that your first comment was correct and the current wording for the lead is now more accurate.
I understand you arguments about the relevance of the article, but I would not agree with deleting it, as it meets our notability requirements. Regards. --Crystallizedcarbon (talk) 15:01, 5 January 2016 (UTC)