SAN JOSE, Calif. - What will happen in the DRAM market in 2011? Here are five trends collected from various reports from analyst C.J. Muse of Barclays Capital:
1. Down on DRAMs
''2011 is likely to be the year of the NAND, not DRAM. What happened is that tablets started cannibalizing the lower end of the notebook/netbook market. In that situation, the per box GB DRAM content of the smaller size notebook of up to 2 GB was replaced by tablets that only housed 256 Mb or 1/8th the amount. Hence, DRAM oversupply.''
2. Going to 3x-nm node
''Overall, all the vendors are moving to various forms of 3x-nm. Both Samsung and Hynix are moving to 3x. Interestingly, Elpida is also trying to make the leap from 65-nm or 65xS (65-nm shrink) chips to 3x-nm, in a bid to drastically lower the costs.''
3. Korean duopoly
''In terms of the market, unlike in NAND, a tier system exists, but there exists a danger that it becomes a Korean duopoly. At present, only Samsung and Hynix can be called as Tier 1 vendors at the forefront (most profitable technology) whereas Micron, Nanya, Inotera (finances, technology slightly lagging) are probably Tier 2 and Elpida, Powerchip, Rexchip (technology, money both lagging but attempting to make a grand leap forward) are Tier 3.
Samsung is right at the top as far as DRAM is concerned. They have also cornered the market on the (ASML's) NXT (lithography) tools which are a necessity as far as making 30-nm DRAM (due to overlay requirements). So this year has been a Samsung success story where they have increased their DRAM market share. In contrast to the world wide DRAM bit growth of likely 45-50 percent, Samsung was able to post a DRAM growth rate of 70 percent.
Hynix is a respectable tier 1 name in DRAM just behind Samsung. The only problem now with Hynix is the debt overhang and the ownership structure which we hope will be resolved in 2011. In their case, they are producing DRAM at 4x and 3x-nm at M10 and Wuxi, (China) and they are only behind Samsung. Next year will see them further gaining steam as they move from an 8F2 to a 6F2 architecture.''
4. Tough times for Micron camp
''For a while it did appear that Micron might make it into Tier 1 and in fact, was rated as Tier 1 for some time in 2009. But that was contingent on perfect execution of the trench to stack DRAM capacitor geometry transition. While it started off well in the beginning of 2010 with a pull-in of orders, the yields have been found wanting and bit growth has been downshifted throughout the year.
The trench to stack transition has not gone smoothly (for Micron's Taiwan DRAM partners). Nanya, Inotera are still struggling with the yield at 50-nm, let alone what we expect will be tough times to come in the 40-nm and 30-nm transitions. But next year may be a different story as at least the trench to stack transition is compete. So perhaps this year has to be written off as a learning curve, while next year will be the transition to 4x and 3x-nm as Micron and its partners Nanya and Inotera do have the financial wherewithal (Formosa Plastics) that they can make progress. One indication is the fact that Micron can afford to guide to spend 2/3rds of their capex to NAND in 2011.''
5. Elpida's gamble
''Elpida has been trying its best to make it back as a premier player. First the Taiwan funding/government bailout of DRAM did not come through but the partnership with Rexchip/Powerchip/Promos remained. Now Elpida is trying to raise money in the financial markets in both Japan and Taiwan. In addition, on the technology end, the company came up with 65-nmXS, a shrink version of the 65-nm DRAM chip which allowed it to make profits in DRAM when the prices were still on an upswing. In addition, the company now is trying to do a leap to 40-nm and 30-nm.''
The historical trend shows that the price of NAND has reduced at a much faster rate than the price of DRAM. I think today cost per GB for NAND is less than the same for DRAM...correct? I don't think that price would be an issue.
The tablet market will definitely reduce the total number of DRAM's that are actually has to be used. But at the same time if the volumes of tablets and the smartphones go up we will see incresing demand for the DRAM. I feel 2011 will be an interesting year for DRAM.
Blog Doing Math in FPGAs Tom Burke 2 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...