Good news has arrived for those watching the development of the next type of storage medium, there has been new information about Spin Transfer Torque MRAM published. One of the major hurdles in the development of the new type of memory, apart from yields, has been predicting the performance of MRAM cells. The Register have linked to an article on IEEE, jointly published by IBM and Samsung, which details how new STT-MRAM materials fabbed at the 11nm behave. We are still a while off of STT-MRAM hitting the market but it continues to draw closer as researchers try to bring us the next generation of storage media.
"IBM and Samsung scientists have published an IEEE paper demonstrating switching MRAM cells for 655 devices with diameters ranging from 50 down to 11 nanometers in just 10 nanoseconds using only 7.5 microamperes. They say it is a significant achievement towards the development of Spin Torque MRAM."
Here is some more Tech News from around the web:
- Capacity limits are utter tosh: Toshiba fattens SSD, disk with flash layers, helium @ The Register
- New DNA 'hard drive' could keep files intact for millions of years @ The Register
- Free Upgrade To Windows 10 Mobile Will Continue Past July 29 @ Slashdot
- Linux Lite 3: The Ideal Platform for Old Hardware and New Users @ Linux.com
- Facebook Messenger is getting end-to-end encryption @ The Inquirer
- Avast woos AVG shareholders with $1.3bn buyout offer @ The Register
- Corsair LAPDOG Gaming Control Center @ techPowerUp
- Asus' RT-AC88U router & RP-AC68U repeater @ The Tech Report
I have seen a huge number of
I have seen a huge number of articles over the years talking about next gen memory or storage tech, and mostly no commercial products have come out of it. While a lot of R&D goes into new types of technology, probably a lot more R&D goes into extending and improving current technologies. This usually leaves the new technologies well behind the curve, such that a commercial product is not viable. I have seen this happen repeatedly in many different sectors. The “old” technology is running out of steam and isn’t going to scale, so a “new” technology is needed. By the time the “new” technology is even close to ready, the “old” technology has been scaled beyond where the “new” technology is. Or it is close enough that a massive switch over is not economical.
I hope some of this technology actually makes it to market soon. We do have the Intel x-point stuff, which is assumed to be phase change memory, I guess. Phase change memory has been talked about for a very long time also, as if a commercial product would be availabile soon, but no commercial products have come out yet. The next revolution seems to be more in packaging technology with 2.5D and 3D die stacking. DRAM access latency hasn’t decreased that much compared to how much everything else has scaled. The latency of an individual cell has been hidden by bringing larger and larger numbers of cells into temporary SRAM buffers. If this type of memory can deliver lower latency at the cell level, then it could get use in some specialized areas that could then lead to wider general use. The silicon interposer or actual 3D stacking technology could lead to making a separate L3 or L4 cache die. This would allow smaller die to increase yields. If this new memory falls somewhere in between SRAM and DRAM, then it could be useful for such devices even if it is not a direct competitor to DRAM (depends on density and power consumption, in addition to access latency). It also would fit well with the ability of die stacking tech to mix die made on completely different process technologies. Given past history though, I wouldn’t expect any commercial products soon.
STT-MRAM is already on the
STT-MRAM is already on the market and has been for years, it just hasn’t reached the densities needed to take on Flash, but its increasingly used in niche roles such as the automotive and enterprise space.
Everspin (used to be a part of Freescale) have been selling STT-MRAM since 2012. They just started selling a DDR3 compatible 256Mb chip, with a 1Gb chip expected to sample by the end of the year. Its definitely not DRAM or Flash like density, but its got other benefits such as no refresh requirement (unlike DRAM) and is word addressable (unlike flash which is block addressable), so random IO performance is extremely high. Access to modern sub-45nm fabs has been the main thing holding Everspin back from releasing higher densities (and no, I don’t work for Everspin, they are just one example of a current STT-MRAM porducer).
Did you seriously just link
Did you seriously just link to an article that is linked to another article? Why couldnt you just link the correct article to begin with rather than expect people to go to one and then click to another? Clumsy feeling.