SiDEvans
Regular
Here’s one for the diary. Not to be missed.
https://us06web.zoom.us/webinar/register/WN_ieaIY14tQ6G0xMP1HBipLg
When I look on twitter there are a ALOT of other organisations retweeting Edge Impulse and Brainchip announcements...View attachment 13847
Here’s one for the diary. Not to be missed.
https://us06web.zoom.us/webinar/register/WN_ieaIY14tQ6G0xMP1HBipLg
The reason for my question mark was the following statement:
Neuromorphic object localization using resistive memories and ultrasonic transducers
https://www.eetimes.com/neuromorphic-device-with-low-power-consumption/ => https://www.nature.com/articles/s41467-022-31157-y.pdf
BRN - WBT and Edge Impulse?
I am an absolute beginner as far as ReRAM is concerned. As much as I would have liked it to be true that BRN, WBT and Edge Impulse somehow work together this cannot be deduced from this technical article. First, I checked the authors to see if there was any overlap. There are not. Then I took a closer look at the ~Latin - which reads like this to me. There a few clear sentences stand out:
"Two different fabricated RRAM circuits are used in this article. The first is a 16,384 (16k) device array (128 × 128 devices) of one-transistor/one-resistor, 1T1R, structures. The second chip is the neuromorphic platform presented in Fig. 4a. The RRAM cell consists of HfO2 5-nm-thick film sandwiched in a TiN/HfO2/Ti/TiN stack. The RRAM stack is integrated into the back end Of Line (BEOL) of a standard 130 nm CMOS process"
I would exclude the first (1T1R) one via this text passage:
"The 1T1R (one transistor one resistor) architecture used in embedded ReRAM arrays is not sufficient to support the large arrays of memory cells needed in discrete memory chips. For this reason, Weebit’s crossbar arrays were developed using a 1S1R (one selector one resistor) architecture that enables the high density needed for discrete chips. Such an architecture also allows..."
https://embeddedcomputing.com/techn...-demonstrates-its-first-crossbar-reram-arrays
I would exclude the second (HfO2) one via this text passage:
"Most ReRAM designs utilise a special phase-changing material (such as HfO2) controlled using a standard NMOS transistor. When a high current is passed through the resistive material, its resistance changes and the material's resistance is also dependent on the direction of the current. Therefore, when the bit is read, a small gate voltage is applied to the transistor which allows current to flow through the phase-changing material, and the current that results determine the state of the bit.
Weebit Nano is an Israeli company formed in 2015 to develop ReRAM for next-generation computing systems. The growing need for non-volatile memory in computing systems that allows for unlimited read and write cycles presents ReRAM a real opportunity to become an established memory technology.
Unlike most other ReRAM technologies, Weebit Nano has focused on using Silicon Oxide as their phase-changing material."
https://www.electropages.com/blog/2021/05/weebitnano-patent-and-what-reram
View attachment 13845
The only consolation is that Edge Impulse with WBT ReRAM would get the better results in any case if I summarize it correctly. They made a conscious decision not to use either manufacturing process. Maybe Weebit should give EI a call.
So I assume that this time it is not yet so far. But as already mentioned, my conclusion is based on my naivety. I only looked for a clear logical connection and found this contradiction. So I can be wrong. @Slymeat please correct me if you see it differently.
Guys do you mind if I revisit the Volterra thing. In the screenshot below at the bottom a bloke makes a link with us and Volterra, and makes a strong statement. Am I flogging a dead horse? If so I apologise and also go straight to the naughty corner.
View attachment 13844
Thank you DiogenesPatents last for 20 years. Our first one was filed in 2008, but we have an ongoing programme of patent filing to protect new developments.
Just saw that.... WOW haven't seen that kind of positive indicative for a LOOOOONG time .... lets hope it stays there!!
Well done @cosors, this is an interesting find and you do seem to be gaining a good understanding of ReRAM. I agree that EI should give Weebit a call, or vice versa, Weebit’s rechnology of sandwitching SiOx between two TiN layers has been proven to work and integrates very well with standard CMOS. The massive advantage is that is is simple for fabs to work with and importantly requires no exotic materials.
Neuromorphic object localization using resistive memories and ultrasonic transducers
https://www.eetimes.com/neuromorphic-device-with-low-power-consumption/ => https://www.nature.com/articles/s41467-022-31157-y.pdf
BRN - WBT and Edge Impulse?
I am an absolute beginner as far as ReRAM is concerned. As much as I would have liked it to be true that BRN, WBT and Edge Impulse somehow work together this cannot be deduced from this technical article. First, I checked the authors to see if there was any overlap. There are not. Then I took a closer look at the ~Latin - which reads like this to me. There a few clear sentences stand out:
"Two different fabricated RRAM circuits are used in this article. The first is a 16,384 (16k) device array (128 × 128 devices) of one-transistor/one-resistor, 1T1R, structures. The second chip is the neuromorphic platform presented in Fig. 4a. The RRAM cell consists of HfO2 5-nm-thick film sandwiched in a TiN/HfO2/Ti/TiN stack. The RRAM stack is integrated into the back end Of Line (BEOL) of a standard 130 nm CMOS process"
I would exclude the first (1T1R) one via this text passage:
"The 1T1R (one transistor one resistor) architecture used in embedded ReRAM arrays is not sufficient to support the large arrays of memory cells needed in discrete memory chips. For this reason, Weebit’s crossbar arrays were developed using a 1S1R (one selector one resistor) architecture that enables the high density needed for discrete chips. Such an architecture also allows..."
https://embeddedcomputing.com/techn...-demonstrates-its-first-crossbar-reram-arrays
I would exclude the second (HfO2) one via this text passage:
"Most ReRAM designs utilise a special phase-changing material (such as HfO2) controlled using a standard NMOS transistor. When a high current is passed through the resistive material, its resistance changes and the material's resistance is also dependent on the direction of the current. Therefore, when the bit is read, a small gate voltage is applied to the transistor which allows current to flow through the phase-changing material, and the current that results determine the state of the bit.
Weebit Nano is an Israeli company formed in 2015 to develop ReRAM for next-generation computing systems. The growing need for non-volatile memory in computing systems that allows for unlimited read and write cycles presents ReRAM a real opportunity to become an established memory technology.
Unlike most other ReRAM technologies, Weebit Nano has focused on using Silicon Oxide as their phase-changing material."
https://www.electropages.com/blog/2021/05/weebitnano-patent-and-what-reram
View attachment 13845
The only consolation is that Edge Impulse with WBT ReRAM would get the better results in any case if I summarize it correctly. They made a conscious decision not to use either manufacturing process. Maybe Weebit should give EI a call.
So I assume that this time it is not yet so far. But as already mentioned, my conclusion is based on my naivety. I only looked for a clear logical connection and found this contradiction. So I can be wrong. @Slymeat please correct me if you see it differently.
Guys do you mind if I revisit the Volterra thing. In the screenshot below at the bottom a bloke makes a link with us and Volterra, and makes a strong statement. Am I flogging a dead horse? If so I apologise and also go straight to the naughty corner.
View attachment 13844
Hats of to this bloke, he seems to really know his stuff - even if he is smoking a spliff! He got a lot of "likes" too.
View attachment 13857
View attachment 13859
Taking a leaf out of @SERA2g book
"He sounds like a fucking idiot"
But in all serious, isn't it a bit embarrassing commiting to the idea without any factual basis?
What impact does this have on the BrainChip brand?
View attachment 13864
Taking a leaf out of @SERA2g book
"He sounds like a fucking idiot"
But in all serious, isn't it a bit embarrassing commiting to the idea without any factual basis?
What impact does this have on the BrainChip brand?
View attachment 13864