RAMBUS DRDRAM

There are good and bad things about DRDRAM. Well, first of all DRDRAM has really fast operation. I mean really, really fast. How about 800 MHz? However, remember how SDRAM transmitted data in 64 bit increments? Well DRDRAM does it with only 16 bits. So it's fast, but has a smaller bus size. This overall increases memory bandwidth, but also latency. Just think of smaller chunks (1/4 size) moving at much higher speed. Is this much better or faster? We'll have to see once it comes out and is tested more thoroughly but initial reports show it to be very fast.

Not everything is milk and honey though

DRDRAM has its share of bad traits, though. First off, this technology is proprietary. This means that Rambus owns the patent on it, and other manufacturers that choose to manufacture this kind of RAM will have to pay royalties. Of course, this is great for Rambus, but for all the other manufacturers, this sucks. Another point on the downside is that initial estimates show that DRDRAM is very expensive. We're talking an estimated $400-$500 for a 128 MB RIMM. Ouch!

Hopefully, as manufacturing processes get better, prices will lower. Also, the it seems that the increased clock speed for this RAM causes a lot of problems as far as signal interference and EMI (Electro Magnetic Interference). What this means is that motherboard manufacturers have to better shield the motherboard's wiring. This is done by a process called POOL, which stands for Pad On Outer Layer, and by using PCB's (Printed Circuit Board) with more layers. POOL is basically adding a shielding/insulating element on the top layer of the motherboard to prevent interference. Obviously this will bump costs up on manufacturing.

Slow to adopt

Several motherboard manufacturers are taking a wait and see policy on DRDRAM. Also, we have word that some mobo manufacturers are considering supporting either DDR SDRAM or DRDRAM, but not both at the same time. The timings and technology is too different for initial support of both of these types of RAM, and supporting both on the same mobo will require a bridge to do so. What's their likelihood of success? Well, did I mention that Intel is backing this as their format of choice? This brings an entirely new factor into the race for technology, the political aspect of business partnerships and economic gains. We'll have to see if Intel's support alone can spur on the growth of the proprietary DRDRAM.

Opposing the forces of Rambus and Intel are a consortium of other RAM manufacturers. Since DRDRAM is proprietary, you can pretty much count on the fact that everybody else is banding together to create a different standard format with which manufacturers will be able to use and produce freely without paying royalties. Well, then we come up with SLDRAM. SLDRAM stands for SyncLink DRAM, but the SyncLink had to be abandoned due to some kind of legal issue. Thus, we're left with SLDRAM. SLDRAM is also a proprietary format, but this one is open to manufacturers. In other words, no royalty payments to manufacture this RAM will hopefully translate to lower cost of production, which in turn translates to lower price to consumers.

The SLDRAM Alternative

SLDRAM looks pretty good on paper, too. Its architecture already puts it up to 200 MHz. This is just a result of the way the module is set up. I haven't learned how the actual architecture differs, but once I find out more you can be sure there will be an addendum to this guide, for all of you who are interested. Now SLDRAM also does what DDR DRAM does- it has two signals per clock cycle. This puts the bus up to 400 MHz, in theory. There is potential for more, too. Reports say that SLDRAM has a maximum bandwidth of 3.2 GB versus a 1.6 GB bandwidth for DRDRAM. This is due in part to lower latency.

Just like the highly hyped and touted upcoming battle between the Katmai and the K7, we will have to wait and see for this battle, DRDRAM and SLDRAM. Both look great on paper and in theory, but theory doesn't mean jack until someone gets their hands on the hardware and we see the true testing.