ISSCC Although some industry observers – as The Reg recently noted – say that flash memory is approaching a technical brick wall, the cofounder and former CEO and chairman of SanDisk sees things differently.
"Industry experts – even the best ones – are often shortsighted, and sometimes outright wrong," said Eli Harari at ISSCC on Monday. "And therefore, perseverance in the face of naysayers sometimes does pay off. And in the case of NAND, big-time."
Emphasizing that he was speaking for himself – he retired from SanDisk in 2010 – Harari made what he identified as "bold predictions" about the future of NAND, namely that it will not only overtake and supplant most if not all hard drive–based storage, but also that DRAM is on its hit list, as well.
Harari did admit that the future won't be a cakewalk. "There's no question that NAND faces significant challenges as we march from where we are today, 19 nanometer, to close to 10 nanometer, we believe, over the next several years."
However, he noted that as the rate of NAND scaling is slowing, the industry is developing new NAND variants such as planar cells, hybrid cells, three-dimensional arrays, and the like.
"The most important element for future scaling," Harari opined, "is availability of cost-effective EUV – extreme ultraviolet lithography." He said that although EUV can be accomplished today, the required wafer-baking throughput hasn't yet been achieved.
That said, he believes that highly scaled NAND with strong support from advances in flash system architecture will "dominate the decade." In the post-NAND world, Harari is putting his money on 3D resistive RAM, which he said "has made excellent progress in recent years, and as a true three-dimensional memory, I believe it has a real shot at becoming the next big game-changer in the second half of this decade."
By 2020, Harari said, three-dimensional resistive RAM (3D-ReRAM) may quite possibly be the "checkmate for the hard disk drive industry."
Although SSDs aren't going to fully replace hard disk drives anytime soon, he says, it's important to note that mobile computing is already ushering in the hard drive's doom. "In smartphones, tablets, and ultrabooks, flash is inside. That game is over."
But flash is also the future in the enterprise and in the cloud, he believes. "Flash SSD is absolutely revolutionizing enterprise storage," Harari said. "Facebook and YouTube and all those apps are just driving IOPS – input/output operations per second – just driving them through the roof. Flash SSD just demolishes hard disk drives in IOPS."
IOPS aren't flash's only advantage, said the man who built his career on flash, citing such other upsides as low latency, massive parallelism, low power, and crash resistance.
Cloud computing, Harari believes, is "uncharted territory." The problem with the combination of the cloud and mobile devices, he said, is wireless bandwidth, which is finite. "They don't make any more bandwidth anymore," he said.
Bandwith usage is also expensive, he said, citing a figure of $10 per gigabyte per month. "You can buy for $10 today 8GB of flash memory – and you don't have to pay every month," he said, noting that with a large amount of flash in your mobile device, you can download content onto it during less-expensive, off-peak hours, or free over Wi-Fi.
"One more thing," Harari concluded. "NAND, I believe, will disrupt DRAM. Today, the cost of NAND per gigabyte is 10 times lower than the cost of DRAM ... and that's not likely to change. If anything, it's going to change in favor of NAND, because DRAM scaling is even more challenging than NAND."
With a 10X price differential, he argued, having ten times as much flash as system memory in a PC is a clear choice. "The question is, can 10 gigabytes of NAND or one gigabyte of DRAM give you a better performance boost? And the answer is NAND does a better job."
Not only does the price differential argue for NAND, Harari said, but so does the fact that it's non-volatile, and therefore requires zero standby power. "These are compelling reasons for system architects to design future platforms that maximize NAND and minimize DRAM," he suggested.
"I believe that by 2020, flash – which is highly scaled NAND and 3D resistive RAM – will be the undisputed king of storage," he predicted.
You may agree or disagree with Harari's predictions, but you can't gainsay his conviction. When talking about the early days of NAND, he reminisced about being "ridiculed by the competition" because of the expense of NAND and its onboard system controller versus Intel's far cheaper software-controlled NOR.
But SanDisk hung on, and NAND is now ubiquitous – and if Harari is right, about to become even more central to systems.
"The lesson here," he said, "is you should stick to your convictions, particularly when your gut tells you that you are right." Certainly, technical problems abound for flash, but from Harari's point of view, its challenges are problems to be solved, not brick walls. ®