Wouldn't it be crazy/awesome if, after years of cost-cutting and right-sizing and "pivoting" and various CEO's gimmicks, the thing that saved HP was something that came out of their R&D department?

If it happens, hopefully people will learn the right lessons.

EE Times has a more detailed article with some interesting statistics:

"Read times are less than 10 nanoseconds and write/erase times are about 0.1-ns. HP is still accumulating endurance cycle data at 10^12 cycles and the retention times are measured in years"

http://www.eetimes.com/electronics-news/4229171/HP-Hynix-to-...

I guess that 0.1 ns must be 0.1 us. As stated (also in eetimes), writing and erasing will be at least an order of magnitude faster than reading. That seems weird as writing typically takes more energy then reading.

The only reason I can think of for this to be reversed would be that reading will be destructive. That, however, would make me wonder whether this can be a good replacement for Flash. Think about it: reading the memory would run the risk of erasing it.

Well, 0.1 us write times wouldn't make sense with the claim that this can replace DRAM and SRAM, but those claims do seem rather dubious anyway...

Don't link to slashdot, link to the original article: http://www.electronicsweekly.com/Articles/06/10/2011/51988/i...

I did, yesterday. It didn't make the front page.

I really hope this works out because I really want to play around with some programmable memristers and this would help make them cheap.

How reliable are memristors, as compared to HDDs and SSDs?

Potentially more reliable than both, but that remains to be proven out. HDDs have a 1E14 unrecoverable harder error rate, flash is higher (1E15) The lack of moving parts is a plus for SSDs and memristors but the susceptibility to tunnelling electrons is higher on flash/memristors than in the magnetic domain of HDDs. When they were first talked about at ISSCC the temperature sensitivity of memristors was also mentioned.

I know, not a lot of useful info, my notes are sparse on these guys. I figured we'd see pattern memory (IBM) sooner. Maybe we'll get both, that would suck since they both need high initial costs to quickly offset their development costs. Hard to do in a strongly competitive environment.

[1] http://spectrum.ieee.org/semiconductors/processors/how-we-fo...

[2] http://news.cnet.com/8301-11386_3-20026553-76.html

[3] http://www.sciencemag.org/content/330/6012/1810.abstract

10^12 cycles with retention measured in years sound much better than flash (i.e. SSD).

this is really encouraging!

This is really exciting, I think there's a lot of potential in memristors, so seeing it used for a practical purposes is great news.