Memory Prices: What to Do? Part III

by Anna Filatova
05/25/2002 | 11:08 AM

Well, it’s high time we released our third part of the analytical story devoted to the situation in the memory market called "Memory prices: What to do?", which was first posted on our site last November when the memory prices suddenly started growing up very rapidly this way waking up a lot of concerns and unpleasant emotions in our souls (see here).

In fact our constant Hardware News reader do know that since the Part II of our story, we’ve been following carefully the memory market that is why the factual info we are going to share with you today will hardly be anything new to you. I will just do my best to sum everything up for your convenience and to draw some conclusions.<%BANNER[article]%>

Let us start with the graphs showing the changes in the memory pricing within the period from January 16 to May 23 (the data is borrowed from DRAMeXchnage: red stands for the max price, blue – for the min price and green – for the average).

In fact, these graphs allow us to tack the major events and tendencies emerged in the memory market during the last 4 months. Here falls the absolutely unexplainable price drop in the end of January (see this news story), the critical rates of $9 and $9.5 for 256Mbit chips and inability to reach the sacred $5 for 128Mbit chips where the memory makers were going to stop, and the smooth price reduction which then turned into a rush price drop down to $3 (see this news story), $2.5 (see this news story) and even $2 (see this news story) for a 128Mbit chip.

To illustrate the mentioned things I will later offer you the current DIMM prices, however, before that let me veer a bit away, to explain why I selected Pricewatch as the pricing reference and not any other price-list. In fact, it is all very simple: my experience with Pricewatch showed that some local prices, for instance, are highly sensitive to any price growth and are hardly that eager to go down as the prices decrease. So, it seems to me that Pricewatch is a more reliable source to base the analysis and conclusions on. :) Especially, since every local market will sooner or later get close to the world market in terms of pricing. Here are the first 20 price offers from Pricewatch:

• 256MB PC2100 DDR SDRAM DIMM modules - $25-35 (January 16 - $60-$80);
• 256MB PC133 SDRAM DIMM modules - $25-35 (January 16 - $36-$60);
• 128MB PC2100 DDR SDRAM DIMM modules - $16-28;
• 128MB PC133 SDRAM DIMM modules - $10-23.

• Forecast 1: "the memory prices for both types will not grow up tangibly in the nearest future".
  The results are displayed on the graphs above: there was no significant price growth. The only exception appeared 256Mbit PC133 chips (their price grew up from $8 to $9.5), but this exception is very logical, because the industry is shifting to DDR SDRAM and 256Mbit chips. therefore, it should be more convenient for the manufacturers to shift directly to 256Mbit DDR266 chips rather than to 256Mbit PC133 (the second version is to continue manufacturing 128Mbit PC133 chips).
• Forecast 2: "by Q2 (and maybe even earlier) DDR SDRAM prices should start going down".
  The result is again displayed on the graphs above: the prices started going down in late Q1 – early Q2.
• Forecast 3: "PC133 SDRAM should also start getting cheaper together with DDR SDRAM, however, this will not last long". The time for this forecast hasn’t come yet, because it is more long-term. The best illustration here is the situation with 256Mbit PC133 SDRAM modules described above.
• Forecast 4, which is also a long-term one: "the price difference between DDR SDRAM and SDRAM in favour of the latter will never again be as huge as it used to be". Well, it looks as if things would take this particular way. Although this forecast was the most evident of all, we will not dwell on it. Please, feel free to make the conclusions yourselves.

• DDR SDRAM shortage, which appeared in the market in the beginning of this year, has been successfully eliminated. At present the largest memory makers have enough chips in stock to last them for 1-2 months even if the whole production suddenly stops.
• The memory manufacturers keep moving from 128Mbit chips to 256Mbit chips, and some of them (such as Samsung) have already started working on 512Mbit chips production.
• Of course, the migration from PC133 SDRAM to DDR SDRAM continues making the latter a new mass standard.
• The today’s most mass memory type is DDR266 SDRAM (PC2100), however, the market share of DDR333 (PC2700) will be growing little by little especially when Intel launches its chipsets with DDR333 support (as we have already found out it is expected to happen in September). DDR-I 400 (PC3200) also starts spreading, although it will be hard for this memory type to become popular unless JEDEC approves its as a standard.
• In addition to all the problems the memory manufacturers had to overcome, there is a "dead season" coming to the memory market, which will last until August 2002.
• The negotiations between Micron and Hynix about the purchase of the latter’s fabs are at a dead end. That is why Hynix will keep hanging between heaven and earth, searching for finance to last them another bit of time (partially by selling some parts of its business, which are not connected with DRAM production). Anyway, the South Korean government will never let this company go bankrupt.
• None of the memory makers considers it possible to reduce the memory production. Moreover, most of them try to think over the possibility to increase the amount of wafers produced (as well as their diameter, we are talking about the migration to 12-inch wafers).
• The memory market stays pretty "full" and nothing signals any changes. Even Microsoft coders are unable to occupy all the memory used in most contemporary PCs.

• During the coming two months DDR SDRAM prices are most likely to get lower in the near future, although just a very tiny bit, if ever.
• DDR SDRAM prices might start going a little bit up in August (when the seasonal demand drop is over) and especially in September-October, after the announcement of Intel i845GE and i845PE with the official support of PC2700 DDR SDRAM. As you understand, the prices of this particular memory type will go up first, however, the increase will undoubtedly touch upon other memory standards as well.
• As for SDRAM memory, it is no so interesting any more to make any forecasts here. The last but one bastion of PC133 SDRAM, Celeron CPUs, started moving to Pentium 4 platform and DDR SDRAM (i845/E and i845G/GL chipsets), and the very last bastion, Duron processors combined with KT133A chipset, will also very soon become a history. So, the price of PC133 SDRAM will most probably be determined by the ability of the memory makers to make their production lines use more progressive technologies, and not by the market situation. It is evident that 256Mbit PC133 SDRAM chips will never become as popular as their 128Mbit brothers, which means that their prices will never be attractive enough for the serious customers. The price of 128Mbit chips is absolutely impossible to predict. We know that their production is reduced little by little, but how fast is this tempo, no one knows except the memory manufacturers. Anyway, PC133 SDRAM is going to the past, so let’s leave the past for the historians. :)

• To cut the long story short, the best times for buying memory have come. And this favourable season will last until the end of summer, I suppose. So, if you were thinking about adding a bit more memory to your PC, it is a great opportunity for you to put your plans into life without much risk for your pocket.
• Going on with this idea, I would like to point out that buying systems with PC133 SDRAM appears a waste of money now. Yes, DDR SDRAM costs a bit more, but firstly, this is a very small price difference, and secondly, systems with DDR SDRAM are much faster and more prospective than its SDRAM counterparts. So, my next advice will be to equip the new systems with DDR SDRAM, if you are not trying to save the last penny.
• As for the optimal memory size for a contemporary PC, I believe nothing has changed since we put up the first story on the memory tendencies: at present 256MB is the most optimal choice for most PCs not intended for heavy loads (work with huge images in Photoshop, etc. – even 2MB may appear too little for applications like that). And 512MB will be more than enough for these monsters as well. I don’t think that it makes sense to increase the memory size. So, I think nothing is going to change here within at least 1.5 years.