When you buy through linkup on our site , we may take in an affiliate military commission . Here ’s how it works .
investigator have developed a fresh eccentric of memory cell that can both store information and do high - speed , gamey - efficiency calculations .
The storage cell enables users to draw high - speed computations inside the computer storage array , researchers reported Oct. 23 in the journalNature Photonics . The faster processing speeds and low muscularity consumption could aid scale up data centers forartificial intelligence(AI ) system .
The magneto-optic memory cell design could one day reduce the energy required to power AI computing farms, researchers said.
" There ’s a tidy sum of index and a lot of energy being put into scaling up data point center or computing farms that have chiliad of GPUs [ nontextual matter processing whole ] that are running simultaneously , " subject co - authorNathan Youngblood , an electric and computer engineer at the University of Pittsburgh , told Live Science . " And the result has n’t inevitably been to make things more efficient . It ’s just been to corrupt more and more GPUs and pass more and more king . So if optics can address some of the same problems and do it more expeditiously and faster , that would hopefully ensue in quash power consumption and higher throughput machine learning system . "
The new cell uses magnetised fields to direct an incoming low-cal signal either clockwise or counterclockwise through a ring - shaped resonating chamber , a component that intensifies Light Within of sure wavelengths , and into one of two output ports . Depending on the intensity of ignitor at each of the output ports , the memory cell can encode a number between zero and one , or between zero and minus one . Unlike traditional remembering cells , which only encode note value of zero or one in one bit of information , the unexampled cell can encode several non - integer values , provide it to hive away up to 3.5 bit per cell .
Related : New ' Pbit - scale ' visual magnetic disc can hive away as much information as 15,000 DVDs
Those counterclockwise and clockwise twinkle signals are akin to " two runner on a track that are running in opposite directions around the trail , and the tip is always in the face of one and to the back of the other . One can go quicker than the other , " Youngblood say .. " You ’re compare the speed at which those two runners are run around the cut , and that set aside you to basically code both positive and electronegative number . "
The number that ensue from this subspecies around the ring resonating chamber could be used to either strengthen or weaken connexion between nodes in artificial neuronal meshing , which are auto learning algorithms that litigate datum in ways like to the human brain . That could help the neuronal web identify object in an picture , for object lesson , Youngblood say .
Unlike traditional information processing system , which make computing in a central processing building block then send results to store , the new memory board cells do gamey - speed computations inside the storage regalia itself . In - retentivity computing is particularly useful for applications programme like artificial intelligence information that need to sue a tidy sum of datum very quickly , Youngblood said .
— ' Quantum CD ' could entertain up to 1,000 times more data than today ’s optical disks
— Google ’s Sycamore quantum computer Saratoga chip can now exceed the fastest supercomputer , new study paint a picture
— Quantum memory discovery ' may head to a quantum cyberspace
The researchers also demonstrated the endurance of the magneto - optic cells . They ran more than 2 billion write and erase cycles on the cellphone without discover any degradation in operation , which is a 1,000 - fold betterment over past photonic memory technologies , the researcher wrote . distinctive flash drive are limited to between 10,000 and 100,000 write and erase round , Youngblood said .
In the future tense , Youngblood and his colleagues hope to put multiple cells onto a computer bit and essay more advanced computations .
Eventually , this technology could aid palliate the amount of power needed to run unreal intelligence systems , Youngblood said .
