I'm not sure of the relevance of the plumbing of the Shuttle to Akida. The Shuttle DH410 is an COTS product to which the Akida PCIe board is designed to plug in. Peter and team did not include the cooling system. It was already included as part of the Shuttle DH410.Well I just finished reading many comments on these and thanks @Diogenese, @JDelekto and @alwaysgreen for your thoughts.
@Fact Finder I saw you had made fun of me again due to the delayed reply but apologies for having a life outside TSEx and BRN. We had some visitors for dinner last night.
I see that you haven't answered my question in 2 but is referring to a statement from Peter. I acknowledge that he was speaking to shareholders and not to a tech audience. So he may have used certain examples or words that meant something similar to make the message more comprehensible, or maybe he just got carried away but I wasn't in there so I don't know in what context he said that.
But I don't believe that and think GPUs are still required in cars in addition to Akida. If I'm not correct I would like to be pointed to at least one presentation slide or other company source that claims that Akida can replace GPUs in cars or other graphics-heavy applications like gaming. My opinion on this is actually perfectly summed up by @JDelekto.
I never said that Akida is only suitable for Edge applications. What I said was it would outperform GPUs in edge applications when it comes to power consumption and cost. I believe that is why the company is targeting edge use cases.
Regarding the use of Akida in data centres, from memory what the company said was they can take some load off data centres by processing data at the edge device and thereby eliminating the need to send data to the data centre for processing. Would love to be corrected here as well and learn how Akida can fit into a data centre (To be used for general data processing like streaming, banking transactions, Office365 data, etc and not for HPC use cases as described in Sandia report because those are specific use cases)
With that in mind I would like to point you to a couple of statements from the same Sandia research you quoted above;
"In optimal cases, neuromorphic computers will solve problems faster and use less energy than conventional computing, he said."
"The bold assertions should be of interest to the high-performance computing community because finding capabilities to solve statistical problems is of increasing concern, Brad said." - So this is not one size fits all and there are specific use cases where SNN may outperform GPUs.
"The claims are not meant to challenge the primacy of standard computing methods used to run utilities, desktops and phones. “There are, however, areas in which the combination of computing speed and lower energy costs may make neuromorphic computing the ultimately desirable choice,” he said." - So we can't just dismiss GPUs and CPUs just because SNN can perform better in specific use cases.
Therefore still you can't say "AKIDA 1.0 out performs GPUs and CPUs." in a general way without mentioning in what way or what use case.
Anil's quote you have given is actually how you should have made your claim as well. There he is explaining that he is comparing the power consumption of Akida and a GPU when performing the same task.
On another note, if you look at Akida Shuttle PC System Information, it has an Intel Core i5, i7 or i9 processor to run the OS. It also includes a Heatpipe cooling system with two fans. Now given that Peter and the team understand the full power of SNN computing (which I don't challenge) why would they go through all that trouble to include a crappy intel CPU in there when it could have been replaced by a superior Akida 1.0?
https://us.shuttle.com/products/dh410/
The Akida Shuttle PCIe development kit is intended for demonstration/familiarization:
https://shop.brainchipinc.com/products/akida-enablement-platform-shuttle-pc
" DEVELOPMENT KITS ARE NOT INTENDED TO BE USED FOR PRODUCTION PURPOSES"
Last edited: