Thursday, November 26, 2020

Wafer Scale Engine Computer: Faster than Reality

The 1st generation WSE 1.2 trillion transistor chip (~8.5 x 8.5 inches) -- 
the 2nd generation chip will have 2.6 trillion transistors, 
850,00 cores and more than twice the memory


An article at the SungularityHub, The Trillion-Transistor Chip That Just Left a Supercomputer in the Dust, describes what seems to be the first actualization of an old idea in computer chips. The idea is to make computer processing chips bigger, not smaller. So far, all the innovation has gone into making chips and components smaller and smaller and smaller. At present, billions of transistors can be put on a small chip as shown above. The new wafer-scale engine (WSE) takes existing miniaturization technology to put a trillion transistors on a big chip. The big chip is made by Cerebras, a California startup company.

“The Cerebras Wafer-Scale Engine is massive any way you slice it. The chip is 8.5 inches to a side and houses 1.2 trillion transistors. The next biggest chip, NVIDIA’s A100 GPU, measures an inch to a side and has a mere 54 billion transistors. The former is new, largely untested and, so far, one-of-a-kind. The latter is well-loved, mass-produced, and has taken over the world of AI and supercomputing in the last decade.

When Cerebras first came out of stealth last year, the company said it could significantly speed up the training of deep learning models.

Since then, the WSE has made its way into a handful of supercomputing labs, where the company’s customers are putting it through its paces. One of those labs, the National Energy Technology Laboratory, is looking to see what it can do beyond AI.

So, in a recent trial, researchers pitted the chip—which is housed in an all-in-one system about the size of a dorm room mini-fridge called the CS-1—against a supercomputer in a fluid dynamics simulation. Simulating the movement of fluids is a common supercomputer application useful for solving complex problems like weather forecasting and airplane wing design.

The trial was described in a preprint paper written by a team led by Cerebras’s Michael James and NETL’s Dirk Van Essendelft and presented at the supercomputing conference SC20 this week. The team said the CS-1 completed a simulation of combustion in a power plant roughly 200 times faster than it took the Joule 2.0 supercomputer to do a similar task.

The CS-1 was actually faster-than-real-time. As Cerebrus wrote in a blog post, ‘It can tell you what is going to happen in the future faster than the laws of physics produce the same result.’

The researchers said the CS-1’s performance couldn’t be matched by any number of CPUs and GPUs. And CEO and cofounder Andrew Feldman told VentureBeat that would be true “no matter how large the supercomputer is.” At a point, scaling a supercomputer like Joule no longer produces better results in this kind of problem. That’s why Joule’s simulation speed peaked at 16,384 cores, a fraction of its total 86,400 cores.

A comparison of the two machines drives the point home. Joule is the 81st fastest supercomputer in the world, takes up dozens of server racks, consumes up to 450 kilowatts of power, and required tens of millions of dollars to build. The CS-1, by comparison, fits in a third of a server rack, consumes 20 kilowatts of power, and sells for a few million dollars. 

Computer chips begin life on a big piece of silicon called a wafer. Multiple chips are etched onto the same wafer and then the wafer is cut into individual chips. While the WSE is also etched onto a silicon wafer, the wafer is left intact as a single, operating unit. This wafer-scale chip contains almost 400,000 processing cores. Each core is connected to its own dedicated memory and its four neighboring cores.”

What does all that mean?
What that means is that there is a new generation computer technology that can do some things better than existing supercomputers. It's simulations of events can be faster than real time, allowing predicting and acting in advance of future events. At present, the things that WSE dowes best relate to solving specific, highly complex problems that require vast amounts of computing power. WSE will not replace existing technology like the ipad or laptop, which are designed for general uses and generally work quite well. As with most or all new technologies, this can be, and probably will be, used for good and bad. 

WSE excels at doing high speed simulations in real time. It can simulate and at least partially automate aircraft landings. It works faster to train artificial intelligence software than current supercomputers. WSE can be used to train neural networks, simulating brain data processing. Since this is still early days in WSE technology development, it is not known how influential this will become. Competing technologies include quantum computers and memristor-based neuromorphic chips, that mimic the brain by putting processing and memory into individual transistor-like components. 

My guess is that WSE will be tested in stock markets to see if the future can be predicted far enough out to trade on a stock before it moves up or down. It also seems reasonable to think that WSE will be tested in medical situations where real time computing can help in diagnosis or predicting future medical problems using artificial intelligence (AI) technology. AI is used in medicine for a growing number of applications.
 
Lots more cores, memory and bandwidth -- more is better!
(A PB is a petabyte = 2 to the 50th power of bytes; 
1,024 terabytes (TB) = 1 petabyte, or 1 million gigabytes; for comparison, 
human brain data processing operates unconsciously at about 
1.4 million bytes/second and about 1-60 bytes/second consciously)


No comments:

Post a Comment