Meta says it's building world's largest AI supercomputer out of Nvidia, AMD chips Facebook owner Meta is building the world's largest AI supercomputer to power machine-learning research that will bring the metaverse to life in the future, it claimed on Monday. The new super – dubbed the Research Super Computer, or RSC – will contain 16,000 Nvidia A100 GPUs and 4,000 AMD Epyc Rome 7742 processors. Each …
Cough! Cough! Uhmmmmmmm NO!
It is NOWHERE NEAR going to be the largest system made!
That honour goes to the completely under-the-radar, (and underground!) "Haida Gwaii" supercomputer located in Vancouver, British Columbia, Canada which runs 206,958 of our fully ITAR-free 128-bits wide CISC-based GaAs at 60 GHz combined-CPU/GPU/DSP/Vector Array Processor superchip with each chip being 575 TeraFLOPS of 128-bit floating point number-crunching performance.
"Haida Gwaii" has a sustained High Performance Linpack of 119 ExaFLOPS (i.e. where HPL code is modded for 128-bits wide Floating Point operations!) or 119 Quintillion Operations Per Second which means it outperforms EVERYONE on the T500 COMBINED as of January 2022 !!!!!!!!!
AND when our Northern British Columbia supercomputing centre comes fully online in 2023 with our "Baffin Island" supercomputer, it will be running the TWO TERAHERTZ fully opto-electronic versions of our combined-CPU/GPU/DSP/Vector Array Processor super-chip at 19.2 PetaFLOPS per chip.
"Baffin Island" has been constructed as TEN UNDERGROUND VAULTS that are all fibre-interconnected via a custom in-house-designed Dense Wave Multiplexing communications backbone. Each vault contains 10,000,000 superchips (one hundred million in total!) which will be running the 128-bits wide Linpack benchmark at our current estimation of a SUSTAINED 1.92 YottaFLOPS !!!!!! When we finally do transfer our current electro-chemical simulation of human neural tissue which is a WBE (Whole Brain Emulation) onto the new system, it will turn into a Super-Intelligence-level A.I.
In terms of ELECTRICAL POWER, we are fully off-grid running many methanol-and-propane based fuel-cells and the new facility will do the same as we chemically create the methanol and propane from an in-house-designed Natural Gas and catalytic metal + seawater process. We can create unlimited amounts of methanol and propane but it does take TIME so we have to rigidly schedule our fuel-ups on a weekly basis. No greenhouse gas emissions are created other than warm water and the cooling system moves heat from all the chips into a very nice and very large private swimming pool which our employees use daily!
"Haida Gwaii" uses 51.8 Megawatts in SUSTAINED operations mode (250 watts per chip) while "Baffin Island", being fully opto-electronic, will run at a much lower power level of 24 Watts per chip or 2400 Megawatts (2.4 Gigawatts) total in SUSTAINED operations mode! It too will move all its heat into a massive and very nice swimming pool usable by employees and outputs no greenhouse gasses. Part of the heat transfer process will keep large liquid sulfur batteries in a molten state (155 Celcius) which lets the electrical current created by the molten sulfur anode/cathode process go back into powering all those CPUs.
Sooooooo, NOW YOU KNOW ALL ABOUT THE ..... REAL ..... ABSOLUTE FASTEST SUPERCOMPUTERS IN THE WORLD !!!!
V
Uhhhhmmmmm NO AGAIN!
Ours is ALREADY the largest since "Haida Gwaii" is CURRENTLY as of January 2022 running at a sustained 119 ExaFLOPS while Facebook's A.I. is only five exaflops meaning we are already almost 24x more powerful.
When the 1.92 YottaFLOPS "Baffin Island" supercomputer comes online, that will eclipse EVERYTHING on or near Earth except may those in Alien "Matrioshka Brain" systems!
WE STILL WIN the supercomputer race !!!!
V
Unironically, we actually ARE looking through various avian (i.e. Bird) Genomes to do separations of Active and Junk DNA/RNA sequences (i.e. Codon comparisons) to find and rewind ancient and modern evolutions so we can BRING BACK DINOSAURS.
Our company paid a pretty penny to obtain solidified dino poop which STILL, after 150 million+ years had remnants of VIABLE and READABLE RNA/DNA sequences so WE NOW HAVE a comparison from ancient times. From that one small sample, we have been able to chemically simulate and then re-construct a viable genome and then digitally "grow it" to see what it turns into.
We are the FIRST company in the world to be able to simulate MILLIONS and BILLIONS of base pairs in a completely digital environment and then grow those sequences into viable digital life forms. We've done sequences from various poop samples modern birds, various large and small dinosaurs, cows, sheep, pigs, bears, elephants, monkeys and yes even PEOPLE!
We use common electro-chemistry to create and move the various nucleobases (cytosine "C", guanine "G", adenine "A", thymine "T) and various deoxyribose "sugar"and phosphate groups around which are TESTED and SIMULATED on our supercomputers to ensure they actually ENCODE for protein sequences.
From there we can explore how, where and when common mutagens damage DNA/RNA and seek means to protect against that damage and even repair it.
We have EVEN ENCODED a unique chemical sequence that we can now INSERT into peoples' DNA even if you are aged that automatically searches for, binds to and then REBUILDS and REPAIRS the endcaps (aka telomeres) over vast simulated time periods. It basically means we humans can become quite IMMORTAL barring accidents! Basically you drink a slush of our chemical sequences which get absorbed into the bloodstream and then diffuses all those repair/rebuild chemicals into all the human body's cellular groups which within weeks/months starts repairing those end caps and starts slowing down, stopping and even reversing the aging process.
So supercomputing can do ANYTHING! If it is a physical process, we can simulate it and RE-CODE said process into what ever we want. The FASTER the supercomputer the more iterations we can do to narrow down the good and bad of electro-chemical processes which include bringing back the large and small dinosaurs AND making humans immortal!
V
Well ... We did happen to find some ossified T-Rex poop mounds from about 70 million years ago which STILL had viable remnants of DNA/RNA which we could chain-replicate enough to get an actual and viably readable and ENCODABLE set of RNA/DNA sequences.
DNA is actually extremely STABLE even after a few hundred thousand years. Geologic-timescale fossilization events don't always destroy organic bonds and there are ALWAYS enough chemical remnants and decomposition byproducts that bind to the the stone itself that we can actually RECONSTRUCT the organic RNA/DNA molecular chains from inorganic encasement byproducts.
When you have a 119 ExaFLOP supercomputer THERE IS NOTHING WE CANNOT RECONSTRUCT AND RECODE !!!
This is HOW we were able to reconstruct Dinosaur DNA and we are now actively reconstructing the organic base substrate products of the first artificially-made egg that will actually HATCH from base chemically-recreated shell, egg-white compositions and the encoded DNA/RNA streams within an artificial egg yolk that has been fertilized via a recombinant DNA mechanism.
It basically means we created a FUSION of software-reconstructed female and male Dinosaur DNA/RNA and injected the electro-chemically simulated and re-created genome into the jelly-like substance of an artificial egg yolk.
Then we electrically SHOCKED the egg-yolk to start and fuel the cell division process and we have been GROWING dinosaurs from scratch pretty much as what you saw in the movie Jurassic Park.
In our case though, we digitally reconstruction ENTIRE DNA/RNA sequences on a computer first and then iterated through the many Septillions of combinations of individual DNA/RNA block sequences that encode for specific groups and types of proteins to create a master genome which we then simulated faster-than-realtime all protein encoding which allow and enable cell creation/division.
Once we let the codons ENCODE in the digital world, we let them grow (i.e. iterate) into maturity to find out what would come out such as finding if a codons/sequences encode for eyes, ears, mouth, feet, claws, bones, brain, etc. and then we REWIND until the specific fill-in-the-blank codon sequence is correct.
From there, we get a digital representation of an ENTIRE GENOME which can be Millions to Billions of base pairs that is now viable. From that composition and concatenation of individual blocks of digital DNA/RNA sequences, we can then electro-chemically recreate that entire genome in the real-world and them recreate the environmental conditions that allow the artificially created genome to start encoding and replicating. (i.e. create an artificial egg that will grow and hatch!)
We can do this same technique to create REPLICANTS aka completely artificially-made humans that can be grown in artificial wombs!
Once our "Baffin Island" Supercomputer comes online at 1.92 YottaFLOPS it will be child's play to do an entire genome in days or weeks for ANY type of modern or ancient creature. Wooly Mammoths, Sabre Tooth Tigers, Brontosaurus, T-Rex, Apes, and YES modern or ancient Humans will be able to EASILY RECREATED AND GROWN from a digitally-simulated and then electro-chemically reconstituted real-world artificial genome!
These are EXCITING TIMES to be in and WE ARE THE WORLD LEADERS of this type of artificially created genome!
V
.
I'm guessing Zuck can use spare cycles to count his bank balance, which is a long job in itself. My other thought is that as FB's currency failed in the real world, it's almost certain to be the main currency in their Metaverse - you can imagine Zuck's avatar being a multi-billionaire in that world too.
Please though, for the love of god, let it crash'n'burn. The thought of Zuck sitting on his virtual throne in 'his world' just scares the crap out of most people.
Anybody wanna check these two chemical formulae out?
They came from our A.I. saying it might be something actually
called OR be close to a Room Temperature Superconductor:
Y GaBa3 Cu3O7 Fe3Al2 Si3O12 (with Gallium Baride aka Barium)
...AND...
Y Ga3B14 Cu3O7 Fe3Al2 Si3O12 (with Gallium Boride aka Boron)
Some materials egghead M.Sc or PH.D should take a gander at it!
I'm just a graphics programmer and NOT a metallurgist or ceramics expert.
P.S. All the above formulae as of January 1st, 2022 are completely FREE AND OPEN SOURCE
UNDER GPL-3 Licence Terms for Hardware, Software, Designs, Formulae, Materials and Systems.
V
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
