iXBT Labs - Computer Hardware in Detail






DDR2 Memory: Near or Distant Future?

June 18, 2004

This material contains no diagrams or DDR2 test results, there is a whole article devoted to it. Because both the test material and the analytical one proved quite copious, we found it reasonable to separate them into two independent articles. Perhaps, some will content with nothing else but the test results — it's all right with us, we don't want to impose our opinion on anyone. Because yes, this article is just our opinion, our view of the standard's future. We deemed it appropriate to compile all we know about DDR2 in order to understand what we can expect of its appearance on the market and what prospects the standard itself has on it.

Speaking of things eternal...

We would like to begin by mentioning a good old marketing principle: if you want to sell a product to a man who has no need for it, then you should create this need in the first place. In our case, we should realise that such categories as "objective needs" or "true/false needs" can't actually be applied to the modern computer market at all. Did users have a need for DDR? No, they didn't. Users need neither DDRs, nor new CPUs, nor new graphic cards or HDDs. All they need is to have programs they deal with working nice and fast on their computers. But memory producers as well as CPU and chipset manufacturers decided that users needed DDR memory. Although, it was, in fact, like this: they needed them to need it.

Therefore, an ideological brainwashing was carried out (with the help of the media as well) under the motto "We all have a great need for higher memory bandwidth!". After that, chipsets, CPUs with fast and wide buses, and memory itself appeared. And then users who had voted for these symbols of progress with their own purses wanted to have a confirmation that it hadn't been in vain. This "second wave" created programs that demonstrated excellent results received due to fast memory. Were these results biased? Not at all, they were quite fair: in most cases, fast memory really improved things. Were these programs useless? Negative again: many of them were really helpful software that could solve serious problems. Would this software have appeared in this particular shape if it hadn't been for the DDR boom? No, but it would have appeared anyway, only in a different form.

If, for example, an increase in the CPU cache size was chosen for the aim instead of memory bandwidth, a similar ideological brainwashing would follow again (along with scores of systems with increased cache size appearing on the market). It would create a huge demand for programs working fast with large cache sizes. Different algorithms, different optimisation, but the same idea: everyone would see the proof that the course was right: an increased cache gives magnificent results. And no cheating! It's just the market chain in action: when there is no unique right way of development, in the end, the right way turns out to be the one that had managed to attract the majority of market subjects. That is why, whenever we estimate the prospects of a new technology, we should remember the following rules:

  1. None of the new technologies gives a qualitative leap forward as soon as it has appeared on the market. Why does this happen? Rule 2 explains it.
  2. You can use a chance only when there is one. It may sound trite, but once you understand it, the following thing becomes obvious: we can't predict the effect of an idea implemented hardwarily as most people have just never thought what can be done with this idea in terms of its practical use.
  3. When there are different ways to achieve the ultimate goal, completely right or completely wrong ways are far less numerous than partially right ones. You can focus on building good roads or constructing mighty ATVs. You can modify correction mechanisms and noise immunity or improve connection quality. At some point, both ways can speed up task fulfilment and increase user's comfort. But sooner or later, they both will become exhausted.
  4. The evaluation of the company's ability to promote a technology onto the market is no less important than the evaluation of the technology itself. Because it is only implemented ideas that can be useful for customers, others are only interesting for historians and a limited number of experts. The thought to reject a Wonderful Progressive Idea basing on doubts about its promoter's strength may sound cynical, but cynicism is just what we, customers, need if we don't want to become sponsors of a yet another Big Beautiful Burst Bubble.

DDR2 from the market point of view

What is DDR2, considering all we have said above? It is a new memory standard promoted by Intel. Potentially, it enables to reach higher frequencies and higher bandwidth. But apart from that, DDR2 has a number of inattractive features that mostly affect time latencies. At this particular moment, its concrete implementations (DIMMs) can't show us the strongs of DDR2, on the contrary, they underline its weaks. The reasons for that are quite material and objective: insufficient (even for DDR2-533) CPU bus bandwidth in today's Pentium 4 and absence of low-timing DIMMs from mass production.

Can these shortcomings be made up for in the future? In theory, DDR2 as a standard doesn't hinder that. But experts predict that most DDR2 modules will have bigger latencies than the actual DDR400 for quite a while. Although certainly, it doesn't deny the fact that DDR2 (beginning with DDR2-533) is better in terms of bandwidth. And we can also expect CPUs with a higher bus frequency to appear in the nearest future, which will eliminate one more shortcoming of actual DDR2 systems.

But what does it all mean for the market? To all appearance, one of the industy's flagships considers [peak] bandwidth more important than latency. Leading memory manufacturers are quite ready to support Intel's innitiative by producing corresponding chips and DIMMs. At the same time, they don't deny that the DDR market still looks attractive to them and they are not going to leave it as long as there's a demand for DDR.

It's easy to see that memory producers' attitude is more than just good for the end user, as it leaves it up to him to make the main decision. Certainly, the fact that "Intel itself" is behind DDR2 makes the standard look very significant. On the other hand, the presence of almost as fast DDR on the market reduces that "mythical attractiveness" of DDR2 and incites the user to pay more attention to the factual attractiveness, that is, the real advantage/disadvantage of DDR2 systems in frequently used applications. In other words, manufacturers can't count too much on gifts from users'. We dare to predict that the main migration to DDR2 will be caused not by leap expectations, but rather by physical evidence that performance growth is real and tangible. Besides, we shouldn't forget that the difference between DDR2 and DDR400 is not so great as was in the case of DDR266 and PC133. Those days, it was a twofold memory bandwidth increase (even though only maximal and theoretical). Today, DDR2-533, the fastest of all supported by chipsets, has just a 33-percent higher bandwidth than DDR400, while DDR2-400 has no advantage whatsoever. And Intel certainly realises that, otherwise their new chipsets wouldn't support both DDR2 and DDR. The company has learnt the Rambus lesson all too well and is not going to introduce the new standard with no alternative, as it can provoke a negative reaction from the market.

On the other hand, suppose AMD won't be too quick to support DDR2. Then the standard will be a handy tool for Intel to exert pressure on its rival, as it will create another obstacle for users' free interplatform transition. Indeed, it is still unclear whether the new AMD Socket 939 platform will be adapted for DDR2, and besides, the integration of the memory controller into the CPU creates problems too. Thus, AMD will have to support only DDR for quite a while. If Intel is quick enough to make DDR2 the main memory for its future Socket 775, then Intel and AMD platforms will become even less compatible (chipsets, boards, CPUs, and now memory types). Should this situation occur, new module and chip manufacturers (not large-scale ones) will appear that will either provide a full support for DDR2 or reject it altogether, thus choosing one of the rival platforms.

But strange as it may seem, this situation will be favourable for users too. Although this time, only for those preferring Socket 775 + DDR2. The thing is, Intel will have to resort to a very intensive DDR2 promotion, and there are only two ways to increase its factual attractiveness: price reduction or a quick frequency increase. In other words, the future of DDR2 modules will depend not only on objective, technological reasons (chipmakers' capabilities, prime cost), but also on how much Intel is ready to do for promotion.

Besides, we shouldn't forget that DDR2 has no alternative now in terms of global prospects. It is very doubtful that JEDEC approve DDR500 specification, as even DDR400 module validation created certain problems for chipmakers. Thus, sooner or later, we all will have to shift to DDR2. It is not so much about WILL it happen as about WHEN it will happen. Another important point is that we don't know which company will benefit the most from the mass introduction of the standard to the market. Intel and AMD have both advantages and disadvantages. The former, being a trailblazer, will have to make all the mistakes related to DDR2 support, although it will also gain a lot of experience. The latter will have no such experience, but it can make products for DDR400 meanwhile, standing aside and watching the events develop. The most importnat thing for AMD to do in this case is not to miss the right moment.

What have we got in the end?

  1. DDR2 as a standard is understandable and predictable: this memory type has rather big latencies (which is bad for specifically optimised software) but feature a twofold bandwidth increase (which is good for specifically optimised software).
  2. We have seen test results of a concrete DDR2 platform, and they are not very optimistic: weaks are visible, strongs aren't. But these results were quite predictable once you considered technical characteristics of the memory and the used CPU.
  3. The actual DDR2 itself will hardly become widespead, it needs a more appropriate entourage: at least, CPUs and chipsets that can make use of the increased bandwidth and get a real advantage from it.
  4. The standard's future depends on how soon technically more attractive modules will appear (e.g. DDR2-533 with timings 3-3-3 or DDR2-667) and how well they will be supported by corresponding CPUs and chipsets.
  5. DDR2 will be the leader tomorrow no matter what, as it has no alternative for the moment. The only scenario that can prevent it is CPU and chipset manufacturers declaring they no longer need memory bandwidth higher than actual DDR400 can offer. But that will hardly ever happen.
Stanislav Garmatyuk (nawhi@ixbt.com)


Write a comment below. No registration needed!

Article navigation:

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.