BIG BLUE BOFFINS claim to have developed a prototype of what they think is the semiconductor industry's smallest, densest and fastest on-chip dynamic memory.
The 32nm silicon-on-insulator (SOI) technology is supposed to go like the clappers, use less juice and be more reliable than the sun coming up.
In a statement, IBM said its SOI technology can provide up to 30 per cent performance improvement with up to 40 per cent power reduction, compared to standard bulk silicon memory technology.
Basically transistors on the chip are protected with a "blanket" of insulation that reduces electrical leakage. That saves power by allowing current to flow through the circuit more efficiently, thereby keeping heat down and warding off the boogie man underneath the bed.
So far IBM has made a test chip with this embedded dynamic random access memory (eDRAM) technology that it says contains the industry's smallest memory cells.
The company claims it offers higher density, speed and capacity better than conventional on-chip static random access memory (SRAM) that has been announced in 32nm and 22nm technology. It thinks its eDRAM is comparable to what would be expected of SRAM produced on a 15nm process, three technology generations ahead of chips in volume production today.
IBM's eDRAM cell is twice as dense as any announced 22nm embedded SRAM cell and up to four times as dense as any comparable 32nm embedded SRAM in the industry.
Common thinking is that higher memory density can lead to chips that are smaller, more efficient and can process more data, improving system performance.
The device apparently manages latency and cycle times of less than two nanoseconds. It also uses four times less standby power and has up to a thousand times lower soft-error rate.
IBM said that embedded memory is the key to controlling multicore processors and other integrated circuits, and that its prototype has many implications for the future of computing technology.
It claims that use of this technology in high-performance server, printer, storage and networking applications can result in improved system performance and energy savings. In mobile, consumer and game applications, it can result in smaller systems, lower cost and energy savings, IBM said.
Big Blue is looking at manufacturing the devices, but has not predicted when it will go into production. µ
MB buffer, for ram and HD!
Is that one quarter the power? If so, say so "four times less" is nonsense.
Dynamic memory is Vast improvement, in few articles earlier, 22 nm is mentioned & Intel is planning to wow world with few 22nm sampling in new 32 nm world, Coming on d' morrow. yet dynamic can outperform 18 nm.
it is so much of improvement, IBM is planning to incorporate 32 nm dynamic memory into ALL its present equipment in BIg change over.
With useages going far beyond todays mortal 'puter, it has added bonus of doing things never done before.
In past computer was limited to dying after brief spurt, which 'puter did often, thus NO off switch, it'd die forever. Now computers are far moore advanced with atx tech, So it makes sense cpu & main start working properly together, too.
DRASHEK
Does it make sense then for AMD to put these into their CPUs and replace the SRAM?
Wasn't AMD and IBM working on Z-RAM using SOI? Last I heard they were waiting for a smaller process to manufacture it (32nm or so).
Heres bit more with diagrams Nathan Brockwood emailed:
http://www.semiconductor.net/article/354546-IBM_Readies_32_nm_eDRAM_With_Low_Latency.php
Yes, there is also going to be 22 nm soi of tech. Although NOT available to Public.
This allows 45 nm 4 core cpu to have 4X more cache in its inards. Power 7 will be 32 Mb of Cache.
Drashek M.D.
Just say 1/4th, -3 doesn't make any sense here even though that's what your statement amounts to.
Had a similar problem in a science(!) magazine a few days ago. It said that the budget on something had been reduced by 100%. So I went to the graph and discovered it had actually gone down by half, indicating that the person who'd done the numbers wasn't as smart as they thought they were. Didn't check the other numbers, but simply assumed that the percentage numbers were crap(rubbish for those of you who speak English east of the Atlantic).
Mathematics is the ONLY universal language, let's try to be consistent, mmkay?