BRN Discussion Ongoing

Diogenese

Top 20
As per previous post, Intel is is supporting UCle (Universal Chiplet Innterconnect) which aims to create full interoperability between chiplets from different manufacturers, allowing chips to mix-and-match chiplets as chip makers see fit. Check out the partners (promoters) below.






View attachment 7340






Universal Chiplet Interconnect Express UCIe 1.0 Launched​


By Patrick Kennedy March 2, 2022



Universal Chiplet Interconnect Express UCIe 1.0 Cover Universal Chiplet Interconnect Express UCIe 1.0 Cover

Today’s big announcement is the Universal Chiplet Interconnect Express (or UCIe) industry effort. UCIe 1.0 is designed to allow silicon providers to focus on building best-of-breed silicon IP, then having the ability to package these chiplets together. We have been discussing some of the early mentions of these efforts at STH for some time, but now we get to put a name to it.

Universal Chiplet Interconnect Express UCIe 1.0 Launched​

As a bit of background here, companies that sell and consume systems need a way to customize chips for unique workloads. The chip industry has become so large that having a small number of monolithic dies servicing all segments does not work. Furthermore, there is now enough worldwide volume to allow more specialization.
A great example of this is the difference between HPC clusters, network firewalls, and web servers. HPC servers need parallel double-precision floating-point processing and high memory bandwidth. Network firewalls need cryptographic offloads, packet processing acceleration, and network interfaces. Web servers can get by with many cores and mostly focus on integer performance while needing to scale core counts. All of these chips are fundamentally different and will have different power, cooling, and networking needs.
Universal Chiplet Interconnect Express UCIe 1.0 Motivation Universal Chiplet Interconnect Express UCIe 1.0 Motivation
Chiplets provide the path to integrate IP blocks that serve various functions and then package them according to an application. When we hear Intel discuss its plans for its Intel Foundry Services (IFS) and how it plans to help customers, this is exactly why. Intel’s packaging technology allows it to not just manufacture its own chips, but also take chips from other foundries and integrate them into solutions, adding value atop of the entire stack by providing the packaging. If you read our Intel Ponte Vecchio is a Spaceship of a GPU piece, this is perhaps the biggest impact here. Intel is getting into multi-fab packaging, and Ponte Vecchio is a packaging showcase.
Intel Ponte Vecchio Suspended Intel Ponte Vecchio Suspended
AMD for its part has been packaging TSMC compute dies with Global Foundries I/O dies for years and we expect to see more complex packaging in the future. Although we have been discussing Intel a lot here, before we get too far into this, it is worth looking at the promoters:
UCIe Promoters UCIe Promoters
We have the four major hyper-scalers in the US, aside from AWS. We also have many of the world largest chip IP vendors. This is going to happen because big customers want it, and big chip companies are going to provide it. Getting back to the customization aspect, the cloud players may eventually decide to do things like integrating networking or even DPU-like features into chips to streamline their architectures. Since these large players make up such a large piece of the ecosystem, they want specialized chips that cater to their needs.
Saying that companies want to integrate is great, however, desire alone does not make chips. Instead, to get IP from different companies there needs to be some set of standards to allow this to happen. UCIe is that standards effort.
Universal Chiplet Interconnect Express UCIe 1.0 Key Metrics Universal Chiplet Interconnect Express UCIe 1.0 Key Metrics
This concept we have actually discussed before. In my discussion with Raja Koduri at Intel (see Raja’s Chip Notes) we discussed Lightbender Silicon photonics chiplets that will come in the Falcon Shores era for supercomputers. Raja’s key insight was that one of the biggest challenges with integrating Silicon photonics at Intel has been getting it into monolithic dies. By instead specing a set of interface and power parameters, Intel could have a chip that would fit in a certain size and power envelope (I am probably not supposed to talk about the power, but it makes a lot of sense), and with a set of interfaces, it basically makes future chip building like building software with API’s then building chips with the API’s provided by different teams. Another software way to think about it is like silicon containerization.
Universal Chiplet Interconnect Express UCIe 1.0 Characteristics And Key Metrics Universal Chiplet Interconnect Express UCIe 1.0 Characteristics And Key Metrics
At the heart of UCIe are some well-known technologies such as PCIe and CXL. This allows chiplets to interface in a common way and transition from external devices to chiplets. UCIe also addresses things like how chiplets are packaged. Of course, in the future, we will see more standardization on things like form factors. After all, a chiplet design becomes more powerful if one can swap new generations of technologies from different vendors as needed. This is just like PCIe cards, OCP NICs, SSDs, and more have standard form factors.
Jumpstarting Universal Chiplet Interconnect Express UCIe 1.0 Jumpstarting Universal Chiplet Interconnect Express UCIe 1.0
When one looks at the usage models for UCIe, it becomes clear what is happening here. Many of the items that used to be add-on boards or devices are going to be packaged together.
The best way I can draw an analogy here is just thinking back to computers of years past. Many of us remember the late 1990’s and early 2000’s when a desktop computer would have a motherboard. There would then be external caches, memory, drives, network cards, GPUs, sound cards, storage adapters, and so forth. If you have seen our TinyMiniMicro series looking at 1L corporate desktop PCs, then you will see systems without these cards sticking out from a large motherboard. Over time, common features get absorbed into larger packages. At the same time, that integration means the amount of connectors on motherboards or those 1L PCs has gone up and that is why adding off-package connectivity is important. Remember, Intel Lightbender will have the ability to reach off-package and connect things like stacks of memory or other chips so Intel needs this capability.
Universal Chiplet Interconnect Express UCIe 1.0 Usage Models Universal Chiplet Interconnect Express UCIe 1.0 Usage Models
The bottom line is that this needs to happen in the industry. While companies are talking about handheld devices, perhaps the big driving force these days is in servers. The large cloud providers want to customize silicon for specific roles in their infrastructure.

Final Words​

Looking ahead, it will impact the way we purchase servers and other devices in the future. Having a Lenovo or a HPE Intel Xeon may be different in the future. Folks have been very charged about Lenovo Vendor Locking Ryzen CPUs with AMD PSB, and both Dell and Lenovo doing the same on servers. Chiplets offer a path for fully locking packages and functionalities to certain systems, and creating either more or less interoperability between vendors’ solutions. To us, the one aspect that needs to be clearly addressed in this is defining models for reusability since that circular economy piece has the ability to create or divert large amounts of e-waste.
Overall UCIe is a step into the chiplet world that is needed. There is still work to do, but having standards is important. Not only will it shift how chips are designed, but it will also shift the value chain for those providing, IP, foundry, and packaging services. One of the exciting developments here is that we are nearing an era with many more chip designs than the era we are exiting.
lg.php


Yes, the computer industry has understood the need for interoperability for many years, eg, internet, USB have standard interfaces, and, looking at that star cast, you'd want to be in the chorus line if you want to be in the production.
 
  • Like
  • Love
Reactions: 10 users

Diogenese

Top 20
Yes, the computer industry has understood the need for interoperability for many years, eg, internet, USB have standard interfaces, and, looking at that star cast, you'd want to be in the chorus line if you want to be in the production.
... not that Akida will be in the chorus line ...
 
  • Fire
  • Like
  • Love
Reactions: 6 users
Hi Ff,

A little while ago I did speculate that BRN could provide the Akida simulation software (MataTF/ADE) as a commercial product which could be installed on a standard CPU/GPU to provide some of the speed and power benefits of the SoC without needing to make any hardware changes ... and now you produce this:

https://www.edge-ai-vision.com/companies/brainchip/
The Company provides software and hardware solutions that address the high-performance requirements in civil surveillance, gaming, financial technology, cybersecurity, ADAS, autonomous vehicles, and other advanced vision systems.

...
so it's obvious they are quite willing to take up my ideas.
If not there would be thousands in the street rioting.

Well maybe not thousands but a lot well at least a few hundred if it’s fine.

Of course rain is predicted all this week in Sydney but I am sure Blind Freddie will wave his stick angrily as he goes past.

What not if it’s wet.

Ok then nobody will actually turn out but their absence will be felt. 😎FF

AKIDA BALLISTA
 
  • Haha
  • Like
Reactions: 5 users

Diogenese

Top 20
The last Brainchip perspectives paper by Rob Telson provides part of the answer to improved battery performance/life due to AKIDA.

My opinion only DYOR
FF

AKIDA BALLISTA
When you read all the news about partnerships an applications for Akida, and then look at the share price, you have to think:

"The whole world is mad - save me and thee, and sometimes I have my doubts ..."

If we are in a parallel universe, I don't want to go back to the old one.
 
  • Like
  • Love
  • Fire
Reactions: 19 users
“…they have to find extra space which I believe is already occurring in Perth.”
Is PVDM looking for more office space?
Yes.
 
  • Like
  • Love
  • Fire
Reactions: 14 users

Diogenese

Top 20
When you read all the news about partnerships an applications for Akida, and then look at the share price, you have to think:

"The whole world is mad - save me and thee, and sometimes I have my doubts ..."

If we are in a parallel universe, I don't want to go back to the old one.
I'm trying to develop a champagne pouring system so that it only dispensed a mouthfull at a time so you get the fizz all the time.
 
  • Haha
  • Like
  • Love
Reactions: 15 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
I'm trying to develop a champagne pouring system so that it only dispensed a mouthfull at a time so you get the fizz all the time.


Sorry, I already beat you to it.


champagne-drinking.gif
 
  • Haha
  • Like
  • Love
Reactions: 21 users

Proga

Regular
It's the way I read it JK. I think the SP will be north of $20 before we see a dividend and that won't happen until 2026/27 at the earliest - Proga
--------------------------------------------------------------------

GEEZE.........$20 SP equals a $36 BILLION MC and no Dividend being paid before that? Seriously hope thats not reality by any stretch of imagination!

Our cost are quite small and while getting bigger quickly would not be capable of outstripping the Revenue being generated so a Dividend could be paid before we have a $20 SP. Something wrong if not so.
At $20 we should well be doing a $2 Billion Revenue (my guestimate) and therefore costs would still be just a fraction.
There is a reason why they chose the ARM business Model of IP Licensing.
98% Profit.

As for comparing Brainchip divs to CSL divs thats just wrong period! Totally different business and company. PLUS BRN is an absolute game changer in its business model and probably will be studied and used as a case model in teaching at Universities.
Also I doubt we could hit $20 SP without solid multi Billion dollar Revenue. Hope to be proven wrong and it does it under $1 Bil revenue.

My opinions only, form your own please! These ones are copyrighted and wrong any way according to the wife!

Yak52
They'll want to build a hefty bank first before paying dividends. It's the prudent thing to do just in case any take over targets emerge or for a rainy day. The reason I speculated on $20 is because once revenues start flowing in, it will have dawned on investors what a unique position BRN is in and the potential for explosive growth ie the only product on the market/default standard and how many companies are working on developing applications using Akida.

Revenue is going to emerge BRN from the shadows and leave the doubters with nowhere to go.
 
Last edited:
  • Like
  • Fire
  • Love
Reactions: 28 users

Diogenese

Top 20
I'm trying to develop a champagne pouring system so that it only dispensed a mouthfull at a time so you get the fizz all the time.
... my reason being that the bubbles lift the alcohol directly to the brain without any diversions.
 
  • Haha
  • Like
  • Love
Reactions: 14 users

Diogenese

Top 20
  • Haha
  • Like
Reactions: 3 users
We knew this already but it is great to see academics catching on to what Peter van der Made proposed over 20 years ago:

“To conclude, the presented results contribute to a growing body of research addressing the open question of how infor- mation can be encoded in the temporal domain of spikes. In contrast to rate-based algorithms, purely spike time-dependent methods have the potential of enabling highly energy effi- cient realizations of artificial intelligence algorithms for edge applications, such as autonomous decision making and data processing in uncrewed spacecraft [41]. Thus, by making novel data structures, methods and applications accessible to SNNs, the presented results are especially interesting for upcoming neuromorphic technologies [40], [42]–[48], unlocking the po- tential of joining event-based computing and relational data to build powerful and energy efficient artificial intelligence applications and neuro-symbolic reasoning systems.
ACKNOWLEDGMENT
The author thanks Alexander Hadjiivanov and Gabriele Meoni for useful discussions and feedback on the paper. He further thanks his colleagues at ESA’s Advanced Concepts Team for their ongoing support. This research made use of the open source libraries PyKEEN [49], libKGE [18], PyTorch [50], Matplotlib [51], NumPy [52] and SciPy [53].”

My opinion only DYOR
FF

AKIDA BALLISTA
This morning at about the same time as posting the above I sent it to Tony Dawe for information no reply required.

Just a short time ago I received the following:


Hi Fact Finder

Peter has responded to your email. Please below.

Regards

Tony
Get Outlook for iOS

From: Peter van der Made
Sent: Sunday, May 22, 2022 1:30:50 PM
To: Tony Dawe <tdawe@brainchip.com>
Subject:Re: BRN

Hi Tony
This is the published paper;

A Hardware-Deployable Neuromorphic Solution for Encoding and Classification of Electronic Nose Data​

by
Anup Vanarse
1,* ,
Adam Osseiran
1 ,
Alexander Rassau
1 and
Peter van der Made
2


1
School of Engineering, Edith Cowan University, 6027 Perth, Australia
2
Brainchip Inc., Aliso Viejo, CA 92656, USA
*
Author to whom correspondence should be addressed.
Link:
https://www.mdpi.com/1424-8220/19/22/4831
Best regards
Peter

Get Outlook for iOS

From: Peter van der Made
Sent: Sunday, May 22, 2022 12:59:43 PM
To: Tony Dawe <tdawe@brainchip.com>
Subject:Re: BRN

Hi Tony
Anup has developed the perfect encoding technique for data called Aero. He has published this in his paper on olfactory sensing and classification.
Best regards Peter”

Not that Tony or Peter needed to reply but the paper was published back in late 2019. For those who have a calculator and can subtract 2019 from 2022 we get 2.5 years. If you read the paper I posted this morning you will see that this paper was only recently published and they are still chasing what Peter and Anup resolved 2.5 years ago.

I think Tony and Peter are a great match they are always engaged with the Brainchip vision and mission.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Love
  • Fire
Reactions: 52 users
D

Deleted member 118

Guest
I'm trying to develop a champagne pouring system so that it only dispensed a mouthfull at a time so you get the fizz all the time.
Have you tried drinking it straight from the bottle? As that seems to work

 
  • Haha
  • Like
  • Thinking
Reactions: 10 users
... my reason being that the bubbles lift the alcohol directly to the brain without any diversions.
Yes in these economic times your invention when perfected will allow more people to get drunk on every bottle of champagne. The developing world has been crying out for this for years. 🧐 FF
 
  • Haha
  • Like
Reactions: 12 users

Tony Coles

Regular
Recent tweet from 10 hrs ago…Followed by BrainChip! Read the article and watch the 2 minute video, using AI and machine learning. Interesting, see what the 1000 👀 think of this.

 
  • Wow
  • Haha
  • Like
Reactions: 6 users

Dozzaman1977

Regular
I think this is now more nuanced than when first stated by the CEO Sean Hehir.

In his last interview that I listened to he was asked in question time about giving income guidance and he said words to the effect ‘That is not something that we are doing but it is something which the Board has agreed to review/discuss at the beginning of 2023.’

My present view is that Sean Hehir has stepped back from ‘break even’ end 2022 which I personally thought was a very bold prediction because of what appears to be explosive growth going on around Brainchip, the major publicity drive orchestrated by Jerome Nadel and the burgeoning market that is opening up as a result of government legislation and the Von Neumann bandwidth bottleneck.

These things are known unknowns which I would consider impossible to properly cost.

A simple example of what I mean is Brainchip Engineering Support for Customers.

In the past Brainchip dealt with this by saying we are putting a cap on the number of EAP customers at about 2 dozen.

If as Brainchip now is bent on becoming ubiquitous and the default standard they cannot cap the number of customer engagements.

So being conservative let’s say between now and beginning of fourth quarter Brainchip has 200 new customer engagements all requiring engineering support then this must blow out the windows on only having about 100 employees by end 2022.

If Brainchip has to find extra employees then they have to find extra space which I believe is already occurring in Perth.

In business when success takes hold it is like hanging onto a bull by the tail. I have been there.

This is not a bad thing, do not get me wrong, but true explosive growth is the wildest ride you can be on in business and trying to do forward budgets is near impossible.

The great thing about the growth is that even though you are paying out money left right and centre and have no idea how much you will need from minute to minute YOU CAN VALUE YOUR WORK IN PROGRESS and start to make accurate future earnings predictions with high accuracy.

If I am reading things correctly not reaching break even by end of 2022 as a result of growth is the best news shareholders could have. Time will tell.

My opinion only DYOR
FF

AKIDA BALLISTA
Hi Fact Finder and Yak52


All in my opinion of course
Sean Hehir interview with the strawman 28/03/2022

At the 42.18 minute mark the Strawman asked a question basically when will brainchip be in a position of breakeven .

Quote Seans answer "Remember when i talked building up to 100 people or so this year, we think at that point we actually at a pretty good size scale and were are going to say after that the revenue growth will start to outgrow the expense growth pretty rapidly, thats the model that were going to start to see in leverage"

Now the captions in the interview at this time says "start to grow the expense growth" but if you listen closely to Sean in the interview he says "start to outgrow the expense growth"

He also states " He has a model"
He also states "Line of sight"

I believe this is the correct time frame for when Brainchip will reach the breakeven (from end 2022 into early 2023) if nothing out of the ordinary has happened internally since
IMO Sean would have a Strategic business plan in place with a timeline of objectives based on internal information including hiring a CMO and business expansion etc etc

Things change thats business, but going to trust what our CEO has stated in the interview which was less than 2 months old.
So for the first time Fact finder I'm going to go with my point of view over yours, sorry!!!!!!!!!!!!!!!!!!!!!!!! (Dont hold it against Me)
 
  • Like
  • Fire
  • Love
Reactions: 36 users

Tony Coles

Regular
“…they have to find extra space which I believe is already occurring in Perth.”
Is PVDM looking for more office space?
YEAH! off cause more office space, its for the billions of chips to distribute to customers, Akida Everything/Everywhere.
 
  • Like
  • Love
Reactions: 7 users

Slade

Top 20
We know that Mercedes adopted Akida and Sonatic for their voice response system in the EQXX. I found it interesting the work sonantic has done to generate Val Kilmer’s voice. Val now struggles to speak after operations for throat cancer.

 
  • Like
  • Fire
  • Love
Reactions: 11 users

Bravo

If ARM was an arm, BRN would be its biceps💪!
Magnums to boot ..
... when's your next party?

It'll be directly after we receive the next price sensitive announcement. If it's a good announcement, I'll drink out of the bottle as demonstrated, but if it's oober-good, I'll pull my leprechaun boots out of the wardrobe and drink out of them instead. 🥳
 
  • Haha
  • Like
  • Love
Reactions: 17 users
Top Bottom