Hey Brother, sometimes I think you know how I'm going to react before asking the question Sir ;-)

So here, have a peek at this lot... It should keep you entertained.. Others may think it's all BS, but I can't help that or them... DYOR

Whether it is 4DS or simply the reram tech derivative (non fil, mott trans, interface switching) it possibly has already identified itself as useful a version of the DoD

See back around 2016 - 2017, the yr you mentioned Ken in the above post Donald J had just passed an edit that the US wasn't going to get beat by the Chinese in the Super computer race...

And although they wanted to win, the changing rhetoric around global warming meant that energy consumption had to be reduced massively in order to be considered a success...

Enter some new players to the HPC arena in the faces of Nvidia & Tesla... Now look at their SP appreciation in the period 2016-17 to 2021 while I put some slides together.

2016 & Oakridge let slip some details of the upcoming Summit supercomputer build..

And note the change in spec between the old Titan and the new Summit builds... Notably the addition of 800GB of nvMe



Also the Processors for the first time were new Tesla Volta units..

Important features available in the “Volta” GPU architecture include:

• Exceptional HPC performance with up to 8.2 TFLOPS double- and 16.4 TFLOPS single-precision floating-point performance.
• Deep Learning training performance with up to 130 TFLOPS FP16 half-precision floating-point performance.
• Deep Learning inference performance with up to 62.8 TeraOPS INT8 8-bit integer performance.
• Simultaneous execution of FP32 and INT32 operations improves the overall computational throughput of the GPU
• NVLink enables an 8~10X increase in bandwidth between the Tesla GPUs and from GPUs to supported system CPUs (compared with PCI-E).
• High-bandwidth HBM2 memory provides a 3X improvement in memory performance compared to previous-generation GPUs.
• Enhanced Unified Memory allows GPU applications to directly access the memory of all GPUs as well as all of system memory (up to 512TB).
• Native ECC Memory detects and corrects memory errors without any capacity or performance overhead.
• Combined L1 Cache and Shared Memory provides additional flexibility and higher performance than Pascal.
• Cooperative Groups – a new programming model introduced in CUDA 9 for organizing groups of communicating threads

Basically it was a new design and the OSTI mob below who reported on it are the same mob in a previous post that said build it this way if you want to succeed...




Which means the Reram stack they refer to as being the preferred option had to have all the characteristics they described,,



And which makes this next point critical to the argument of why I believe that A... 4DS is far more advanced than any give it credit for & why the cone of silence was so great that nothing could escape the supposed NDA..

Quote: Timothy Morgan
"if Nvidia can reach its original goal of 32 GB of HBM memory per GPU accelerator it hoped to hit with Pascal on the Voltas, that works out to 192 GB of HBM memory with 6 TB/sec of bandwidth. That is a lot more than the 64 GB of HBM memory and aggregate 2.8 TB/sec of GPU memory bandwidth in the current “Minsky” Power Systems LC precursor to the Summit’s Witherspoon node. There is another 800 GB of non-volatile memory in the Summit node, and we are pretty sure it is not Intel’s 3D XPoint memory and we would guess it is flash capacity (probably NVM-Express drives) from Seagate Technologies but Oak Ridge has not said. The math works with this scenario, with 512 GB of DDR4 main memory, a total of 192 GB of HBM memory on the GPUs and 800 GB of flash, across 4,600 nodes that is a total of 6.9 TB of aggregate memory. (By the way, that chart has an error. The “Titan” supercomputer has 32 GB of DDR3 memory plus 6 GB of GDDR5 memory per node to reach a total of 693 TB of aggregate memory.)"

And here is the first challenge brother, for you or for anyone... Find 8tey the name of the supplier of the 800GB of nvMe quoted and I will never mention the theory again & will apologize to Hash for doubting his claims..

Maybe start at the Oakridge official page and check out the latest Summit Computer stats... It's still described as a Hybrid build using both IBM & NVIDIA GPU's via an NV link and still with 800GB of non volatile RAM !! per node.. Although it does also quote 1.6Tb nvm/node in the specs list so I assume its per IBM or NVIDIA node to produce a combined total.. Anyone know otherwise?


So you have the OSTI mob saying build it like this and that they think Reram is the best bet, of which Mott Transition showed the greatest promise but was the least known, if any care to remember an earlier post... And this is where being on the spectrum (Autism not F;wit hehe) helps..

You remember a lot of seemingly trivial info at the time that often appears relevant later...

You can also pick BS artists a mile away but that is a story for another day..

So in light of what we know the gov has done and in absence of any idea what 4DS has in the same time period the speculative conspiracy theorist might ask was it an NDA or NSA order the company was sworn to abide by... ?? A race against the Chinese would constitute as of national importance one would guess... hehe



Anyway who cares who bankrolled the thing and/or made Tesla/Nvidia two very wealthy companies in the process.. But do yourselves a favor and have a look at the Oakridge website.


The thing is that at least it proved the new memory design concept using burst buffer, or extended memory concept worked..

So how fast do we think the burst buffer would need to be to carry out the task it was built to perform? It would have to be at the same speed as the GPU or very close to it wouldn't it?

Which leaves Mram or Reram as the prime candidates and if we recall the OSTI boffins seemed to think MRAM wasn't quite the best choice..

Now tell me who has developed and tested the fastest known Reram product that does not require error correction tech and can still theoretically scale to terabit on chip at the smallest geometries. ??

And which now is formed using a fully crystalline PCMO material... Similar to the one the Summit computer was used to research the qualities of last yr.. Probably nothing to do with what we are discussing but they do sound promising don't they? A formula check on diff types could prove interesting if anyone is bored?

4DS has for the first time demonstrated fabrication of fully crystalline Pr1-xCaxMnO3 (“PCMO”) at temperatures compatible with the advanced processes run in today’s leading-edge high-volume memory DRAM and NAND factories;

From Oakridge Website..

What are perovskites and their applications

Named after 19ty century Russian mineralogist Lev Perovski, perovskite materials have a crystal-lattice structure of inorganic molecules like that of ceramics, along with organic molecules that are interlaced throughout.

Perovskites are a family of crystals that show promising properties for applications in nanotechnology, especially nanostructured solar cells. ‘Perovskite’ is a term used to describe a group of materials that have a distinctive crystal structure of cuboid and diamond shapes. They have long been of interest for their superconducting, electronic and ferroelectric properties.

Since it was discovered that they are also remarkably efficient at absorbing photons of light and that this can be converted into an electric current, research interest in perovskite solar cells has skyrocketed.

PowerPoint Presentation (epfl.ch)

So yeah Brabs.... I guess there is enough coincidence for me, personally, to believe that there are plenty of uses for the tech the company holds..

& that the US Gov funded think tanks for 10 years to arrive at the same conclusion

& that someone bankrolled the development of Summit and fitted it with an unknown burst buffer non volatile RAM

& that this proved the high speed theory of the new memory stack using an array memory macro in support of some high bandwidth computing

& was compatible with Nvidia & IBM GPU's at full speed

& that whoever supplied the tech or the product has a possible jump on the market should they ever want to venture outside of official US gov operations.

& that maybe it already has cause Elon has an even bigger better faster computer to support his AI aspirations which mean maybe Von Neuman aint such a block anymore.

But other than that my friend... I don't know shite... ;-)

I mean how could anyone ?

But for what I do know I will thank you for you have prompted a lot of the curious egg hunts I have been on along the way..

WD License, Samsung Purchase.

Calling it now, happening soon...

IMHO

8tey