Computer Memory Can Be Read With a Flash of Light 69
ananyo writes "A new kind of computer memory can be read 10,000 times faster than flash memory using pulses of light, taking advantage of principles used in solar panel design. Researchers built the prototype device using bismuth ferrite. In conventional computer memory, information is stored in cells that hold different amounts of electric charge, each representing a binary '1' or '0.' Bismuth ferrite, by contrast, and can represent those binary digits, or bits, as one of two polarization states, and, because of its photovoltaic properties, can switch between these states in response to visible light."
The NSA! (Score:5, Funny)
Re: (Score:2)
Re:The NSA! (Score:5, Funny)
Depending on the data, some computer memory is read with a Fleshlight.
Re: (Score:2)
Re:Call it "Photonic Memory". (Score:4, Funny)
In any case : don't call it Flash Memory... that would be dumb...
Re: (Score:1)
Sir, we have another dissident here. He replied with 'no comment' when he was supposed to toe the party line.
Re: (Score:3, Informative)
No. SRAM and DRAM are not particularly faster than flash for read operations. The bigger impact on flash vs. SRAM is that SRAM is often on chip whereas flash is stuck behind a slow interface.
Flash is many times slower for erase and write operations.
It works on light flashes? what do we call it? (Score:4, Funny)
>Hey, lets use light flashes and solar tech to store memory!
Nah, we can't do that, the name Flash is already taken for memory; try another idea.
Re:Know what else is 10,000x faster than flash? (Score:5, Informative)
No. SRAM and DRAM are not particularly faster than flash for read operations.
A typical flash chip has a read latency of around 50us (MLC) or 25us (SLC) and can operate one transfer per cycle at about 50-100MHz. A typical DRAM chip has a read latency of around 15ns and can operate two transfers per cycle at about 266-333MHz. A typical SRAM chip has a latency of about 10ns and can operate two transfers per cycle at similar rates to the DRAM.
Depending on the measure you use, Flash is between 7 and 3,000 times slower at reading than DRAM, and up to 5,000 times slower than SRAM.
Re: (Score:3, Funny)
Re: (Score:2)
See my other response. Those facts are lacking factuality.
Re: (Score:2)
Guards!
Re: (Score:1)
ONFI NAND is a DDR interface with transfer speeds of > 300MT/s. Otherwise spot on. Do not forget that DRAM and SRAM have effectively unlimited write cycles while NAND is in the 10's of thousands.
Re: (Score:2)
50us? 50-100MHz? 1/50E-6 = 20KHz, not 50MHz.
http://download.micron.com/pdf/datasheets/flash/nand/2_4_8gb_nand_m49a.pdf [micron.com] -> Sequential READ: 30ns
I.E. half the speed of 15ns DRAM.
Parallel read stuff is a bit slower, but not a lot. You can pay more for faster and you can always wire it up in parallel.
http://download.micron.com/pdf/datasheets/flash/qflash/MT28F640J3.pdf [micron.com]
SRAM speed depends entirely on the context, of which there are many. The on chip ones I use take less than 1ns to read on a modern silicon p
Re: (Score:2)
Old school flash is the best though.
http://www.xilinx.com/products/boards/ml410/datasheets/M29F040B.pdf [xilinx.com]
Re: (Score:1)
A simple microcontroler (like Microchip's PIC16xxx) has FLASH memory and has access to it at the same speed as RAM. Running at clocks from 20 to 40MHz.
Not to mention ARM's and other similar devices who work at much more speed... smarphones maybe ?
Access to flash may be crippled by interface in the case you mention.
Re:Know what else is 10,000x faster than flash? (Score:5, Informative)
...
Uhm, Flash is orders of magnitude slower than DRAM, which is orders of magnitude slower then SRAM.
SRAM is on chip because its expensive and its interface is expensive so you'd waste a lot of money and effort to have a SRAM discreet component on the other side of the mobo. Designing an interconnect to put the L1/L2 cache in an external chip is a REAL PITA by itself for consumer devices.
Looks like modern SLC NAND goes at roughly 100ns access times.
DDR3 - 2000 has a 9ns access time.
That DDR3 is roughly on par with the SRAM cache on my old 486, which was about 10ns. I have no clue what the internal L1 latency is on something like an i7, but considering that the MHZ is a couple orders a few orders of magnitude higher, I can assume its off the scale faster in comparison to NAND.
NAND is fast compared to a slow spinning platter disk that has to move a head and a platter into the right position to even get started. Its not really fast otherwise unless you do massively parallel reads. NAND has to do parallel (multiple chips) reads just to keep up with current SATA speeds. The interface isn't the issue.
Re: (Score:2)
You just explained why the interface is a significant issue, then your last sentence was that the interface isn't the issue.
Re: (Score:2)
> Flash is orders of magnitude slower than DRAM
You will find that the datasheets I posted elsewhere in this thread show flash at a semiconductor technology level to be 2-10 times slower than modern DRAM. That is not 'orders of magnitude'.
Re: (Score:3)
> Flash is orders of magnitude slower than DRAM
You will find that the datasheets I posted elsewhere in this thread show flash at a semiconductor technology level to be 2-10 times slower than modern DRAM. That is not 'orders of magnitude'.
It is for powers of 2.
Re: (Score:3)
Re: (Score:2)
Re:Know what else is 10,000x faster than flash? (Score:5, Informative)
Bismuth memory? (Score:5, Funny)
Re:Bismuth memory? (Score:5, Funny)
Re: (Score:2)
I zinc some people just like to throw a tantalum and there's no curium.
Re: (Score:1)
There's no bismuth like show bismuth
New and improved Pepto now treats cache misses!
Will be hard to work into the "Heartburn, nausea, indigestion, upset stomach, diarrhea!" jingle, though.
10 micrometres wide (Score:5, Funny)
From the article: "10 micrometres wide"
So move on. There's nothing to be seen here.
Re:10 micrometres wide (Score:4, Insightful)
Re: (Score:1)
Yeah, no kidding. If I could see something ten micrometers wide I'd be working for Intel.
Re: (Score:2)
That would be funny. A lab technician standing next to you instead of a microscope(scanning electron... whatever they use) and asking you what you see and you trying to describe it to them in words, or drawing a little picture, lol. Not trying to be mean, just a funny image to me.
Re: (Score:2)
That would be funny. A lab technician standing next to you instead of a microscope(scanning electron... whatever they use) and asking you what you see and you trying to describe it to them in words, or drawing a little picture, lol. Not trying to be mean, just a funny image to me.
Almost... From The Far Side: It's a Mammoth [chemistry-blog.com]
Re: (Score:2)
Right but it was the first and only transistor at the time, so it was competitive.
However a 100Kbit, quite fast memory that is expensive because it's made on a low volume manufacturing line, without the benefit of the billions of dollars spent developing silicon based manufacturing equipment is hardly going to be competitive in the market and won't be able to generate the cash necessary to build up a manufacturing infrastructure that can compete with cheaper, higher volume memories that are a 1000 times mor
Re: (Score:2)
Re: (Score:2)
LOL You don't hear the grapevine! You hear IT THROUGH the grapevine. I.e. if there were a grapevine wall and people talking on the otherside that didn't know you were there. The grapevine itself doesn't talk, it's a freaking plant!
Re: (Score:2)
...and I'm just about to loooose my mind, honey, honey yeah.
Bipolar memory? (Score:5, Funny)
Do we really want bipolar memory with its constant cycling between manic and depressive states?
Comparison to PCM (Score:5, Informative)
The link to the actual Nature Communications paper is here: Non-volatile memory based on the ferroelectric photovoltaic effect [nature.com].
This somehow resembles Phase-Change Memory [wikipedia.org] (PCM). PCM devices are composed of a material which, under a high current, there is a thermal fusion and changes to a different material status, from amorphous to crystalline. This changes two properties: light reflectivity (exploited in CDs and DVDs) and electrical resistance (exploited in emerging non-volatile PCM memories). The paper cites PCM and other types of emerging non-volating memories.
In this case, it is the polarization what changes, without requiring a thermal fusion, therefore increasing the endurance of the device, one of the main shortcomings of PCM. The other main shortcoming of PCM is write speed due to the slow thermal process, in the paper they claim something like 10ns. If this can be manufactured with a large scale of integration and low cost, it will probably be a revolution in computer architecture.
Re:Comparison to PCM (Score:4)
Re: (Score:1)
What I find intriguing about the method is that it seems to imply that since the write and read methods are different, you could achieve asynchronous reads and writes -- which could be good or bad, depending on what's happening. Definitely a boost to some custom applications, and a possible revolution in some niche processor architectures. Not sure if it will be viable for generic computation systems.
Flash! (Score:1)
Bismuth = Radioactive (Score:1)
Bismuth is very slightly radioactive, not sure I'd trust memory that is generating it's own bit-rot via alpha decay.
Re: (Score:1)
Bismuth is very slightly radioactive, not sure I'd trust memory that is generating it's own bit-rot via alpha decay.
For very slight values of radioactive. It's half-life time is ~10^20 years. I think I can we can have enough error correction bits to deal with that.