The biggest question on everyone's mind (ok, maybe no the biggest but it is definitely an important one) is "what is up with the sharp increase in RAM prices?"  Although I don't have a single "it's his fault" finger pointing answer for you all, I do have quite a few sources that I managed to talk to and quite a bit of information that they provided me with.  There are potentially a number of causes of this DRAM price hike we've been forced to endure and what I'd like to touch on are a few of the most likely (although not necessarily true *cough* disclaimer *cough*) reasons you have to pay over twice as much for a 128MB DIMM now than you did just a few weeks ago.

I mentioned in a front page update a while ago that most vendors and distributors I talked to pointed their fingers at one company in particular as the cause of the memory price hikes, and no, it's not Intel (they've caused other things, leave them alone this time, they didn't do anything ;)...)   That company seems to be Micron, now I'm not pointing any fingers nor am I making any accusations, I'm only passing on the information that I've received from more than one generally reliable source.  It seems as if Micron's semi-recent acquisition of Texas Instruments has made Micron quite the player in the US market, let's take a look at a little quote about the acquisition made back in June '98:

"This strategic acquisition will enhance Micron's position as the most cost-effective memory producer in the world, by leveraging our leading-edge technology into existing fabs without significantly increasing R&D, administrative and operating costs," said Steve Appleton, chairman, CEO and president of Micron. "The additional global capabilities, including participation in a unique joint-venture manufacturing strategy, positions Micron to take advantage of future markets."

It seemed innocent enough at the time, however there has been quite a bit of finger pointing recently. Many say that due to Micron/TI's strong presence in the US, with no reason to competitively price DRAM products due to pressure from Taiwanese/Korean makers (which would be courtesy of the tariffs placed on Taiwanese chip manufacturers), they have been simply letting the prices go up mainly because they have no reason to remain competitively priced - aka the classic monopoly theory in action.  Now keep in mind, that if this is the case, it is only contributing to a portion (if any part at all) of the recent hikes.  So, what else could be causing them?

With the transition to RDRAM (Rambus DRAM) being pushed for by Intel (ok, you can blame them for that, but really, they didn't do anything wrong this time) in time for their September 27th release of the Camino chipset - which will support SDRAM - memory manufacturers are currently trying to juggle more than a few balls here.  Just look at what the industry is asking them to do: produce PC66 modules (seemingly not that great of a demand, but there are still PC66 compliant systems out there guys), produce PC100 modules, produce PC133 modules (just in case RDRAM fails miserably), and now produce RDRAM.  That's a lot of balls they're being asked to juggle, and frankly, it makes perfect sense for them to have trouble cranking out four different flavors of memory. 

A shortage of regular SDRAM modules due to the pressure for a shift to RDRAM is another possibility for such a drastic price hike, however on the other side of the issue we have the OEMs and distributors that actually sell the stuff.  They are claiming the opposite, in that quite a few OEMs and distributors don't understand the price hike at all, stating that there couldn't possibly be a shortage as they are currently sitting on more than enough DRAM parts just waiting to be sold.  While this doesn't apply to everyone, it could very well be a contributing factor to the rocketing prices.

Going along with the shortage theory, is the theory that the move to smaller than 0.25 micron fabrication processes (0.22 micron, 0.18 micron) is leaving a number of chips unfit for sale.  This is a highly likely theory as venturing below the 0.25 micron mark is uncharted territory, even Intel has yet to prove whether or not they can produce high enough yields on their 0.18 micron Coppermine processors to make the switch profitable.  How does this affect 0.25 micron parts that aren't going to be made using the new fab processes?  Well, to upgrade the fab plants you naturally have to close down the fab, and what does that contribute to?  A chip shortage. 

And we have yet another possibility, this is a report I haven't heard too much about however one vendor is stating that they've been told of some very large OEM purchases (*cough* Dell/Compaq *cough* first they [Compaq] take our Athlons, now they take our memory) of memory thus removing the modules we all want from the hands they need to be in (ours of course).  Most of these items of speculation seem to point at the same conclusion, a shortage, is that the case?  I don't think anyone but the manufacturers themselves can offer an answer to that question, but just for your own information, I've managed to put together a quick graph of the price of a 128MB SDRAM module (PC100) from August 3, 1999 up to September 8, 1999 below from a particular vendor. 

Not too appealing at all is it?   So the big question is, what do you do if you are in the market for memory today?   The best thing to do is to wait, the memory prices will go back down, it is only a matter of time. 

Some of the most extreme reports indicate that the chips alone for a 128MB memory module are going for around $14 a piece (8x8 chips), multiply that number by 16 (the number of chips on a 128MB SDRAM module) and you get the cost of the 128MB of memory on a stick of SDRAM, this is not including the cost of the PCB nor the manufacturing costs either.  Surprising, isn't it? 

Stay away from memory right now, it is not a good time to buy, but I'm sure you are already well aware of that.  After the initial batches of RDRAM begin making their way into the hands of the vendors/distributors from the manufacturers, we should see some order restored to the SDRAM market.  Memory manufacturers currently in quite a bit of a bind in terms of what they can do right now, Intel is pressuring them to produce RDRAM, while the rest of the industry wants PC133 compliant SDRAM.  What is a company to do?

RDRAM - 800MHz of crap?

I'm sure you've read all about the benchmarks on RDRAM and how the i820 (Camino) is inferior in every single way to the BX chipset and the usual rants and raves.  I'm here to tell you that not all of these benchmarks are unfounded, in fact, there is quite a bit of truth to the current performance of RDRAM on the 820. 

On the one hand, Intel is claiming that RDRAM's greater bandwidth allows it to perform much better than SDRAM, and on the other VIA (and family) is claiming that SDRAM's lower latency allows it to outperform the more expensive RDRAM technology.  I'm not here to answer that question (as that would involve publishing benchmark results which I'm not yet ready to do, however I will say that after quite a few weeks of toying with the 820 and RDRAM, you will see a nice comparison of technologies on the 27th of September here on AnandTech), however I will offer you some words of advice (and some pictures of RDRAM as well) as you're looking at all of these documents either supporting or denouncing RDRAM as a memory solution be sure to pay attention to who is publishing the benchmarks. 

Of course Intel is going to be making sure that their benchmarks represent RDRAM's strengths, and of course VIA is going to be boasting about PC133 over RDRAM, it is up to you, the consumer, to make sure that you understand this and take a look at independent benchmark results.  The ideal comparison (in my mind) would be an Intel BX based motherboard and SDRAM running a Pentium III 600 (100MHz FSB) versus an Intel 820 based motherboard and RDRAM running a Pentium III 600 (100MHz FSB), then the same comparison using the 133MHz FSB.  And guess what I have running on two separate test beds in the lab? :)

Click to Enlarge

Here we have a Samsung RDRAM RIMM, notice the blue heatsink.  I bet you never thought RAM would come with heatsinks now did you? 

Click to Enlarge

And here we have a CRIMM module, which is essentially a terminator for any unoccupied RIMM slots on the motherboard.  Motherboards will ship with at least two of these CRIMM modules (depending on how many RIMM slots are present on the motherboard).

The High Density 256MB DIMM

Here's something the guys at Mushkin sent me, a 16 chip 256MB SDRAM DIMM using Toshiba SDRAM chips (16MB chips).  I'm messing with compatibility testing on this chip, but so far the outcome seems to be that this little (pun intended) puppy is much more compatible than the big monstrous 256MB modules we're used to. 

Take a look at the picture comparison below to see exactly what I'm talking about. 

Click to Enlarge

Here we have the "old" way of making 256MB modules, basically stacking two 128MB modules on top of each other.  This provided for quite a few incompatibilities with motherboards and did result in some system instability, since these modules were primarily used in servers, well, you can guess how great of an idea that was :)

Click to Enlarge

...and here is the consumer level 256MB module that Mushkin sent over, you'll start seeing more and more of these appear once DRAM prices drop to more reasonable levels. 

Click to Enlarge

Kingston also has had a module like that for quite some time, this is a registered DIMM and retails for considerably more however it provides the same basic functionality (18 chips on the Kingston due to an extra ECC bit on the Kingston chips).