back to article Resistance is not futile: Here's a cookie sheet of luke-warm RRAM that proves it

Boffins devising Resistive RAM (RRAM) have found that using porous silicon oxide makes the devices easier to manufacture, longer lasting and less power hungry. The researchers, at Rice University, Texas, stumbled upon silicon oxide RRAM five years ago. The stuff works as computer storage by having differing resistance levels …

Bronze badge

Au or PT

That is EXPENSIVE

I just bought a platinum ring for my GF and I can tell.. 1500$/Oz -> about 50$ per GRAM.

0
5

Re: Au or PT

And a typical RRAM die will probably use a tiny fraction of a gram of Pt/Au.

Anyway, what matters here is scale of manufacture, and hence the cost of the product, as well as the capacity offered, and the speed it can operate at and bandwidth it can offer.

7
0
Silver badge

re. dielectric material

A dielectric material is not defined by its lack of ability to conduct electricity; the property of being dielectric refers to its ability to be polarised by an electric field. It may be that the silicon dioxide has dielectirc propertes that are useful in this application, or it may be that its properties as an insulator are why it is used. Either way, some stable, reversible and non-destructive change is taking place.

3
0

"excellent cycling endurance"

Unless that's expressed in the form Ax10^B cycles, where B is a very large number I'm not interested. I've had quite enough of these "solid state" technologies that manage to succumb to wear.

1
1
Gold badge
Go

Sounds potentally very good.

Reasonable materials (for semiconductor mfg)

Reasonable voltages.

Density and cycle life however....

It's V 0.1 tech but it might have some legs.

3
0
Bronze badge

Re: Sounds potentally very good.

Abstract of actual paper gives sub-50ns switching speed which is very good, but IIRC (corrections welcome) memristors are promising sub-10ns latencies. While sub-50ns is excellent, sub-10ns is revolutionary, as it would allow CPUs to drop cache memory.

Of course both technologies are in early stages. Perhaps RRAM will put a little more pressure on HP to develop memristors right (or provide them alternative to switch into, if they fail to do it), so this development is encouraging even if one is cheering HP efforts.

0
0
Silver badge

Re: Sounds potentally very good.

No it wouldn't. Even if the RAM chips could achieve that type of speed, the long wires of the memory bus could not. At those frequencies, a wire isn't a wire - it's a capacitor.

There's an exception to this though - it would let you skip the cache in embedded systems. It's commonplace in that field to use SoCs where the memory is on the same die as the processor, or at least in the same package. Faster performance, and save some real estate, which translates to cheaper chips. The mobile phone industry would be very happy.

0
0
Bronze badge

Re: Sounds potentally very good.

No it wouldn't. Even if the RAM chips could achieve that type of speed, the long wires of the memory bus could not

... hence HP work on photonic connections between memory and CPU, alongside with work on memristors.

0
0
Anonymous Coward

9 bits per cell?

That would require the writing and detection of 512 discrete levels per cell. NAND Flash seems maxed out at 3 bits per cell, which only (!) requires 8 levels.

0
0

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Forums