Memory, Evolution or Revolution?

Memory
Evolution or Revolution?

(Summary)

Does implementation of DDR signal the end of Rambus?

We don’t believe it does, but what it may be is a clear signal to Rambus, Inc. to clean up the nasty licensing problems and end the needless litigation over them and concentrate on quality and pricing. Indeed, DDR SDRAM and DDR-II look extremely promising as the next generation memory, but isn’t that what the memory makers said about SLDRAM and Rambus three years ago?

Keep in mind that even though some of the biggest memory manufacturers such as Micron, Samsung, Hyundai and Kingston for example, are pushing DDR technology really hard, they are also manufacturing and selling DRDRAM!

When Extended Output DRAM (EDO) was thought to be the latest and greatest, it was quickly overshadowed by the latest SDRAM developments, PC 100 and PC 133. Has EDO died? not hardly. There are literally hundreds of thousands of computers still using EDO and it should be obvious to everyone that it will be here for quite some time. Today, you can pickup a well configured EDO based computer for approximately the same money as one of the cheap $500-$800 units using Celeron processors. The only difference is the Celeron may only have 32Mb of memory, while the EDO based machine with a Pentium II may have 128Mb or more. For word processing, light to medium graphics (Adobe Photoshop), Internet surfing and email, they make great machines.

Now that everyone is using PC 100 and PC 133 SDRAM, the pundits are saying that it too is on its way into the history books, and that it was being replaced by DRDRAM. Shortly after issuing those statements, the pundits started griping about Rambus licensing, the high cost of Rambus and their perception of an illegitimate relationship between Rambus Inc. and Intel. If this weren’t bad enough, a few exclusionary tests were run comparing the new DDR DRAM and Rambus and these same pundits stated Rambus was trash and would be replaced by DDR DRAM.

Conclusion

When reviewing any statements regarding the performance of any new technology, especially when there is testing involved and that testing involves comparisons, there are a few issues to keep in mind:

1. Is software involved in the testing process, such as the comparisons between DDR and Rambus? If so, is the software optimized to take advantage of both forms of technology equally? If not, were different software test packages used in order to obtain an objective result? Think about this. How can you run objective tests on the latest motherboard chipsets, processors and memory technologies released in 2000 or 2001 using Winstone^® 99 or other similar package developed a year or more before the release of the technology. This by no means reflects on the software, but rather it reflects on the common sense of those running the tests and putting them in print.
2. Are the components being tested appropriately matched? Motherboards with specific chipsets should be appropriately mated with the processors and memory components around which they were designed to support. Although we will catch some flack for saying this, but some of the comparative tests performed early in 2000 comparing Rambus on an Intel I820 motherboard against motherboards optimized for the latest version of DDR were obviously slanted towards DDR. Furthermore, everyone in the industry knew that the i820 chipset had problems, and it Rambus had nothing to do with it. Ironically, tests involving Rambus were run on motherboards that had Intel’s i840 chipset (Intel’s OR840), and the results were amazing.
3. Are the results and comments being published those of the writer or is he/she echoing the opinions of others? Are the opinions that you are reading qualified? More important though, are the comments and results verifiable? Benchmarking results are important factors on which people rely to make decisions. We take benchmarking very seriously! So seriously in fact, that any benchmark that we run must be repeatable within 2% over ten (10) separate tests on randomly selected production machines. Are the results you read about verifiable and repeatable on equipment that has been randomly selected as opposed to a specific machine tweaked for testing?
4. When you read reviews about new technologies and some of the performance testing that occurs, make sure that the comparison testing is one of apples to apples. Computers have several different classifications, and range from very basic machines through to power-house workstations. Many motherboards will work reliably with either Celeron processors or the hottest Pentium III, or AMD’s Duron or Thunderbird, but the question is what processor was the motherboard optimized for? The same principle applies to motherboard chipsets and memory. This means that in order to make an intelligent decision about the review or comparison, you may need to do a little research to see if the review or comparison makes sense.
5. Lastly, if you remember anything at all from what we have published here, remember this. There is no single motherboard, processor or memory solution that fits every possible circumstance. To say otherwise is absolute foolishness and is nothing more than advertising hype. Most computer manufacturers that sell their mass produced computers expect you to either guess what your needs are, or to design your needs around their computer. That’s a bit of backward thinking now, isn’t it? Why not design your computer around your needs!

Memory technology is presently in an ever-changing state at the moment. Manufacturers are trying desperately to catch up with the speed of processor development, and it seems that they may have difficulty gaining real ground unless they are willing to address new technologies rather than just improving the present ones. If you would like to review what may lay ahead in memory technology, review our Memory Trends for 2001.

Please be sure to review our companion memory topics, and if you need help with visually identifying memory modules, try this link. Visually identifying your memory type.

Want to dig deeper into memory issues, including how memory speed is determined? If so, visit this review.

Is your PC-100 memory really PC-100 compliant? Want to know for sure? PC-100 Compliant Memory.

Is your PC-133 memory really PC-133 compliant? Want to know for sure? PC-133 Compliant Memory.

Do you know How Much Memory Do You Need?