So what exactly determines the arbitrary RAM limits on a MB?

[DanielJamesDavis]

I've been poppling around this forum, and others like it, and I keep noticing something-

Whenever RAM limitations are mentioned, people simply shrug that XXXGB is all that this motherboard or that motherboard will "recognize".

So I understand the limitations of a 32-bit processor- they can't physically address more than 4GB.

But when confronted with, for example, a pair of DDR2 PC2-4200 sockets, does anyone know exactly what makes a motherboard decide, "well, that's it. No more RAM for you"? I have laptops that, for all practical purposes, should be able to have a nice 4GB set of PC2-4200 DDR2 stuffed in there, but they can't.

Just because "the motherboard limit" has made it so. It just simply "won't recognize" it.

Bear in mind, I'm really just curious about this, about these hardcoded limitations that companies have embedded in this motherboard designs.

[RealBlackStuff]

See if this gets you anywhere: http://www.pchelpforum.com/xf/threads/w ... mit.55820/

[DanielJamesDavis]

It was getting interesting, and the topic creator got exactly to the question I had...

And then the thread got locked.

I guess I just never have understood why an artificial RAM limitation exists on a motherboard; if a SODIMM of the correct size exists, it just is kind of baffling that the computer should just throw up its hands and ignore it.

Of course, I grew up with computers like the Mac SE/30; it was a machine built during an era where 1MB to 4MB of RAM was considered "standard"... and yet it supported up to 128MB of RAM.

Artificial limitations make me itchy.

[automobus]

That is a brilliant question. Thank you for giving me something to ponder.

The discussion which RBS mentioned is a let-down indeed. Not worth reading.

My only guess, is the address bus width of the memory controller. This is the only example I have in mind: Pentium III chips with 36-bit addressing. Although Intel positioned Xeon CPUs for the server market, there were many third party server boards designed for the regular Pentium III. In a consumer desktop board, the northbridge would put out a meager PCI 32/33 bus, and it was completely passively cooled (no heatsink/fan). The northbridge in a server board was considerably more capable, often hosting dual 64/66 PCI or PCI-X busses. Such a chip likely had a heatsink on it, and was not limited to a consumer's amount of memory!

I do not mean to imply that one does not still today find server boards for consumer CPUs. (I have no idea.)



However they accomplish memory limits, I suspect it is completely intentional. Chipzilla knows exactly what it is doing. It could be a case of planned obsoletion. Or it could be perfectly calculated for maximum business/success. They look at all data, consider everything: sales, customer input (referring to Apple and Dell, not you), history and projection of market size, Moore's law and such. Make one of those statistical curves, of specs-vs-usage (ports, speeds, power states, memory channels and sizes). The layperson, using only graphics integrated in CPU and no add-in cards, in not taking advantage of even half of those ports which the chipset can supply. Naturally, they plan to sell the most, make the most money.

Chipzilla makes a lot of chips. Wafers are bigger than ever, and total production numbers are higher than ever. The number of a mainstream northbridge/hub (or whatever that part of the Core i-prime number platform is called) produced could be one million or even higher. Thousands of brain-hours plan the platform, maybe tens of thousands of hours design that IC of one million transistors. They do not produce a full spread of chipsets with varying capability. It is not like the range of clothing sizes in department stores. They make a chipset with six (or so) SATA ports, twelve (or so) USB ports, and that one size fits ninety-nine percent of all consumers.

Need more from the computing platform? The step-up in class is a magnitude higher in price.

I am making this all up. No proof, no evidence, no insider knowledge, and I did not even look at Xeon prices and specs of the past few years. The theory makes sense to me.

[Cigarguy]

The CPU and chipset are factors. I've got an Atom netbook that I'd love to upgrade the memory to 4 GB but neither the chipset nor N270 CPU can handle more than 2 GB.

[ajkula66]

I guess I just never have understood why an artificial RAM limitation exists on a motherboard; if a SODIMM of the correct size exists, it just is kind of baffling that the computer should just throw up its hands and ignore it.

While I'm cynical enough to believe that "planned obsolescence" is an important part of today's industry of any kind, one small observation on my end...

The oldest ThinkPad currently in use in my household (A31p) which houses twice the amount of RAM that IBM had originally claimed it was capable of handling (2GB vs. 1GB) is running happily, and its bottlenecks are the CPU's 512KB L2 cache and aging GPU. The RAM limit would not even come close to the top the improvements its design would call for if one were to re-engineer it a decade later...with the hindsight of everything that happened in the meantime...

[jdrou]

One aspect of this is that you need to separate the advertised limits from the "actual" limits. It's very common for a laptop or motherboard to be released and advertised with a certain maximum memory capacity based on the fact that no higher-density modules existed when it was being designed. In many cases new modules are later released which allow more than the official capacity (sometimes requiring a BIOS update for best results). Even when a BIOS update specifically mentions adding support for larger modules (as in the Thinkpad 600E) the published specs for the model might never be updated to match.

This is somewhat related to the situation with the Thinkpad W520 which has a chipset designed to run memory at up to DDR3-1600 but no modules of that speed were available to Lenovo to test while designing the laptop so they only support DDR3-1333. DDR3-1600 modules actually do run at full speed though unless you install the recent BIOS update that capped it.