HPC’s lost stories will drive the future of technology • The Register

opinion The first indications of the fall of the empire only become clear in hindsight. By the time they appear, they can seem like they’re just part of existing trends. Crypto Chaos as FTX falls apart like a giant rotting peach? The whole scene stinks. Everyone knows it, except the brands.

Nvidia having a hard time? Who isn’t in the tough silicon supply chain slog party? Check out Intel. look at arm Oh, and the world just hit 8 billion people.

What has made Intel, Nvidia and ARM so prominent in the first place is the economies of scale where their products are connected to life around the world. More than 85 percent of people on earth with a smartphone, all connected to a globe-spanning infrastructure is as big as you can get. While Moore’s Law makes faster-equals-better – FEB – the easiest path, the designs with the largest market share will pull away, leaving everyone else in the lurch. Then suddenly you hit the 8 billion body barrier.

There are no more worlds for conventional technology to conquer. As a species, we may no longer need personal CPU cycles, nor megabits per second, nor terabytes of memory. As Moore’s Law fades, the FEB will become more expensive even if its attractiveness decreases. There is no new mass market driver to adopt.

Take the crypto experiment that gorges on spicy silicon like a hungry school kid given a bowl of ramen. Turns out it’s not tech-economy brawn, it’s a tumor hogging resources in exchange for criminal toxicity. And if Krypto is a tumor, then the best thing to say about the metaverse is just a nasty little rash.

Economies of scale have hit falling yields. For example, there will always be a market for top-end GPUs, but they are far less compelling now, as imperceptible increases in performance are bought at the expense of ruthless increases in power consumption. Joules per frame may drop with each generation, but the battle for dominance in the spec sheet is pulling overall performance back up and up and up. Nvidia’s flagship RTX 4090 can draw almost 800 watts during peak chat. For what? The smell of melting plastic? It looks like a bad place.

These things are unsustainable and undesirable and there is no other way. Like Concorde, the model where ever-increasing performance trumps all other considerations is flawed. In an economy and environment where resource constraints are both inevitable and desirable, the model quickly falls apart.

This has some interesting and exciting consequences. While our personal FEB equation is reversed, it remains the absolute rule for high performance computing. This plays a very important role in the future of all of us.

HPC has had a free ride on the back of many decades of consumer and enterprise FEB. As industry-standard architectures became all but better than the various technologies that powered older HPCs, the sector saw prices fall, software support swell, and development times shrink. The core compute, communications, and storage components could be side-ordered, making it worthwhile dedicating resources to scalability optimization and larger architectural issues.

That’s great, but it leads to a monoculture where innovation is an easily avoided risk. The latest top 500 supercomputers list is 95 percent Intel/AMD X86, because that’s where almost all of the world’s CPU power budget has gone – the hunt for FEB for everyone. The same goes for data center GPUs, which wouldn’t have happened without gamers choking on 4k@60fps.

This driver is what opens the door to competitive advantage through radical reinvention of core technologies. We can see signs of this from companies like Cerebras, which view wafer-scale integration — WSI — as some kind of supersteroidal SoC with commensurate efficiency and speed benefits. How very 21st century – only WSI first existed in the 1980s at Sinclair Research. Yes, this Sinclair Research.

Go back to the 80’s and 90’s, before the X86 hordes rolled across the world, and you can find a hundred promising new technologies on the edge of performance computing. To cite just one example among many, Cray’s massively parallel XMT architecture originated in Spooksville in the early 2000s, just as the Intel/AMD juggernaut was gaining momentum. Unable to match the scale and R&D spending, XMT quietly disappeared. But that R&D has now run out and the factories are waiting. What could XMT do in today’s clothing?

It’s not just silicon; Superconducting logic using quantum engineering for classic combinatorial designs was hot. In a very cold way. Whole device families chilled in the laboratory. We’re much better at this type of technique now, so what could be rediscovered today to have practical benefit?

Expect a revival of the ’80s and ’90s as researchers apply today’s techniques to the most promising ideas of the past. Check out reports from years past supercomputing conferences where many exotic ideas were presented. When empires stop growing, they stagnate and die, and then revolutionaries use what’s left to remake the world. HPC will be the starting point. ®

https://www.theregister.com/2022/11/21/hpc_opinion_column/ HPC’s lost stories will drive the future of technology • The Register

Rick Schindler

World Time Todays is an automatic aggregator of the all world’s media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, all materials to their authors. If you are the owner of the content and do not want us to publish your materials, please contact us by email – admin@worldtimetodays.com. The content will be deleted within 24 hours.

Related Articles

Back to top button