LONDON – A team of students and researchers from the University of California, San Diego, is about to demonstrate a phase-change memory solid-state storage system that it claims is up to seven times faster than current solid-state drives.
The device, called Moneta, was developed in the Computer Science and Engineering department at the University of California San Diego Jacobs School of Engineering and is set to be demonstrated June 6 to June 8 at the Design Automation Conference, which is being held in San Diego, this week.
Unfortunately, UCSD did not disclose the size of the SSD or how many memory modules or 128-Mbit chips are in use, or the connection methods used. According to a paper that appears to be earmarked for presentation at the Hot Storage 2011 event later this month, the Onyx memory module, contains 80 128-Mbit phase-change memories and connects to the host system using the PCIe bus standard. UCSD did say that the team has received help from Micron Technology Inc., which has provided non-volatile phase-change memories, BEEcube and Xilinx Inc.
Although faster than magnetic hard disk drives, flash memory based solid-state drives (SSDs) are still too slow to meet high performance computing demands where there is the necessity to sift through data quickly, the University of California San Diego (UCSD) said. Examples where PCM solid-state storage could help include analyzing scientific data collected by environmental sensors and search engines, UCSD added.
Phase-change memory stores data in terms of the difference resistance of a chalcogenide material depending on whether it is a crystalline and amorphous state. Data is written by the application of heat through an electrical current and read by use of a smaller current to determine which state the material is in.
@dylan.mcgrath: The companies have very little to gain from techno-Ponzis in the long run. In the short run, however, the motivation is stealing investor money. For example, Intel got investor money and bank loans for the money-losing Numonyx by hyping PCM. Samsung tried to mask its failure to gain market share in the NOR business by lying about PCM in its 2004 investor presentation. BAE was gunning for military/space contracts. IBM was losing share in the storage market and had to come up with some fairy tale. ECD and Ovonyx have been complete frauds from the get go. But really, the motivation combo is universal - ego/money. Take Mr. Lai of Intel, for example. He had done good work on NOR, but totally missed NAND and the fact that Flash can scale, so he had to hype PCM for almost 10 years to deflect attention from his failure and keep his job. The motivation for the other individuals is similar (the details vary a bit, though).
I understand that PCM has its share of detractors and that it has been one of those technologies that has for a very long time always seemed to be just a few years away. I know plenty of people feel as you do that it will never be commercialized in volume, and for all I know you may well be right about that. But what would companies like Intel, Samsung, IBM, etc. have to gain by throwing precious R&D dollars at a technology that they know is not viable? How do the orchestrators benefit from the alleged "techno-Ponzi scheme" as you see it?
Well, Mr. Clarke, Fusion-io's SLC Flash drive is 640GB. To replicate such a storage capacity, one needs about 40,000 128Mbit PCM chips, which will be good enough to cook a few meals (and, no, even the huge heat sinks won't be able to help). And, of course, nobody would be able to afford it, given that each PCM chip sells for about $4 a piece. Good luck with that!
@dylan.mcgrath: I have been following the PCM "developments" over the years. I can prove that most of the key "players" (corporations and individuals) in the PCM "space" have consistently deceived the public and investors regarding the performance, commercialization status, and market demand for PCM virtually since its "invention" in the late 60s, early 70s. That evidence covers actions and statements by Intel, Numonyx, Samsung, STM, BAE, IBM, Ovonyx, and ECD, and their officers and employees. In short, PCM is the longest-running techno-Ponzi I am aware of - the scammers drain R&D and investor money with promises that the commercialization (in volume) is just around the corner (typically, the drain lasts 10 years, after which they take a break of a few years and start again).
PCM will never be commercialized in volume. It is simply too slow and power-hungry in write, unstable, costly, and lacks density and scalability (and, again, note that it was invented before the first DRAM or Flash went on the market). You may want to read Mr. Neale's series of articles in EETimes Design to understand why PCM cannot scale.
Old news--Intel beat them to the punch. "Intel Corp. and STMicroelectronics reached a key industry milestone today as they began shipping prototype samples of a future product using a new, innovative memory technology called Phase Change Memory (PCM)." http://www.eetimes.com/electronics-products/other/4101982/Intel-and-STMicroelectronics-Phase-change-memory-is-now-sampling
Oh, wait, that article is from EETimes, 2/12/2008.
How many articles can this magazine post that make similar claims about PCM but with different casts of characters?
The team of students and researchers from the UCSD are either woefully clueless or they are simply aspiring fraudsters. Commercial Flash SSDs (such as Fusion-io's) can write and read at 1.5 gigabytes per second, something that the PCM-based Moneta can't match (it can get only to 1.1 gigabytes per second read, best case, and less than 500 megabytes per second write, best case). PCM sucks as a server storage memory - it is too slow and too power-hungry in write, has horrible density, tolerates higher temperatures poorly (note the huge sinks in the Moneta "device"), is too expensive, and cannot scale. PCM will never be commercialized in volume, in any application. It is a techno-Ponzi scheme that has run its course.
Mr. Clarke, of course, knew all that, yet he continues to perpetuate the techno-Ponzi. Great job, EETimes!
A fundamental problem with PCM is its performance including retention is very sensitive to temperature.
I would be curious how this group would interface the PCM. Is it still like Flash, with page- or block-based access?
Also, as I mentioned elsewhere, the Micron product is a NOR drop-in, so it would suggest a NOR Flash could also be the basis of SSD.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.