Talk:Registered memory

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Computing / Hardware (Rated Stub-class, Mid-importance)
WikiProject icon This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
Stub-Class article Stub  This article has been rated as Stub-Class on the project's quality scale.
 Mid  This article has been rated as Mid-importance on the project's importance scale.
Taskforce icon
This article is supported by Computer hardware task force.
 

Unnamed section =[edit]

If Registered memory has a clock delay, the bandwidth is comparable, but latency is WORSE. So, there is a performance hit... :-? Jcea 19:08, 7 March 2007 (UTC)

It's random-access-memory. There will be a one-clock delay as you "randomly" go out and read or write one word (typically 64 bits). If you are in sequential or burst mode then you only get the performance hit on the first memory cycle. Marc Kupper (talk) (contribs) 03:14, 26 October 2007 (UTC)

Possible Errors RE Article Factual Information[edit]

The article at present states in part, "Although most server-grade memory modules are both ECC and registered, there are both registered non-ECC modules and non-registered ECC modules."

According to Kingston, a major manufacturer of RAM, "Registered memory includes ECC functionality but not all ECC is registered." [cf: (http://www.kingston.com/us/support/technical/valueram.aspx?questionid=148&afmid=41239) {FAQ: KTM-012711-GEN-03} in RE: "What is the difference between registered, unbuffered, ECC and fully buffered memory?"]

If Kingston's claim is true for all RAM, then the article's claim is false in that there are no registered non-ECC modules; however, Kingston's claim could be construed to apply to only their own products. Please cite at lease one example of registered non-ECC modules.

From http://www.edaboard.com/thread8144.html come the following quotes: [Q]"Do all registered modules come with ECC?" [A]"All Corsair registered modules have ECC. But, strictly speaking, this does not have to be the case. But, since most registered memory configurations have large amounts of mission-critical memory, it only makes sense to use error correction. (Read my bulletin on ECC for more info)."

The current Memory FAQ at Corsair (http://www.corsair.com/support/faq/) does not address the instant issue; therefore, it seems the quote either may have been wrongly attributed to Corsair, or may have been featured in an official blog or similar publication on behalf of Corsair, or may have been the product of independent research; moreover, the information may yet be readily available from Corsair either upon request or as an archived feature or on a web page that I simply didn't visit.

It seems reasonable that registers should be differentiated from ECC hardware; however, despite the independent functioning of these technologies, it is reasonable to imagine that there may in fact be no registered non-ECC modules, if only for the reason identified in the quote attributed to Corsair. In other words, while it may be technically possible to make such a module, the reasons to not make it have thus far outweighed the reasons to make it: if so, then it seems unreasonable and probably inaccurate to state that there are registered non-ECC modules.

If there are in fact registered non-ECC modules, then the article's verbiage should remain unchanged and a citation should be added referencing a source indicating the existence of registered non-ECC modules; else, the article's verbiage should be changed from "Although most server-grade memory modules are both ECC and registered, there are both registered non-ECC modules and non-registered ECC modules" to "Although non-registered ECC modules are possible, most server-grade memory modules are both ECC and registered; additionally, there are registered non-ECC modules."

Thanks. — Preceding unsigned comment added by 98.95.41.185 (talk) 16:37, 14 March 2012 (UTC)

Unverified claim[edit]

The line that based upon the previous commenter used to say "Although most server modules are both ECC and registered", now says "Although most memory modules are both ECC and registered". Is there data to support that? I would think there was more non-ECC Desktop memory around than Server memory, but perhaps the balance is changing.

Also note that the description of buffered last paragraph first says it's not the same as registered, then goes on to give the same definition of buffers on the control lines. Mike163 (talk) 16:13, 29 April 2013 (UTC)

Information outdated? Registered memory only sold for marketing reasons?[edit]

I suppose since SDRAM there has been no non-theoretical reliability difference and such claims as memory integrity and scalability about ECC+registered memory seem to be as old as 30 years plus, from the time when memory modules were still soldered by hand in 'electronics manufactures', and have vanished as much since then. In fact the lines between server/consumer technology have blurred and disappeared as much as that the main difference that distinguishes server hardware from consumer hardware in terms of reliablity now is price and thus component quality alone, and a few artifacts such as incompatible ECC memory modules and incompatible CPU socket formats persist until this day, apparently for no good reason (the remaining differences being, which make sense in terms of a specialized purpose: rack mountability, unit compactness/height and onboard LOM). Consider here that most servers that use registered/ECC memory modules do not simultaneously have more memory slots, i.e. 8 or 16, where it would make sense to argue that registered memory was needed. And even if it were 8 or 16, would the 'load on the memory controller' really be that much intolerably higher in modern servers than with 4 modules? Or would the underlying theory explained in the article, turn out to only make a difference after ridiculous amounts of modules, let's say 64 or 128? Since 10 years now, most consumer CPUs have simply been re-branded and re-socketed as Xeons and sold at twice the price. There is no real justification for those socket format mismatches and fashioning sockets to be as future-unproof as possible is anyway a known trick to drive manuracturer sales up in the consumer world alone. In the same way, couldn't the so called 'advantages and disadvantages' of ECC or whatever claims about necessities with registered memory, be entirely void in modern times and only a remnant kept alive by price trickery driven by the manufacturers rather than real and factual reasons to implement and keep this technology/format in place? It would be great if someone could actually put this up to date by supporting it with information that relates directly to modern technologies, such as DDR(3) RAM. Otherwise it seems to me so far that we are rather keeping a myth alive by recursively restating and requoting now irrelevant facts throughout scholar articles that probably would trace back to originate from now non-existent hardware and computing environments, and are as such nonsensical in implicitly assumed non-historical contexts. 37.201.226.165 (talk) 23:21, 14 September 2014 (UTC)

Hello there! Are you questioning only the registered ECC memory, or unregistered ECC memory as well? — Dsimic (talk | contribs) 00:10, 15 September 2014 (UTC)