This is getting more hilarious than the PCM/PRAM scam! So Meg Whitman's company and Hynix, two companies that can't get their own houses in order, are going to have a chip in 18 months? If this is a ploy for Mr. Williams to somehow keep his job a few more months, surely it won't work.
What is interesting is that for PCM and now for Memistors both IBM and now HP/Hynix have abandoned the dreaded product road map and replaced it with specific NV memory targets. IBM has 2016 for PCM in its servers and now we have HP/Hynix claiming they will have devices that will challenge Flash in 2013.
This allows the question, what will need to happen as the intermediate steps in order to achieve those particular target dates. In house, chip design, mask sets, development and proving of the fabrication process, full characterization and full reliability testing. Plus all the other steps that allow cost competitive memory product to be put on the market
What will it take to be competitive with Flash? Will there be a continuity of chip bit capacities for HP/Hynix or will they go for the big bang-that now needs to be 1G-bit plus. My view with respect to the 2013 target for HP/Hynix Memistor, is unless they are well advanced with respect to most of the in-house steps mentioned above the target is unlikely to be met and the motive behind making such an announcement now must be questioned.
Unless, in the very near future (early 2012) Samsung, Micron or Hynix announce a performance and price competitive scaled 1G-bit PCM product, I cannot find any way to be optimistic that even the long-range claim of 2016 by IBM for their PCM application will actually be met by a PCM based device.
Somebody is going to make one of these variants work commercially... the basic physics is sound and reproducible. Whether that will be HP/Hynix on their claimed schedule .. I wouldn't bet on that.
My guess is that we'll see it first in really high-end server hardware, or possibly "big iron," for cache. IBM rather aggressively added eDRAM L3 cache to the Power7s ... which I thought was a bit gutsy since they are sold as Hi_Rel. IBM might be the first to roll out memristor cache.
Kudos to HP Labs and Stan for getting to a point where a product is on the--albeit distant--horizon. Clearly there are many pitfalls between now and 2013 they will have to avoid, but this is (literally) promising.
"However, in 2008 researchers from HP published a paper in Nature that tied the hysterical I-V characteristics of two-terminal titanium oxide devices to the memristor prediction of Chua."
It's hysterical, all right.
Blog Doing Math in FPGAs Tom Burke 24 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 ...